Great news...I have a machine with an ATI card and another with a Nvidia. Either way I should be good to go.

HCC1 will not see quite that dramatic of performance improvement. It is faster though.

The goal of credit is represent the amount of 'research' done. Thus if a given workunit should result in the 15 credits - then regardless of the platform it is run on (linux, windows, gpu, mac) it should be awarded 15 credits. Credits purpose is to show how much 'compute capability' a user has contributed to World Community Grid.

For GPU processes, we will utilize the 'elapsed time' metric for how much 'runtime' is added to their statistics for the purposes of earning badges. If a user has a graphics card that is able to run the workunits, that card will then about 24 hours of time per day (assuming it is fully utilized for running research applications).

Maybe you could share what you know already and what BOINC shows if multiple jobs run on 1 card, as an education for the WCG volunteers to whom GPU crunching is entirely new.

Ah, an edit while typing: Whether there's a duo or an octo CPU, BOINC as said records the elapsed time for the GPU job. 1 job, 1 elapsed time recording in BOINC, so I understand from knreed's post.

--//--

If 2 tasks run simultaneously you can see both in BM. Both will share GPU resources, so doing this will use more GDDR, CPU time and bus width. Works on Fermi's and high end AMD cards, but not all cards.
If GPU utilization by the app is low (say <75%) using an altered app_info.xml file can improve overall task return. If GPU utilization is high however (say >85%) it either makes no difference or can even reduce overall task throughput (as both tasks compete for the same resources). Obviously this varies from project to project, and is app dependent. So depending on utilization the anonymous platform could be useful, or could be a bad thing.

Elapsed time would rise normally for both tasks. So my concern is that the anonymous platform could be misused, just to increase Run Time Rankings. On some projects people run 3 tasks on the one GPU, and some people have systems with 3 or 4 GPU's. They are doing this for the extra points - most projects don't have a Run Time credit system. If these HCC tasks are very small (<100MB), someone could conceivably be running 10 per GPU, just for the Run Time credit.

- So GPU count by Run Time, and then by SM count ;P:
The formula uses,
<coproc> <count>0.5</count> </coproc>, for 2 tasks or <coproc> <count>0.33</count> </coproc> for 3 tasks per GPU...

- I still think you should just award a separate Badge for GPU's based on credit, rather than Run Time.

i think equal credit for equal work sounds good.

i am currently running 5 cores (just fixed one of my computers)

i dont think any of my gpu's will run any of the projects but if they do great if not then fine.

the goal is to get work done. the badge process helps make it fun

if the gpu's come online as well then i am happy for whatever gpu's do the work

right now lets get everything going on the projects we have with what we have... if the HCC2 starts in the new year i will be happy but we need to make sure we get hcc1 finished.

any and every project on WCG is working for good things it doesn't matter what anyone runs we are helping get things cured .

i am 1 1/2 years out from thyroid cancer I survived... hopefully it will remain in remission.

cancer research is one of my main goals but i love working on all the other diseases as well

so basically
if i dont care how they reward badges for gpu... if someone manages to have a gpu that counts as 1 or 200 cores equivalent who cares... the goal is to get the projects done ok so they will get there sapphire badge quicker but we will also get the research done quicker.

lets cure everything before we die from it