mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Software

Reply
 
Thread Tools
Old 2013-02-19, 19:07   #1
dtripp
 
Feb 2013

1 Posts
Default AMD vs Intel

I'm running p95v279 on two types of nodes, a dual 6 core opteron and a 4 core Xeon. The Xeons are wiping the floor with the Opterons even though the Opteron has 3 times as many cores and 96 Gigs of RAM compared to 16 for the Xeons. I'm guessing the pre-compiles binaries strongly favor Intel.

Has anyone tried compiling from source for AMD and seen better performance?
dtripp is offline   Reply With Quote
Old 2013-02-19, 20:09   #2
sdbardwick
 
sdbardwick's Avatar
 
Aug 2002
North San Diego County

23·5·17 Posts
Default

For Prime95 LL tests, Intel silicon really is that much better than AMD.

Rough performance comparison based on my observations:
4 bulldozer cores (Opteron 4280) @ 2.8 GHz =
2 K10 cores (Opteron 6128) @ 2.0 GHz <
1 i5-2500 core @ 3.4 GHz.

Intel's implementation of SIMD (AVX) along with better cache performance (and I believe better memory controllers) really kills AMD for Prime95.
sdbardwick is offline   Reply With Quote
Old 2013-02-19, 20:15   #3
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

160358 Posts
Default

Recompiling with different compiler options won't help: all the speed critical parts are written directly in assembly anyways, and despite George Woltman's best efforts, the AMD parts just aren't that fast.

Memory size doesn't play a role in speed (around 7-8MB required per test), and memory speed only plays a role insofar as nothing is bottlenecked.
Dubslow is offline   Reply With Quote
Old 2013-02-19, 20:20   #4
ewmayer
2ω=0
 
ewmayer's Avatar
 
Sep 2002
Rep├║blica de California

22·31·79 Posts
Default

See here for details on how AMD screwed the pooch when it comes to SIMD (SSE2 badly, AVX catastrophically).

There may be some gains to be realized from use of fused mul/add (in SSE2 mode only, though - AMD AVX is literally worse than useless here) on the more recent chips, but from the perspective of someone like George who needs to choose his coding time carefully, that's a lot of work just to get a poor chip design to "suck less".

Intel's forthcoming implementation of AVX2 will only make the performance disparity that much greater.
ewmayer is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Unknown Intel fidelitas Information & Answers 4 2014-08-20 18:36
Intel NUC nucleon Hardware 2 2012-05-10 23:53
Intel RNG API? R.D. Silverman Programming 19 2011-09-17 01:43
AMD or Intel mack Information & Answers 7 2009-09-13 01:48
Intel Mac? penguain NFSNET Discussion 0 2006-06-12 01:31

All times are UTC. The time now is 13:35.

Sat Oct 31 13:35:11 UTC 2020 up 51 days, 10:46, 2 users, load averages: 2.45, 2.50, 2.36

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.