• A Great friend to the HardForum with a great kid that he is trying to get a scholorship to continue his schooling. Please give hime a vote! Only 24 hours left! Thanks.
    If you have an VOTE FOR KEENAN!

New extra large -bigadv on Beta

Schro said:
So maybe my head math is off, but it would seem to me that a fast 8 thread machine (i.e. a 2600K) could complete these with plenty of time to spare before the preferred deadline. I could see the i7-920's of the world getting left in the dust, but the Sandies should in theory be able to finish.... Thoughts?
Click to expand...
I think your rite.

Yeh... I would think 4 day for bonus and 5 day credit would have been a better choice.
 
Mtnduey said:
does that mean the normal -bigadv flag is all thats required for these now?
Click to expand...

Kasson didn't specify, but yes... edit I as wrong about the rest. /edit
 
Last edited:
Mtnduey said:
does that mean the normal -bigadv flag is all thats required for these now?
Click to expand...

That should be exactly what it means. What did that take? 5 days? I am not sure that having -bigbeta open to the masses is such a bad thing...
 
Mtnduey said:
does that mean the normal -bigadv flag is all thats required for these now?
Click to expand...

Someone please let us know... I will need to play with the boxen if so. So long as -Bigbeta will continue to pull from the same pool I will just leave it on...

Schro said:
So maybe my head math is off, but it would seem to me that a fast 8 thread machine (i.e. a 2600K) could complete these with plenty of time to spare before the preferred deadline. I could see the i7-920's of the world getting left in the dust, but the Sandies should in theory be able to finish.... Thoughts?
Click to expand...

You are right with regards to compute power, their deadlines are massive and I assume they are structure so as to allow 2p rigs with HT to complete them (think a couple of 2ghz quads running 5-6 days). In most units my single 12 core dowrecka at 3.0ghz is 20-30% faster than a 4.8-5.0ghz 2600k when running linux (e.g. 6901: 74k on g34 vs 55-60k on 2600k). This chip is currently crunching a 6903 at 42:35 tpf for ~119k ppd, a sandy bridge should definitely make the deadline from a compute standpoint.

IIRC though, there are some assumptions for thread counts built into different work units that makes such an argument more difficult to validate. Case in point would be new bigadv units crashing on SMP-7 because of the prime divisible, it seems unlikely this is as big an issue with the non-prime 8 and 12 core counts but it could simply be Stanford futureproofing the WU class. I think as long as they continue to offer stratified units that the client can download based on your thread count, it will be a pretty fair system which gives bigger rewards for the disproportionately more expensive server class hardware.
 
Last edited:
amdgamer said:
That's it? I would have thought that it would have been a much bigger upload considering how much time it takes.
Click to expand...

It takes a bit over 2x the time to finish, so a bit over 2x the data is right on target.
 
R-Type said:
It takes a bit over 2x the time to finish, so a bit over 2x the data is right on target.
Click to expand...

Ohh, I didn't realize that they were that quick. My Gulftown does a 6900 in just under 2 days, which means that I could easily do a 6903 in just under 4 days then. If that is the case, I am curious why they arn't opening up superbigadv to 8 cores as we probably know very well that the upcoming Zambezi 8 cores will be more than capable of crunching one of these.
 
amdgamer said:
Ohh, I didn't realize that they were that quick. My Gulftown does a 6900 in just under 2 days, which means that I could easily do a 6903 in just under 4 days then. If that is the case, I am curious why they arn't opening up superbigadv to 8 cores as we probably know very well that the upcoming Zambezi 8 cores will be more than capable of crunching one of these.
Click to expand...

Sounds about right, my single 12 core g34 is right on 3 days to complete 6903.

As for the restrictions, I can only hope they have a technical reason for it (12 highly interleaved threads with multiple dependencies and formulas that break at lower thread counts or lose a ton of efficiency or something) and not simply arbitrary to get us to buy more hardware..
 
R-Type said:
Sounds about right, my single 12 core g34 is right on 3 days to complete 6903.

As for the restrictions, I can only hope they have a technical reason for it (12 highly interleaved threads with multiple dependencies and formulas that break at lower thread counts or lose a ton of efficiency or something) and not simply arbitrary to get us to buy more hardware..
Click to expand...

I didn't know that you were running a single 12 core Magny in your rigs. Which one are you running, and are you overclocking it at all? 3 days seems to be pretty fast for a 12 core Magny even though they are all clocked pretty low.
 
amdgamer said:
I didn't know that you were running a single 12 core Magny in your rigs. Which one are you running, and are you overclocking it at all? 3 days seems to be pretty fast for a 12 core Magny even though they are all clocked pretty low.
Click to expand...

Just a little 50% overclock to 3ghz...;)
 
R-Type said:
Just a little 50% overclock to 3ghz...;)
Click to expand...

That must be sweet, as 3 days sounded fast for a stock 12 core Magny.

If I may ask, what kind of cooling solution are you using? Running a full real 12 cores at 3ghz must not be easy due to the insane amount of heat it would produce. I dream of a time when I will have the spare cash to invest in a dual or quad G34 setup.
 
amdgamer said:
That must be sweet, as 3 days sounded fast for a stock 12 core Magny.

If I may ask, what kind of cooling solution are you using? Running a full real 12 cores at 3ghz must not be easy due to the insane amount of heat it would produce. I dream of a time when I will have the spare cash to invest in a dual or quad G34 setup.
Click to expand...

Its truly not that bad... I cool 2 of them currently... water :D lots o water....
 
KMac said:
Kasson didn't specify, but I would add -advmethods just to be sure. Just don't mix -advmethods with -bigbeta.
Click to expand...
-advmethods and -bigadv should never be mixed. They contradict each other. -bigadv is all that is needed.
 
amdgamer said:
That must be sweet, as 3 days sounded fast for a stock 12 core Magny.

If I may ask, what kind of cooling solution are you using? Running a full real 12 cores at 3ghz must not be easy due to the insane amount of heat it would produce. I dream of a time when I will have the spare cash to invest in a dual or quad G34 setup.
Click to expand...

It does make a fair bit but you would be surprised how reasonable it is. If you keep in mind its the same lithography process as the desktop phenom II's, it kind of follows the same voltage/speed curve (except you are top end limited by the amount of current you can draw). This chip can do 3.1 at 1.3v and 3.2 somewhere around 1.35v but I run it at 3ghz with just 1.25v. I'm actually cooling it with an old Freezer 7 pro I had kicking around and modified the 775 mounting to fit g34 (corner to corner 775 is just about identical to point to point mounting on G34). This keeps the chip at 57-58c folding, not the best temps but no worse than people are getting with stock clocks on 4p boards with stock dynatrons.
 
Putting 1.4V into an i7 920 would melt a Freezer 7. That's very cool compared to what 45nm Intel CPU's used to do, now SB is much more reasonable.
 
Yeah I agree. Freezer 7s aren't great. Those chips must be cooler than I thought.
 
I just picked up a 6903 using just the 'smp' and 'bigbeta' flags
 
Zero82z said:
From my understanding, using the -bigbeta flag should allow you to receive both beta and non-beta -bigadv units.
Click to expand...
I can confirm that much. I've been picking up 6901s since I added the -bigbeta flag.
 
So when are they going to update HFM.net?

I wana see that wondrous PPD.
 
Kendrak said:
So when are they going to update HFM.net?.
Click to expand...

HFM>Edit>Preferences>Web Settings.
HFM%20pummary%20update.JPG
 
so whats the bottom line? -smp -bigadv is enough to get these on qualifying rigs or do we need something else?
 
I can confirm -bigbeta still does as of today. I cannot confirm -bigadv does.
 
Bigadv might do it but The only confirmed way to get 6903 is "-bigbeta" at this point
 
Zink said:
Putting 1.4V into an i7 920 would melt a Freezer 7. That's very cool compared to what 45nm Intel CPU's used to do, now SB is much more reasonable.
Click to expand...

jebo_4jc said:
Yeah I agree. Freezer 7s aren't great. Those chips must be cooler than I thought.
Click to expand...

Honestly there is much more to it than that. The biggest difference is die size; a 45nm nehalem has a die 263mm2 while a G34 dodeca has a total die area of 692mm2. Both of these have heat spreaders but the rest of the IHS is nowhere near as efficient at transferring heat as the area directly over the die. So while my dodeca may be putting out a bit more heat than an overclocked nehalem (although honestly not much more, nehalem is a hog), it has almost 3x as much optimum heat transfer area.

I used to run that freezer 7 on a C2Q that would hit the low 70's but the cooler wouldn't even feel warm. With it installed on the dodeca though, that heatsink is very warm and my IR thermometer reads around 40C on the lower fins. This is just a case where the physical layout of the dies makes it much more efficient at conveying that heat to a heatsink.
 
[08:25:22] - Digital signature verified
[08:25:22]
[08:25:22] Project: 6903 (Run 7, Clone 14, Gen 0)
[08:25:22]
[08:25:22] Assembly optimizations on if available.
[08:25:22] Entering M.D.
:) G R O M A C S (-:

Groningen Machine for Chemical Simulation

:) VERSION 4.5.3 (-:

Written by Emile Apol, Rossen Apostolov, Herman J.C. Berendsen,
Aldert van Buuren, Pär Bjelkmar, Rudi van Drunen, Anton Feenstra,
Gerrit Groenhof, Peter Kasson, Per Larsson, Pieter Meulenhoff,
Teemu Murtola, Szilard Pall, Sander Pronk, Roland Schulz,
Michael Shirts, Alfons Sijbers, Peter Tieleman,

Berk Hess, David van der Spoel, and Erik Lindahl.

Copyright (c) 1991-2000, University of Groningen, The Netherlands.
Copyright (c) 2001-2010, The GROMACS development team at
Uppsala University & The Royal Institute of Technology, Sweden.
check out http://www.gromacs.org for more information.


:) Gromacs (-:

Reading file work/wudata_04.tpr, VERSION 4.5.4-dev-20110530-cc815 (single precision)
[08:25:29] Mapping NT from 24 to 24
Starting 24 threads
Making 2D domain decomposition 6 x 4 x 1
starting mdrun 'Overlay'
250000 steps, 1000.0 ps.
[08:25:33] Completed 0 out of 250000 steps (0%)
[08:57:46] Completed 2500 out of 250000 steps (1%)
[09:31:53] Completed 5000 out of 250000 steps (2%)
[10:06:57] Completed 7500 out of 250000 steps (3%)
[10:42:30] Completed 10000 out of 250000 steps (4%)
[11:17:04] Completed 12500 out of 250000 steps (5%)
[11:51:16] Completed 15000 out of 250000 steps (6%)
[12:24:20] Completed 17500 out of 250000 steps (7%)

Exciting!
 
HFM is showing no PPD for me. Has anybody gotten it working properly?
 
I had 5 6903s going when I left the house this morning, so they are definately out there.
 
jebo_4jc said:
HFM is showing no PPD for me. Has anybody gotten it working properly?
Click to expand...

Make sure you have psummaryC, and update the rpojects list. It worked for me as soon as these units came out, and updated correctly when the base point value dropped as well.
 
You must log in or register to reply here.
Back
Top