• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

How to use SSE with AMD!!!!!

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
gtsimmo said:
nice find kendan. got the SSE turned on and folding away. But is it ok with stanford to use the beta client in this way? And dont Beta clients expire after a certain point?

It is a quick fix for the problem of Stanford changing things and not making a way to use SSE without the beta client. Personally I think they should have waited to disable SSE till they had a non beta client out for it or made a new flag for turning it on. But they didn't:mad: So we have to do what we have to do to get it to work. I do not know if this beta client will expire but they should have an official release before then anyways that will hopefully make it easier to turn on SSE for the rigs that can handle it:D
 
anyone knwo how to change the service installed thing so it runs SSE? i know you overwrite the thing but where does it dothe tags stuff or where in teh reg etc.. i have checked in services and it appears to start up a program called srvany.exe wich probably starts up F@H after it... aany ideas?
 
CrashOveride said:
anyone knwo how to change the service installed thing so it runs SSE? i know you overwrite the thing but where does it dothe tags stuff or where in teh reg etc.. i have checked in services and it appears to start up a program called srvany.exe wich probably starts up F@H after it... aany ideas?

just click the link I just posted: http://forum.oc-forums.com/vb/showthread.php?s=&threadid=179298


the tags go in the registry and you still need follow the steps I posted in the first post:D
 
I am running the Linux console on a XP1600. I tried to get it to use SSE with no luck. Is there a way to do this? Maybe I am at the wrong forum?
 
you can use the flags with linux. you can also use them with Wine and linux. I had problems with wine flags and the beta windows client. I suggest using wine with the windows client and just holding out until they get this issue taken care of before trying to make SSE work:D Using wine should give you a boost in speed. Good luck!!!
 
I am fairly certain that most of our team is made up of AMD processors, and, if not most, at least a substantial portion. I suspect that other teams have very substantial numbers of AMD cpu's as well.

The question then becomes, why is Stanford making it so difficult to take advantage of our AMD cpu's? At a minimum, they don't seem to care; worst case, they seem to be actively supporting the Intel platform.
 
Last edited:
flounder43 said:
I am fairly certain that most of our team is made up of AMD processors, and, if not most, at least a substantial portion. I suspect that other teams have very substantial numbers of AMD cpu's as well.

The question then becomes, why is Stanford making it so difficult to take advantage of our AMD cpu's? At a minimum, they don't seem to care; worst case, they seem to be actively supporting the Intel platform.

Well the stability problem was just with AMD processors. There fix is workable but a pain. hopefully they will have a better solution soon:D I know you are thinking the Intel connection and the thought has crossed my mind. I am just going to think positively and hope that that is not the case. Until then turn on SSE but make sure your rig is stable using SSE.
 
Does this mean i did it right?

[17:25:21] Folding@home Gromacs Core
[17:25:21] Version 1.48 (May 7, 2003)
[17:25:21]
[17:25:21] Preparing to commence simulation
[17:25:21] - Assembly optimizations manually forced on.
[17:25:21] - Not checking prior termination.
[17:25:22] - Expanded 511206 -> 2524929 (decompressed 493.9 percent)
[17:25:22]
[17:25:22] Project: 909 (Run 48, Clone 0, Gen 2)
[17:25:22]
[17:25:22] Assembly optimizations on if available.
[17:25:22] Entering M.D.
[17:25:43] (Starting from checkpoint)
[17:25:43] Protein: p909_vill_str0
[17:25:43]
[17:25:43] Writing local files
[17:25:46] Completed 120000 out of 250000 steps (48)
[17:25:46] Extra SSE boost OK.

Edit: Shaved about a minute off my frame time. Last frame was 8:21 on the p909_vill_str0 protien.
 
Last edited:
flounder43 said:
The question then becomes, why is Stanford making it so difficult to take advantage of our AMD cpu's? At a minimum, they don't seem to care; worst case, they seem to be actively supporting the Intel platform.
They really aren't. We should consider ourselves LUCKY that they provided a workaround so we can enable SSE again. The problem was/is that T-Breds were/are failing on certain WU's with SSE enabled, but NOT with 3DNow. So, in order for Stanford to get USEABLE results, they needed to (by default) have all AMD CPU's NOT run SSE. Now this problem does not exist with Pally XP's, only with the T-Breds. In a perfect world, the client could differentiate between the 2 and only USE 3DNow on the WU's that are known to crash, or even better, just not send T-Bred CPU's the WU's that they have problems with with SSE enabled. It is very simple to re-enable the SSE.
 
nikhsub1 said:

They really aren't. We should consider ourselves LUCKY that they provided a workaround so we can enable SSE again. The problem was/is that T-Breds were/are failing on certain WU's with SSE enabled, but NOT with 3DNow. So, in order for Stanford to get USEABLE results, they needed to (by default) have all AMD CPU's NOT run SSE. Now this problem does not exist with Pally XP's, only with the T-Breds. In a perfect world, the client could differentiate between the 2 and only USE 3DNow on the WU's that are known to crash, or even better, just not send T-Bred CPU's the WU's that they have problems with with SSE enabled. It is very simple to re-enable the SSE.

There fix sucks!!!

They should have made a new flag not used one that was for something else. That is just my first complaint. they made it so you have to jump thru too many hoops to get it turned back on. I really hope they find a way to make it easier in the future.
 
Mizzery said:
how soon before i know wether my comp is stable for sse?


Just keep an eye on the log and make sure it is completing workunits and not getting errors and starting over.
 
Kendan said:


There fix sucks!!!

They should have made a new flag not used one that was for something else. That is just my first complaint. they made it so you have to jump thru too many hoops to get it turned back on. I really hope they find a way to make it easier in the future.
Am I missing something here? Hoops? All you need to do (as you know) is download the beta client and delete the gromacs core. That's it. What is so hard about that? Most already have the -forceasm flag if they run gromacs anyway, I know I do on ALL my machines.
 
Seems to speed things up here on my side.. But i'll keep an eye out for the errors mentioned previously. Folding, folding away!
 
Back