H.W. Stockman
04-08-2004, 12:11 PM
Would you help me by running a benchmark on AMD64 and P4 systems with
dual-channel DDR?
I am trying to decide between AMD64 and P4 (875 or 865xx) dual-channel PC3200
systems. Most of the machine's CPU cycles will be used to run just one program,
which happens to be very memory-intensive. This benchmark is a scaled-down
version of that particular program, with a realistic calculation load.
Two years ago, the available single-channel DDR solutions, with low-speed FSB,
didn't do too well on this benchmark, relative to RDRAM; I'm hoping the
situation has changed, since RDRAM doesn't seem to be going anywhere.
=====the test========
Here are some rather dated, painfully explicit instructions for running the test
under windows.
The Linux version is very similar in usage:
http://users.viawest.net/~hwstock/bench/3d0/how_3d0.doc
....If you try to open this file directly from the web, and you have Norton
Antivirus, you may get a complaint that this file has a "malicious script" (it
doesn't), probably because it has embedded web URLs. In any case, you may want
to download the file and scan it before you read it.
And here are the zipped benchmark executables, for windows (95 and up) and Linux
x86.
http://users.viawest.net/~hwstock/bench/3d0/3d0.zip
http://users.viawest.net/~hwstock/bench/3d0/3d0_linux.zip
The how_3d0.doc file explains what I need returned to me. BUT, if you would
rather just post results, you can use the "MUPS" number from the bottom of file
lb_data.txt. Two years ago, a 1.4 GHz P4 with PC800 RDRAN scared about 5.94
MUPs.
Please specify the mboard chipset, CPU, FSB, and RAM type/channel.
==================
If you feel more comfortable communicating via an edu or gov address, I can be
reached at:
hwstock _AT_ alum dot mit dot edu
or
hwstock _AT_ sandia dot gov
dual-channel DDR?
I am trying to decide between AMD64 and P4 (875 or 865xx) dual-channel PC3200
systems. Most of the machine's CPU cycles will be used to run just one program,
which happens to be very memory-intensive. This benchmark is a scaled-down
version of that particular program, with a realistic calculation load.
Two years ago, the available single-channel DDR solutions, with low-speed FSB,
didn't do too well on this benchmark, relative to RDRAM; I'm hoping the
situation has changed, since RDRAM doesn't seem to be going anywhere.
=====the test========
Here are some rather dated, painfully explicit instructions for running the test
under windows.
The Linux version is very similar in usage:
http://users.viawest.net/~hwstock/bench/3d0/how_3d0.doc
....If you try to open this file directly from the web, and you have Norton
Antivirus, you may get a complaint that this file has a "malicious script" (it
doesn't), probably because it has embedded web URLs. In any case, you may want
to download the file and scan it before you read it.
And here are the zipped benchmark executables, for windows (95 and up) and Linux
x86.
http://users.viawest.net/~hwstock/bench/3d0/3d0.zip
http://users.viawest.net/~hwstock/bench/3d0/3d0_linux.zip
The how_3d0.doc file explains what I need returned to me. BUT, if you would
rather just post results, you can use the "MUPS" number from the bottom of file
lb_data.txt. Two years ago, a 1.4 GHz P4 with PC800 RDRAN scared about 5.94
MUPs.
Please specify the mboard chipset, CPU, FSB, and RAM type/channel.
==================
If you feel more comfortable communicating via an edu or gov address, I can be
reached at:
hwstock _AT_ alum dot mit dot edu
or
hwstock _AT_ sandia dot gov