They are still tweaking the issues. They discovered they have to rewrite a lot of things because everything was based on the older architecture. The 30% increase was with using Fermi on the older GTX2xx architecture.
From GDF
Current performance on the GTX480 is 35% faster than the fastest version of ACEMD on a GTX285.
There is little more we can do I think. The cuda3 compiler seems much worse at optimizing the code than the 2.2 version. If this is the case, then 3.1 could bring some improvement.
Tomorrow, we clean up and submit the new application in beta.
gdf
PS: no doubts this is much less than expected
Last edited by =[PULSAR]=; 04-26-2010 at 07:35 PM.
Very disappointed in the performance figures quoted by GDF Looks like the 295 will still be the king of cards for a while. Glad I waited before ordering now
Update
we have now a good performance model of the application. The maximum we should have expected from fermi is a reduction in time of 60% compared to a gtx275. Actually, we now think that this will be achievable in the future. The problem is that cuda 3 is slower even on gtx200 by almost 15%. Indeed, we are 60% faster compared to both applications compiled for cuda3 but compared to cuda2.2 only 35%.
On our machine the running temperature is 91 degrees. Equivalent to a gtx275.
The only real problem is the price. Later i will report the time of the gtx470 and running temperature.
Gdf
Last edited by YukonTrooper; 04-27-2010 at 07:50 PM.
For my part I know nothing with any certainty, but the sight of the stars makes me dream.
..
Okay well fermi is up and running now just waiting on the gpugrid to release beta they said sometime tomorrow.
@Pulsar ... have you run any Collatz to get a feel for your OC? It runs a higher on the GPU utilization than GPUGrid so I think we will be stable if we max Collatz. Nice way to get a feel for the card in general, it's OC capabilities and the temps. Currently I am rock solid at 1672 but can push it to 1732 at stock v. I have not had much sucess OCing higher with more volts but I am sure I will be working on that again once GPUGrid is ready for us.
I'll give it a shot later on this evening trying to hammer out the last kinks in my WCG farm, almost had it finish and then HDD died. Back to using a prehistoric IDE
Isn't it a better idea to buy used GTX295s off eBay? Some of them are going under 300 dollars already.
just got mine today and been really frustrated with it . look at all these errors it's giving. i've been trying to fix it for over an hour already and still no luck.
it seems that nvidia only has one driver to cover all cards except for the new gtx400 series, which has it's own drivers. i updated it, and still doesn't work. it keeps going through a loop of downloading the WU, and instantly report it back as error, and DL's a new one, and same thing happens. any idea how i can just get it running or is fermi still not yet supported?
GDF says it is beta is supported for linux. He says it may or may not work on windows yet, some users are getting it to work on XP. I tried running it today on win7 and no luck.
yeah running win7 64 here and as soon as a unit start it errors and loads another.
maybe you could for the time being run gpugrid on linux using vmware?
I'm currently to addicted to the supersonic sled demo haha
That was happening to me also on Win7 64 ... then I turned my OC down until I started processing properly. I think I'm at high 15XX and stable.yeah running win7 64 here and as soon as a unit start it errors and loads another.
I also set the environmental variable they posted (SWAN_SYNC=0) which forces the GPUGrid app to use a full core per WU, they say you get a substantial (20-30%) performance increase.
Bookmarks