• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Linux A2 Core -advmethods

EvilAlchemist

2[H]4U
Joined
Jan 11, 2008
Messages
2,732
New A2 Core (v2.00) Linux & OSX
Using -advmethods flag

Here are some Numbers from that Folding Forum.

Project 2662
Q6600 @ 3.276GHz (364x9): 4:38/frame, 5967PPD
E6600 @ 3.303GHz (367x9): 9:27/frame, 2925PPD

Those are the only solid number as of yet.
Most other on that forum are from running VM - and only using 2 cores, which A2 need 4 to run full speed and it shows in the numbers for A2 .. only getting about 500PPD increase.

I can tell you that my Q660 @ 3.4 only gets 4100 PPD with 2 x SMP A1.
So this is a big jump in production

Most users are also reporting 97% CPU Usage for the A2 core with 1 SMP client.
Numbers look great so far.
 
This is very good news for those of us with a number of quad core machines. Any reports yet on how well it scales to 8 cores?
 
Um...what forum is "that" forum?
 
Any [H]orde members running A2 core yet?
All quad machines I have are VM, need more info
before any os replacement.
 
I never pay attention to the stupidity posted on the official folding forum.
The team fora are a far better resource for folding info.
 
Project 2662
core A2
Running in a VM across 2 cpus in a Q6600 @ 3.15GHz (350x9)
SLES 10 SP2 x64
PPD 2502 based on the first 3 frames.

Not on a native linux install across 4 cores, but not bad.
It's at 12% with an ETA of about 15.5 hours

The box has a 8800GT/256 and this one VM and its putting out 7700PPD with one whole CPU idle. Think I need to add another GPU
 
VM are gonna hold this core back.

Looks like you got the 500 PPD boost most are reporting running in VM.

Native Quad is where the A2 core really shines.

For all of you who run "notfred" diskless clients .. he will be adding a feature where you can have the different commands put in like -advmethods ...

But I think this will go full client very soon
 
Preparing a Quad boxen this afternoon to try out A2 core.
Running natively in 64bit *nix, I'll report back later.
 
Is this coming out for windows?


Unfortunately no :( too much problems right now with Windows. That sucks since I'm running Windows on all boxens (beside 1) due to either GPU2 running or Windows Home Server.

 
None of my quads have pulled down the A2 core yet, tried using -advmethods flag and everything.
 
None of my quads have pulled down the A2 core yet, tried using -advmethods flag and everything.

If you get assigned a 2662 WU, you'll get the A2 core. Not all WU will use that core unfortunately.

 
Unfortunately no :( too much problems right now with Windows. That sucks since I'm running Windows on all boxens (beside 1) due to either GPU2 running or Windows Home Server.


Same! Would really be nice to get the boost and truly max the resources of each of my machine I know the CPU is not maxed on each right now.
 
I know these questions probably rank right up there with being "some of the dumbest questions ever asked", but please try to bear, bare, bar, beer (just kinda' go along with me :rolleyes:) if you would.

D'oh, how do you get the A2 core? Does it only work for certaian WU's? Will the folding fairies automatically download the core when I need it?

The below, thanks to Evil posting it, looks mighty good to me :p

Project 2662
Q6600 @ 3.276GHz (364x9): 4:38/frame, 5967PPD
E6600 @ 3.303GHz (367x9): 9:27/frame, 2925PPD

FOLD ON!

 
The A2 is only for certain wu's, for now at least. I imagine eventually it'll become the default core over the A1.
 
Looking Good so far... the quick results
Running natively on 64bit Suse 10.3
Q6600 @3.3GHz 2GB RAM
1st couple of steps:
p2662

[22:57:15] Completed 2500 out of 250000 steps (1%)
[23:01:38] Completed 5000 out of 250000 steps (2%)
[23:06:09] Completed 7500 out of 250000 steps (3%)


edit: approx: 4:30 a step or about 6100ppd, not shabby :D
based on 1920 point wu
 
[22:52:36] Project: 2662 (Run 0, Clone 207, Gen 0)
[22:52:36]
[22:52:36] Assembly optimizations on if available.
[22:52:36] Entering M.D.
[22:52:45] (Run 0, Clone 207, Gen 0)
[22:52:45]
[22:52:46] Entering M.D.
NNODES=4, MYRANK=0, HOSTNAME=linux-orlt
NNODES=4, MYRANK=3, HOSTNAME=linux-orlt
NNODES=4, MYRANK=1, HOSTNAME=linux-orlt
NNODES=4, MYRANK=2, HOSTNAME=linux-orlt
NODEID=0 argc=19
NODEID=2 argc=19
:) G R O M A C S (-:

Groningen Machine for Chemical Simulation

:) VERSION 3.3.99_development_200800503 (-:


Written by David van der Spoel, Erik Lindahl, Berk Hess, and others.
Copyright (c) 1991-2000, University of Groningen, The Netherlands.
Copyright (c) 2001-2008, The GROMACS development team,
check out http://www.gromacs.org for more information.


:) mdrun (-:

NODEID=3 argc=19
Reading file work/wudata_01.tpr, VERSION 3.3.99_development_20070618 (single precision)
NODEID=1 argc=19
Note: tpx file_version 48, software version 56
Making 1D domain decomposition 1 x 1 x 4
starting mdrun 'IBX in water'
250000 steps, 500.0 ps.
[22:57:15] Completed 2500 out of 250000 steps (1%)
[23:01:38] Completed 5000 out of 250000 steps (2%)
[23:06:09] Completed 7500 out of 250000 steps (3%)

Writing checkpoint, step 8430 at Sat Jul 26 19:07:52 2008
[23:10:40] Completed 10000 out of 250000 steps (4%)
[23:15:10] Completed 12500 out of 250000 steps (5%)
[23:19:33] Completed 15000 out of 250000 steps (6%)

Writing checkpoint, step 16820 at Sat Jul 26 19:22:52 2008
[23:24:09] Completed 17500 out of 250000 steps (7%)


2662WU 1920 points for 6144 PPD. Smoking
Running natively on 64bit Suse 10.3
Q6600 @3.3GHz 2GB RAM
 
A2 Core Running on q6600@3.3Ghz 2GB RAM
VM Ware server SUSE 10.3 64 Bit 2 Cores
OS WinXP 32

[00:16:58] Completed 2500 out of 250000 steps (1%)

Writing checkpoint, step 253780 at Sat Jul 26 20:22:03 2008
[00:26:56] Completed 5000 out of 250000 steps (2%)

Writing checkpoint, step 257460 at Sat Jul 26 20:37:03 2008
[00:37:13] Completed 7500 out of 250000 steps (3%)
[00:47:04] Completed 10000 out of 250000 steps (4%)

Writing checkpoint, step 261270 at Sat Jul 26 20:52:04 2008
[00:56:55] Completed 12500 out of 250000 steps (5%)
 
Looking Good so far... the quick results
Running natively on 64bit Suse 10.3
Q6600 @3.3GHz 2GB RAM
1st couple of steps:
p2662

[22:57:15] Completed 2500 out of 250000 steps (1%)
[23:01:38] Completed 5000 out of 250000 steps (2%)
[23:06:09] Completed 7500 out of 250000 steps (3%)


edit: approx: 4:30 a step or about 6100ppd, not shabby :D
based on 1920 point wu

Now those are some AWESOME numbers.
Based on your results (& other I have seen) about a 2000 PPD boost from my 2 X SMP A1 to 1 x SMP A2 ... awesome!
 
Wondering if we Windows people did a single VM with all 4 cores dedicated if there would be enough room for the gpu to run on! Cuz honestly its killing me that my quads aren't producing what they should be!

I assume they accomplished this from no longer utilizing the network connection to send data along, its now fully processed within the CPU?

 
Wondering if we Windows people did a single VM with all 4 cores dedicated if there would be enough room for the gpu to run on! Cuz honestly its killing me that my quads aren't producing what they should be!

I assume they accomplished this from no longer utilizing the network connection to send data along, its now fully processed within the CPU?


Tell me how to get a VM software which can tap into 4 cores and we'll talk ;) Only VMWare (and 1-2 obscur ones) can use 2 cores. I'm sure there is other versions but we're talking $$$$$$.

 
Tell me how to get a VM software which can tap into 4 cores and we'll talk ;) Only VMWare (and 1-2 obscur ones) can use 2 cores. I'm sure there is other versions but we're talking $$$$$$.


I'll look at teh version I had I swear it was 4 core capable!
 
I'll look at teh version I had I swear it was 4 core capable!

I think that the server version and workstation version top out at 2 cores. I have a quad xeon running ESX with a 4 core sles10 x64 VM running. I'll post something if it ever gets a 2662 and a2 work unit.
 
Starting to see a steady diet of 2662 WUs on my VM 64bit linux boxen.
Out of the 8 SMP clients I have running 4 of them are running 2662s.
3.3GHZ averages about 10:00 a step, 3.4ghz 9:45 a step for
2760-2830ppd. Usually I see 2100-2200ppd with 2065 WUs.
Nice increase, to run the new work units you need either 64 Bit Linux
or MAC OSX. set -advmethods flag on.
 
Tell me how to get a VM software which can tap into 4 cores and we'll talk ;)
How about this one: http://bochs.sourceforge.net/

Claimed support up to 8 processors! http://bochs.sourceforge.net/doc/docbook/user/bochsrc.html#BOCHSOPT-CPU

I have no idea how good this is but looks to be open source. I've been wanting to give it a try for a while but had too many other computer-related stuff take up my time. If this works well enough, it could be what GPU users with quads or higher core systems are looking for.

 
How about this one: http://bochs.sourceforge.net/

Claimed support up to 8 processors! http://bochs.sourceforge.net/doc/docbook/user/bochsrc.html#BOCHSOPT-CPU

I have no idea how good this is but looks to be open source. I've been wanting to give it a try for a while but had too many other computer-related stuff take up my time. If this works well enough, it could be what GPU users with quads or higher core systems are looking for.


I might give that a try this week. Once I have my current problem child up and running which should be tonight then I'll work on this on a test box!

 
I might give that a try this week. Once I have my current problem child up and running which should be tonight then I'll work on this on a test box!


I will love to hear the results of your testing. Can you make a new thread about it when it's time so we can follow.
 
D'oh, how do you get the A2 core? Does it only work for certaian WU's? Will the folding fairies automatically download the core when I need it?



It looks like you just have to config the client with the -advmethods flag. Then if you sacrifice a chicken over your boxen you might indeed get a visit from the folding fairies. When they bring tthe right WU they will indeed bring the A2 core along ;)

Luck.

 
One of my dual cores keeps getting 2662's but neither of my quads have yet :(


 
It looks like you just have to config the client with the -advmethods flag. Then if you sacrifice a chicken over your boxen you might indeed get a visit from the folding fairies. When they bring tthe right WU they will indeed bring the A2 core along ;)

Luck.


Yeah, I tried all of that (of course I used a plastic tarp to keep chicken blood off my boxen), I've said numerous prayers to the computer Gods and I left cookies and milk out for the folding fairies (really good chocolate chip jobbers too), but still no A2 core yet :rolleyes:

Only kiddin' (not about the tarp though ;)) I'm gonna' give the ole' -advmethods trick another try when I complete a few more A1 WU's :D

FOLD ON!

 
I added an 8 core box today 2x E5345's and it grabbed a A1 core WU. Here's to hoping stanford starts putting out more A2 WU's.

 
Back
Top