Personal computing discussed
Moderators: renee, farmpuma, just brew it!
Pegasus wrote:I have a Project 5509 my 8800GT 256MB is working on. FahMon says 3200 PPD. Worth 430 points.
http://foldingforum.org/viewtopic.php?f=52&t=5452
Looks like there is more things to simulate in these WUs. But less time passed in the simulation?
"This isn't due to different benchmarking or a different way to assign points, but rather we anticipate NV to be slightly less efficient with larger proteins than it has been with the smaller ones."
thecoldanddarkone wrote:I'm still getting about 5000ppd on my gtx 260 on these work units.
thecoldanddarkone wrote:I'm still getting about 5000ppd on my gtx 260 on these work units.
texaslabratOnce again..it's not a "deduction"...it's a consequence of cold-hard math using the rules that have been in place from the beginning of the project.
When a new WU is ready to be assigned points, it is run on the reference machine. That machine has a ati 3850 in it. They let it run, and note how long it took to complete the WU. They multiply the time in days by 1500ppd to get the points the WU is worth. That's the way it has always been (well, they used to multiply by 1000ppd until they assigned the 50% bonus). So, for example, if the reference machine took exactly 24 hours to complete a WU being benchmarked, it would be worth 1500 points for all clients that run that WU from that point forward.
So, when the tiny proteins were the norm, the ati client choked on them in a cpu-starved fashion...its gpu would run 50% or less in utilization. It took additional time to complete the WU...thus the overall score assigned to each WU of that family was higher. When an nvidia client ran that same WU, it was able to run at 100% gpu utilization and destroy it...earning a very very high PPD (perhaps 5 times faster than the reference machine).
With the new proteins, the ati client is *not* cpu-bound...thus its card runs at 100% utilization. It finishes the WU relatively quickly...and a score is computed. That WU is run on an nvidia client...and now the nvidia client only runs twice as fast as the reference machine on these new proteins. Because the points are set based on the performance of the reference machine...that means that the nvidia client earns only twice the ppd of the reference machine on that particular WU.
At the end of the day, the PPD is a measure of the amount of science accomplished *IN COMPARISON TO THE REFERENCE MACHINE". It is *not* a measure of strict nano-seconds of simulation completed. Those have been the rules of the game from the beginning..this has not been some "bait and switch" routine. As I mentioned before in a previous post...the PPD gained from the tiny proteins on the nvidia clients up til now have been an inadvertent gift due to the cpu-bound condition of the ati-client on those proteins. The proper response is "thank you for the free points".
Last edited by texaslabrat on Tue Sep 09, 2008 11:15 pm, edited 2 times in total
HurgyMcGurgyGurg wrote:Good news, just got an HD 4870 for my new system, should be folding very soon on it, but 64 bit driver support from linksys is next to nothing so internet is so far worthless on it...
just brew it! wrote:HurgyMcGurgyGurg wrote:Good news, just got an HD 4870 for my new system, should be folding very soon on it, but 64 bit driver support from linksys is next to nothing so internet is so far worthless on it...
NICs are dirt cheap these days. Maybe pop another NIC (with better 64-bit support) in there until Linksys gets their act together?
