NvNews + OSNN + SP Folding Team
- Thread starter Sazar
- Start date
- Status
- Joined
- 5 Apr 2002
- Messages
- 13,314
Mainframeguy
Debiant by way of Ubuntu
- Joined
- 29 Aug 2002
- Messages
- 3,763
NOTE - this is available in beta NOW....
OK now I should like to hear back from you guys who is running this client and on what GFX card? I realise some of you may be Nvidia users, but I am all ATI (9800 AGP :dead: ) so I just want to know if I'll be the first...
OK now I should like to hear back from you guys who is running this client and on what GFX card? I realise some of you may be Nvidia users, but I am all ATI (9800 AGP :dead: ) so I just want to know if I'll be the first...
- Joined
- 11 Feb 2004
- Messages
- 3,406
damn, I might need to get a ps3 just to fold on. right before the ps4 come out.
I'll be installing it this afternoon to run it on my x1900xtx. From what I've read, only the X1900 series is supported right now with support for the x1800 cards coming a bit later.
Actually, it's not silly at all. The folks @ Stanford have detailed exactly why it can't be done on current Nvidia hardware.
http://folding.stanford.edu/FAQ-ATI.html
"The R580 (in the X1900XT, etc.) performs particularly well for molecular dynamics, due to its 48 pixel shaders. Currently, other cards (such as those from nVidia and other ATI cards) do not perform well enough for our calculations as they have fewer pixel shaders. Also, nVidia cards in general have some technical limitations beyond the number of pixel shaders which makes them perform poorly in our calculations."
^^ And those limitations are detailed in this interview with Mike Houston from Stanford.
http://www.beyond3d.com/forum/showthread.php?t=34122
What ATI can do that NVIDIA can't that is currently important to the folding code being run is that we need to dynamically execute lots of instructions per fragment. On NVIDIA, the shader terminates after 64K instructions and exits with R0->3 in Color[0]->Color[3]. So, on NVIDIA, we have to multi-pass the shader, which crushes the cache coherence and increases our off-chip bandwidth requirements, which then exacerbates the below.
The other big thing for us is the way texture latency can be hidden on ATI hardware. With math, we can hide the cost of all texture fetches. We are heavily compute bound by a large margin, and we could actually drive many more ALUs with the same memory system. NVIDIA can't hide the texture latency as well, and perhaps more importantly, even issuing a float4 fetch (which we use almost exclusively to feed the 4-wide vector units) costs 4 cycles. So NVIDIA's cost=ALU+texture+branch, whereas ATI is MAX(ALU, texture, branch).
While it would be possible to run the code on the current NVIDIA hardware, we would have to make pretty large changes to the code they want to run, and even past that, the performance is not great. We will have to look at their next architecture and re-evaluate. The next chips from both vendors should be interesting.
Mainframeguy
Debiant by way of Ubuntu
- Joined
- 29 Aug 2002
- Messages
- 3,763
I'll actually be looking into one of these - but not until those cards are what I perceive as value (and also with a helluva decent case for cooling - because when I have a Duo or Quatro CPU and a couple of those crossfiring and folding - that baby is going to be my winter heating for the office I reckon! Might be a folding monster though :twisted:
Alrighty, here's my results from my first night of Folding with my x1900xtx.
First of all, running this will use one core on your CPU completely as well as stressing your card 100%. The video card does not speed up to 3D clocks automatically when running the app. It stresses it 100%, but keeps it @ it's 2D clocks. I was able to run the card @ it's 3D clocks simply by using ATITOOL and there was no problem and the speed gains were linear, as one would expect.
The folks over @ Stanford said the program would stress the video card much less than 3D apps.......very misleading. The temps on my GPU were exceeding those I see while playing F.E.A.R. for hours or looping 3DM06 with stock cooling and fan speeds. Of course I cranked up the fan to 100% overnight to keep the temps down and it ran flawlessy all night. So when they say it stresses the GPU less it's simply because it runs @ 2D clock speeds by default.
I would be very weary of running by video card 24/7 on stock cooling, especially if you don't manually set your fan speed to 100%. That being said, I will be running mine full time because I'm putting it back on water this weekend so my temps won't be a concern.
Now, for what you guys care most about......how many points did I get for one WU. I got 353 points for a unit that took 8:45 to finish @ 3D clocks and would have taken 11:10 to finish @ 2D. Which would obviously be ~700 points in a 24 hour period while running @ 2D clocks and 25% - 30% more @ stock 3D clocks. Of course, you can also continue to run one more instance of folding on the second core of your CPU so the returns from one just machine folding full-time could be huge.
So from just these rough numbers I think it's safe to conclude that this one PC can get me close to a thousand a points a day if I run one GPU and one CPU client 24/7. It seems pretty good to me.
I'm off to the dentist......but I'll check back later today to see if you guys have any questions for me.
First of all, running this will use one core on your CPU completely as well as stressing your card 100%. The video card does not speed up to 3D clocks automatically when running the app. It stresses it 100%, but keeps it @ it's 2D clocks. I was able to run the card @ it's 3D clocks simply by using ATITOOL and there was no problem and the speed gains were linear, as one would expect.
The folks over @ Stanford said the program would stress the video card much less than 3D apps.......very misleading. The temps on my GPU were exceeding those I see while playing F.E.A.R. for hours or looping 3DM06 with stock cooling and fan speeds. Of course I cranked up the fan to 100% overnight to keep the temps down and it ran flawlessy all night. So when they say it stresses the GPU less it's simply because it runs @ 2D clock speeds by default.
I would be very weary of running by video card 24/7 on stock cooling, especially if you don't manually set your fan speed to 100%. That being said, I will be running mine full time because I'm putting it back on water this weekend so my temps won't be a concern.
Now, for what you guys care most about......how many points did I get for one WU. I got 353 points for a unit that took 8:45 to finish @ 3D clocks and would have taken 11:10 to finish @ 2D. Which would obviously be ~700 points in a 24 hour period while running @ 2D clocks and 25% - 30% more @ stock 3D clocks. Of course, you can also continue to run one more instance of folding on the second core of your CPU so the returns from one just machine folding full-time could be huge.
So from just these rough numbers I think it's safe to conclude that this one PC can get me close to a thousand a points a day if I run one GPU and one CPU client 24/7. It seems pretty good to me.
I'm off to the dentist......but I'll check back later today to see if you guys have any questions for me.
Mainframeguy
Debiant by way of Ubuntu
- Joined
- 29 Aug 2002
- Messages
- 3,763
darn it ! I hit 796 and no more - so near and yet so far from brown!
- Status
Members online
No members online now.
Affiliates
Latest profile posts
Xie wrote on Electronic Punk's profile.
Impressed you have kept this alive this long EP! So many sites have come and gone. 
Just did some crude math and I apparently joined almost 18yrs ago, how is that possible???
Just did some crude math and I apparently joined almost 18yrs ago, how is that possible???
Electronic Punk wrote on Sazar's profile.
Rest in peace my friend, been trying to find you and finally did in the worst way imaginable.