HCMD2 is not even the last project to choose for making performance analysis, it is definitely the project to avoid for this purpose.
This is due to the way WUs are built with the parent/child/grand-child/grand-grand-child concept, and this concept itself is due to the fact that if WUs were built traditionally they would range from a few hours to several days as we have seen during the beta tests.
So forget about HCMD2 if you want to build runtime comparisons. Even the average of one day of results in the same machine/OS is barely meaningful from one day to the next.

As has been said HCC is the best project for such analysis because it is the most homogeneous as far as runtimes are concerned.
HFCC and FAAH are less homogeneous but they can still give interesting information as long as you work on a big enough number of results (several tenths).

Sorry for not mentioning boinc versions.
For windows was 6.6.36 , and for linux Boinc was 6.4.5.
As for power profiles , in win i had High Perfomance profile active.

I know use Vista Sp2 32bit, with boinc version 6.6.38 and after several HCC projects i will post again the cpu times.
Cheers. Harris Anastopoulos

Well, projects such as GPUGrid are working with CUDA, and while they are being very successful, it can be hard for some people to get it working. Sometimes it works great first try, sometimes it takes a little tinkering, and that's not WCG's style. They work to make helping as simple as downloading a single program and forgetting about it. You can do more, but they don't want you to have to.

I'm sure WCG is working on implementing CUDA (as they have said many times), but they need time to get a project well suited to it, get the scientists on board, and work out all the kinks so it meets the set and forget feature.

So many people argue that we should be pushing the envelope on tech, yet when problems occur because they were trying something new (CEP) there is a riot! We can't expect to be on the bleeding edge AND not have problems. WCG has taken the route of erroring on the side of not having any problems, other projects take the opposite approach.

I have safely been running both WCG and GPUGrid on my computer for two months now with no impact on WCG to speak of. As soon as WCG supports GPUs I'll be full blast here, but until then I'm supporting another Bionic GPU project in the hopes that the more work we do, the faster the bugs will get fixed. That's all I can do at the moment, so that's what i'm doing.

And Fermi looks very promising. I really want to see it actually released, but we'll have to wait a little while as the demos are all still in the prototype stage. Can't wait!

P.s. - for the ATI users, the newest bionics version supports ATI, so it's comming, just give it some time.

Well actually WCG has said little on CUDA or GPU for that matter, the last 'official' was that initially WCG was looking only at CUDA and then the plan was expanded to include OpenCL, in infant shoes. CUDA is only looked at for the HCC project was included in the post by bbover3. Further to remind that GPU computing was according a WCG tech still not considered ready for the big time as per the definitions of WCG... impaired device usability, or nighttime crunching. Boils down to, yes it will come, don't know where, don't know when. If you know anything more 'official', I'm all ears.

As for Fermi... if it does OpenCL who knows will one of the science project developers jump on that wagon.

My mistake!

I just remember reading about it on the Forums a while ago, and over time the "being looked at for HCC" translated into WCG is working on it in my memory...

On a semi related note - what language do the scientists use for their programs? I had the chance to talk to a recruiter from IBM about internships, (which apparently IBM doesn't do internships for WCG. ) but when I asked what languages they used and the like, he responded that was more the domain of the scientists...