Hey all, I'm trying to understand why I'm seeing such a huge difference in the credit rate awarded between my HD 6970 and my GTX 590.

Here are the links for comparison:

HD 6970

GTX 590

The following times and credits are relatively consistent:

Card / Run time(sec) / CPU time(sec) / Credit
HD 6970 / 2,232.75 / 13.04 / 13,173.48
GTX 590 / 963.40 / 963.40 / 3,763.20

Why would my GTX 590 receive only 25% of the credit of the HD 6970?

My GTX 590 finishes a task in less than 50% of the time, but receives 25% of the credit as my AMD card.

Please don't think this is a flame. I love the credits I get for my AMD card, but my more powerful GTX 590 isn't getting any love. Note that the GTX 590 is a dual GPU card and so the credit discrepancy seems even worse considering Moo! uses both GPUs per task.

Can someone explain how the credit works here?

Your packets and stat units are higher with your HD 6970. But I'm not sure if this is the only reason. Generally speaking, I've found a few clients prefer (mathematically that is) one particular GPU over the other.

---

It's because your AMD is MUCH faster for this project than your NVidia. The NVidia WU sizes are much smaller and therefore get fewer credits.

Can someone explain how the credit works here?

The credits per Stat Unit is the same for both NVidia and AMD cards, the difference - as Beyond pointed out above - is the AMD application works faster here than the NVidia application. So the more Stat Units processed per WU, the more credits awarded. You can see the numbers of Stat Units processed each time in the Task Name column of each individual WU inside Task Details for the PC, which shows the contents of the stderr file.

View it as the reverse of the Prime Grid Applications, where NVidia app is written in the native NVidia CUDA, and the AMD app is written in OpenCL (here the AMD app is written in native AMD Stream and therefore flys). At Prime Grid, the NVidia cards fly compared to AMD because of this. So at Prime Grid your 590 will bring in the credits by the bucketload, whereas the 6970 would crawl with below average rate of credits.

Regards
Zy

Also the faster ATI GPUs process the huge WUs because of their speed. The nVidia GPUs process the tiny and normal WU where smaller credit payouts are made.

See the Project Server status

Thanks for the explanations on this fine New Years Day :)

I definitely had some wrong assumptions as to how the credit was determined. Some projects appear to just give flat rates based on a per wu or per elapsed time basis and I had continued that thought here.

My intent is to spread my GPUs across the available projects efficiently, so I'll put my 590 somewhere else.

Maybe I'll get a 7970 when it comes out and see what it can do here. :)

Thanks and Happy 2012!

Thanks for the explanations on this fine New Years Day :)

I definitely had some wrong assumptions as to how the credit was determined. Some projects appear to just give flat rates based on a per wu or per elapsed time basis and I had continued that thought here.

My intent is to spread my GPUs across the available projects efficiently, so I'll put my 590 somewhere else.

Maybe I'll get a 7970 when it comes out and see what it can do here. :)

Thanks and Happy 2012!

A Happy 2012 to you springdus[MM].

Others put their ATI GPUs to ATI biases projects, like - the defunct DNETC, Moo! Wrapper, Milkyway, Collatz, etc.

Others place their nVidia GPUs on nVidia biased projects like PrimeGrid, etc.

Those with several PCs and mixed ATI and nVidia GPUs, not in the same box, can pick and choose their projects.

Those with several PCs and mixed ATI and nVidia GPUs, not in the same box, can pick and choose their projects.

Actually project flexibility is greater if the machine does have mixed ATI & NVidia GPUs.