The new project: 10127

Moderators: Site Moderators, PandeGroup

Re: The new project: 10127

Postby Datsun 1600 » Mon Oct 31, 2011 2:16 am

As soon as I saw a 26 minute frame time for these WUs, I totally lost interest in SMP folding. I prefer WUs to finish in less than 24 hours.

The 11,600ppd output of these WUs, when I average 16,000ppd on this comp is really quite miserable, but then over the nine years of folding it has always been the same.

I now have another top end games comp that won't be folding, thanks Stanford for the lower electricity bill.
Datsun 1600
 
Posts: 51
Joined: Mon May 05, 2008 2:42 am

Re: The new project: 10127

Postby 7im » Mon Oct 31, 2011 4:01 pm

The performance difference of some projects as compared to other projects has been explained many times. If your hardware varies significantly from the benchmark system, your PPD, at times, will also vary somewhat because the chip or memory architecture in your system differs in some way from the benchmark. And some projects do not perform as well because of that difference, while other projects are unaffected.

Nothing has changed in that regard. And since you didn't report any hardware specs, there's nothing more that can be done, other than to ask for more information about your system.
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: The new project: 10127

Postby codysluder » Mon Oct 31, 2011 4:24 pm

"Large" WUs can be measured by a number of different criteria. A WU can require a lot of RAM to process properly (often because it has a lot of atoms). They're properly called [/u]big[/u]. Because the protein is big, there are lots of calculations in every step so they require more time to process or more cores to work in parallel or both. The -bigadv setting is typically required and they require a minimum of 8 cpu cores.

The amount of simulated time in a single WU can vary simply by packaging more simulation steps into a single WU. These are often called big, even though they might contain smaller proteins. These might run just as long and be worth a lot of points, but they probably shouldn't be called big.

People also call WUs that earn a lot of points "big" but that's obviously a misuse of the word.
codysluder
 
Posts: 2128
Joined: Sun Dec 02, 2007 12:43 pm

Re: The new project: 10127

Postby artoar_11 » Mon Oct 31, 2011 8:24 pm

Increased OC of SB 2500K to 4,50 GHz/1333 MHz. This morning I see: TPF - 23:24 min; total time - 39h.
Then I increased memory @ 1600 MHz (9-9-9 by default). Now I have: TPF - 22:06 min; total time - 37h.
Other SMP WUs I have not noticed to react (so much) to increasing the frequency of memory.
User avatar
artoar_11
 
Posts: 631
Joined: Sun Nov 22, 2009 8:42 pm
Location: Bulgaria/Team #224497/artoar11_ALL_....

Re: The new project: 10127

Postby Datsun 1600 » Mon Oct 31, 2011 9:35 pm

7im wrote:Nothing has changed in that regard. And since you didn't report any hardware specs, there's nothing more that can be done, other than to ask for more information about your system.


The comp was an i7-860 @ 3.652GHz with HT on, RAM @ 1333MHz with 7.7.7.24 1T timings and had done 2,178 smp WUs without an eue, with the odd bad WU, absolute stability over anything else, was what I went for at stock voltage.
Datsun 1600
 
Posts: 51
Joined: Mon May 05, 2008 2:42 am

Re: The new project: 10127

Postby GreyWhiskers » Tue Nov 29, 2011 7:50 pm

There were some recent posts on the beta forum for this project, even though it is in the general circulation. I also want to post some of my recent numbers, and attest that these projects seem to garner far lower ppd awards than I usually get from my i7 2600k 4.1 GHz system running non-bigadv SMP8 work units. I typically get ~25,000 ppd for those,but on the 1012x projects I get on v6 (obvious since these are HFM stats):

Project ID: 10125
Core: GRO-A3
Credit: 1186
Frames: 100


Name: DigiStormSMP
Path: \\AL-PC\Users\Public\Documents\FAH SMP Client\
Number of Frames Observed: 300

Min. Time / Frame : 00:06:57 - 18,501.6 PPD
Avg. Time / Frame : 00:07:07 - 17,855.5 PPD


Project ID: 10127
Core: GRO-A3
Credit: 3262
Frames: 100


Name: DigiStormSMP
Path: \\AL-PC\Users\Public\Documents\FAH SMP Client\
Number of Frames Observed: 98

Min. Time / Frame : 00:18:27 - 19,513.9 PPD
Avg. Time / Frame : 00:19:04 - 18,574.9 PPD


Project ID: 10128
Core: GRO-A3
Credit: 1200.79
Frames: 100


Name: DigiStormSMP
Path: \\AL-PC\Users\Public\Documents\FAH SMP Client\
Number of Frames Observed: 100

Min. Time / Frame : 00:07:02 - 18,521.1 PPD
Avg. Time / Frame : 00:07:10 - 18,006.6 PPD


Project ID: 10130
Core: GRO-A3
Credit: 1426.67
Frames: 100


Name: DigiStormSMP
Path: \\AL-PC\Users\Public\Documents\FAH SMP Client\
Number of Frames Observed: 298

Min. Time / Frame : 00:08:09 - 19,221.8 PPD
Avg. Time / Frame : 00:08:24 - 18,370.1 PPD


Project ID: 10134
Core: GRO-A3
Credit: 1264.12
Frames: 100


Name: DigiStormSMP
Path: \\AL-PC\Users\Public\Documents\FAH SMP Client\
Number of Frames Observed: 300

Min. Time / Frame : 00:07:28 - 18,282.5 PPD
Avg. Time / Frame : 00:07:36 - 17,803.5 PPD


EDIT: to contrast, here are some other SMP8 WUs (non bigadv)
Project ID: 7138
Core: GRO-A3
Credit: 585
Frames: 100


Name: DigiStormSMP
Path: \\AL-PC\Users\Public\Documents\FAH SMP Client\
Number of Frames Observed: 200

Min. Time / Frame : 00:02:31 - 28,780.1 PPD
Avg. Time / Frame : 00:02:35 - 27,673.3 PPD


Project ID: 6990
Core: GRO-A3
Credit: 552
Frames: 100


Name: DigiStormSMP
Path: \\AL-PC\Users\Public\Documents\FAH SMP Client\
Number of Frames Observed: 200

Min. Time / Frame : 00:02:40 - 25,280.3 PPD
Avg. Time / Frame : 00:02:44 - 24,361.1 PPD
User avatar
GreyWhiskers
 
Posts: 766
Joined: Mon Oct 25, 2010 5:57 am
Location: Saratoga, California USA

Re: The new project: 10127

Postby Grandpa_01 » Wed Nov 30, 2011 4:07 am

As I have stated before there are some huge variations in PPD on smp projects when it comes to upper end and top end machines. I just put a sandy bridge 2700k running at 4.8Ghz rig together a couple of days ago and it is worse than my Gulftowns when it comes to point spreads between WU's the worst I have seen so far is 18,000 PPD the best has been 50,000 PPD that is a 32,000 PPD spread on the 2700k I really do not think that is acceptable no matter what the difference is in hardware. Maybe the standard of the benchmark machine needs to be raised or user input should carry more gravity when it comes to final points assignments for WU's.

I really do not care for the standard answer that your mileage may vary depending on your hardware specks that is a very poor answer for a problem that has been going on for a very long time. With the new requirements coming out in January for bigadv folders those people that were running bigadv with mid to upper range single processor rigs are going to get a very large point spread when running smp wu's and there is really no reason for that to happen. I am sure it is not Stanford's intention to punish folders for folding with faster hardware but that is what is happening when it come to folding smp on it. I myself will continue on with whatever happens but I can foresee a future of unhappy folders and dumped wu's.
Image
2 - SM H8QGi-F AMD 6xxx=112 cores @ 3.2 & 3.9Ghz
5 - SM X9QRI-f+ Intel 4650 = 320 cores @ 3.15Ghz
2 - I7 980X 4.4Ghz 2-GTX680
1 - 2700k 4.4Ghz GTX680
Total = 464 cores folding
User avatar
Grandpa_01
 
Posts: 1757
Joined: Wed Mar 04, 2009 7:36 am

Re: The new project: 10127

Postby ChasR » Wed Nov 30, 2011 2:44 pm

The difference isn't just in high end machines. on a mere Q6600 @ 3.0 running Windows, p10127 makes 5600 ppd, while p11050 makes 11,000 ppd. Its hard to see why the discrepancy is so large, almost 100%, on a processor so similar in performance to the benchmark machine. It does appear to me that p10127 is closer to the expected output on the Q6600 than p11050 is.
User avatar
ChasR
 
Posts: 698
Joined: Sun Dec 02, 2007 5:36 am
Location: Atlanta, GA

Re: The new project: 10127

Postby DocJonz » Sat Dec 03, 2011 6:38 pm

I also find the same issue on mid-CPU's - I have two identical i7 930's, one is currently running P6941 @ 9,935 PPD and the other is running P10129 @ 6,804 PPD .... the P101xx always seem to generate lower PPD on these machines compared with other WU's ....
User avatar
DocJonz
 
Posts: 347
Joined: Thu Dec 06, 2007 6:31 pm
Location: United Kingdom

Re: The new project: 10127

Postby bruce » Sat Dec 03, 2011 8:56 pm

Is this a bonus issue or a baseline points issue? Can you compare the baseline points for these projects with the baseline points for other projects and report your findings. I know that nobody looks at baseline points any more, just totals after the QRB is applied, but both values need to be considered if there's ever going to be an equitable fix.
bruce
 
Posts: 20828
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: The new project: 10127

Postby ChasR » Sat Dec 03, 2011 9:49 pm

I did that in another thread. It's a base line issue on the q6600.

viewtopic.php?f=16&t=19978

+ 0r - 16% from the mean, 32% base ppd variability.
User avatar
ChasR
 
Posts: 698
Joined: Sun Dec 02, 2007 5:36 am
Location: Atlanta, GA

Re: The new project: 10127

Postby DocJonz » Sun Dec 04, 2011 8:59 am

bruce wrote:Is this a bonus issue or a baseline points issue? Can you compare the baseline points for these projects with the baseline points for other projects and report your findings. I know that nobody looks at baseline points any more, just totals after the QRB is applied, but both values need to be considered if there's ever going to be an equitable fix.

Is this what you need?
Same two i7 930 machines now running -
P10128 - base credit of 1200.79 with k-factor of 3.0 ( PPD @ 6,781)
P7147 - base credit of 585 with k-factor of 3.23 (PPD @ 11,062)
User avatar
DocJonz
 
Posts: 347
Joined: Thu Dec 06, 2007 6:31 pm
Location: United Kingdom

Re: The new project: 10127

Postby k1wi » Thu Dec 08, 2011 10:18 pm

i7 920 @ 4GHz:

P10130: tpf 9min 47s: 14,803.8 ppd (1426.67 base points, KFactor 3.0)
P7147: tpf 2min 56s: 22,801.6 ppd (585.00 base points, KFactor 3.23)
PPD of 7147 = 54.0% higher

Therefore, PPD without bonus (base points) =
P10130: 2117.9 ppd (148 frames per day @ 1426.67 per 100 frames)
P7147: 2871.8 ppd (490.9 frames per day @ 585 per 100 frames)
ppd of 7147 = 35% higher

Therefore, in my reading, the data would suggest that for an i7 920 overclocked to 4Ghz perspective, the base points of the p101xx projects are 35% undervalued relative to p7147 and the K factor exacerbates this issue. I understand that the 101xx projects have a smaller atom count and therefore *may* benefit less from larger caches, but the cache on the i7 930 is the same as the i5 750. Bloomfield has more RAM channels than Lynnfield, so does this mean we are dealing with p7147 benefiting 'unfairly' from the better i7 RAM subsystem?

Does anyone have a i5 750 that can compare the differences?
k1wi
 
Posts: 1152
Joined: Tue Sep 22, 2009 10:48 pm

Re: The new project: 10127

Postby artoar_11 » Fri Dec 09, 2011 10:39 pm

SB 2500K @ 4.40GHz (+GPU GTX460); Win7/64b
Q9650 @ 3.50GHz; Win7/32b

p6940-6947(bp-552.00; Kf-3.30); p6952-6997 (bp-552.00; Kf-3.33)
2500K; 03:24 min TPF; 17.5k PPD
Q9650; 04:46 min TPF; 10.5k PPD

p7100-7154 (bp-585.00; Kf-3.23)
2500K; 03:20 min TPF; 18.8k PPD
Q9650; 04:46 min TPF; 11.0k PPD

p10124 (bp-1705.00; Kf-3.00)
2500K; 10:40 min TPF; 16.7k PPD
Q9650; 14:46 min TPF; 10.3k PPD

p10125 (bp-1186.00; Kf-3.00)
2500K; 08:18 min TPF; 14.2k PPD
Q9650; 11:48 min TPF; 8.4k PPD

p10126 (bp-1208.00; Kf-3.00)
2500K; 08:15 min TPF; 14.7k PPD
Q9650; 11:42 min TPF; 8.7k PPD

p10127 (bp-3262.00; Kf-3.00)
2500K; 22:30 min TPF; 14.5k PPD
Q9650; 33:10 min TPF; 8.1k PPD

p10128 (bp-1200.79; Kf-3.00)
2500K; 08:30 min TPF; 13.9k PPD
Q9650; 11:38 min TPF; 8.7k PPD

p10130 (bp-1426.67; Kf-3.00)
2500K; 10:04 min TPF; 14.0k PPD
Q9650; 14:00 min TPF; 8.5k PPD

p10131(bp-1173.33; Kf-3.00)
2500K; 08:12 min TPF; 14.2k PPD
Q9650; 11:27 min TPF; 8.6k PPD

p10132 (bp-1145.36; Kf-3.00)
2500K; 08:06 min TPF; 14.0k PPD
Q9650; ---

p10133 (bp-1357.17; Kf-3.00)
2500K; 09:36 min TPF; 13.9k PPD
Q9650; 13:40 min TPF; 8.2k PPD

p10134 (bp-1264.12; Kf-3.00)
2500K; 09:00 min TPF; 13.8k PPD
Q9650; ---
User avatar
artoar_11
 
Posts: 631
Joined: Sun Nov 22, 2009 8:42 pm
Location: Bulgaria/Team #224497/artoar11_ALL_....

Previous

Return to V6.34Beta SMP2 with passkey [Not Bigadv]

Who is online

Users browsing this forum: No registered users and 2 guests

cron