core_a4 projects with silly ppd

[Amaruk]

My guess is the majority of these are being done on -smp rigs thus should be benchmarked to reflect that.

Posts in this thread would seem to indicate they are benched as SMP.

The real question is how does project 8xxx compare to other uniprocessor cores when it's assigned to a uniprocessor client.

Here are the benchmarks from one of my uni clients before the adjustment.

Code: Select all

 Project ID: 6892
 Core: GROMACS
 Credit: 136
 Frames: 100


 Name: Chow
 Path: C:\Users\Chow\Documents\Folding\634Console\
 Number of Frames Observed: 300

 Min. Time / Frame : 00:10:45 - 182.2 PPD
 Avg. Time / Frame : 00:10:48 - 181.3 PPD

-----

 Project ID: 8001
 Core: GRO-A4
 Credit: 88.59
 Frames: 100


 Name: Chow
 Path: C:\Users\Chow\Documents\Folding\634Console\
 Number of Frames Observed: 296

 Min. Time / Frame : 00:07:33 - 760.4 PPD
 Avg. Time / Frame : 00:07:37 - 750.5 PPD

-----

 Project ID: 8004
 Core: GRO-A4
 Credit: 88.59
 Frames: 100


 Name: Chow
 Path: C:\Users\Chow\Documents\Folding\634Console\
 Number of Frames Observed: 300

 Min. Time / Frame : 00:07:31 - 765.5 PPD
 Avg. Time / Frame : 00:07:36 - 753.0 PPD

-----

 Project ID: 10722
 Core: GRO-A4
 Credit: 2244
 Frames: 100


 Name: Chow
 Path: C:\Users\Chow\Documents\Folding\634Console\
 Number of Frames Observed: 176

 Min. Time / Frame : 02:20:51 - 475.4 PPD
 Avg. Time / Frame : 02:21:44 - 471.0 PPD

Revised PPD numbers (calculated) are:

8001 at 00:07:37 is 375.24 PPD instead of 750.5 PPD

8004 at 00:07:36 is 376.48 PPD instead of 753.0 PPD


That's still more than double the PPD for non-QRB, and more in line with the other a4 project. (10722 at 471.0 PPD)

[ChasR]

Add p7014 to the list of overly high producing a4 Wus. Perhaps, all the p7000 to p7028 are over producers.

[Nantes]

On a 2600K @ 4.8 GHz running v7 in a VM (Ubuntu 10.10), performance on p8004 went from 59,500 ppd to 29,400 ppd. THe average a3 SMP WU on version 6.34 produces about 36,500 ppd. So we've gone from about 63% more than the average a3 WU to 20% less, at least on a 2600K.

How? How??? I have a 2500K @ 4.8 GHz running smp:4 and it only gets around 11k PPD :,-(

@topic: I am running project 7611 and getting ~5k PPD rather than the usual ~11k. This project should get balanced...

[MtM]

Haven't really been paying much attention but I finally figured out why this project seemed so low to me..


Except from the scaling from smp:2 to smp:4 which seems low ( tpf wise ), the only way the ppd for the smp:4 rig isn't exponentially bigger then smp:2 is if I'm not recieving any bonus. The most recent entries are shown as not receiving one as well.

My systems are configured with a passkey and my completion rate's are well over 80%. I did get qrb before. Could a mod check my last submissions for these projects and confirm I'm not getting it now, and then maybe suggest something as to why I'm not getting it right now ( excluding the passkey and completion rate's which I already ruled out... unless an client upgrade messed up my configuration and changed the passkeys which I doubt is possible )

[ChasR]

On a 2600K @ 4.8 GHz running v7 in a VM (Ubuntu 10.10), performance on p8004 went from 59,500 ppd to 29,400 ppd. THe average a3 SMP WU on version 6.34 produces about 36,500 ppd. So we've gone from about 63% more than the average a3 WU to 20% less, at least on a 2600K.

How? How??? I have a 2500K @ 4.8 GHz running smp:4 and it only gets around 11k PPD :,-(

@topic: I am running project 7611 and getting ~5k PPD rather than the usual ~11k. This project should get balanced...

A 2600K is about 25% faster than a 2500K at the same clock. FAH running under Linux is about 15% faster than in Windows, even when run Linux is run as guest in a VM. 2133 MHz CL9 ram is about 7% faster than 1600 MHz CL9 ram. My machine is running without a gpu folding, so the SMP instance isn't getting slowed by the loss of cpu resources. It all adds up (more or less) and then gets the QRB applied.

[Nantes]

On a 2600K @ 4.8 GHz running v7 in a VM (Ubuntu 10.10), performance on p8004 went from 59,500 ppd to 29,400 ppd. THe average a3 SMP WU on version 6.34 produces about 36,500 ppd. So we've gone from about 63% more than the average a3 WU to 20% less, at least on a 2600K.

How? How??? I have a 2500K @ 4.8 GHz running smp:4 and it only gets around 11k PPD :,-(

@topic: I am running project 7611 and getting ~5k PPD rather than the usual ~11k. This project should get balanced...

A 2600K is about 25% faster than a 2500K at the same clock. FAH running under Linux is about 15% faster than in Windows, even when run Linux is run as guest in a VM. 2133 MHz CL9 ram is about 7% faster than 1600 MHz CL9 ram. My machine is running without a gpu folding, so the SMP instance isn't getting slowed by the loss of cpu resources. It all adds up (more or less) and then gets the QRB applied.

Omg I had no idea setting up a passkey made you eligible for extra points. I thought handing a WU faster gave you more points regardless! Thanks, I have one set up now.

So GPU folding takes away from SMP folding, but I would assume it's much more efficient to have a GPU + SMP than SMP alone, as long as the GPU isn't much more inferior than the CPU (not my case as I have a GTX 560).

Is the 25% difference due to hyperthreading being enabled?

[MtM]

Omg I had no idea setting up a passkey made you eligible for extra points. I thought handing a WU faster gave you more points regardless! Thanks, I have one set up now.

But it does since you complete more work unit's in the same amount of time

[ChasR]

Hyperthreading and perhaps a larger cache account for the difference.

[Nantes]

Omg I had no idea setting up a passkey made you eligible for extra points. I thought handing a WU faster gave you more points regardless! Thanks, I have one set up now.

F@h is now estimating 48k PPD with the 7611 project as opposed to less than 5k before. Holy casseroles!
Edit: It seems to be constantly declining, it's at 35k atm.

[SomeStones]

@topic: I am running project 7611 and getting ~5k PPD rather than the usual ~11k. This project should get balanced...

I've got a 7611 now and it shows about 1/2 the PPD of most other units.

[Joe_H]

F@h is now estimating 48k PPD with the 7611 project as opposed to less than 5k before. Holy casseroles!
Edit: It seems to be constantly declining, it's at 35k atm.

Which client are you using to monitor the PPD? If it is from the V7 beta client, that is known to have a bug with its PPD calculation, especially after restarting in the middle of a WU. In any case, if you have not been using a passkey it will take completing 10 QRB eligible WU's before you get bonus points.

P.S. On my 2.8 i7 Project 7611 WU's get about 8,000 PPD. 7610 WU's get about 10-15% better PPD.

[sortofageek]

Could a mod check my last submissions for these projects and confirm I'm not getting it now ...

In response to your report, I sent a PM with your results for the three WUs you mentioned in your report. Perhaps someone else will help explain these results.


========================
Project: 8001 (Run 79, Clone 13, Gen 66)

Hi Marvin_The_Martian (team 11314),
Your WU (P8001 R79 C13 G66) was added to the stats database on 2012-01-24 01:09:53 for 236.18 points of credit.

-----------------
Project: 8001 (Run 65, Clone 4, Gen 63)

Hi Marvin_The_Martian (team 11314),
Your WU (P8001 R65 C4 G63) was added to the stats database on 2012-01-21 14:10:05 for 259.66 points of credit.

-----------------
Project: 8001 (Run 11, Clone 0, Gen 96)

Hi Marvin_The_Martian (team 11314),
Your WU (P8001 R11 C0 G96) was added to the stats database on 2012-01-24 10:11:11 for 178.86 points of credit.

=====================

[Nantes]

Which client are you using to monitor the PPD? If it is from the V7 beta client, that is known to have a bug with its PPD calculation, especially after restarting in the middle of a WU. In any case, if you have not been using a passkey it will take completing 10 QRB eligible WU's before you get bonus points.

What can I use then? I tried HFM but I don't know how use that thing.

[bruce]

The author of HFM (and the authors of other 3rd party monitoring programs) are busily working on revising their software to work with V7. The developers of V7 are working hard on correcting bugs (including this one) prior to an upcoming Milestone called "Open Beta Phase 2" when they consider V7 ready for a wider release. Until either of those events happen, you'll just have to ignore what's being reported and either calculate the information manually, do without it, or revert to V6 temporarily.

[Joe_H]

Could a mod check my last submissions for these projects and confirm I'm not getting it now ...

In response to your report, I sent a PM with your results for the three WUs you mentioned in your report. Perhaps someone else will help explain these results.

========================
Project: 8001 (Run 79, Clone 13, Gen 66)

Hi Marvin_The_Martian (team 11314),
Your WU (P8001 R79 C13 G66) was added to the stats database on 2012-01-24 01:09:53 for 236.18 points of credit.

-----------------
Project: 8001 (Run 65, Clone 4, Gen 63)

Hi Marvin_The_Martian (team 11314),
Your WU (P8001 R65 C4 G63) was added to the stats database on 2012-01-21 14:10:05 for 259.66 points of credit.

-----------------
Project: 8001 (Run 11, Clone 0, Gen 96)

Hi Marvin_The_Martian (team 11314),
Your WU (P8001 R11 C0 G96) was added to the stats database on 2012-01-24 10:11:11 for 178.86 points of credit.

=====================

Those points credited are all consistent with getting bonus through the QRB as the base points are 89. As I can complete one of the 8001 WU's in just over 4 hours and get around 300 points on my 2.4 Core 2 Duo, these obviously took a bit longer.