F@H Holiday Racers Welcome!

[biodoc]

That's great Sid! I know how much you :heart: ubuntu but here's fahmon benchmarks for one of my rigs running 2xSMP @3.2GHz The min. time/frame is one SMP.

-- ubuntu3 --

Min. Time / Frame : 7mn 38s - 3320.17 ppd
Avg. Time / Frame : 11mn 12s - 2262.86 ppd
Cur. Time / Frame : 11mn 23s - 2226.41 ppd
R3F. Time / Frame : 11mn 23s - 2226.41 ppd



-- ubuntu4 --

Min. Time / Frame : 7mn 30s - 3379.20 ppd
Avg. Time / Frame : 11mn 13s - 2259.49 ppd
Cur. Time / Frame : 11mn 35s - 2187.97 ppd
R3F. Time / Frame : 11mn 35s - 2187.97 ppd

So there's still a slight advantage to 64 bit ubuntu! over XP!

:beer:

 

[MadMurph]

This thing (race) has been a hoot. It really drives home the idea that this forum is family. I know I'm a relative n00b here, but I was over at the Brand-X forum for almost 8 years, and have folded since 1998 (SETI), until new owners treated us "oldies" really poorly. I had been using Anandtech as a reference a lot, anyway, so decided to move myself and my crunching over here. It has been pure joy from the very first post, literally and truly.

To all of you, esp. the vets, THANK YOU! :beer::wine::thumbsup:

Won't be here for the end of the contest, have to trust the farm to the whim of powers that be, as I head to the mountains to heal my soul, center my spirit, and recharge the batts.

Happy New Year to all!!!!! :wine::beer::wine::beer: Stay safe, chat upon my return. To TAS!!!

Cheers, MadMurph

edit: Bummer. Looks like I'm not even going to be here for my best day yet. 51 processors, Mark???? How'd that happen. Insane.:laugh:

 

[Assimilator1]

Glad that you came over to us ,their bad behaviour has been to our benefit

Oh btw you don't fold SETI you crunch it & it started in '99.................who said I was being picky?? heh
So how many SETI classic WUs did you do? ,I did about 51k.

Have fun on your trip:thumbsup:

Originally posted by: Insidious

Originally posted by: Assimilator1

And you're getting 12m20s/step on each SMP client?

What clock speed is it again?

Yeah, that's on each client here's a paste from the FAHMon benchmarks
(My Quads are running at 3.2GHz.)

-- Q6600-main SMP1 --

Min. Time / Frame : 12mn 23s - 2046.62 ppd
Avg. Time / Frame : 12mn 23s - 2046.62 ppd
Cur. Time / Frame : 12mn 25s - 2041.13 ppd
R3F. Time / Frame : 12mn 25s - 2041.13 ppd
Eff. Time / Frame : 12mn 43s - 1992.98 ppd

-- Q6600-main SMP2 --

Min. Time / Frame : 12mn 25s - 2041.13 ppd
Avg. Time / Frame : 12mn 25s - 2041.13 ppd
Cur. Time / Frame : 12mn 26s - 2038.39 ppd
R3F. Time / Frame : 12mn 26s - 2038.39 ppd
Eff. Time / Frame : 12mn 46s - 1985.17 pp

When I ran a single SMP, it was ~8.5 minutes (IIRC)
-Sid

OMG that's a massive difference!:Q

In theory IF the SMP client efficiently used all 4 cores then 2xSMP clients should give you about 17mins/step(2x8.5), yet you get ~12.5mins/step!, so by affixing affinity and running 2xSMP clients you've cut processing times by about 28%!
Man this info should be in a folding sticky!
Can you confirm that 8.5mins/step is correct for a single SMP? (you said 'IIRC').
I've got try this when the current WU is finished.

Originally posted by: Fullmetal Chocobo

Originally posted by: Assimilator1
lol no probs ,P3s don't have Hyperthreading either ,you mean P4s
What clock speed is it btw?

Hmmm. I know four cores show up in windows (not sure if they are logical or actual). If it isn't HT, then it might be a quad proc P3. I know they are 500MHz P3 Xeons though.

Ah ,in that case it will be a 4 socket mbrd then!:Q , that would of been an awesome rig in its time!. Are they S370 or Slot1?

 

[caferace]

So knuckle-headed me waited until this morning to run two consoles instead of one on my Q6600. That was.... dumb.

Old ppd: 2246.14
New ppd: 3141.82

Those are with stock clock, by the way. I'll compile the affinity changer this weekend and see what it does then...

-jim

 

[Insidious]

Originally posted by: Assimilator1
OMG that's a massive difference!:Q

In theory IF the SMP client efficiently used all 4 cores then 2xSMP clients should give you about 17mins/step(2x8.5), yet you get ~12.5mins/step!, so by affixing affinity and running 2xSMP clients you've cut processing times by about 28%!
Man this info should be in a folding sticky!
Can you confirm that 8.5mins/step is correct for a single SMP? (you said 'IIRC').
I've got try this when the current WU is finished.

I'll let a single run for a couple frames and let you know for sure on those times AS1.

Everybody remember, the reason this affinity changer works so well for me is because WXP does such a terrible job of managing threads. Linux does this pretty well and I don't think much of a gain from the affinity changer should be expected. Read the notes on that page I linked to.... it sheds some light on this.

This thing is going to help the people running two SMP clients on a Quad core under windows more than anyone else.

Really all it is doing is making up for the WXP ineffeciency and closing the gap between the rates WXP and Linux crunch.


-Sid

 

[Insidious]

AS1, I let a single client run for 4 steps and have this result (all else the same as when two were running)

-- Q6600-main SMP1 --

Min. Time / Frame : 8mn 50s - 2869.13 ppd
Avg. Time / Frame : 8mn 50s - 2869.13 ppd
Cur. Time / Frame : 8mn 50s - 2869.13 ppd
R3F. Time / Frame : 8mn 50s - 2869.13 ppd
Eff. Time / Frame : 13mn 14s - 1915.16 ppd

I think I'd see that drift closer to 8.5 min/step with more steps, but it wouldn't change too much.

-Sid

 

[Assimilator1]

Cool ,thanks for the info
I'm going to try it now.

[update]
Ok I've got 2xSMP clients running along with A Ch ,when I bring up task manager i see 8 FaHcore_a1.exe processes!:Q ,is this normal? or have I messed something up?
I mut of done something wrong ,20mins/step.............oh wait I forgot I've got F@H GPU running! lol ,will that mess things up?

Jim
Keep us posted

 

[Insidious]

AS1,

You definitely don't want to have 2 SMPs and the GPU clients running. If you do the 2 SMP clients, you need to stop the GPU (it hogs one core to do it's thing)

Secondly, yes you should see 8 of the FahCore_A1 processes running (4 for each client)

-Sid

 

[Assimilator1]

I wonder which I'd be better off doing?
FaHmon says Im getting 455ppd from GPU client atm.

 

[Insidious]

LOL

-- Scott-Quad1-SMP2 --

Cur. Time / Frame : 11mn 33s - 2194.29 ppd


-- Scott-Quad1-SMP1 --


Cur. Time / Frame : 11mn 37s - 2181.69 ppd

I think I'd opt for the 4,375 from a (WXP) quad with 2 SMPs

:beer:

 

[GLeeM]

Sorry about barging this in here, but I didn't want the other thread to go too far off topic.

Originally posted by: Assimilator1
Btw A Stanford University protein study called Folding@home now gets the majority of its processing work from the consoles

Is that true??

That statement would be based on total TFLOPs contributed.

You can read this and decide for yourself.

• Windows, Linux & Mac do 22.4% of TFLOPs with 87% of CPUs

• The GPU client does 3.4% of total TFLOPs with only 0.2% of CPUs!

• The PS3 client does 74.2% of total TFLOPs with 12.7% of CPUs!

Remember that PCs include all of the old P3, P4 and other slow CPUs. If all PCs were quads the percent would only be a little closer. The PS3 really crunches, even if it can only do one type of simulation core (or is it two now?), it does it very fast and there a lots of them.

Notice that the GPU client does better than the PS3 per CPU

PS: After thinking about all this, read the FAQs linked in above link, before asking questions :roll:

 

[biodoc]

Interesting reading over there GLeeM.

Seems the GPU client/core is very limited in the types of science it can do but is very fast!

The PS3 is more flexible and they seem to be putting more effort in the types of science it can do. It's also fast but the numbers out there are that are crunching are mind blowing! I wonder if the moms and dads know their kids PS3s are using 200W power 24/7

Looks like streaming processors are the wave of the future science-wise. As usual, it takes a lot of software development!

Thanks GLeeM!!!

PS: your link is broken...I found my way to the right places

 

[biodoc]

Jeez...how much new hardware was purchased for this RACE?

I bought 2 SLACR/MB combos:shocked:

Sid, Markfw, Nate, Dog, FMC, Petrus, etc others I seem to remember pulled out the ol' plastic!

Post if you opened your wallet for the RACE!

 

[Insidious]

Originally posted by: biodoc
Jeez...how much new hardware was purchased for this RACE?

I bought 2 SLACR/MB combos:shocked:

Sid, Markfw, Nate, Dog, FMC, Petrus, etc others I seem to remember pulled out the ol' plastic!

Post if you opened your wallet for the RACE!

I've got to qualify mine.... I was planning on making the switch to Intel (quad) this Spring, but the race gave me the (fun!) reason to do it early. Er..... and I seem to have forgotten to sell the two X2 systems I've 'replaced'

But nonetheless,,,, Two new C2Q PCs (from the case up) :roll:

-Sid

 

[Markfw]

2 new Q6600's (most of the systems), but 5 new video cards that allowed the install of ubuntu on other boxes, which upped my ppd 50% on those boxes. A total of at least 15k ppd was added. And cost ? Probably about $2000 for December alone.

 

[xpblackdragon]

1 new Q6600 system@3500
1 new X2 3600@3000 got a good deal on it
Edit: And upgrade my old opty 165@2600 to a Q6600@3500

 

[Neurodog]

1 Q6600 running stock. mobo will not O/C
1 x2 6000@3300
Overclocked my Opteron 165 from 1.8ghz to 2.6ghz

 

[petrusbroder]

One new AMD x2 6000+ (Barebones system + processor + RAM)

Will be a good BOINC-cruncher later!

 

[Fullmetal Chocobo]

I didn't buy anything more, but I pulled as much favors as I could to get machines running. I figure I did pretty well. I'll be prepping for the race next year.

 

[GenHoth]

F'n A. I'm drunk as a skunk for this race, does that count? We should total how much I've spent on alcohol this race and compare to how much others have spent on hardware!

 

[Slider76]

I'm having heat problems. I can't seem to cool the house enough atm, so the pc has been getting turned off more lately. Today it got to 106f (41.5c), infact over the last week the coolest day was 97f (36c). Its meant to be hotter tomorrow its going to be a crazy new years eve!

 

[Insidious]

A tin o' :beer: and a block o' ice for our....... poor hot friend. :sun:

(I visited your end of the world when I was in the USN.... I love your home!)

-Sid

 

[Pokey]

Originally posted by: biodoc
Jeez...how much new hardware was purchased for this RACE?

I bought 2 SLACR/MB combos:shocked:

Sid, Markfw, Nate, Dog, FMC, Petrus, etc others I seem to remember pulled out the ol' plastic!

Post if you opened your wallet for the RACE!

Well, I replace a couple of slower crunchers with dual cores. One from Compusa going out of business sale and one from Dell Refurb site.
Their production is so much greater I will actually be able to cut down or get rid of all the slower units in time. I am using the SMP client on both with very good results (relative to the older machines). I'm getting over 800 points a day from each compared to 100 or less before. And to my knowledge haven't botched one yet.............. knock on wood.

 

[Insidious]

Pokey, (edit: This isn't limited to Pokey.... )

I really was serious about selling my two X2 systems (2.5GHz and 2.4GHz). If it turns out that you might want to consider used, but nicely treated... please fire me off a PM. I can all but promise that even with shipping, I can do you better than CompUSA or Dell Refurb.

-Sid

PS: To the thread Police:
I won't be putting these in FS:FT under any circumstances... so please... I know you mean well, but don't direct me there.

(edit: OK, so I wish they were 3.5 and 3.4...... fixed )

 

[caferace]

New P180b case and Q6600 (M0 stepping)
Enermax 650 PSU (80+0 with modular plugs, 2x1GB Geil DDR3 @ 12,800 7-7-7-20, Gigabyte X38 DQ6 mobo, 500Gb SATA3 HD, floppy , Radeon 3580 Vid w512Mb ... other stuff...

oof.

-jim