F@H: CLI becomes competative?!

[Insidious]

This could be the best news we've had in a while when it comes to the growing disparity of scoring that is obsoleting GOOD hardware in the Folding@Home project.

I just noticed one of my X2 machines showing a PPD of 1144.62 (310 points, ~7.4 minutes/step) for one of it's CLI clients (it runs two at a time)

WoW :shocked:

I double checked.... I have finished one and have been assigned another so maybe we'll see a lot of these beauts! (the run of the mill work units this PC was getting were benching around 140 PPD.... this is a Vista Opteron 170 @ 2.6GHz)


for everyone.....

-Sid

from fahlog.txt:

[15:16:47] Completed 0 out of 500000 steps (0)
[15:16:47] Extra SSE2 boost OK
[15:23:03] Completed 5000 out of 500000 steps (1)

from fahmon benchmark:

Project : 3906
Core : Double Gromacs B
Frames : 50
Credit : 310


-- Dan Opteron Vista CLI 2 --

Min. Time / Frame : 7mn 06s - 1257.46 ppd
Avg. Time / Frame : 7mn 38s - 1169.61 ppd
Cur. Time / Frame : 7mn 46s - 1149.53 ppd

So this is optimized for SSE2 and it only has 50 frames. :shocked: :thumbsup:
That probably explains the speed! (of course, NOTHING is going to explain how the scoring was derived :laugh: )

 

[biodoc]

Is this the 7b core?

Apparently the double gromacs and double gromacs B cores support SEE2 optimizations.

:beer:

 

[Insidious]

Yeah, it's 7b

just got another one a few minutes ago, (I only have one machine on CLI because it goes to school with my son and needs the service install).

-Sid

 

[biodoc]

I just switched my stock quad from SMP to 4 CLI clients.

All four got the new 3906 WUs and fahmon says I'll get 6600 PPD!!

I was getting 2200 PPD with SMP.

I hope these keep loading up!!!!:shocked:

Boy am I


Thanks for the heads up Sid! :beer:

 

[Markfw]

Where is this CLI client on the download page ???? I don't see it.

 

[biodoc]

Sorry, I meant the text-only console client.

For a quad...4 directories, 4 different Machine IDs and run each with -local switch. I also turned on advmethods in the config.

I hope these puppies keep coming for a while or else I go back to SMP.

Edit: I do have affinity changer running now but I doubt if that makes a difference in this case.

 

[Insidious]

I hope some of the people we lost because they had single core machines see this...
Your relevance has been returned. :thumbsup:

It's about time Stanford realized they had "priced single core PCs out" with their outrageously generous scoring of the newer clients.

-Sid

 

[caferace]

Originally posted by: Insidious
I hope some of the people we lost because they had single core machines see this...
Your relevance has been returned. :thumbsup:

It's about time Stanford realized they had "priced single core PCs out" with their outrageously generous scoring of the newer clients.

-Sid

Maybe it's time to un-retire my FX-55 proc that's been sitting in a box the last six months?

-jim

 

[Insidious]

OK,

This is embarrassing....

All that excitement.... all that joy......

Well it turns out that this work unit really does have the normal 100 frames instead of the 50 that fahmon thinks it does...

So, that means the fahmon reported PPD is twice what you'll really get...

here's an excerpt from fahlog.txt for one of these:

[15:16:41] - Expanded 238970 -> 1203061 (decompressed 503.4 percent)
[15:16:41]
[15:16:41] Project: 3906 (Run 460, Clone 2, Gen 0)
[15:16:41]
[15:16:41] Assembly optimizations on if available.
[15:16:41] Entering M.D.
[15:16:47] Working on Lig in water
[15:16:47] Completed 0 out of 500000 steps (0)
[15:16:47] Extra SSE2 boost OK
[15:23:03] Completed 5000 out of 500000 steps (1)
[15:30:09] Completed 10000 out of 500000 steps (2)
[15:37:40] Completed 15000 out of 500000 steps (3)
[15:45:27] Completed 20000 out of 500000 steps (4)
[15:53:13] Completed 25000 out of 500000 steps (5)
.
. (snip)
.
[21:42:12] Completed 245000 out of 500000 steps (49)
[21:49:59] Completed 250000 out of 500000 steps (50) (<--- it's still going :shocked: )
[21:57:47] Completed 255000 out of 500000 steps (51)
[22:05:34] Completed 260000 out of 500000 steps (52)

yeah, I'm disappointed too..

OH well, 550 is still A LOT better than the normal 140...... right? :roll:

-Sid

 

[biodoc]

Yeah I just notice that too but still 3200 PPD is not bad for a stock quad so I'll keep running it for now.

Fahmon got fooled on this one!

Good time for my first glass of :wine:

 

[Slider76]

3906 171.64.122.88 p3906_ 6355 59.00 86.00 310.00 points 50 frames DGROMACSB

hmmm...it is only meant to be a 50 frame WU according to the standford site

 

[Markfw]

Well, good... I don't want to switch back over 30 cores config... Sounds like it won't buy me anything.

 

[Insidious]

Originally posted by: Slider76
3906 171.64.122.88 p3906_ 6355 59.00 86.00 310.00 points 50 frames DGROMACSB

hmmm...it is only meant to be a 50 frame WU according to the standford site

I have one crunching now at frame 58 and the other one is at frame 75. I think they just typo'd the entry in the database (which is the same one that fahmon uses for it's info). but the thing definitely goes to frame 100

-Sid

edit: from fahlog.txt

[14:57:50] Completed 490000 out of 500000 steps (98)
[15:05:41] Completed 495000 out of 500000 steps (99)
[15:13:35] Completed 500000 out of 500000 steps (100)
[15:14:36]
[15:14:36] Finished Work Unit:
[15:14:36] - Reading up to 7928172 from "work/wudata_06.trr": Read 7928172
[15:14:36] - Reading up to 1145524 from "work/wudata_06.xtc": Read 1145524
[15:14:36] logfile size: 28262
[15:14:36] Leaving Run
[15:14:36] - Writing 11589578 bytes of core data to disk...
[15:14:41] Done: 11589066 -> 9584255 (compressed to 82.7 percent)
[15:14:42] ... Done.
[15:14:42] - Shutting down core
[15:14:42]
[15:14:42] Folding@home Core Shutdown: FINISHED_UNIT

 

[GLeeM]

Originally posted by: biodoc
I also turned on advmethods in the config.

According to server status page you must have -advmethods set to recieve these (for the time being).

 

[Slider76]

Sid, typo's don't you love them. Still sounds like they will be great PPD.

Its nice to see that its "offical" that there will be a client for the new ATI GPU's http://folding.typepad.com/
however still no release date set.

 

[caferace]

Originally posted by: Slider76Its nice to see that its "offical" that there will be a client for the new ATI GPU's http://folding.typepad.com/
however still no release date set.

That *is* cool news. The new Q6600 has a ATI 640Mb 3750 "just because", and I'm not gaming with it. Heck, it's never but running 1280*1024 on a CRT, so it'd be nice to put it to a test.

-jim

 

[biodoc]

Originally posted by: Insidious

Originally posted by: Slider76
3906 171.64.122.88 p3906_ 6355 59.00 86.00 310.00 points 50 frames DGROMACSB

hmmm...it is only meant to be a 50 frame WU according to the standford site

I have one crunching now at frame 58 and the other one is at frame 75. I think they just typo'd the entry in the database (which is the same one that fahmon uses for it's info). but the thing definitely goes to frame 100

-Sid

edit: from fahlog.txt

[14:57:50] Completed 490000 out of 500000 steps (98)
[15:05:41] Completed 495000 out of 500000 steps (99)
[15:13:35] Completed 500000 out of 500000 steps (100)
[15:14:36]
[15:14:36] Finished Work Unit:
[15:14:36] - Reading up to 7928172 from "work/wudata_06.trr": Read 7928172
[15:14:36] - Reading up to 1145524 from "work/wudata_06.xtc": Read 1145524
[15:14:36] logfile size: 28262
[15:14:36] Leaving Run
[15:14:36] - Writing 11589578 bytes of core data to disk...
[15:14:41] Done: 11589066 -> 9584255 (compressed to 82.7 percent)
[15:14:42] ... Done.
[15:14:42] - Shutting down core
[15:14:42]
[15:14:42] Folding@home Core Shutdown: FINISHED_UNIT

Sid, are you getting credit for the 3906 WUs? I'm not and others are not either

Check this thread

 

[Insidious]

I just checked.....nope. No creds given. :|

/me throws snowballs at Stanford windows

So let's see.... this thread has gone from 'hey look you get 1100PPD", to..... "aw crap, you only get 550PPD", to...... :shocked: you get NO ppd.

:laugh: Rofl :laugh:

-Sid

And tell me it doesn't figure.... the ONLY machine I have that is remote (with my son) is the one with the CLI client installed. Well hopefully it'll keep his dorm room warm. Apparently that's all it's going to be good for right now.

 

[scottv67]

Insidious wrote:

This could be the best news we've had in a while when it comes to the growing disparity of scoring that is obsoleting GOOD hardware in the Folding@Home project.

I hope some of the people we lost because they had single core machines see this...

It's about time Stanford realized they had "priced single core PCs out" with their outrageously generous scoring of the newer clients.

Whine much, Sid? Are you comparing the PPD output of a five year old machine against a system with a motherboard and CPU that were released in 2007? My Q6600-based systems can turn-around an SMP WU in less than twelve hours. That's a lot of information for Stanford in a very short time (so they "pay" a higher bounty to the machines that can do it).

Are you implying that people stopped contributing to F@H because their single core PCs couldn't "run with the big dogs"? That's just silly. There are still WUs for the older single core machine to process.

If you are "obsoleting" perfectly good hardware that could still be contributing to F@H just because that older machine can't produce the same PPD as a something I ordered from Newegg last month, that's your problem.

I don't get it. The bottom line is that the information that Stanford gets from the project is used to research certain diseases. That's why I contribute to this project and not SETI. *This* project is actually helping our fellow humans on this planet (which is very important to me). The "PPD race" is just a game that we play while the real scientific work is being done at Standford. We all joke about reaching 10,000 PPD, 20,000, or 30,000 PPD but at the end of the day, it is the quantity of scientific data that is returned to Stanford that makes this whole effort worth doing.

I don't understand why someone would feel that they need to stop contributing to the F@H project (which helps research diseases that affect our fellow humans) just because some arbitrary "carrot on a stick" is not as big as the carrot in front of another horse/donkey/mule.

Your relevance has been returned.

Those machines never stopped being relevant.

-s :beer:

 

[Insidious]

Anybody got a shovel?

 

[GLeeM]

Originally posted by: scottv67
Those machines never stopped being relevant.

In a sense they have.

The following is not me nor most of us, but is many folders.

If I could comfortably afford to pay for five computers running 24/7 I would prefer to run Quads. I would look at my investment and the return F@H gives and say there is not enough return for the electricity to run a P4 or older Athlon. If I get 3,000 for each Quad and 100 for a P4, I would not run the P4. If I got 1,000 for the P4 I might run it for a while.

Me, I got to use what I have, until I can upgrade :roll:

 

[Markfw]

Gleem, exactly. I have 2 X2's I don't even use except lan parties. My electric bill went down a lot when I turned off one dual 248 opteron box that was getting 1000 ppd. I replaced it with 2 Q6600's for 10k ppd, and the electric bill is still less (and the heat is less)

So to be "green" we should all replace out single core and X2 boxes with Q6600's !

 

[caferace]

Originally posted by: Insidious
Anybody got a shovel?

:laugh:

/me hands Sid keys to backhoe.

-jim

 

[biodoc]

I'm back to SMP

No word from Stanford on the p3906 WU credit yet

They probably took the weekend off

 

[biodoc]

Originally posted by: Markfw900
Gleem, exactly. I have 2 X2's I don't even use except lan parties. My electric bill went down a lot when I turned off one dual 248 opteron box that was getting 1000 ppd. I replaced it with 2 Q6600's for 10k ppd, and the electric bill is still less (and the heat is less)

So to be "green" we should all replace out single core and X2 boxes with Q6600's !

This is why I got back into DC. I had a few single core rigs but the output vs electricity costs didn't make much sense these days. I replaced 3 with quad cores and retired the fourth and I'm using much less electricity and my DC output went through the roof.

A stripped quad core doesn't use that much juice It'll pay for itself in the long run!

If you're considering a dual core Intel or AMD, take a look at quad core! 2 rigs will use more electricity than one. Use free linux software and save more $$.

Now I get off the soap-box

More :wine: