Preliminary numbers for Nvidia GPU client

Now I'm curious. FahSpy says I'm running 5001 (5-135-1) work unit, 34secs per %, credit listed as 98, PPD 2490, on a dedicated core.

This is with a Q6600 @ 3.6GHz and a 8800GT OC @ 625/900MHz
 
Now I'm curious. FahSpy says I'm running 5001 (5-135-1) work unit, 34secs per %, credit listed as 98, PPD 2490, on a dedicated core.

This is with a Q6600 @ 3.6GHz and a 8800GT OC @ 625/900MHz

Which drivers are you running? With 177.35 my 8800 GTS 512 at 770 is averaging around 15-16 seconds per % on 5001 and 5002.


 
Anyone know what the highest performing single-slot NVidia card is currently?

 
Off hand, my 8800 GT is single slot....the 512 GTS may be, though I think that would be the highest if indeed there are single slots of that one.

 
Too bad there aren't any 200-series singles, probably way too hot.
 
Which drivers are you running? With 177.35 my 8800 GTS 512 at 770 is averaging around 15-16 seconds per % on 5001 and 5002.



Hmm.. I must be doing something wrong.. I just can NOt get the 177.35 driver to work for me, and I too have an 8800GTS 512..

I used Driver Sweeper on the Nvidia drivers... Then I uninstalled the Nvidia drivers from the Programs list... Then I restarted... Next I installed the drivers setup, then put the .inf file into the main 177.35 folder (Right..?). Then I installed the actual drivers, but it does that thing where it takes only 2 seconds then says "The system has not been modified" etc etc.

What am I doing wrong? :( I want to fold on my video card! (And hopefully one day, both cards :p)
 
8600GT (256mb @ 691/891 & my shader clock is @ 1898) and using the 174.55 drivers
Opty 165 @ 2.2GHz
2gb DDR
approx. 2117 ppd!!!!!!!!!!!!!!!!!
 
And now for the dumb @ss question of the year or maybe even the century :rolleyes: (nahh.., but obviously you can tell I don't know "jack" about graphics cards et al). :D

Will the EVGA 384-P3-N851-AR GeForce 8800 GS 384MB 192-bit GDDR3 PCI Express 2.0 x16 HDCP Ready SLI Supported Video Card (copied from NewEgg as you probably can tell ;) ) do any justice to the GPU folding deal, because if it will, I found it on NewEgg for the low price of just a $119 smackers. :p

Before I make any decisions or get the "clicking finger syndrome" I'll be sure to check with Crazy Larry because I don't think anyone has a catalog of deals as good as his. :eek:

Thanks for any input from a "graphics card information challenged" person :p

You know, up until a little while ago I didn't pay much attention to the pernts. I was more concerned with the science and finding a cure, of course I still am, but the pernts have attracted more of my attention than before. Perhaps this is a phychological method of Stanford's to get more people involved in F@H or give those already involved, a new method of increasing their production and point standing ? :confused:

Edit: :eek: I think I'm morphing into a male pernts w[H]ore :eek: (nahh... IMHO it's gotta' be all for the cure no matter how many pernts you have)

 
Oh, and this is what FahSpy is saying.

[04:28:30] Trying to unzip core FahCore_11.exe
[04:28:31] Decompressed FahCore_11.exe (1818624 bytes) successfully
[04:28:36] + Core successfully engaged
[04:28:41]
[04:28:41] + Processing work unit
[04:28:41] Core required: FahCore_11.exe
[04:28:41] Core found.
[04:28:41] Working on Unit 01 [June 18 04:28:41]
[04:28:41] + Working ...
[04:28:53] CoreStatus = C0000135 (-1073741515)
[04:28:53] Client-core communications error: ERROR 0xc0000135
[04:28:53] This is a sign of more serious problems, shutting down.
 
And now for the dumb @ss question of the year or maybe even the century :rolleyes: (nahh.., but obviously you can tell I don't know "jack" about graphics cards et al). :D

Will the EVGA 384-P3-N851-AR GeForce 8800 GS 384MB 192-bit GDDR3 PCI Express 2.0 x16 HDCP Ready SLI Supported Video Card (copied from NewEgg as you probably can tell ;) ) do any justice to the GPU folding deal, because if it will, I found it on NewEgg for the low price of just a $119 smackers. :p

Before I make any decisions or get the "clicking finger syndrome" I'll be sure to check with Crazy Larry because I don't think anyone has a catalog of deals as good as his. :eek:

Thanks for any input from a "graphics card information challenged" person :p

You know, up until a little while ago I didn't pay much attention to the pernts. I was more concerned with the science and finding a cure, of course I still am, but the pernts have attracted more of my attention than before. Perhaps this is a phychological method of Stanford's to get more people involved in F@H or give those already involved, a new method of increasing their production and point standing ? :confused:


The thing about the GPU2 client is that it puts out a ridiculous amount of points and science. I'm pushing 4700 PPD and 48 WUs per day at the rate I'm rolling my 8800 GT. As for the GS, I don't see why it wouldn't hit at least 3000 PPD, but I would recommend searching here or on the folding forums for people using that card first.

 
GPU: XFX Geforce 8600GTS (Using 174.55)
CPU: AMD Athlon 64 3200+ (939)
CPU load while folding: 100%
Board Connection (PCI-E, AGP, etc.): PCI-E
Memory used by the client: 15MB
Temps of the GPU (NVMonitor): 66-69 C
GPU Clocks: 675/1000
PPD (as stated by FAHMON or other program): 1690-1700 PPD
 
I'm getting 5644.8 PPD with the card in sig at said speeds.

My first three WU all ended up being EUE though.
 
GPU: Leadtek 8800GTS 640MB
CPU: AMD X2 3800+ @ 2,53GHz
CPU load while folding: 80-100%
Board Connection (PCI-E, AGP, etc.): PCI-E
Memory used by the client:~15 megs for the core, 4 megs for the client
Temps of the GPU (Rivatuner): 78C GPU, 59C GPU Ambient
GPU Clocks: Core/Mem 648/945
PPD ~2822
 
And now for the dumb @ss question of the year or maybe even the century :rolleyes: (nahh.., but obviously you can tell I don't know "jack" about graphics cards et al). :D

Will the EVGA 384-P3-N851-AR GeForce 8800 GS 384MB 192-bit GDDR3 PCI Express 2.0 x16 HDCP Ready SLI Supported Video Card (copied from NewEgg as you probably can tell ;) ) do any justice to the GPU folding deal, because if it will, I found it on NewEgg for the low price of just a $119 smackers. :p

Before I make any decisions or get the "clicking finger syndrome" I'll be sure to check with Crazy Larry because I don't think anyone has a catalog of deals as good as his. :eek:

Thanks for any input from a "graphics card information challenged" person :p

You know, up until a little while ago I didn't pay much attention to the pernts. I was more concerned with the science and finding a cure, of course I still am, but the pernts have attracted more of my attention than before. Perhaps this is a phychological method of Stanford's to get more people involved in F@H or give those already involved, a new method of increasing their production and point standing ? :confused:

Edit: :eek: I think I'm morphing into a male pernts w[H]ore :eek: (nahh... IMHO it's gotta' be all for the cure no matter how many pernts you have)




READ THE FIRST POST! That is what I am using. I got it on sale from Fry's after a MIR for about $90.
 
GPU: eVGA 8800GT
CPU: X3350(Q9450) @3.56
CPU load while folding: 100 cause i'm running a winSMP as well
Board Connection (PCI-E, AGP, etc.): PCI-E
Memory used by the client: about 15meg for the core
Temps of the GPU (NVMonitor): forgot to check but low, its WC'ed
GPU Clocks: Core/Mem stock for now 600/900
PPD (as stated by FAHMON or other program): 2420 by fahmon :(

I haven't played with this much yet. Only had about 20 mins last night with the stock 174.55 driver. Do we have issues here where it could be running 2D clocks instead of 3D? Cause I feel like even with the winSMP i should be getting more than 2400PPD.
 
GPU: eVGA 8800GT
CPU: X3350(Q9450) @3.56
CPU load while folding: 100 cause i'm running a winSMP as well
Board Connection (PCI-E, AGP, etc.): PCI-E
Memory used by the client: about 15meg for the core
Temps of the GPU (NVMonitor): forgot to check but low, its WC'ed
GPU Clocks: Core/Mem stock for now 600/900
PPD (as stated by FAHMON or other program): 2420 by fahmon :(

I haven't played with this much yet. Only had about 20 mins last night with the stock 174.55 driver. Do we have issues here where it could be running 2D clocks instead of 3D? Cause I feel like even with the winSMP i should be getting more than 2400PPD.

Hmm, that's low. Try killing the SMP client and see if it improves. I'm guessing if you installed the GPU client with idle priority and aren't using something to assign affinity, the SMP and GPU clients could very well be fighting for cycles.
 
GPU: BFG Geforece 8800GTS OC 640MB
CPU: Q9450
CPU load while folding: 25%
Board Connection (PCI-E, AGP, etc.): PCI-E
Memory used by the client: 4MB
Temps of the GPU (NVMonitor): 43-44 C
GPU Clocks: 702/1000
PPD (as stated by FAHMON or other program): 2919 to 3024PPD

I also run a single VM instance with Xubuntu 8.04 on 2 cores and a console client on the 3rd for a total of about 6000 ppd.



Xilikon,

im using X3350
running single instance VM Ubuntu SMP on 2 cores, and 8800 GT,
Question for you, how do we set the console client? What cpu core should we assign to the console...

thanks

btw ppd now, around 6700, 2500 on SMP vmware, and 4200 on GPU2
 
Xilikon,

im using X3350
running single instance VM Ubuntu SMP on 2 cores, and 8800 GT,
Question for you, how do we set the console client? What cpu core should we assign to the console...

thanks

btw ppd now, around 6700, 2500 on SMP vmware, and 4200 on GPU2

You set the console client as normal but make sure to configure it so the cpu usage is set to idle. Set the GPU2 to "slighty higher" so it will grab a core by itself. Same with SMP by setting to idle. It's basically setting everything to idle excepted the GPU2 and let them sort themselves.

4200ppd with a 8800GT is normal :)

 
You set the console client as normal but make sure to configure it so the cpu usage is set to idle. Set the GPU2 to "slighty higher" so it will grab a core by itself. Same with SMP by setting to idle. It's basically setting everything to idle excepted the GPU2 and let them sort themselves.

4200ppd with a 8800GT is normal :)


wooohooo, thanks man
im so excited... :D
 
Hmm, that's low. Try killing the SMP client and see if it improves. I'm guessing if you installed the GPU client with idle priority and aren't using something to assign affinity, the SMP and GPU clients could very well be fighting for cycles.

Yeah that's what its looking like. I'll play with it more tonight. I just checked the stats, our WU per update count went through the roof because of the new nvidia client.
 
Imitation, a easy fix is to set the SMP to idle and GPU2 to "Slighty higher". If things works fine, it should claim a core by itself, leaving the rest to SMP to eat ;)

 
So tempted to purchase this and put it in my low profile comp at home...ppdpd (points per day per dollar) looks to be fantastic ;)
 
So tempted to purchase this and put it in my low profile comp at home...ppdpd (points per day per dollar) looks to be fantastic ;)

A 8800GT is a better value (4200 ppd for about 100-150$ depending on deals) vs the average of 1000-1500 ppd for 75$.

 
I'm holding out to see what sort of numbers the 9600GT cards are going to put out. At full load they run a TON cooler than the 8800 series but - if the point gap is too big it may be worth it to jump on the 8800 bandwagon. As much as I'd love to throw a newer 200 series card at folding I simply cannot justify the cost as I'm not a gamer and thinking $500+ for a 7000ppd card or $450ish for 3 4000ppd cards really has me considering picking up a few cards that are cheaper.
 
There are 9600GT results at the Folding Forum.

toTOW wrote:Well it's time for numbers from my nVidia boards :

9600 GT @ 775 MHz on a Q6600 @ 3 GHz : 4032 PPD

A used 8800GT would probably be a better idea.
 
READ THE FIRST POST! That is what I am using. I got it on sale from Fry's after a MIR for about $90.


I appologise (I think) :confused: Not knowing much about graphics cards I take it the card in the first post (no need to yell by using caps) is the identical card I mentioned in my post :confused: :rolleyes:

You see I'm not a gammer, I've never said I was and IMHO anything to do with graphics cards is the domain of a gammer. I was just trying to find the cheapest card for the job and the whole thing just might be some Stanford BS anyway like the illustrious GPU1. Currently I'm waiting on more results for the GPU2 client and scanning many articles, posts, opinions, etc. :)


IMHO one of the reasonable and faily accurate ways to learn about anything computer wise is to ask questions on this forum and one of the many other forums on the "Net". ;)


 
Sorry for using caps. Was working at work and we have to use caps for certain apps.
 
The real question is what will a GT280 do when all 280SP are used. Right now I don't think the cards are truly showing their potential. Am I wrong? Seeing as a stk GTX = 4300 and a GT280 is seeing 6500... I would have expected the GT to be upward of 8-9k...

Thoughts?

 
Sunin, that's exactly why Scott Le Grand commented the numbers is low on the GTX 280. I guess a future driver update will improve the performance.

 
Does Stanford have preliminary data on the 4000-series Ati cards? They're rumored to have 800 stream processors, even though that's not exactly the same as Nvidia SP.

Also, why is there such a wide variance in performance between the 8800GT cards? I'm seeing anywhere from mid-2000s to over 5000PPD... :confused:

 
Does Stanford have preliminary data on the 4000-series Ati cards? They're rumored to have 800 stream processors, even though that's not exactly the same as Nvidia SP.

Also, why is there such a wide variance in performance between the 8800GT cards? I'm seeing anywhere from mid-2000s to over 5000PPD... :confused:


Some numbers I saw on 4850 testing gave about 4000-5000 ppd. Mike Houston said they need to tweak the driver to improve performance since they are right now not fully utilizing (ATI is using not enough power and NVIDIA is using too much, causing some display issues and lag).

As for the 8800GT, there is 2 factors : GPU overclocking and CPU utilization. If you are using SMP and GPU2 with the same priority, they will battle for cpu cycles and usually the GPU2 will suffer.

 
Also, why is there such a wide variance in performance between the 8800GT cards? I'm seeing anywhere from mid-2000s to over 5000PPD... :confused:

Agreed. My GT was showing 5600PPD when I left for work this morning.

I wonder if it's driver related. I'm currently using the leaked 177.23 drivers. Given, that's also with core/shader/memory clocks of 750/1850/1050 and a 3.5GHz core feeding it.

But I'm hanging even with g92 GTS cards with similar clock speeds. I would have thought with 16 more stream processors, the GTS cards would perform noticeably better.

I must be getting the sweet WUs. Projects 5001 and 5002 are giving me these results:

GPU: BFG Geforce 8800GTS OC 512MB
CPU: E6750 @ 3.52
CPU load while folding: 50+%
Board Connection (PCI-E, AGP, etc.): PCI-E
Memory used by the client: 5MB or 15MB (Am I looking for the Fahcore or [email protected]?)
Temps of the GPU (RivaTuner): 63 C
GPU Clocks: 675/972
PPD (as stated by FAHMON or other program): 4980PPD

edit: OC'd the card to 756/1053. FahMon showing 5292PPD.

Same CPU as me. He's running at roughly the same speed. We're showing the same CPU load. His GPU clocks are very similar to mine. I was showing about 5200-5600PPD on WUs, so pretty similar there. Difference...his g92 GTS to my g92 GT.
 
Some numbers I saw on 4850 testing gave about 4000-5000 ppd. Mike Houston said they need to tweak the driver to improve performance since they are right now not fully utilizing (ATI is using not enough power and NVIDIA is using too much, causing some display issues and lag).

As for the 8800GT, there is 2 factors : GPU overclocking and CPU utilization. If you are using SMP and GPU2 with the same priority, they will battle for cpu cycles and usually the GPU2 will suffer.


Yeah From what I have heard the new ATI cards will have independently clocked shaders (like the Nvidia cards) so I think we should see a decent jump in folding power.

I agree. It seems that not only the number of SPs but the speed of the SPs have a big effect on performance. As well as the CPU power to feed all the SPs. OC'ing the SPs should have a performance gain more so than OC'ing the core or RAM.
 
Yes, I just read a thread about overclocking and GPU2 effect. It seems that the core or memory speed isn't affecting the folding speed that much but playing with the shader speed gave a big boost. You can technically lower the core and memory to make it run cooler but push the shaders clock high in the 2000 mhz area.

 
Same CPU as me. He's running at roughly the same speed. We're showing the same CPU load. His GPU clocks are very similar to mine. I was showing about 5200-5600PPD on WUs, so pretty similar there. Difference...his g92 GTS to my g92 GT.
That's really strange. I guess Xilikon is right and it may have to do with the way the clients are set up, and as you say possibly the drivers. It just makes it hard to decide which card is best to run on a bang for the buck basis when the data is all over the place. So far it's looking like the 8800GT, but what if we get one and it proves to be on the lower end of the performance scale? Too bad there's no way of knowing for sure before committing to the product.
 
I see no indication that folding bang for the buck is going to differ from gaming bang for the buck.
 
Let things macerate for a few days and we will have a clearer picture ;)

 
I just wanted to note that everyone that is talking about performance, these are beta drivers and a beta client for nvidia folding.

Everything takes time to get working the right way, we just need to give the folding guys some time to get the client up to par. After all it is the very first release for us to test out for them and to find and report any problems that we find so that they can get everything working. Then we can compare stats and drivers and PPD's, but for now we should just enjoy that we can finally fold with our speedy nvidia cards. :D

And I can under stand wanting to make sure that you have the best possible hardware to run this on, I cant wait for native 64bit windows support as well as sli support. Then we can all say that today is a good day to fold.:D
 
Same CPU as me. He's running at roughly the same speed. We're showing the same CPU load. His GPU clocks are very similar to mine. I was showing about 5200-5600PPD on WUs, so pretty similar there. Difference...his g92 GTS to my g92 GT.

The only question is what are his SPs clocked at? You said yours are running at 1850 which is OC'ed quite a bit. If his are running at stock they are probably more in the 1500 - 1600 range.

A OC on the core and RAM gave Dalemite a 300 PPD gain. He didnt say whether he OCed the shaders as well.

It would be interesting to know what kinda of PPD you would get if you ran your 8800GT at the same core and RAM clocks but used a 1600 SP clock.
 
Back
Top