That could very well be explained by the Linux code being re-compiled for Linux on the GNU C compiler, which -though less sophisticated- does not cripple non-Intel CPUs.
It might not have to mean that it's a virtue of Linux itself.
Senior Post Whore
That could very well be explained by the Linux code being re-compiled for Linux on the GNU C compiler, which -though less sophisticated- does not cripple non-Intel CPUs.
It might not have to mean that it's a virtue of Linux itself.
Motivator & Enthusiast
Moderator
How many of these brainiac souls out there doing the testing actually understand that using Intel based testing tools is demoralizing the efforts of AMD projects. Just how many?
Challenge me, or correct me, but don't ask me to die quietly.
…Pursuit is always hard, capturing is really not the focus, it’s the hunt ...
Senior Post Whore
A lesson to be learned could be to use only non-Intel compilers when compiling for non-Intel CPUs and to use non-Intel libraries (e.g. Yeppp!) while compiling for those CPUs.
Anyone willing to re-compile a BOINC application -and perhaps even the very BOINC client- for his/her AMD CPU, using a non-Intel compiler and those Yeppp! libraries to test this?
I do predict big gains to be made for FX-8350 CPUs running Windows....
Last edited by Dirk Broer; 11-24-2013 at 10:44 PM.
Senior Member
Intel i3. Currently, it's the only model I've been able to try. Might play with some Amazon cloud servers to try it a bit more.Originally Posted by Dirk Broer
The linux distributable is downloaded already compiled, probably with Intel Compilers from what I gather. I've only once compiled it myself and that was on a Raspberry Pi.
Senior Post Whore
Can the CPU benchmarks of Pyrosim be helpful to provide aditional comparisons?
From what I gather, the Linux distributable was compiled using GCC -which might explain the difference in performance between Windows and Linux.
But then a rather poor C compiler (MicroSoft's?) must have been used for the Windows version that even Intel suffers...
Last edited by Dirk Broer; 11-24-2013 at 11:59 PM.
Senior Member
Those are the benchmarks I'm using - the Pyrosim ones.
Pyrosim is a java based front end for Fire Dynamics Simulator. Which is a FORTRAN based program, compiled using, AFAIK, the Intel Fortran compiler.
Senior Post Whore
Have more people collected data on Pyrosim like you did, per CPU/OS? Would they be willing to share those with you?
Last edited by Dirk Broer; 11-25-2013 at 02:50 PM.
Senior Member
Not sure - I think I'm the only one I know of.
If others have done it, they might have done it as part of their companies and kept it within that. I know one of our competitors has their own cluster for processing models. Most people that run the software don't really care about the nitty gritty - even researchers will tend to just run the models and suck it up on how long it takes. However, in industry, time is money as it were and we try to optimise the time it takes, whilst retaining accuracy.
AMD Users Alchemist
Moderator
Drezha, didn't we run a benchmark simulator a while ago with you? It may not have been Pyrosim, but I remember sending a few results from my machines here. Was it a different application?
2009 MULTI CORES CONTEST STATS
http://neogen.amdusers.com/contest2009/contestoverall.htm
CHRISTMAS QUEST CONTEST STATS
http://neogen.amdusers.com/contest2008/xmascontest.htm
http://neogen.amdusers.com/contest2008/stats.png (snapshot)
FEBRUARY '08 RACE STATS
http://neogen.amdusers.com/contest2008/racefeb08b.htm
http://neogen.amdusers.com/contest2008/racefeb08nb.htm
Senior Member
No that was the same (the results of which can be viewed here.)
Pyrosim is just a graphical front end for FDS - FDS was what we ran and what is affected by the compiler. As you can see from the results, I only have one machine that I've managed to run Linux and Windows Benchmarks on. I can't even be sure that the Windows one wasn't affected by Windows aggressive power saving features stopping the machine hitting the max CPU speed.
Posting Permissions
Reply With Quote