p2665 points/deadline?

[AllGold]

The project summary is currently listing GRO-SMP p2665 at 1275 points but with a 6.0 day preferred and final deadline. Also, it appears to be running at about half the PPD of similar WUs. Is the project summary incorrect on this one?

[v00d00]

No it is correct. For the time being anyway.

Points may be changed in the future, but not at present.

[7im]

The project summary is currently listing GRO-SMP p2665 at 1275 points but with a 6.0 day preferred and final deadline. Also, it appears to be running at about half the PPD of similar WUs. Is the project summary incorrect on this one?

Half the PPD on what kind of hardware?

[photo123]

I have the same issue.

I'm using a Core 2 Quad @ 2.4 GHZ with 4 GB of RAM on windows vista 32 "I'm getting around 3.2 GB of RAM because of the 32 bit os".

for other projects I would get an average of 2100 PPD, I got this project 6 Hrs ago and FahMon shows that I will get only 1054.16 PPD.

[AllGold]

My point was the preferred deadline is twice as long as most SMP units (and it appears to be taking twice as long to complete) yet the points are lower. It seems like a mistake.

7im, the hardware is a Q6600 at stock speed running Vista with 2GB RAM. When I say half the PPD, I mean on that specific machine compared to the average of hundreds of other SMP work units it has completed. By the way, Task Manager shows 1.09GB or 55% of the physical memory in use.

[toTOW]

And here come the first complaints ...

[ChasR]

Complaints are coming in on the OC Forum (and probably all the team forums) as well. Not too surprising when you consider p2665 produces 51% of the ppd of p2653 on a Q6600 running Linux SMP. It is the same for everybody though and one could argue the points for p2653 are too high...

[Mactin]

Half the PPD on what kind of hardware?

I'm not complaining about points, I'm only presenting facts about p2665

On my stock Q6600,
My overall average is 2063 PPD across 314 WUs (all Vista/5.91/a1)
I'm averaging 2083 PPD (+1%) across 264 p2653's.
The best I have seen so far is 2496 PPD (+21%) with a single p2609.
The worst so far is an average of 1637 PPD (-21%) across twelve p2652's.

At this moment, after 10% of my first p2665 it's folding at 15min29secs per frame for a projected 1186 PPD.

1186 PPD is -43% under my overall average.
This is way below any reasonable statistical variation.

[7im]

I'm not complaining about points, I'm only presenting facts about p2665.

And I am only trying to get ALL the facts. Without a point of reference, posting the PPD had dropped by half doesn't mean a lot. Thanks for adding more detail.

Like with all discussions about points, the past performance is not an indication of future performance. This is a beta client, and the FAQ clearly states the SMP Bonus can be adjusted at any time. And unless you are getting significantly less points than the benchmark of 1760 PPD (on similar hardware), then there isn't much to discuss.

Please post as much detail as possible so we can see what is going on. Thanks.

[Sunin]

Considering on a 2653 I have a quad with Dual SMPs getting 4400PPD @ 3.2Ghz and now it is barely getting 1500PPD with the 2665 you need to reconsider the points. I mean I didn't invest $500 on nicely overclocked quad folding box for 1500PPD... the wattage to points at this level doesn't even make it attractive! 1500PPD / 175w = 8.5 PpW... I'm used to in excess of 20PpW. I know its about science, but sheesh at some point you guys need to be consistent!

[Xilikon]

It's also bad for science to have this such discrepancy because people will tend to kill those WU until they get a different one. This mean a lot of them would always break the deadlines and stall the whole thing.

My opinion seems to be that this is surely a case of a erroneous assignment of the amount of points. If the deadline is 6 days, it mean it's a very big WU, bigger than those 3065 and the points should reflect this fact (probably around 2500 points for that WU) to keep a correct points average within 10%. I know that WU is always benched on a reference machine to determine the amount of points but sometime, it's so badly benchmarked that it's bad for everyone overall

[v00d00]

Considering on a 2653 I have a quad with Dual SMPs getting 4400PPD @ 3.2Ghz and now it is barely getting 1500PPD with the 2665 you need to reconsider the points. I mean I didn't invest $500 on nicely overclocked quad folding box for 1500PPD... the wattage to points at this level doesn't even make it attractive! 1500PPD / 175w = 8.5 PpW... I'm used to in excess of 20PpW. I know its about science, but sheesh at some point you guys need to be consistent!

P2653 is probably going to be phased out shortly, so dont rely on that work unit.

[toTOW]

With these new projects, I think we'll have to follow Stanford's recommendations : one fahcore per physical core

[Sunin]

With these new projects, I think we'll have to follow Stanford's recommendations : one fahcore per physical core

Am I not correct that their base benchmark system is a dual core system and thus a Quad should be able to handle 2 SMPs?

If that has not changed then I'd say a quad should be able to handle dual SMPs. All my systems run 4gig so memory should not be an issue.

[toTOW]

No you're wrong :

How do you decide the credit value of SMP work units?

Points are determined by the performance of a given machine relative to a benchmark machine, similar to the CPU client benchmark process. Before releasing any new project (series of work units), we benchmark it on a dedicated Macintosh Pro with 2 - 2.33 GHz Dual Core Xeon processors. (more specifically, 2 Woodcrest 5140 processors with 4 MB cache (each), 5 GB FBDIMM Memory (667 MHz DDR2), 1.33 GHz Bus)

We plug the results of this benchmark test into the following formula:

points = 1760 * (daysPerWU)

where daysPerWU is the number of days it took to complete the work unit.

Please note the very concept of a reference machine will mean that some WU performance will vary from the performance on your machine. Even between various Xeon processors, there are significant differences in architectures. Moreover, there are variations between WUs within a given project which can lead to speed differences.

Our goal is consistency within a given definition of a reference machine setup (described above), but beyond that, the natural variation from machine to machine and WU to WU will never allow any point system to perfectly predict what you get on your machine.