![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
At the moment the nVidia client has around twice the PpD of the ATI client.
But that will change as the clients get tweaked more and new protiens come out.
Because the nVidia card has a small number of quick shaders, it does better on small protiens, like the test protien it started with.
The ATI cards have a larger number of slower shaders, so it does better on the bigger protiens that are coming out.
Because all protiens are benched on ATI cards expect the PpD of the mid range ATI cards to stay fairly level, maybe a 50% increase in the top range cards.
On the other hand nVidia cards may have a 0-60% drop in PpD with new protiens.
So at the moment nVidia is better but tomorrow it may all change.
Luck .............
Note I am a beta tester for F@H so I can't give you state on production)
While I agree with you in part you have to have been here a number of years to totally understand my whole position.
This whole thing started out with VJs program under Stanford asking for our spare idle computer cycles. I was fine with that as was the team as a whole. It didnt take long for some teams to be formed and of course with the competition many people added a second or third or more machines. That too was ok and I am as guilty as anyone, I used to have between 26 and 29 boxen running at any one time.
At some point a few short years ago VJ and or someone on the crew figured out how stupid we who fold really are and the game changed. Suddenly if you wanted to play you were going to have to pay. Again, I was willing to pay to play right up to the point where it hit me that this was becoming not only abusive on the part of Stanford but many of us found it changing our lifestyles.
So, at the moment Im down to three machines and I have drawn a line in the sand so to speak where I will no longer throw good hardware after bad hardware away just to satisfy VJs lust for power. If and when the gear I have now becomes totally inefficient to provide a good cost to work unit ratio I will be done.
With three quads alone that is 40 freaking GHZ of computing power considering all four cores of all three quads are running all out. That doesnt even include the GPU power. Add up everyone folding Stanford has the largest computer in the world. If they cant use it properly, fuckem.
Call me crazy but I guess I'm not sure what your intentions are with this post. Are you interested in purchasing an ATI card or did you just want to create a post that you could (yet again) bash the f@h project's management?
So you drew a line in the sand that you werent donating any more hardware to the cause, then made a post asking about how the ATI cards were doing, and finally decided to derail your own post by ranting about how poorly this DC project has been run?
Bill I dont mean any disrespect here, I know you are a long time contributor to the project and I think we all appreciate that contribution. However to me this forum should be about encouraging current and potential folders to donate to medical research and further human knowledge. I cant help but feel that you constant negative posts about VJ and the f@h project can only be discouraging people from contributing to the project. I know that you are just trying to warn people not to dump boatloads of money into DEDICATED folding hardware but it just seems to me like many of your posts are just coming off as quite negative posts toward the project in general. Just something to think about.
Let me see if I can get this post back on track by talking about some ATI hardware. Apollo - I have a 3870 and I am getting right at about 2000 PPD with these lambda units. These units perform better than the older units for me as the older units were more CPU intensive and I only have a 2.4GHz Athlon X2 in that rig.
Seems like the ATI clients are still a bit more CPU intensive than the Nvidia units. My new 4850 is "only" doing about 2700 PPD where I have heard other estimates on the forums about 3000+ PPD with these cards. I can only assume my couple year old processor is to blame.