IT Forums - Hypertransport group gains TI, IBM

Keith R. Williams

08-12-2003, 07:02 PM

In article <Ggh_a.178308$rsJ.121502
@news04.bloor.is.net.cable.rogers.com>,
bbbl67.spam@yahoo.com.nospam says... "Keith R. Williams" <krw@attglobal.net> wrote in message news:MPG.19a2d73658d8acc7989a29@enews.newsguy.com... -IBM is already using Hypertransport in its PowerPC 64-bit chip on the Apple Macs, inside a chip-to-chip connection inside the chipset Apple is using... The hypertransport bus is used from the G5's "northbridge" (Apple's chip) to the "southbridge". It doesn't touch the processor. The PPC970 bus is "Elastic I/O". <shrug> Maybe after now, Power5 will become Hypertransport? </shrug>

Perhaps wishful thinking? Why bother with Hypertransport on the
processor? What would hypertransport give that Elastic I/O
doesn't? Hypertransport is the I/O bus in the G5. As such it's
plenty fast enough for some time to come.

Power5?
-They join existing members: Apple, Cisco, Sun, Nvidia, Broadcom, Altera, Xilinx, PMC-Sierra, Agilent, etc. -No word whether Intel will join yet Yeah, right. I'm sure Intel can at least get a free membership with Hypertransport's new educational institution membership, through Intel University. :-)

What would make you think that Intel would ever "lower" itself to
accept someone else's standard? That's never been their style.

Again... Yeah, right.

--
Keith

George Macdonald

08-14-2003, 03:38 PM

On Tue, 12 Aug 2003 23:02:37 -0400, Keith R. Williams <krw@attglobal.net>
wrote:
In article <Ggh_a.178308$rsJ.121502@news04.bloor.is.net.cable.rogers.com>,bbbl67.spam@yahoo.com.nospam says... "Keith R. Williams" <krw@attglobal.net> wrote in message news:MPG.19a2d73658d8acc7989a29@enews.newsguy.com... > -IBM is already using Hypertransport in its PowerPC 64-bit chip on the > Apple Macs, inside a chip-to-chip connection inside the chipset Apple is using... The hypertransport bus is used from the G5's "northbridge" (Apple's chip) to the "southbridge". It doesn't touch the processor. The PPC970 bus is "Elastic I/O". <shrug> Maybe after now, Power5 will become Hypertransport? </shrug>Perhaps wishful thinking? Why bother with Hypertransport on theprocessor? What would hypertransport give that Elastic I/Odoesn't? Hypertransport is the I/O bus in the G5. As such it'splenty fast enough for some time to come.Power5? > -They join existing members: Apple, Cisco, Sun, Nvidia, Broadcom, > Altera, Xilinx, PMC-Sierra, Agilent, etc. > -No word whether Intel will join yet Yeah, right. I'm sure Intel can at least get a free membership with Hypertransport's new educational institution membership, through Intel University. :-)What would make you think that Intel would ever "lower" itself toaccept someone else's standard? That's never been their style.

Nah they'd rather pay for external IP and use it to hijack the entire
architecture? Talking of which.... I haven't seen much "comparison" of how
HT measures up against the RaSer stuff. Am I right in thinking our
"friends" (YKW) are in the process of patenting the PCI-Express
implementation?... with Intel's umm, collusion? Is it in fact possible to
do a PCI-Express without paying "them" their license fees? Am I being
paranoid here?:-)

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

Robert Myers

08-17-2003, 08:11 PM

George Macdonald wrote: Nah they'd rather pay for external IP and use it to hijack the entire architecture? Talking of which.... I haven't seen much "comparison" of how HT measures up against the RaSer stuff. Am I right in thinking our "friends" (YKW) are in the process of patenting the PCI-Express implementation?... with Intel's umm, collusion? Is it in fact possible to do a PCI-Express without paying "them" their license fees? Am I being paranoid here?:-)
You know they're up to something, and you would be reasonable in
imagining they're up to no good. They've been all over the place in
terms of I/O technology and schedules. Where was "communication
streaming architecture", the Intel-only Gigabit connection that suddenly
sprang from the ICH5 memory controller in all their press hype? Nowhere
that I can tell. Infiniband? Forget it. It will get 10Gigabit access
to the memory controller when hell freezes over.

That's why I can never understand all the animus against Rambus. If
Intel weren't Intel, nobody would do business with them. They play
hardball just as hard as Rambus ever did. They're just a little
smoother about it.

RM

Yousuf Khan

08-18-2003, 03:01 PM

"Robert Myers" <rmyers@comcast.net> wrote in message
news:AlY%a.43647$2x.16340@rwcrnsc52.ops.asp.att.net... That's why I can never understand all the animus against Rambus. If Intel weren't Intel, nobody would do business with them. They play hardball just as hard as Rambus ever did. They're just a little smoother about it.

That's exactly the reason why, you just answered your own question. Because
you have to do business with Intel, you have to. Because you don't have to
do business with Rambus, you can ignore them. Rambus just got a little ahead
of its actual influence.

Let's face it Intel started out as a nice guy when it was starting out too.
Weren't we all glad when good-guys Intel and Microsoft took away the
decision making power of the PC industry from big-bad meanie IBM, back in
the late 80's/early 90's? You actually wanted to do business with Intel at
that time.

Yousuf Khan

Robert Myers

08-18-2003, 07:26 PM

On Mon, 18 Aug 2003 23:01:13 GMT, "Yousuf Khan"
<bbbl67.spam@yahoo.com.nospam> wrote:

<snip>Let's face it Intel started out as a nice guy when it was starting out too.Weren't we all glad when good-guys Intel and Microsoft took away thedecision making power of the PC industry from big-bad meanie IBM, back inthe late 80's/early 90's? You actually wanted to do business with Intel atthat time.

And now it's IBM to the defense of Linux against former Linux
distributor SCO.

Not hardware-related, but I met a really nice person from M$ with some
degree of responsibility at a high-profile event recently. I told the
M$ staffer that I wouldn't even consider using an M$ product for the
topic of the conference, and the M$ rep asked for my business card.

Soon I got a really nice e-mail from the M$ staffer asking if I would
share my concerns in more detail. I sent back a polite but blunt
e-mail: would M$ please act as if it were a member of a community
rather than the proprietor of it, and would M$ please stop trying to
obstruct peaceful interaction with Linux by sharing details of NT file
format and SMB so that both could be used properly by Linux networking
tools. I also said that, as soon as M$ file formats were open and did
not compromise the rights of owners of intellectual property, I would
stop advising people not to store intellectual property in M$ formats.

I got no reply.

RM

George Macdonald

08-19-2003, 01:11 AM

On Mon, 18 Aug 2003 04:11:12 GMT, Robert Myers <rmyers@comcast.net> wrote:
George Macdonald wrote: Nah they'd rather pay for external IP and use it to hijack the entire architecture? Talking of which.... I haven't seen much "comparison" of how HT measures up against the RaSer stuff. Am I right in thinking our "friends" (YKW) are in the process of patenting the PCI-Express implementation?... with Intel's umm, collusion? Is it in fact possible to do a PCI-Express without paying "them" their license fees? Am I being paranoid here?:-)You know they're up to something, and you would be reasonable inimagining they're up to no good. They've been all over the place interms of I/O technology and schedules. Where was "communicationstreaming architecture", the Intel-only Gigabit connection that suddenlysprang from the ICH5 memory controller in all their press hype? Nowherethat I can tell. Infiniband? Forget it. It will get 10Gigabit accessto the memory controller when hell freezes over.That's why I can never understand all the animus against Rambus. IfIntel weren't Intel, nobody would do business with them. They playhardball just as hard as Rambus ever did. They're just a littlesmoother about it.

The big difference here is that Intel is a bona fide manufacturer. They
cannot possibly own all the IP required to design and make all their chips.
The hardware industry has forged ahead as quickly as it has, in large part
due to broad cross-licensing agreements - the hardware mfrs have a strong
interest in continuing this... including Intel... though Intel and others
in strong positions, occasionally have their little excursions into
architecture "proprietarization".

The trouble with the IP companies, like Rambus, is that they are not in the
patent trading business. Of course if the Intel/Rambus/DRDRAM saga had
played out as they planned, Intel, through Rambus warrants, would have
owned by proxy, the memory interface... but would not have been in a
position to "trade" it. Maybe it was all partly Intel's fault but Rambus
is easy to hate for the above reasons - parasites are just fugly.

There's an interesting situation with Qualcomm at the moment which is worth
looking at, as a sort of parallel exercise to Rambus, in terms of how bad
things could have gotten if DRDRAM had become as dominant as planned.
Qualcomm is a good ways further along their corporate developmental path
(than Rambus) and is now becoming more of a fabless mfr than a pure IP
company... while, at the same time, not wanting to lose its IP taxes.
There are some nasty aspects to their business model though - they are
hated outside the U.S., particularly in Europe, for the "double pricing" on
their IP and chips, whether made by their contracted fab or some
licensee's.

To go forward from GSM, it looks like the global cell/wireless market is
going to have to adopt some aspects of CDMA, since it has some strong
technical points. Of course Qualcomm would like the entire industry to
just fallover and adopt their latest incarnation - CDMA2000 or whatever -
but the industry would rather that Qualcomm join the "community" of cross
licensers and play nice. The conferences on the subject sound like things
are very messy and just recently Qualcomm has filed to rescind its cross
license agreements with TI and is suing TI for breach of the agreement.

It seems to me that this is the exact same direction that Rambus wants to
take the entire computer industry - sounds ugly to me.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

Tim Sullivan

08-19-2003, 02:29 PM

fammacd=!SPAM^nothanks@tellurian.com (George Macdonald) wrote in message news:<3f41b45d.238156682@news.tellurian.com>... On Mon, 18 Aug 2003 04:11:12 GMT, Robert Myers <rmyers@comcast.net> wrote:George Macdonald wrote: Nah they'd rather pay for external IP and use it to hijack the entire architecture? Talking of which.... I haven't seen much "comparison" of how HT measures up against the RaSer stuff. Am I right in thinking our "friends" (YKW) are in the process of patenting the PCI-Express implementation?... with Intel's umm, collusion? Is it in fact possible to do a PCI-Express without paying "them" their license fees? Am I being paranoid here?:-)You know they're up to something, and you would be reasonable inimagining they're up to no good. They've been all over the place interms of I/O technology and schedules. Where was "communicationstreaming architecture", the Intel-only Gigabit connection that suddenlysprang from the ICH5 memory controller in all their press hype? Nowherethat I can tell. Infiniband? Forget it. It will get 10Gigabit accessto the memory controller when hell freezes over.That's why I can never understand all the animus against Rambus. IfIntel weren't Intel, nobody would do business with them. They playhardball just as hard as Rambus ever did. They're just a littlesmoother about it. The big difference here is that Intel is a bona fide manufacturer. They cannot possibly own all the IP required to design and make all their chips. The hardware industry has forged ahead as quickly as it has, in large part due to broad cross-licensing agreements - the hardware mfrs have a strong interest in continuing this... including Intel... though Intel and others in strong positions, occasionally have their little excursions into architecture "proprietarization".

Do you think the memory industry would have forged ahead as quickly in
the last few years if Rambus did not exist? Isn't it good to have
innovation and competition instead of cartels controlling the pace of
change?
The trouble with the IP companies, like Rambus, is that they are not in the patent trading business. Of course if the Intel/Rambus/DRDRAM saga had played out as they planned, Intel, through Rambus warrants, would have owned by proxy, the memory interface... but would not have been in a position to "trade" it. Maybe it was all partly Intel's fault but Rambus is easy to hate for the above reasons - parasites are just fugly.

A parasite takes resources and gives nothing in return. Has not Rambus
made any contributions to advances in memory technology?
There's an interesting situation with Qualcomm at the moment which is worth looking at, as a sort of parallel exercise to Rambus, in terms of how bad things could have gotten if DRDRAM had become as dominant as planned. Qualcomm is a good ways further along their corporate developmental path (than Rambus) and is now becoming more of a fabless mfr than a pure IP company... while, at the same time, not wanting to lose its IP taxes. There are some nasty aspects to their business model though - they are hated outside the U.S., particularly in Europe, for the "double pricing" on their IP and chips, whether made by their contracted fab or some licensee's. To go forward from GSM, it looks like the global cell/wireless market is going to have to adopt some aspects of CDMA, since it has some strong technical points. Of course Qualcomm would like the entire industry to just fallover and adopt their latest incarnation - CDMA2000 or whatever - but the industry would rather that Qualcomm join the "community" of cross licensers and play nice. The conferences on the subject sound like things are very messy and just recently Qualcomm has filed to rescind its cross license agreements with TI and is suing TI for breach of the agreement. It seems to me that this is the exact same direction that Rambus wants to take the entire computer industry - sounds ugly to me. Rgds, George Macdonald "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

Paul Tiseo

08-20-2003, 07:34 AM

In article <1d62d428.0308191429.1d663030@posting.google.com>,
timsullivan2003@yahoo.com says... Do you think the memory industry would have forged ahead as quickly in the last few years if Rambus did not exist? Isn't it good to have innovation and competition instead of cartels controlling the pace of change?

Considering they stole their presumed "innovation" from JEDEC
meetings and from prior art?

I will admit, it is appealing to consider they might have been an
impetus to other memory technologies to get to market faster, possibly.
This point could be debated. One could argue that changes in memory
technology would have happened anyways due to the growing gap between
CPU performance and memory subsystem performance. Rambus was one such
change...others were brewing in parallel, if not beforehand.
A parasite takes resources and gives nothing in return. Has not Rambus made any contributions to advances in memory technology?

Where? Patenting double-clocking? Patenting voltage levels?

Right...

----------------------------------------
Paul Tiseo, Systems Programmer
Research Computing Facility, Mayo Clinic
tiseo128.paul23@mayo.edu
(please remove numbers to email me)

daytripper

08-20-2003, 11:50 AM

On Wed, 20 Aug 2003 19:52:27 GMT, Tony Hill <hilla_nospam_20@yahoo.ca> wrote:
On 19 Aug 2003 15:29:32 -0700, timsullivan2003@yahoo.com (TimSullivan) wrote:fammacd=!SPAM^nothanks@tellurian.com (George Macdonald) wrote in message news:<3f41b45d.238156682@news.tellurian.com>... On Mon, 18 Aug 2003 04:11:12 GMT, Robert Myers <rmyers@comcast.net> wrote: The big difference here is that Intel is a bona fide manufacturer. They cannot possibly own all the IP required to design and make all their chips. The hardware industry has forged ahead as quickly as it has, in large part due to broad cross-licensing agreements - the hardware mfrs have a strong interest in continuing this... including Intel... though Intel and others in strong positions, occasionally have their little excursions into architecture "proprietarization".Do you think the memory industry would have forged ahead as quickly inthe last few years if Rambus did not exist? Isn't it good to haveinnovation and competition instead of cartels controlling the pace ofchange?If it hadn't been for Rambus (or more to the point, Intel's claim thatthey would exclusively support DRDRAM), the memory industry would haveforged ahead with it's SLDRAM plans and we would all be using that.Instead they abandoned those plans and started working on RDRAM, untilthey realized that it wasn't all it was cracked up to be technicallyand Rambus was such a dirty and litigious company that no one wantedto deal with. Instead they had to backtrack to their fall-back planof DDR and now DDR2.So, to answer the question, yes I think that the memory industry wouldhave forged ahead just as quickly if not more so if Rambus did notexist. It's not like Rambus is the only company developing new memorytechnology. In fact, they aren't even all that big of a player in theR&D front. Their technology focuses exclusively on one piece of thememory puzzle. The trouble with the IP companies, like Rambus, is that they are not in the patent trading business. Of course if the Intel/Rambus/DRDRAM saga had played out as they planned, Intel, through Rambus warrants, would have owned by proxy, the memory interface... but would not have been in a position to "trade" it. Maybe it was all partly Intel's fault but Rambus is easy to hate for the above reasons - parasites are just fugly.A parasite takes resources and gives nothing in return. Has not Rambusmade any contributions to advances in memory technology?One could certainly argue that they haven't really, or at least notfor PC memory technology. The small bit of technology that they didbring to market has been of some use for certain embedded markets,though even there other solutions do exist.

Not only embedded markets - they sucked in the folks who designed Marvel ;-)

Tim Sullivan

08-20-2003, 01:30 PM

Paul Tiseo <tiseo128.paul23@mayo.edu> wrote in message news:<MPG.19ad5ef226a16e809896aa@news.easynews.com>... In article <1d62d428.0308191429.1d663030@posting.google.com>, timsullivan2003@yahoo.com says... Do you think the memory industry would have forged ahead as quickly in the last few years if Rambus did not exist? Isn't it good to have innovation and competition instead of cartels controlling the pace of change? Considering they stole their presumed "innovation" from JEDEC meetings and from prior art?

What did they steal? The original patent that describes all Rambus'
inventions was filed in 1990, well before Rambus joined JEDEC. It
looks more like JEDEC borrowed ideas from Rambus, made slight changes
and called the technology their own. As for prior art you may be
right, however the USPTO who had an examiner doing a full time job of
researching prior art found none. Also courts in Europe have
researched prior art and have found none.
I will admit, it is appealing to consider they might have been an impetus to other memory technologies to get to market faster, possibly. This point could be debated. One could argue that changes in memory technology would have happened anyways due to the growing gap between CPU performance and memory subsystem performance. Rambus was one such change...others were brewing in parallel, if not beforehand. A parasite takes resources and gives nothing in return. Has not Rambus made any contributions to advances in memory technology? Where? Patenting double-clocking? Patenting voltage levels? Right...

Rambus didn't patent the technologies you mention, but the application
of the technologies in memory systems. Using known technology(eg
double-clocking) to solve a new problem or in a novel way(eg on a
memory chip) is patentable.
Maybe you say they are obvious but most solutions to hard problem are
obvious once someone describes the solution to you.
You must hate Rambus alot if you think that they have no advances in
memory technology.
---------------------------------------- Paul Tiseo, Systems Programmer Research Computing Facility, Mayo Clinic tiseo128.paul23@mayo.edu (please remove numbers to email me)

Keith R. Williams

08-20-2003, 05:44 PM

In article <ei53kvk2ucce7l2lsjfae717l04ol8kijb@4ax.com>,
rmyers@rustuck.com says... On Mon, 18 Aug 2003 23:01:13 GMT, "Yousuf Khan" <bbbl67.spam@yahoo.com.nospam> wrote: <snip>Let's face it Intel started out as a nice guy when it was starting out too.Weren't we all glad when good-guys Intel and Microsoft took away thedecision making power of the PC industry from big-bad meanie IBM, back inthe late 80's/early 90's? You actually wanted to do business with Intel atthat time. And now it's IBM to the defense of Linux against former Linux distributor SCO. Not hardware-related, but I met a really nice person from M$ with some degree of responsibility at a high-profile event recently. I told the M$ staffer that I wouldn't even consider using an M$ product for the topic of the conference, and the M$ rep asked for my business card. Soon I got a really nice e-mail from the M$ staffer asking if I would share my concerns in more detail. I sent back a polite but blunt e-mail: would M$ please act as if it were a member of a community rather than the proprietor of it, and would M$ please stop trying to obstruct peaceful interaction with Linux by sharing details of NT f1147 format and SMB so that both could be used properly by Linux networking tools. I also said that, as soon as M$ file formats were open and did not compromise the rights of owners of intellectual property, I would stop advising people not to store intellectual property in M$ formats. I got no reply.

Do you think you would have gotten a reply from RMBS or INTC
management with such a pointed question? I don't think anyone
here is saying that M$FT is a saint. The issue is RMBS and their
patent shenanigans. Come on" DDR data recovery by inverting a
clock? Programmable latency registers? Please!

Frankly, I can't understand anyone (other than the usual shills)
supporting RMBS' actions. The USPTO has problems, but RMBS makes
them stand out like a big toe laced with gout!

--
Keith

Keith R. Williams

08-20-2003, 05:51 PM

In article <07k7kvk11qs15kr13d9qh28bbact9ea5e0@4ax.com>,
day_trippr@REMOVEyahoo.com says... On Wed, 20 Aug 2003 19:52:27 GMT, Tony Hill <hilla_nospam_20@yahoo.ca> wrote:On 19 Aug 2003 15:29:32 -0700, timsullivan2003@yahoo.com (TimSullivan) wrote:fammacd=!SPAM^nothanks@tellurian.com (George Macdonald) wrote in message news:<3f41b45d.238156682@news.tellurian.com>...> On Mon, 18 Aug 2003 04:11:12 GMT, Robert Myers <rmyers@comcast.net> wrote:> The big difference here is that Intel is a bona fide manufacturer. They> cannot possibly own all the IP required to design and make all their chips.> The hardware industry has forged ahead as quickly as it has, in large part> due to broad cross-licensing agreements - the hardware mfrs have a strong> interest in continuing this... including Intel... though Intel and others> in strong positions, occasionally have their little excursions into> architecture "proprietarization".Do you think the memory industry would have forged ahead as quickly inthe last few years if Rambus did not exist? Isn't it good to haveinnovation and competition instead of cartels controlling the pace ofchange?If it hadn't been for Rambus (or more to the point, Intel's claim thatthey would exclusively support DRDRAM), the memory industry would haveforged ahead with it's SLDRAM plans and we would all be using that.Instead they abandoned those plans and started working on RDRAM, untilthey realized that it wasn't all it was cracked up to be technicallyand Rambus was such a dirty and litigious company that no one wantedto deal with. Instead they had to backtrack to their fall-back planof DDR and now DDR2.So, to answer the question, yes I think that the memory industry wouldhave forged ahead just as quickly if not more so if Rambus did notexist. It's not like Rambus is the only company developing new memorytechnology. In fact, they aren't even all that big of a player in theR&D front. Their technology focuses exclusively on one piece of thememory puzzle.> The trouble with the IP companies, like Rambus, is that they are not in the> patent trading business. Of course if the Intel/Rambus/DRDRAM saga had> played out as they planned, Intel, through Rambus warrants, would have> owned by proxy, the memory interface... but would not have been in a> position to "trade" it. Maybe it was all partly Intel's fault but Rambus> is easy to hate for the above reasons - parasites are just fugly.A parasite takes resources and gives nothing in return. Has n[made any contributions to advances in memory technology?One could certainly argue that they haven't really, or at least notfor PC memory technology. The small bit of technology that they didbring to market has been of some use for certain embedded markets,though even there other solutions do exist. Not only embedded markets - they sucked in the folks who designed Marvel ;-)

Judging by the recent RMBS price, they've suckered in a whole new
generation of shills. ...though I think RMBS may be having some
competition with SCO, on this front.

--
Keith

Tony Hill

08-20-2003, 06:45 PM

On Wed, 20 Aug 2003 19:50:47 GMT, daytripper
<day_trippr@REMOVEyahoo.com> wrote:On Wed, 20 Aug 2003 19:52:27 GMT, Tony Hill <hilla_nospam_20@yahoo.ca> wrote:One could certainly argue that they haven't really, or at least notfor PC memory technology. The small bit of technology that they didbring to market has been of some use for certain embedded markets,though even there other solutions do exist.Not only embedded markets - they sucked in the folks who designed Marvel ;-)

What's perhaps even more amazing though is that HP is actually
SHIPPING those systems now!

-------------
Tony Hill
hilla <underscore> 20 <at> yahoo <dot> ca

daytripper

08-20-2003, 07:53 PM

On Thu, 21 Aug 2003 02:45:58 GMT, Tony Hill <hilla_nospam_20@yahoo.ca> wrote:
On Wed, 20 Aug 2003 19:50:47 GMT, daytripper<day_trippr@REMOVEyahoo.com> wrote:On Wed, 20 Aug 2003 19:52:27 GMT, Tony Hill <hilla_nospam_20@yahoo.ca> wrote:One could certainly argue that they haven't really, or at least notfor PC memory technology. The small bit of technology that they didbring to market has been of some use for certain embedded markets,though even there other solutions do exist.Not only embedded markets - they sucked in the folks who designed Marvel ;-)What's perhaps even more amazing though is that HP is actuallySHIPPING those systems now!

It is amazing considering the last couple of years had to be hell for the
walking dead that were involved.

In any case, I believe the current owners of the Alpha legacy - like the
owners before - were contractually obligated to ship some number of Marvel
systems, lest the humongous penalties be invoked.

They'll likely never recoup the development costs, but rendering those
penalties moot will save a huge pot of HP bucks...

jack

08-20-2003, 11:45 PM

Robert Myers <rmyers@rustuck.com> wrote:
::
<snip>
: The same cannot be said for the thugs in the software business. Make
: no mistake about it, some of the major players are acting like thugs,
: and they are not confining their thuggery to players in the business.
: Ask anyone who has been audited by the Software Business Alliance or
: who has been contacted by SCO. In another post you patronizingly tell

Funny how you should say that. Read this interesting clip:
http://news.com.com/2008-1082_3-5065859.html

Talk about thuggery!

J.
--
--------
The end to "Personal Computing" as we know it is just around the corner.
TCPA will take away ALL rights from you, the consumer. Learn more
about it here: http://www.againsttcpa.com/what-is-tcpa.html and
here: http://www.againsttcpa.com/tcpa-faq-en.html

David Winter

08-21-2003, 12:45 AM

I did, enough to vote with my $$s. I would NOT and did NOT buy a P4 until
non-RDRAM options were available. I'm sure I was not and am not alone. I
will NOT buy any product which I have knowledge provides them with any
revenue. Thus speaketh this corner of the market.

DW

"Tim Sullivan" <timsullivan2003@yahoo.com> wrote in message
news:1d62d428.0308201330.328e457a@posting.google.com...
: Paul Tiseo <tiseo128.paul23@mayo.edu> wrote in message
news:<MPG.19ad5ef226a16e809896aa@news.easynews.com>...
: > In article <1d62d428.0308191429.1d663030@posting.google.com>,
: > timsullivan2003@yahoo.com says...
: > >
: > > Do you think the memory industry would have forged ahead as quickly in
: > > the last few years if Rambus did not exist? Isn't it good to have
: > > innovation and competition instead of cartels controlling the pace of
: > > change?
: >
: > Considering they stole their presumed "innovation" from JEDEC
: > meetings and from prior art?
: >
:
: What did they steal? The original patent that describes all Rambus'
: inventions was filed in 1990, well before Rambus joined JEDEC. It
: looks more like JEDEC borrowed ideas from Rambus, made slight changes
: and called the technology their own. As for prior art you may be
: right, however the USPTO who had an examiner doing a full time job of
: researching prior art found none. Also courts in Europe have
: researched prior art and have found none.
:
: > I will admit, it is appealing to consider they might have been an
: > impetus to other memory technologies to get to market faster, possibly.
: > This point could be debated. One could argue that changes in memory
: > technology would have happened anyways due to the growing gap between
: > CPU performance and memory subsystem performance. Rambus was one such
: > change...others were brewing in parallel, if not beforehand.
: >
: > > A parasite takes resources and gives nothing in return. Has not Rambus
: > > made any contributions to advances in memory technology?
: >
: > Where? Patenting double-clocking? Patenting voltage levels?
: >
: > Right...
: >
:
: Rambus didn't patent the technologies you mention, but the application
: of the technologies in memory systems. Using known technology(eg
: double-clocking) to solve a new problem or in a novel way(eg on a
: memory chip) is patentable.
: Maybe you say they are obvious but most solutions to hard problem are
: obvious once someone describes the solution to you.
: You must hate Rambus alot if you think that they have no advances in
: memory technology.
:
: > ----------------------------------------
: > Paul Tiseo, Systems Programmer
: > Research Computing Facility, Mayo Clinic
: > tiseo128.paul23@mayo.edu
: > (please remove numbers to email me)

David Schwartz

08-21-2003, 02:22 AM

"Tim Sullivan" <timsullivan2003@yahoo.com> wrote in message
news:1d62d428.0308201330.328e457a@posting.google.com...
What did they steal? The original patent that describes all Rambus' inventions was filed in 1990, well before Rambus joined JEDEC. It looks more like JEDEC borrowed ideas from Rambus, made slight changes and called the technology their own.

There is strong evidence that Rambus deliberately and fraudulently
manipulated JEDEC so that their standards would infringe on Rambus' patents.

DS

chrisv

08-21-2003, 04:58 AM

On Mon, 18 Aug 2003 23:26:28 -0400, Robert Myers <rmyers@rustuck.com>
wrote:
I sent back a polite but blunte-mail: would M$ please act as if it were a member of a communityrather than the proprietor of it

Ha! The evil businessman cares NOTHING for you. He only wants to
line his greedy pockets. End of story.

Tim Sullivan

08-21-2003, 06:10 AM

"David Schwartz" <davids@webmaster.com> wrote in message news:<bi26gr$5gp$1@nntp.webmaster.com>... "Tim Sullivan" <timsullivan2003@yahoo.com> wrote in message news:1d62d428.0308201330.328e457a@posting.google.com... What did they steal? The original patent that describes all Rambus' inventions was filed in 1990, well before Rambus joined JEDEC. It looks more like JEDEC borrowed ideas from Rambus, made slight changes and called the technology their own. There is strong evidence that Rambus deliberately and fraudulently manipulated JEDEC so that their standards would infringe on Rambus' patents. DS

And this evidence is? Seeing how Rambus was cleared by the CAFC of
committing fraud I would say this supposed evidence wasn't very strong
at all.
At least you have a understanding of what Rambus has been accused of
as opposed to the ill imformed "Rambus stole it" crowd.

Robert Myers

08-21-2003, 07:05 AM

On Thu, 21 Aug 2003 09:45:18 +0200, "jack"
<EKKXEGAUUXOP@spammotel.com> wrote:
Robert Myers <rmyers@rustuck.com> wrote:::<snip>: The same cannot be said for the thugs in the software business. Make: no mistake about it, some of the major players are acting like thugs,: and they are not confining their thuggery to players in the business.: Ask anyone who has been audited by the Software Business Alliance or: who has been contacted by SCO. In another post you patronizingly tellFunny how you should say that. Read this interesting clip:http://news.com.com/2008-1082_3-5065859.htmlTalk about thuggery!
Besides which, there is a certain entertainment value, at least to
people who read this group, in the endless Rambus saga. None of the
players have clean hands, they all can afford their lawyers, and they
all have their lobbyists.

I've got a computer with one of the goofiest chipsets (i820) Intel
ever sold. The super-duper memory runs as slow as molasses in
January. I was pretty irritated with myself at the time, but the
computer does what it needs to do, I own what I regard as a
collector's item, and it all feels kind of funny in retrospect.

I already don't feel that way about Microsoft and SCO, and I just
*know* we haven't seen the worst from either of them yet.

RM

Robert Myers

08-21-2003, 07:09 AM

On Thu, 21 Aug 2003 12:58:35 GMT, chrisv <chrisv@nospam.invalid>
wrote:
On Mon, 18 Aug 2003 23:26:28 -0400, Robert Myers <rmyers@rustuck.com>wrote: I sent back a polite but blunte-mail: would M$ please act as if it were a member of a communityrather than the proprietor of itHa! The evil businessman cares NOTHING for you. He only wants toline his greedy pockets. End of story.

They could at least try to fake it. I only asked them to *act* as if
they were a member of the community. IBM has done a reasonably good
job of faking it for a very long time, and they have gotten *so*
community-minded now that being so lines up exactly with their
corporate objectives.

I don't think that it has quite sunk in on Bill Gates that robber
barons are remembered as robber barons, no matter how many buildings
they endow.

RM

Tim Sullivan

08-21-2003, 01:48 PM

Paul Tiseo <tiseo128.paul23@mayo.edu> wrote in message news:<MPG.19aeb35ee1cd0e4e9896ac@news.easynews.com>... In article <1d62d428.0308201330.328e457a@posting.google.com>, timsullivan2003@yahoo.com says... What did they steal? The original patent that describes all Rambus' inventions was filed in 1990, well before Rambus joined JEDEC. I admit that "steal" is a simple word and perhaps inaccurate or too broad. In another post, you state: "Seeing how Rambus was cleared by the CAFC of committing fraud I would say this supposed evidence wasn't very strong at all." Again, hinting that out-and-out stealing was not something Rambus did. I guess the impression we are to have from that spinmeistering is that Rambus didn't "walk in, listen and leave patenting ideas presented at JEDEC". This is true, but is not the real story as I know it. They steered JEDEC towards its undisclosed patents. This was well- displayed in court and the reason they were initially found guilty. I believe Rambus was "exonerated" on a technicality, in that JEDEC's "rules of conduct" weren't thorough or specific enough to force/demand complete patent disclosure or to make what Rambus did explicitely frowned upon. Rambus was able to manipulate ideas towards their patents, file divisonals and other extensions and then claim ownership after conclusion, thus submarining the industry with their IP. Masterful in its execution, wrong and unethical in its intent.

I think, "steered" and "manipulated" are the wrong words to describe
what Rambus did. Rambus remained silent and knowingly let JEDEC make a
standard that Rambus believed they could obtain patents over. This is
what the CAFC described as the unethical(although not illegal)
business practice and I agree with them. Not that it makes Rambus
actions right, what they did was in reaction to some JEDEC members
trying to destroy Rambus and make Rambus' technology open domain. I
believe that many of JEDEC member companies were far more unethical
then Rambus.
It looks more like JEDEC borrowed ideas from Rambus, made slight changes and called the technology their own. But, that is the point of a standards body as I understand it. Many players come to a round table with ideas, show them off and a selection process picks some of the ideas or otherwise rolls its own to fufill the requirements of the standard. Usually, the patent holder must relinquish the patent in some way to maintain open access, or all parties involved in the standards body agree to pay the patent holder. JEDEC "borrowed" because that's what a standard body does. In fact, Rambus abused this fact by steering JEDEC towards it divisionals. However, in some fairness, this was partly due to JEDEC's lack of oversight and thoroughness. If you don't discipline a brat in your house, don't be surprised if they knock over a vase or two...

Again I believe "steering" is the wrong word. Also JEDEC as well as
not being vigilant wrongly believed that many of Rambus' patents would
not issue due to prior art.
As for prior art you may be right, however the USPTO who had an examiner doing a full time job of researching prior art found none. Also courts in Europe have researched prior art and have found none. Has the USPTO been shown to be very capable arbitrer in determining prior art? Seems to go contrary to what a majority of people claim about the current state of the USPTO vis-a-vis technology IP and patents. There are multiple patent cases tying up court systems that shouldn't even be there. Rambus didn't patent the technologies you mention, but the application of the technologies in memory systems. Using known technology(eg double-clocking) to solve a new problem or in a novel way(eg on a memory chip) is patentable. So why were they ready to hold hostage not only memory makers, but the entire semiconductor industry? Basically anyone who used double- clocking? I distincly remeber the Rambus CEO stating this in his early sabre-rattling on their IP. I don't have links anymore but I remember an interview with Avo from Rambus stating this very same thing. Maybe you say they are obvious but most solutions to hard problem are obvious once someone describes the solution to you. You must hate Rambus alot if you think that they have no advances in memory technology. I have no love/hate for technologies. I have a mild dislike for companies I consider unethical and for people who spread misinformation, and that's far from hate! Sometimes, I feel like airing my own opinion in an attempt to temper the thread and to show new and/or less-knowledgeable readers that your viewpoint may or may not be right. It's called discussion and the fact that I take a contrary position to your opinion doesn't mean I "hate" a technology or a company. ---------------------------------------- Paul Tiseo, Systems Programmer Research Computing Facility, Mayo Clinic tiseo128.paul23@mayo.edu (please remove numbers to email me)

David Schwartz

08-21-2003, 02:07 PM

"Tim Sullivan" <timsullivan2003@yahoo.com> wrote in message
news:1d62d428.0308211348.790d44bb@posting.google.com...
I think, "steered" and "manipulated" are the wrong words to describe what Rambus did. Rambus remained silent and knowingly let JEDEC make a standard that Rambus believed they could obtain patents over. This is what the CAFC described as the unethical(although not illegal) business practice and I agree with them. Not that it makes Rambus actions right, what they did was in reaction to some JEDEC members trying to destroy Rambus and make Rambus' technology open domain. I believe that many of JEDEC member companies were far more unethical then Rambus.

You know that a jar contains gasoline, you know that your friend thinks
it contains water. You let your friend go and 'water' your plants with the
jar, your house catches on fire, and you sue your friend for the damage he
did.

No, you are not legally required to tell people that they are mistaken
about the contents of a jar they're carrying. But no, you cannot then blame
them for damages caused to you that you had advance notice of and could
easily have avoided by just saying that you would be damaged if the conduct
continued.

If I show up at my house and find a wrecking crew about to tear it down
and I investigate and find out that they got the wrong street number, aren't
I obligated to *tell* them that I know something that they don't, or can I
just sit by, let them knock my house down in error, and then sue them for
damages later?

Rambus sat by and deliberately allowed someone else to unknowingly
damage them.

DS

Never anonymous Bud

08-21-2003, 02:34 PM

Separating himself from Baghdad Bob, timsullivan2003@yahoo.com (Tim Sullivan)
whined:
I think, "steered" and "manipulated" are the wrong words to describewhat Rambus did. Rambus remained silent and knowingly let JEDEC make astandard that Rambus believed they could obtain patents over.

And ONLY a lawyer could possibly think it was OK.

To reply by email, remove the XYZ.

Lumber Cartel (tinlc) #2063. Spam this account at your own risk.

It's your SIG, say what you want to say....

Keith R. Williams

08-22-2003, 05:23 PM

In article <tvk8kv49jltlls1unn9nfpvjk5cl98os5k@4ax.com>,
rmyers@rustuck.com says... On Wed, 20 Aug 2003 21:44:27 -0400, Keith R. Williams <krw@attglobal.net> wrote:In article <ei53kvk2ucce7l2lsjfae717l04ol8kijb@4ax.com>,rmyers@rustuck.com says... <snip> Soon I got a really nice e-mail from the M$ staffer asking if I would share my concerns in more detail. I sent back a polite but blunt e-mail: would M$ please act as if it were a member of a community rather than the proprietor of it, and would M$ please stop trying to obstruct peaceful interaction with Linux by sharing details of NT f1147 format and SMB so that both could be used properly by Linux networking tools. I also said that, as soon as M$ file formats were open and did not compromise the rights of owners of intellectual property, I would stop advising people not to store intellectual property in M$ formats. I got no reply.Do you think you would have gotten a reply from RMBS or INTCmanagement with such a pointed question? I don't think anyonehere is saying that M$FT is a saint. The issue is RMBS and theirpatent shenanigans. Come on" DDR data recovery by inverting aclock? Programmable latency registers? Please!Frankly, I can't understand anyone (other than the usual shills)supporting RMBS' actions. The USPTO has problems, but RMBS makesthem stand out like a big toe laced with gout! Since this is a hardware group, I can understand that people would pay more attention to the actions of someone in the hardware business than to someone in the software business, but let me put it to you this way: I own memory involving all of the disputed technologies. I have not the slightest fear that someone will knock on my door and ask to inspect my computers or ask me to hand any memory back. It isn't going to happen. Nor will I be sent letters threatening legal action, nor will someone attempt to extort fees from me ex post facto. That also isn't going to happen. They may be thugs, but they keep their thuggery within the business.

Let me, as an employee of someone serious in the hardware
business, put it another way. The disruption of a RMBS (or
anyone pulling the like) hijacking the USPTO would put this
economy in a tail-spin. I would be unemployed and you wouldn't
have any new toys. ...not good.

YOU may not be worried about having your computer confiscated
(foolish optimism), but I *AM* worried about the future of the
computer business. The RMBS model is a disaster waiting to
happen. The only thing that's saved us from this disaster is
that INTC got a very bloody nose.

The same cannot be said for the thugs in the software business. Make no mistake about it, some of the major players are acting like thugs, and they are not confining their thuggery to players in the business.

That makes RMBS a saint somehow? Please! Don't be so silly!
Ask anyone who has been audited by the Software Business Alliance or who has been contacted by SCO. In another post you patronizingly tell me that I don't understand that that Microsoft king of the hill. By some measure, Microsoft is king of the hill, but it's a relatively recent phenomenon, and it isn't one that's going to last.

Recent? Well, if 20-years of thuggery is "recent", I'd agree.
However with the cash they've built as a *MONOPOLY*, they can
stand, and extend, their challengers. Yes, I do see M$ as evil.
Microsoft is going to take a fall,

YOu have great confidence here. I'm an optimist (Republicans are
;-), but you've gone way past me here. I only wish.
and they're going to hit the canvas hard, just the way that SCO has. When they do, they'll pull out all those onerous EULA's that people have been clicking on and declare ownership of everything on earth.

....which is why I refuse to go past Win2K SP2. I'll be going to
either Linux or a Mac soon. Win2K has been my only M$ OS. It
was goodness, but it's beyond is useful life too.

No, I don't think anyone has read the EULA. I don't think they
really would understand it if they did. No, I still don't think
M$ is going to take the fall. I know people who are Pi$$ed at
AOL, yet continue to use them, year after year.
Defend the actions of Rambus, much less support them? Not me. Keep things in perspective, you bet.

Good grief, you have!

--
Keith

Robert Myers

08-22-2003, 07:47 PM

On Fri, 22 Aug 2003 21:23:32 -0400, Keith R. Williams
<krw@attglobal.net> wrote:
In article <tvk8kv49jltlls1unn9nfpvjk5cl98os5k@4ax.com>,rmyers@rustuck.com says...

<snip>Let me, as an employee of someone serious in the hardwarebusiness, put it another way. The disruption of a RMBS (oranyone pulling the like) hijacking the USPTO would put thiseconomy in a tail-spin. I would be unemployed and you wouldn'thave any new toys. ...not good.YOU may not be worried about having your computer confiscated(foolish optimism), but I *AM* worried about the future of thecomputer business. The RMBS model is a disaster waiting tohappen. The only thing that's saved us from this disaster isthat INTC got a very bloody nose.
The world just doesn't work that way, and in particular the United
States doesn't work that way. Money and power respect the free market
and the rule of law only insofar as it suits the interests of money
and power. If you think a piddling little Rambus is going to be
allowed to come along and hijack something in a way that would put the
computer business in a tailspin, you need to step back and take a
couple of deep breaths. It isn't going to happen.
The same cannot be said for the thugs in the software business. Make no mistake about it, some of the major players are acting like thugs, and they are not confining their thuggery to players in the business.That makes RMBS a saint somehow? Please! Don't be so silly!
Who said anything about anybody being a saint? They aren't, and
neither are the other players in the business. This is about power
and money, not about good and evil.
Ask anyone who has been audited by the Software Business Alliance or who has been contacted by SCO. In another post you patronizingly tell me that I don't understand that that Microsoft king of the hill. By some measure, Microsoft is king of the hill, but it's a relatively recent phenomenon, and it isn't one that's going to last.Recent? Well, if 20-years of thuggery is "recent", I'd agree.However with the cash they've built as a *MONOPOLY*, they canstand, and extend, their challengers. Yes, I do see M$ as evil.
You remember history much differently from me. IBM's OS/2 was still
twitching as late as 1994, Windows was a POS OS, and it was still
possible for Netscape to throw a (very real) scare into Microsoft.
So, from my perspective, Microsoft as someone who is likely to have
been able to give orders to an Intel is a relatively recent
phenomenon.
Microsoft is going to take a fall,YOu have great confidence here. I'm an optimist (Republicans are;-), but you've gone way past me here. I only wish.
Linux has almost done to the world of software what you were afraid
that Rambus might do to the world of hardware. I say *almost*. IBM,
Oracle, and others have not only learned how to play in this brave new
world, they have learned to profit from it. Microsoft is toast.

....at least as we currently know it. Even leaving Linux aside,
monopolists have a very basic business model problem, which is that
there is no market share left for them to grow into. The third world,
which provided new markets to Pepsi and Coke, is closing its doors to
Microsoft (that *is* due to Linux), and Microsoft needs desperately to
refashion itself to justify its stock price. They might refashion
themselves successfully and they might not, but their ownership of the
x86 world is rapidly coming to a close.

<snip>
Defend the actions of Rambus, much less support them? Not me. Keep things in perspective, you bet.Good grief, you have!

Not quite sure whether you're saying that I've kept things in
perspective or that I've defended/supported Rambus, but I think you
mean the latter.

Everything I've heard about the Rambus principals make them sound like
very unlikable people, and I don't go around defending unlikable
people unless I think an injustice has been done to them. I don't
understand the complications of the Rambus affair well enough to pass
judgment one way or the other. George Macdonald sounds not only
grounded and sensible on the matter, but persuasive, so if I had to
pick someone's version to believe, I'd believe what he has told me.

Here's the problem Keith: I need bandwidth. Others who do the kind of
work I do need bandwidth. Pins are expensive and real estate for
traces is limited. It isn't clear to me that the people who have run
Rambus out of town have helped me.

I would not for one second propose that anyone be deprived of their
rights or their property to make my work easier or less expensive,
but it makes me suspicious that people have turned what should be an
economic, technical, and legal question into a personal and emotional
question.

RM

chrisv

08-25-2003, 07:58 AM

On Fri, 22 Aug 2003 23:47:40 -0400, Robert Myers <rmyers@rustuck.com>
wrote:

Here's the problem Keith: I need bandwidth. Others who do the kind ofwork I do need bandwidth. Pins are expensive and real estate fortraces is limited. It isn't clear to me that the people who have runRambus out of town have helped me.

Motherboards with dual 64-bit DDR-400 channels are under $100.
There's your bandwidth.
I would not for one second propose that anyone be deprived of theirrights or their property to make my work easier or less expensive,but it makes me suspicious that people have turned what should be aneconomic, technical, and legal question into a personal and emotionalquestion.

People get angry when someone tries to screw them.

Paul Tiseo

08-25-2003, 10:35 AM

In article <1d62d428.0308211348.790d44bb@posting.google.com>,
timsullivan2003@yahoo.com says... Not that it makes Rambus actions right, what they did was in reaction to some JEDEC members trying to destroy Rambus and make Rambus' technology open domain. I believe that many of JEDEC member companies were far more unethical then Rambus.

How? If Rambus attended the standards setting body, they were expected
to play fair. They are expected to disclose. Are you saying that the so-
called "Dramurai" dragged Rambus to the table unwillingly? Rambus could
have chosen not to attend. The fact that they attended and had plans to
covertly benefit from attending shows a sort of "corporate pre-
mediation", much like the difference between manslaughter and murder.

Rambus attended, manipulated and queitly took from the group to
submarine the industry later. In your own words, Rambus "remained
silent" as JEDEC walked down a path that Rambus had and was patenting.
Again I believe "steering" is the wrong word. Also JEDEC as well as not being vigilant wrongly believed that many of Rambus' patents would not issue due to prior art.

I'm curious, how do you know this?

I find that assertion hard to believe. If JEDEC knew of Rambus' on-going
patent binge, I doubt they would have gone through the hassle of
specifying a standard that matched it, if only to avoid the current
legal expenditures on both side. Companies *do not* go willy-nilly into
a legal morass, even if they think they can eventually win. It's just a
financial black hole.

Let's face it, the only reason Rambus got away with it on appeal is the
lack of clear contractual agreement between JEDEC members and the lack
of clear-cut federal law covering such corporate behavior. IOW, it was
definitely un-ethical and it probably should be illegal.

JEDEC was simply doing what all standards bodies do. Yes, some stadards
bodies effectively co-opt IP from [willing] companies and put them into
the public domain. However, not all standards bodies seek to have
companies completely give up IP. Some do, like W3C. Others don't. I
don't know what JEDEC expected of its members that way. I think, for
example, PCI must be licensed from Intel, but it is open to anyone. (I
could be wrong here.)

What if Rambus hadn't boasted of holding the industry hostage? What if
they presented a more cooperative stance? Could they have not had to go
through what they did? I would think so.
Again I believe "steering" is the wrong word. Also JEDEC as well as not being vigilant wrongly believed that many of Rambus' patents would not issue due to prior art.

Really? Where does that assertion come from?

----------------------------------------
Paul Tiseo, Systems Programmer
Research Computing Facility, Mayo Clinic
tiseo128.paul23@mayo.edu
(please remove numbers to email me)

Keith R. Williams

08-25-2003, 05:57 PM

In article <MPG.19b4212f784ec8c09896ae@news.easynews.com>,
tiseo128.paul23@mayo.edu says...

I think, for example, PCI must be licensed from Intel, but it is open to anyone. (I could be wrong here.)

PCI is licensed from the "PCI Special Interest Group" (a.k.a.
PCISIG). Intel licensed it's patents to the PCISIG. Pay the
$10K to be a member (or buy the solution from a member) and all
is wonderful. The PCISIG isn't perfect (if you're not a member
the specs are difficult to get), but it works.
What if Rambus hadn't boasted of holding the industry hostage? What if they presented a more cooperative stance? Could they have not had to go through what they did? I would think so.

First of all, as an IP *only* company they have nothing else to
offer than threats. There is no product there. The real problem
with RMBS came when they tried to hijack *ALL* SDRAM (and pretty
much all I/O channels). If they'd stuck with the RMBS channel
there would have been no knotted panties. However, when one
tries to patent air, be prepared for a collective gasp!

--
Keith

chrisv

08-26-2003, 04:51 AM

On Tue, 26 Aug 2003 04:39:18 GMT, "Felger Carbon" <fmrfne@jps.net>
wrote:
"Robert Myers" <rmyers@rustuck.com> wrote (Intel) can go back to their original (and correct from a technical point of view) plan to go with Rambus.Intel got burned badly in the Rambus saga. They are unlikely to jumpinto that frying pan again IMHO.

I'd say "no freaking way" would they have jumped completely off the
Rambust wagon, if they were planning on going with it long-term.
There's no way in hell the memory manufacturers will support Rambus,
either. It's dead, Jim.
Remember the CPU 'Timna', which hadto be cancelled outright because it would only work with Rambusmemory?

OMG that was a good one - a super low-end, highly-integrated design
that either required expensive RDRAM or the dreaded MTH's. Yuck.

Robert Myers

08-26-2003, 05:49 AM

On Tue, 26 Aug 2003 04:39:18 GMT, "Felger Carbon" <fmrfne@jps.net>
wrote:
"Robert Myers" <rmyers@rustuck.com> wrote in messagenews:mrukkvsoum72tuognfqr3fm50714hiuat4@4ax.com... <snip> As far as I can determine without becoming more of an expert onbuses and memory technology than I care to be, the selling points of DDRare that it has lower latency (and I have never seen any advantage actually quantified, either by demonstration or analysis), thatRambus channels require more design attention and tighter tolerances thanDDR channels, and that the memory relies on IP owned by Rambus, Inc.The last reason always seems to be decisive.Robert, the lower latency of standard DRAM (including DDR) was welldemonstrated and documented at the time of the original Rambusbrouhaha.

I figured somebody would come back on that one. Yes, the raw latency
numbers are lower, but how does it actually work out in practice?
There is probably a careful study of that at places like IBM and
Intel, but I haven't seen anything I would regard as credible
publicly.

I've guessed publicly at how expensive latency is to make up for in
terms of using on-die resources (probably exponential, at least
superlinear in latency), but other things about Rambus might well make
up for it (for example, the unexploited capacity to keep many pages
open at once).

Asking to see some kind of analysis is not merely to cavil. Intel
seems to have made some strategic decisions about coping with latency
that were aggressive in terms of what they thought was possible, but I
don't know that they were unrealistic, and I'm unwilling to guess.
At the time, standard DRAM was far lower in price thanRambus, a critical parameter for most of us. So standard DRAM bothhad lower latency _and_ lower price!

Now you are defending/explaining the reasons for market reaction and
popular reaction to RDRAM vs. DRAM. Yes, the market liked DRAM. It
liked it so much that OEM's were selling P4 boxes with SDRAM like
hotcakes for a short period. If anyone asked me, I told them they
would be better off with a Tualatin Celeron and SDRAM, and they would
have been. But you sure couldn't beat the price, and I guess the
theoretical latency was probably lower, too.

If Intel made an official benchmark submission for the P4 with
anything other than RDRAM (before the 800MHz FSB), I never saw the
results. If you wanted performance, you paid for the RDRAM, and I
have.
You are correct about thetighter tolerances required by Rambus; Intel made more than onemass-market mobo that simply did not work because of dataunreliability. Proving that threads in this NG drift, the reason oneof those mobos failed was because it didn't work when all four slotsof the multidrop Rambus' bus were filled. I seem to remember Intels'solution was "Only use 3 Rambus modules (RIMMs)"!

I've never seen one with more than two per channel. I know they tried
to get 3-RIMM configurations to work, but I didn't know that anybody
actually succeeded (and I believe there was a board recall for one
failed attempt).
Standard DRAM won on latency, price, and data reliability. RambusRIMMs ran pretty hot, too. Pretty convincing advantages to me.

Data reliability? You mean they had a hard time designing
motherboards that worked, I think. Is there something about i850
motherboards actually in the market I haven't heard?

The space in which I run my computers is not air conditioned, and we
get some pretty hot weather. My computers with Rambus are the least
troublesome of my computers.
(Intel) can go back to their original (and correct from a technical point of view) plan to go with Rambus. What do people think Intel means when they say they don't want an on-chip memory controller because it would tie them to a particular memory technology?Intel got burned badly in the Rambus saga. They are unlikely to jumpinto that frying pan again IMHO. Remember the CPU 'Timna', which hadto be cancelled outright because it would only work with Rambusmemory?
They certainly don't want anyone to be reminded of their part in the
history of this sordid affair. Whether that will keep them away from
Rambus forever is another matter.
DDR is not a scalable replacement for RDRAM as far as I candetermine from my own limited ability to digest information. Felger Carbonhas brought up QDRAM. One way to make a technology scalable is to double its capacity on paper. Where can I buy some?Robert, I have been under the impression that QDRAM _modules_ (notchips) are, or were, available in the marketplace. QDR by way ofadditional interface chips on the module. Not commonly found becausethe mobo makers failed to salute. Meanwhile, back at the ranch, folksdiscovered that it was much cheaper to use two DDR memory channels.DDR400 x2 x64bits gives, uh wait a minute, carry the 3, uh,6.4GBytes/sec. Cheaply. Works great with P4s having an 800MHz FSB,as most of them do these days.
Got any links? A quick run through google for pages in english
doesn't turn up anything credible.
(Hoo boy. I've been smart enough to stay out of this Rambus II threadup 'til now. :)
Gotcha. :-).

When people start throwing furniture, I'll just slip out the back
door.

RM

David Schwartz

08-26-2003, 11:20 AM

"Robert Myers" <rmyers@rustuck.com> wrote in message
news:mrukkvsoum72tuognfqr3fm50714hiuat4@4ax.com...
Yes, Rambus filled a niche before DDR SDRAM was available. Yes, it
stillfills a niche in embedded systems and game consoles.
Either you are not following the "Opteron, the Chip for Supercomputers", or you are following it and not drawing the same conclusions from it I am:
To get around the too-many-drops-to-run-fast issues with the currentarchitecture, Next-generation Intel ia32 quads will be plumbed as two
FSBswith pairs of processors tied to the north side of the MCH du jour, twoDDR2-400 memory hoses out the sides of the MCH, multiple PCI-Express
hosesemitted out the south side of the MCH to the PCH(s), and also a legacy
hose tothe ICH.

This is, IMO, a much less elegant design than the Opteron design.
And this MCH is going to have HOW many pins?! Geez, doesn't soundlike fun designing something around that, or at least not designing aboard using this chipset and having a deadline to meet! :>

AMD and the various manufacturers of Opteron motherboards don't seem to
be having this problem.
No problem. Already done. EV7. Each Rambus channel hastwo references for signaling to and fro. Just add as manychannels as you want. EV7 has 10 channels to memory, anda bunch more for CPU to CPU communications as well. They'reall pretty wide IIRC. The signals will automatically syncwithin each channel. You just have to take care of inter-channelskew on your controller, and viola, you've got a big wide bus.

One issue with this design is the latency stinks. But does anyone really
think this is any better than HT?
As far as I can determine without becoming more of an expert on buses and memory technology than I care to be, the selling points of DDR are that it has lower latency (and I have never seen any advantage actually quantified, either by demonstration or analysis), that Rambus channels require more design attention and tighter tolerances than DDR channels, and that the memory relies on IP owned by Rambus, Inc. The last reason always seems to be decisive.

There's also cost. The lower pin count of RDRAM was supposed to reduce
system cost, but that doesn't seem to have ever happened. And to get more
throughput with RDRAM, you still have to add more memory modules.
On the other hand, the memory chip for a four way P4 described in the "Opteron, the Chip for Supercomputers" thread sounds to me like it is pressing the limits of sanity in terms of pin count and board layout.

Contrast that with a 4-way Opteron machine with one DDR SDRAM controller
already on each Opteron. Add to that the point-to-point ccHT links. Faster,
more elegant, not proprietary.
That's why I think Intel is just waiting for all the brouhaha to settle down or at least for them to come up with a CYA explanation so that they can go back to their original (and correct from a technical point of view) plan to go with Rambus. What do people think Intel means when they say they don't want an on-chip memory controller because it would tie them to a particular memory technology?

Then they can use any high-speed interconnect they like, just add an
extra port from the CPU to go to the memory controller. Even with AMD's
Opteron, you aren't required to use the on-chip memory controller. You have
4 HT busses you can connect memory through.

DS

Felger Carbon

08-26-2003, 01:01 PM

"Robert Myers" <rmyers@rustuck.com> wrote in message
news:lkmmkv4ct9v8snpc1egpmt05h71sc9vh1a@4ax.com... "Felger Carbon" <fmrfne@jps.net> wrote:Robert, I have been under the impression that QDRAM _modules_ (notchips) are, or were, available in the marketplace. QDR by way ofadditional interface chips on the module. Not commonly found becausethe mobo makers failed to salute. Meanwhile, back at the ranch,
folksdiscovered that it was much cheaper to use two DDR memory channels.DDR400 x2 x64bits gives, uh wait a minute, carry the 3, uh,6.4GBytes/sec. Cheaply. Works great with P4s having an 800MHz FSB,as most of them do these days. Got any links? A quick run through google for pages in english doesn't turn up anything credible.

I googled on {Intel NEAR "800MHz FSB" NEAR "dual channel DDR"} and came
up with these hits as the second and fourth articles:

http://www.xbitlabs.com/articles/memory/print/qbm.html

"Bandwidth = Bus width x Bus frequency x Number of packs transferred per
clock cycle. " (this article refers to DDR400 memory as "PC3200 DDR"
and is about the Kentron QDR chip).

http://www.anandtech.com/chipsets/showdoc.html?i=1811

"Then we actually get to the extensive feature list of the 875P; 800MHz
FSB, dual channel DDR400"

I even found an Aussie review of a mobo with dual DDR400 and the P4 with
800MHz FSB:

"Gigabyte GA-8IPE1000MK INTEL 865PE FSB800
DUAL4DDR400/2SATA/AC97/LAN/8USB2/AGP8X/M-ATX"

BTW: some articles refer to DDR400 as "PC3200 DDR". Same thing.

jack

08-27-2003, 01:29 AM

Felger Carbon <fmrfne@jps.net> wrote:
<snip>

: I googled on {Intel NEAR "800MHz FSB" NEAR "dual channel DDR"} and
: came up with these hits as the second and fourth articles:

Hmmm. I searched Google's site high and low for a description of the
"NEAR" operator, and found nothing. I didn't know you could use it, and
it's not described on the web site. Thanks Felger....ya learn something
every day here!
http://www.google.com/help/operators.html

<more snipage>

J.
--
--------
The end to "Personal Computing" as we know it is just around the corner.
TCPA will take away ALL rights from you, the consumer. Learn more
about it here: http://www.againsttcpa.com/what-is-tcpa.html and
here: http://www.againsttcpa.com/tcpa-faq-en.html

George Macdonald

08-27-2003, 09:52 PM

On Tue, 26 Aug 2003 12:51:19 GMT, chrisv <chrisv@nospam.invalid> wrote:
On Tue, 26 Aug 2003 04:39:18 GMT, "Felger Carbon" <fmrfne@jps.net>wrote:"Robert Myers" <rmyers@rustuck.com> wrote (Intel) can go back to their original (and correct from a technical point of view) plan to go with Rambus.Intel got burned badly in the Rambus saga. They are unlikely to jumpinto that frying pan again IMHO.I'd say "no freaking way" would they have jumped completely off theRambust wagon, if they were planning on going with it long-term.There's no way in hell the memory manufacturers will support Rambus,either. It's dead, Jim.

Seems to me that PCI-Express is Rambus' next concerted effort to take
patent infringement in commodity PC hardware to a higher level. It would
not surprise me in the least to see them go after AMD for HT and Nvidia,
VIA, SiS, ALi, et.al. for PCI-Express. Intel, of course has already paid
their dues - no sweat there. Again, maybe I'm being paranoid - nobody
answered that the last time :-) - but I have a feeling that the path to the
next generation is "fraught with impending disaster".

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

Tony Hill

08-28-2003, 09:39 AM

On Tue, 26 Aug 2003 09:49:53 -0400, Robert Myers <rmyers@rustuck.com>
wrote:On Tue, 26 Aug 2003 04:39:18 GMT, "Felger Carbon" <fmrfne@jps.net>wrote:Robert, the lower latency of standard DRAM (including DDR) was welldemonstrated and documented at the time of the original Rambusbrouhaha.I figured somebody would come back on that one. Yes, the raw latencynumbers are lower, but how does it actually work out in practice?There is probably a careful study of that at places like IBM andIntel, but I haven't seen anything I would regard as crediblepublicly.

I don't know of any studies off hand, but I do remember the i820
chipset, which performed quite poorly in general. The i815 chipset
with PC133 SDRAM was definitely the faster of the two chipsets which
were nearly identical aside from their memory controllers. The i820
had the bandwidth, the i815 had lower latency.

Of course, it didn't help that the i820 was one of the buggiest
chipsets in recent memory, often being trickier to work with than the
worst that VIA brought to market.
I've guessed publicly at how expensive latency is to make up for interms of using on-die resources (probably exponential, at leastsuperlinear in latency), but other things about Rambus might well makeup for it (for example, the unexploited capacity to keep many pagesopen at once).

It is true that we never got to see the maximum theoretical
performance of RDRAM, though I don't know that it would have made much
of a difference. I'm guessing that there is a reason why Intel didn't
bother keeping a huge number of pages open at the same time in their
chipsets, it probably just didn't add anything significant to the
performance.

As a bit of a side note, that Digital/Compaq/HP ALpha EV-7 chip that
was supposed to be such a super-dooper high performance chip using
RDRAM is finally out, and it's nothing special. It performs ok, but
even with it's huge amount of bandwidth it still has trouble
outperforming P4 and Opteron chips using DDR SDRAM (SPEC CFP scores
are a bit higher but CINT scores are much lower).
You are correct about thetighter tolerances required by Rambus; Intel made more than onemass-market mobo that simply did not work because of dataunreliability. Proving that threads in this NG drift, the reason oneof those mobos failed was because it didn't work when all four slotsof the multidrop Rambus' bus were filled. I seem to remember Intels'solution was "Only use 3 Rambus modules (RIMMs)"!I've never seen one with more than two per channel. I know they triedto get 3-RIMM configurations to work, but I didn't know that anybodyactually succeeded (and I believe there was a board recall for onefailed attempt).

They "recalled" the i820 chipset two or three days before it was
supposed to have been released because it wouldn't work reliably with
3-RIMMs on it's single channel. Their solution was simply to limit
the specifications to 2-RIMMs per channel and to hell with all the
motherboard manufacturer's that had already produced 3-RIMM boards.
The MTH chip also died a similar death because it caused too much
noise on the RDRAM channel to work reliably and had to be killed. The
combination of these two things very effectively made the i820 a
rather useless chipset and left Intel out to dry in the chipset market
at that time. All they had was the very low-end i810 chipset
(integrated graphics with no AGP support and only PC100 SDRAM support)
and the too-expensive i840 chipset for workstations (dual-RDRAM). VIA
destroyed Intel in chipset sales, jumping from being a bit-player in
the PIII chipset market to ~%60 market share in a matter of months.
Meanwhile AMD was also chipping away Intel's processor market share,
largely due to these chipset problems as well.

As was mentioned earlier, Intel got burned BAD by devoting all their
energy to RDRAM. It just didn't work right for the design and
price-point of PC equipment, never did. Even the P4 was hampered
early on by it mainly only having an RDRAM chipset. Sure, performance
was fine, but the cost was higher than most were willing to pay, and
AMD and VIA were willing to come out with some very fast and much
cheaper alternatives.
Standard DRAM won on latency, price, and data reliability. RambusRIMMs ran pretty hot, too. Pretty convincing advantages to me.Data reliability? You mean they had a hard time designingmotherboards that worked, I think. Is there something about i850motherboards actually in the market I haven't heard?

The first RDRAM chipsets had two separate data reliability issues, so
it had a foot in the ground to start with. The heat thing wasn't
really a big issue. RDRAM did use a bit more power than SDRAM and
definitely more than DDR SDRAM, but not by a significant amount. The
real problem was that the heat production could become very localized
in RDRAM, so one individual chip could become much hotter than it's
neighbors, causing some potential localized overheating and data
reliability problems. This was easily solved though with the use of a
heatspreader.
Intel got burned badly in the Rambus saga. They are unlikely to jumpinto that frying pan again IMHO. Remember the CPU 'Timna', which hadto be cancelled outright because it would only work with Rambusmemory?They certainly don't want anyone to be reminded of their part in thehistory of this sordid affair. Whether that will keep them away fromRambus forever is another matter.

Intel lost a LOT of money because of RDRAM. They aren't going back
any time soon. Besides which first generation RDRAM is getting rather
long-in-tooth these days. Even when pushed to PC1200 speeds it
doesn't offer as much bandwidth as PC3200 DDR SDRAM, and higher speed
DDR/DDR2 is on the horizon. Rambus does have a new version of it's
main memory chips on the blackboard, but they're going to have a HELL
of time getting it manufactured. Rambus Inc. burned their bridges in
a HUGE way, pissing off the manufacturers that you are dependant on to
produce your products is generally a BAD idea.

---------------
Tony Hill
hilla <underscore> 20 <at> yahoo <dot> ca

Thomas Edison

08-28-2003, 10:35 AM

In comp.sys.intel Tony Hill <hilla_nospam_20@yahoo.ca> wrote:
As a bit of a side note, that Digital/Compaq/HP ALpha EV-7 chip that was supposed to be such a super-dooper high performance chip using RDRAM is finally out, and it's nothing special. It performs ok, but even with it's huge amount of bandwidth it still has trouble outperforming P4 and Opteron chips using DDR SDRAM (SPEC CFP scores are a bit higher but CINT scores are much lower).

This is because EV7 was itself a stopgap solution of sorts. The
"original" design goal was to have an SMT core that can generate
tons of memory references sitting in the processor that can provide
the amount of bandwidth and seamless interconnection fabric.

Unfortunately that project was delayed, so someone came up with the
idea that the schedule for this product can be kept in place, and only
minimal resources needed to be diverted to create a new project that
reused the EV6 processor core with the super duper off chip interface
that the processor under design has.

The "original" project became known as EV8 and the intermediate
project became known as EV7. What you're seeing is a very mature
EV6 core connected to an over engineered memory and interconnection
system. I think EV7 is still on 0.18um process, and 0.13um will
arrive shortly. So despite the memory and interconnection system,
with the EV6 core and the process lag, it's difficult to compete
(performance wide in scalar code) with processors fabbed on bleeding
edge processes.

Not that this makes a lot of difference, but the cross-process
generation comparison is always difficult to make. There are all
sorts of ways to justify different comparisons. i.e. It may be
"fair" to compare the EV7 against P4 and Opteron, because they're
"what's shipping now". That is true, however, I would caution
against using this comparison as a basis for making statements
about the performance of the memory system. A difference of a
process generation can really skew the performance comparison.

--
davewang202(at)yahoo(dot)com

Felger Carbon

08-28-2003, 12:40 PM

"Tony Hill" <hilla_nospam_20@yahoo.ca> wrote in message
news:9258e6bb472fb8db3a6cb408a4bed5f2@news.1usenet.com... <much snippage> It is true that we never got to see the maximum theoretical performance of RDRAM, though I don't know that it would have made much of a difference. I'm guessing that there is a reason why Intel didn't bother keeping a huge number of pages open at the same time in their chipsets, it probably just didn't add anything significant to the performance.

When Rambus demonstrated its boxes, they left lotsa pages open for
performance, and as a result had to have special fans blow on the heat
spreaders. Intel didn't want to use more fans for a fairly trivial
boost in performance, so only kept a coupla pages open.

Slightly off-topic: when VCM, or virtual channel memory, came out, I
thought it was a great idea. I even posted to this NG explaining to the
lurkers how it worked. But when they actually tested the parts, the
performance boost just wasn't there, so VCM quietly died.

If John Doe thinks a slide rule is faster than a modern desktop PC, then
John is not going to be happy with anything but a slide rule. I would
never want to interfere with John's happiness. ;-)

Tim Sullivan

08-28-2003, 03:50 PM

Tony Hill <hilla_nospam_20@yahoo.ca> wrote in message news:<9258e6bb472fb8db3a6cb408a4bed5f2@news.1usenet.com>...<snip> Tony Hill hilla <underscore> 20 <at> yahoo <dot> ca

F*ck, even I would be embarrassed to write such a biased and spun
message as this one. Over-emphasize the negatives and under-emphasize
the positives, a classical piece of propaganda.
Me, you and George MacDonald("Let me, as an employee of someone
serious in the hardware business") are shills, and anyone reading our
messages should realize our huge self-interests in how well Rambus
does.

The little lost angel

08-28-2003, 06:07 PM

On 28 Aug 2003 16:50:54 -0700, timsullivan2003@yahoo.com (Tim
Sullivan) wrote:
F*ck, even I would be embarrassed to write such a biased and spunmessage as this one. Over-emphasize the negatives and under-emphasizethe positives, a classical piece of propaganda.Me, you and George MacDonald("Let me, as an employee of someoneserious in the hardware business") are shills, and anyone reading ourmessages should realize our huge self-interests in how well Rambusdoes.

And of course, the rest of us are supposed to be convinced by the your
"abundance" of technical facts and "minimal" use of rants that Tony's
post was biased and overly negative?

Sheesh, either you put up real world technical merits to prove he was
under-emphasizing the positive, or as the saying goes: put up or shut
up.

--
L.Angel: I'm looking for web design work.
If you need basic to med complexity webpages at affordable rates, email me :)
Standard HTML, SHTML, MySQL + PHP or ASP, Javascript.
If you really want, FrontPage & DreamWeaver too.
But keep in mind you pay extra bandwidth for their bloated code

daytripper

08-28-2003, 06:16 PM

On Fri, 29 Aug 2003 02:07:20 GMT, a?n?g?e?l@lovergirl.lrigrevol.moc.com (The
little lost angel) wrote:
On 28 Aug 2003 16:50:54 -0700, timsullivan2003@yahoo.com (TimSullivan) wrote:F*ck, even I would be embarrassed to write such a biased and spunmessage as this one. Over-emphasize the negatives and under-emphasizethe positives, a classical piece of propaganda.Me, you and George MacDonald("Let me, as an employee of someoneserious in the hardware business") are shills, and anyone reading ourmessages should realize our huge self-interests in how well Rambusdoes.And of course, the rest of us are supposed to be convinced by the your"abundance" of technical facts and "minimal" use of rants that Tony'spost was biased and overly negative?Sheesh, either you put up real world technical merits to prove he wasunder-emphasizing the positive, or as the saying goes: put up or shutup.

He's had way more than a reasonable amount of time to shill his shares.
How 'bout he just skips to the ending?

/daytripper (It's *over*. Get used to it already!)

Robert Myers

08-28-2003, 08:44 PM

On 28 Aug 2003 16:50:54 -0700, timsullivan2003@yahoo.com (Tim
Sullivan) wrote:
Tony Hill <hilla_nospam_20@yahoo.ca> wrote in message news:<9258e6bb472fb8db3a6cb408a4bed5f2@news.1usenet.com>...<snip> Tony Hill hilla <underscore> 20 <at> yahoo <dot> caF*ck, even I would be embarrassed to write such a biased and spunmessage as this one. Over-emphasize the negatives and under-emphasizethe positives, a classical piece of propaganda.Me, you and George MacDonald("Let me, as an employee of someoneserious in the hardware business") are shills, and anyone reading ourmessages should realize our huge self-interests in how well Rambusdoes.

I had to go back to read Tony Hill's post to find out what it was that
had bothered you so much.

Tony generally sticks to the facts and presents a credible point of
view, and I don't see the post to which you are objecting so
strenuously as being an obvious exception to that pattern. I might
question as you have as to whether Tony's selection of facts
represents what I would regard as a balanced summary of the situation,
but I don't know that he labelled his post as such.

He was responding to points I had made with particular points that the
supported the point of view he wanted to present. That's fairly
normal for human discourse.

RM

Tim Sullivan

08-29-2003, 02:12 AM

Robert Myers <rmyers@rustuck.com> wrote in message news:<ulktkvc4c911rfvhfrtt97p14a85knqbj8@4ax.com>... On 28 Aug 2003 16:50:54 -0700, timsullivan2003@yahoo.com (Tim Sullivan) wrote:Tony Hill <hilla_nospam_20@yahoo.ca> wrote in message news:<9258e6bb472fb8db3a6cb408a4bed5f2@news.1usenet.com>...<snip> Tony Hill hilla <underscore> 20 <at> yahoo <dot> caF*ck, even I would be embarrassed to write such a biased and spunmessage as this one. Over-emphasize the negatives and under-emphasizethe positives, a classical piece of propaganda.Me, you and George MacDonald("Let me, as an employee of someoneserious in the hardware business") are shills, and anyone reading ourmessages should realize our huge self-interests in how well Rambusdoes. I had to go back to read Tony Hill's post to find out what it was that had bothered you so much. Tony generally sticks to the facts and presents a credible point of view, and I don't see the post to which you are objecting so strenuously as being an obvious exception to that pattern. I might question as you have as to whether Tony's selection of facts represents what I would regard as a balanced summary of the situation, but I don't know that he labelled his post as such. He was responding to points I had made with particular points that the supported the point of view he wanted to present. That's fairly normal for human discourse. RM

His post was not just replying to points in your post. It was just a
blatant RDRAM bashing post, something that an employee of Micron,
Infineon or Hynix might produce in a press release. I am sure someone
with the technical knowledge could write an equally negative and
biased post about SDRAM and DDR SDRAM.
On RDRAM, yes the original i820 was a piece of sh*t, but only just
recently did DDR-SDRAM match the performance and stability of the i850
which was released many years ago. Also why compare bandwidth of
PC1200 RDRAM to PC3200 DDR-SDRAM, when most motherboards do not have
only a single channel of each, it doesn't seem like a valid comparison
to me.

George Macdonald

08-29-2003, 04:30 AM

On Thu, 28 Aug 2003 20:59:50 +0000 (UTC), David Wang <foo@bar.invalid>
wrote:
In comp.sys.intel Felger Carbon <fmrfne@jps.net> wrote: "Tony Hill" <hilla_nospam_20@yahoo.ca> wrote in message news:9258e6bb472fb8db3a6cb408a4bed5f2@news.1usenet.com... It is true that we never got to see the maximum theoretical performance of RDRAM, though I don't know that it would have made much of a difference. I'm guessing that there is a reason why Intel didn't bother keeping a huge number of pages open at the same time in their chipsets, it probably just didn't add anything significant to the performance. When Rambus demonstrated its boxes, they left lotsa pages open for performance, and as a result had to have special fans blow on the heat spreaders. Intel didn't want to use more fans for a fairly trivial boost in performance, so only kept a coupla pages open.I've never spoken to Intel's chipset designers about this, but Ithink the "open pages" issue has more to do with sharing the samedesign characteristics as other "800" series chips than the powerconstraint. The way Intel handles open/close banks/pages is to keeptrack of what's open, and for how long etc. That has to cost a bitof hardware per open bank/page. Keeping state information aroundfor 16 or 32 banks per device and 32 devices per channel wouldquickly get to be really expensive in terms of hardware. I thinkEV7 just keeps a simpler policy of "always open", and it doesn'thave to keep a lot of state information around. This would ofcourseperhaps consume too much power for the desktop market, but I thinkthat the 4d rambus, combined with a chipset that could handle up to128 open pages per channel would have done really well (performancewise). But that will have to remain a conjecture ofcourse.Unless someone write some code to simulate it. ;)

I think you're probably right about Intel possibly having a mindset on page
management. Then again, the i440BX is doc'd as being able to manage 32
simultaneous, open pages across 8 ranks- why stop at 8 per channel with
DRDRAM?

I think you mean 4i Rambus - was it not?... vs. the 2x16d or 32d whatever
you want to call it. 128 open pages per channel would at least have put to
rest - one way or the other:-) - whether the layers of power management
came into play... again something we never really found out with the Intel
chipsets. More "state information" of course... which I assume the EV7
must carry?

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

Robert Myers

08-29-2003, 06:59 AM

On 29 Aug 2003 03:12:59 -0700, timsullivan2003@yahoo.com (Tim
Sullivan) wrote:
Robert Myers <rmyers@rustuck.com> wrote in message news:<ulktkvc4c911rfvhfrtt97p14a85knqbj8@4ax.com>... On 28 Aug 2003 16:50:54 -0700, timsullivan2003@yahoo.com (Tim Sullivan) wrote:Tony Hill <hilla_nospam_20@yahoo.ca> wrote in message news:<9258e6bb472fb8db3a6cb408a4bed5f2@news.1usenet.com>...>><snip>>> Tony Hill> hilla <underscore> 20 <at> yahoo <dot> caF*ck, even I would be embarrassed to write such a biased and spunmessage as this one. Over-emphasize the negatives and under-emphasizethe positives, a classical piece of propaganda.Me, you and George MacDonald("Let me, as an employee of someoneserious in the hardware business") are shills, and anyone reading ourmessages should realize our huge self-interests in how well Rambusdoes.

<snip> He was responding to points I had made with particular points that the supported the point of view he wanted to present. That's fairly normal for human discourse.
<snip>His post was not just replying to points in your post. It was just ablatant RDRAM bashing post, something that an employee of Micron,Infineon or Hynix might produce in a press release. I am sure someonewith the technical knowledge could write an equally negative andbiased post about SDRAM and DDR SDRAM.

Perhaps they could, but that doesn't mean that Tony's motivations
would be the same. Making assumptions and especially accusations
about the motivations of others is a risky enterprise; more often than
not it impugns the speaker more than it does those to whom the speaker
is referring.
On RDRAM, yes the original i820 was a piece of sh*t, but only justrecently did DDR-SDRAM match the performance and stability of the i850which was released many years ago. Also why compare bandwidth ofPC1200 RDRAM to PC3200 DDR-SDRAM, when most motherboards do not haveonly a single channel of each, it doesn't seem like a valid comparisonto me.

That's a better response to Tony than your first response. It's more
likely to get a hearing and a productive response.

I agree with you that the i820 isn't relevant to current concerns
about Rambus, but Tony's bringing up that bit of relatively ancient
history provides some insight as to where some of the deep-seated
resentment against Rambus is coming from and an opportunity to point
out that the fault can't all be laid at the door of Rambus, Inc.

It astonishes me that *Intel* has succeeded in portraying itself as
the victim in this matter. Intel has engineers, lots of them. Intel
knows how to make motherboards and it makes money selling them, lots
of it.

If the principals of Rambus made claims about the number of RIMM's
that could be put on a single Rambus channel and Intel accepted them,
especially when they were under pressure from Intel to do so, and
Intel accepted those claims without proof, the fault lies with Intel,
not with Rambus.

Furthermore, if Intel rushed a chipset into production with a
less-than-wonderful design, the fault lies with Intel, not with
Rambus.

If Intel ham-handedly shoved a memory technology onto the public when
there was not enough support from memory manufacturers to make the
memory available to the public at a reasonable price, the fault lies
with Intel, not with Rambus.

None of that explains or excuses other actions by Rambus that might
have been unattractive, less than ethical, or even illegal, but it
does help to explain part of people who *didn't* have a vested
interest in memory technology took such a strong (and perhaps
unjustified) posture in opposition to Rambus.

In particular, a member of the consuming public would have looked at
the price of RDRAM as compared to DDRAM, looked at the announced plans
of Intel, and decided that they were being set up. What the exercise
really proved is that a narrow manufacturing base for a new product
with little manufacturing experience will result in very high prices.
Intel tried to create demand for Rambus by forcing the issue with
i820, and the effort backfired--badly.

I've asked myself why Intel would have been so stupid as to have tried
to force Rambus onto a FSB that couldn't really use its capabilities.
I think the answer is that they realized that they were faced with two
tough sells at one time in the combination of the P4 and Rambus and
decided to make the transition to Rambus with a CPU, the P3, that had
good market acceptance. It dien't work. They wound up with just as
tough a sell for the P4, and went into it at a disadvantage on the
Rambus sell, having to make up for the bad impression the i820 had
created.

Now *I've* attributed motives to Intel that can't be proven. It's all
a matter of speculation, but if Intel doesn't engage in the kind of
market strategizing that I have speculated about, it's astonishing
they're still in business. That they would have such enormous
resources at their disposal to analyze and plan and blundered so badly
proves that experienced managers with lots of information can make bad
decisions. It doesn't prove much about Rambus memory technology or
even about Intel.

Most people are tired of the discussion. It's interesting to me
because I like to try to glean lessons from episodes like the Rambus
fiasco. Beyond that, it's history.

RM

Tony Hill

08-29-2003, 08:03 PM

On 29 Aug 2003 03:12:59 -0700, timsullivan2003@yahoo.com (Tim
Sullivan) wrote:Robert Myers <rmyers@rustuck.com> wrote in message news:<ulktkvc4c911rfvhfrtt97p14a85knqbj8@4ax.com>...His post was not just replying to points in your post. It was just ablatant RDRAM bashing post, something that an employee of Micron,Infineon or Hynix might produce in a press release.

If I could write press releases for companies like those, I would be
earning a whole heck of a lot more money than I do now! :>
I am sure someonewith the technical knowledge could write an equally negative andbiased post about SDRAM and DDR SDRAM.

Possibly, but that wouldn't have added anything useful to the original
post.
On RDRAM, yes the original i820 was a piece of sh*t, but only justrecently did DDR-SDRAM match the performance and stability of the i850which was released many years ago.

Very true, though that was in a large part due to lack of effort.
Also why compare bandwidth ofPC1200 RDRAM to PC3200 DDR-SDRAM, when most motherboards do not haveonly a single channel of each, it doesn't seem like a valid comparisonto me.

Ok, so compare dual channel PC1200 RDRAM (which, by the way, is not
supported in any chipset out there to the best of my knowledge, I was
using it assuming that if RDRAM development had continued it would be
supported now) to dual channel PC3200 DDR-SDRAM. Intel's top chipset
support two channels of DDR SDRAM, their top RDRAM chipset supported
two channels. nVidia's top chipset for AMD chips supports two
channels of DDR, ATI's newly announced chipset for the P4 will have
dual-channel DDR support, AMD's Opteron supports a 128-bit wide
channel of DDR, etc. etc. ad nauseum. Sure, you could point to the
Alpha EV7 that supports 10 channels of RDRAM, but then one could
equally point to the IBM Power4 which supports some monstrous width of
SDRAM bandwidth.

The simple fact of the matter is that for PC's available today, DDR
SDRAM offers more bandwidth than RDRAM. RDRAM offers more bandwidth
per pin, and theoretically it could be extended to a larger number of
channels, but I am skeptical that it could be done cheaply enough
(just from a chipset/motherboard side) to match dual-channel DDR
SDRAM.

---------------
Tony Hill
hilla <underscore> 20 <at> yahoo <dot> ca

The MyLounge.com forum is intended for informational use only and should not be relied upon and is not a substitute for any advice. The information contained on MyLounge.com are opinions and suggestions of members and is not a representation of the opinions of MyLounge.com. MyLounge.com does not warrant or vouch for the accuracy, completeness or usefulness of any postings or the qualifications of any person responding. Please consult a expert or seek the services of an attorney in your area for more accuracy on your specific situation. Please note that our forums also serve as mirrors to Usenet newsgroups. Many posts you see on our forums are made by newsgroup users who may not be members of MyLounge.com Term of Service