core_a4 projects with silly ppd

[rwh202]

Hi,

I've seen previous discussion and comment about inconsistent ppd across projects. Although I'd observed it on the Linux 6.34 client (core_a3), variances were small and averaged out.

However, I've just started up a machine on windows 7, using the v7 client and getting a4 smp WUs.

The variability in ppd across WU is shocking!

Rather than post in another thread and be accused of derailing, or making unsubstantiated accusations, here is one for people to post their own observations.

Hardware:
i7 2600k @ 4.6GHz
8GB @ 1333 9-9-9-24
Win7 64bit running smp (dedicated - only other software installed is CoreTemp and intel burn test)

Project: 8004 (Run 114, Clone 10, Gen 38)
0:28 tpf
49485 ppd (inc bonus)
2750 ppd (excl bonus)

Project: 8001 (Run 12, Clone 74, Gen 40)
0:28 tpf
49485 ppd (inc bonus)
2750 ppd (excl bonus)

Project: 7611 (Run 0, Clone 5, Gen 114)
5:21 tpf
13564 ppd (inc bonus)
2161 ppd (excl bonus)

A 27% différence in base ppd and a 265% difference in bonus ppd.

My similar rigs in linux make a pretty steady 28k ppd, so accounting for OS differences, I was expecting to see 25k ppd.

OK, so its a small sample so far, but I hope others can add to it. I'll add if I get any of my more familiar a3 units.

[ChasR]

The p800x series makes far more ppd than any other WU on C2Qs as well. Over 12,000 ppd on a Q6600 @ 3.0 GHz while also running a GPU instance.

[Grandpa_01]

If you really want to make some PPD on the 80XX series just install Linux v7 I get 100,000+ PPD on the 8011 running on a 980X at 4.6Ghz VS 85,000 PPD on Windows

[ChasR]

I've got the -oneunit flag on two 2600K 6.34 VMs now. We'll see what they will do @ 4.8 GHz.

[rwh202]

If you really want to make some PPD on the 80XX series just install Linux v7 I get 100,000+ PPD on the 8011 running on a 980X at 4.6Ghz VS 85,000 PPD on Windows

Yeah, i'm not exactly upset by the high ppd

I'll be installing linux next week with my normal v6 smp and wine gpu, but just wanted to see what the deal was with v7.

However, it needs highlighting that whatever system is allocating points is demonstrably broken

We'll have to see how this issue fits in the list of priorities

[ChasR]

If you really want to make some PPD on the 80XX series just install Linux v7 I get 100,000+ PPD on the 8011 running on a 980X at 4.6Ghz VS 85,000 PPD on Windows

What assignment flag, if any is needed to catch these WUs, advanced, beta, big, normal...?

[Grandpa_01]

I do beta testing but they have been released to general population so it is random now any flag should work. They come in spirts 8001 for a while 8004 for a while and 8011 for a while.

[diwakar]

I am running the p80xx series on F@H. Thanks rwh202 and Grandpa_01 for bringing this up and posting the comparison of PPD and bonus point differences between projects.
It seems that we are using inconsistent set of formulas for calculating the PPD and bonus points for 80xx and 7611.
I will change the settings so that p80xx is consistent with other SMP WU's or vice versa but it would only happen on Tuesday (first working day next week).

[Grandpa_01]

In my opinioon The a3 projects need to be raised to better alighn with the the a4 projects and bigadv. The a3 value is very low compared to what I can make in PPD folding bigadv the a3 projects avg about 45% of a 6903 on the same rig. If Stanford wants people to fold -smp projects they should make there value closser to what a person can make PPD wise to other options.

[ChasR]

2600K @ 4.85 GHz running in Linux in a VM:
p8001
TPF: 0:24.54
PPD: > 59,000 ppd

[7im]

In my opinioon The a3 projects need to be raised to better alighn with the the a4 projects and bigadv. The a3 value is very low compared to what I can make in PPD folding bigadv the a3 projects avg about 45% of a 6903 on the same rig. If Stanford wants people to fold -smp projects they should make there value closser to what a person can make PPD wise to other options.

Before we jump the gun and vote for any kind of artificial / additional points inflation, let's see what happens after diwakar actually follows the correct benchmark settings.

And compliments to everyone posting hard facts and helping to find and fix a serious points problem.

[ChasR]

Another one on a 2600K @ 4.8 GHz in a Linux VM.

P7015
TPF: 1:23
PPD: > 57,000

[rwh202]

I am running the p80xx series on F@H. Thanks rwh202 and Grandpa_01 for bringing this up and posting the comparison of PPD and bonus point differences between projects.
It seems that we are using inconsistent set of formulas for calculating the PPD and bonus points for 80xx and 7611.
I will change the settings so that p80xx is consistent with other SMP WU's or vice versa but it would only happen on Tuesday (first working day next week).

OK, thanks for looking into this
Rob

[Fahrenheit451]

... I will change the settings so that p80xx is consistent with other SMP WU's or vice versa but it would only happen on Tuesday (first working day next week).

Please correct me if I'm wrong, but projects 8001, 8004 and 8011 are not SMP WU's. I folded all with the v6.23 Uni-processor clients. According to HFM I got the following results on my Core 2 Duo E6850 (@3GHz):

8001:
Min. Time / Frame : 00:04:24 - 1.709,3 PPD
Avg. Time / Frame : 00:04:48 - 1.500,1 PPD

8004:
Min. Time / Frame : 00:04:36 - 1.599,0 PPD
Avg. Time / Frame : 00:05:12 - 1.330,4 PPD

8011:
Min. Time / Frame : 00:05:38 - 1.759,2 PPD
Avg. Time / Frame : 00:06:14 - 1.511,4 PPD

[PantherX]

... I will change the settings so that p80xx is consistent with other SMP WU's or vice versa but it would only happen on Tuesday (first working day next week).

Please correct me if I'm wrong, but projects 8001, 8004 and 8011 are not SMP WU's. I folded all with the v6.23 Uni-processor clients....

These FahCore_a4 WUs are special as they can be folded by Classic (Uniprocessor) Clients as well as SMP Clients.