0
Quote
Subject: AMD benchmarks
Paul be very careful how you phrase the Intel advantage over AMD. It exists only for applications that have data sets of 4k (for 32 bit apps) or effectively 2Kfor 64 bit apps.
http://www.hpcwire.com/hpc/815242.html
When you move into the more serious apps like Unigraphics NX3 Intel starts to drag up the rear. Fujitsu Seimens had the V830 single core 256 actually outscore the R50 with the same graphics card.
http://www.spec.org/gwpg/Dec1_06/apc.data/...x3_summary.html
Fujitsu withdrew the benchmark. If you look at the current benchmarks a AMD dual core beats any comparable Intel offering. It depends on how difficult the benchmark is. NX3 is based on automotive design.which an order of magnitude more difficult than the routine benchmarks used in the trade press.
http://www.spec.org/...sh/nx3_rel.html
If you move on to something more challenging like DARPA's HPC Challenge Intel's best are something like half as fast in Fast Fourier Transforms and AMD actually beats Blue Gene L .
Last week's 2007 awards http://www.hpcchallenge.org/
The fact that Intel can't beat a 5 year old chip IBM P5 says a lot about the failures of Intel's design in complex computing. Results IBM 7 awards AMD 5 Intel 0. So for very dumb desktop apps Intel is significantly ahead fro more complex workstation apps AMD is slightly ahead, In complex matters like fast fourier series or "button to bang" design of nuclear weapons simulations, AMD takes 14-21 days less time on a 90 day simulation. Since DARPA alone buys twice as many high performance chips per year as all the gamers in the world, AMD has skewed its design to meet that market. It is like comparing a 100m dash runner with a marathon runner. Up to and including 400m, the 100m dash runner will win. Beginning at 800 meters he gets beat and anything beyond 1600meters he gets wiped out and it gets worse with the distance.
Ed Hinders.
Paul be very careful how you phrase the Intel advantage over AMD. It exists only for applications that have data sets of 4k (for 32 bit apps) or effectively 2Kfor 64 bit apps.
http://www.hpcwire.com/hpc/815242.html
When you move into the more serious apps like Unigraphics NX3 Intel starts to drag up the rear. Fujitsu Seimens had the V830 single core 256 actually outscore the R50 with the same graphics card.
http://www.spec.org/gwpg/Dec1_06/apc.data/...x3_summary.html
Fujitsu withdrew the benchmark. If you look at the current benchmarks a AMD dual core beats any comparable Intel offering. It depends on how difficult the benchmark is. NX3 is based on automotive design.which an order of magnitude more difficult than the routine benchmarks used in the trade press.
http://www.spec.org/...sh/nx3_rel.html
If you move on to something more challenging like DARPA's HPC Challenge Intel's best are something like half as fast in Fast Fourier Transforms and AMD actually beats Blue Gene L .
Last week's 2007 awards http://www.hpcchallenge.org/
The fact that Intel can't beat a 5 year old chip IBM P5 says a lot about the failures of Intel's design in complex computing. Results IBM 7 awards AMD 5 Intel 0. So for very dumb desktop apps Intel is significantly ahead fro more complex workstation apps AMD is slightly ahead, In complex matters like fast fourier series or "button to bang" design of nuclear weapons simulations, AMD takes 14-21 days less time on a 90 day simulation. Since DARPA alone buys twice as many high performance chips per year as all the gamers in the world, AMD has skewed its design to meet that market. It is like comparing a 100m dash runner with a marathon runner. Up to and including 400m, the 100m dash runner will win. Beginning at 800 meters he gets beat and anything beyond 1600meters he gets wiped out and it gets worse with the distance.
Ed Hinders.
(odd, my "old" big single core opties, even though not benching as high as newer xeons, ended up wiping the floor w/ the intel system. now i just need the dual quadro gpus...
)
u and me both MOman ......AMD still the rockin king imo 
