SMP2 And Core A3 Released

[LCB001]

Stanford has released the new A3 core out into the wild for testing under -advmethods with new SMP2 WU's to got with it. Announcement by Kasson; http://foldingforum.org/viewtopic.php?f=24&t=13038

They have also redone the points system and developed a new benchmark system to go with it. Posted here; http://foldingforum.org/viewtopic.php?f=24&t=13039.

Like the -bigadv WU's you need a passkey and to have completed 10 WU's with an 80% success ratio to qualify for the Bonus.

Reports on other forums are looking good with ppd equal or better than running a VM on quads with a decent OC, as an example a Q9550 @3.8 ~8000ppd...

 

[Disparia]

LOL figures I set up a new VM yesterday and now the Windows client is up to par.

Oh well, will let it finish this A3 unit it just grabbed and will work on it in the morning...

 

[theonedub]

This is pretty big news, no?

 

[mstenholm]

It seems like 32 bit Windows users are left out.

Edit: http://folding.stanford.edu/English/DownloadWinOther has the beta client

 

[El Fiendo]

This is pretty big news, no?

Yea, its pretty big. If I'm reading this right, this client makes VM's obsolete as of today if it proves faster and more efficient. In theory it should be faster because there isn't extra overhead of the VM. The VM's will likely update their image if they haven't already, and then we'll have to wait for benchmarks on which is faster to know for certain. I can't see why the VM's would be though if everyone has access to the same cores now. Also, all CPU folding now can get bonuses with this release, regular CPU and -bigadv folding alike.

 

[Easy Rhino]

is this available for download yet?

 

[msgclb]

is this available for download yet?

That was answered above.
High Performance Clients

Go down to the 'Windows: V6 Beta SMP/CPU clients'.

I believe that we should use the Windows XP/2003/Vista/2008 SMP client console version 6.29 beta for MPICH (32-bit or 64-bit) or the Drop-in binary for current Windows SMP console client (6.29) if we just want to replace v5.92.

The question I have is can I use -smp 8 on a Core i7 rig or do I need two clients using -smp 4?

 

[Easy Rhino]

i thought i was looking at the older clients. anyway, this works out of the box for me. should i use any special flag for my q9650 ?

 

[BUCK NASTY]

Yea, its pretty big. If I'm reading this right, this client makes VM's obsolete as of today if it proves faster and more efficient. In theory it should be faster because there isn't extra overhead of the VM. The VM's will likely update their image if they haven't already, and then we'll have to wait for benchmarks on which is faster to know for certain. I can't see why the VM's would be though if everyone has access to the same cores now. Also, all CPU folding now can get bonuses with this release, regular CPU and -bigadv folding alike.

El, early this morning I switched over a Phenom II quad @ 3.2Ghz to the new A3 core. Did a clean install I entered my passkey and started my run of 10 WU's. I have the -smp and -advmethods flags set. I am only getting 1000ppd with cpu fully loaded. Have you seen any feedback from others on the ppd of the first 10 WU's?

 

[Easy Rhino]

hrm, one of the links said q9550 were getting 8000 ppd. ive got a faster and meaner q9650 and im barely pulling 1000ppd. btw, using the -smp flag produces an error. something about cant connect on a certain port and an error running mpiexec.

edit: nvm i didnt follow the directions! f mondays!

 

[Easy Rhino]

nope, still working very slowly with mpich installed and the -smp flag added to the executable. it works but it is crawling. maybe i am missing something.

 

[Disparia]

P6012 / 470 point unit, getting about 450ppd from a C2D @ 2.8Ghz.

 

[Easy Rhino]

P6012 / 470 point unit, getting about 450ppd from a C2D @ 2.8Ghz.

ok so im prolly pulling a tad bit more. oh well, i was hoping to boost production by 5000 ppd. i hope they put out another beta version soon!

 

[BUCK NASTY]

ok so im prolly pulling a tad bit more. oh well, i was hoping to boost production by 5000 ppd. i hope they put out another beta version soon!

I'm guessing after the initial 10 WU run, we will get the work unit that has the early return bonus. It will take me 4.5 days to finish the required 10 WU's.

 

[mstenholm]

Have you seen any feedback from others on the ppd of the first 10 WU's?

From the foldingforum.org : The base point values for SMP2 work units will appear low; the benchmarked points values **include bonuses.**

To get the good stuff (the bonus):

Again from foldingforum.org -

1. Must use a passkey to receive bonus points
2. Must successfully return >=10 A2 or SMP2 work units with their passkey to receive bonus points
3. Must successfully return >80% of A2 or SMP2 work units to receive bonus points

Edit: Buck beat me to this post. See above.

 

[El Fiendo]

El, early this morning I switched over a Phenom II quad @ 3.2Ghz to the new A3 core. Did a clean install I entered my passkey and started my run of 10 WU's. I have the -smp and -advmethods flags set. I am only getting 1000ppd with cpu fully loaded. Have you seen any feedback from others on the ppd of the first 10 WU's?

From what I've seen on foldingforum.org, that's to be expected. The low PPD will apparently be corrected by the bonus points.

I'm not certain if this allows for -bigadv folding yet, I've yet to see anything about it. I can't see why it wouldn't, as I could've sworn it was only an additional drop-in required to get -bigadv working.

 

[thebluebumblebee]

I'm seeing MUCH lower than expected PPD from my S939 4400+. Was used to 800-1100 on my FX60, so I was expecting 80% of that. So far, at 2% it's at 308PPD!

 

[theonedub]

This is a an easy question, can someone tell me how to setup a passkey?

 

[El Fiendo]

This is a an easy question, can someone tell me how to setup a passkey?

http://fah-web.stanford.edu/cgi-bin/getpasskey.py

Input your name as F@H would see it (El_Fiendo in my case) and your e-mail. They will send it to you, and then you enter it at client setup.

 

[theonedub]

Will I have to delete and resinstall my current GPU Folding clients to input the passkey?

 

[Easy Rhino]

ok so i ran configonly and put in the passkey. now do i execute with the -smp flag and -bigadv ?

 

[msgclb]

I switched over my Gigabyte GA-EX58-UD3R @ 2.66Ghz to the new A3 core. Someday I hope to get this rig in shape to switch to the -bigadv WUs.

I first removed the v5.92 files and did a clean install of the new v6.29 client.

Then using Buck's How to set up Windows SMP CPU Client I started from Step 2b.

I then set these switches in my shortcut...
-configonly -local -advmethods -smp 8

I then started the client and after entering my user name and team number I entered my passkey. I continued with the options in step 6.

I then removed -configonly and started the client. The smp client loaded all 8 cores and I haven't got any problems yet.

In fact I just completed 2%

18:37:41 Competed 0 out of 500000 steps (0%)
18:42:33 Completed 5000 out of 500000 steps (1%)
18:47:21 Complete 10000 out of 500000 steps (2%)

I have an ETA of 7h 50min running P6014 484 points and 1452 PPD.

 

[Easy Rhino]

if you set advanced methods to yes in the configonly i dont think you have to set the flag. i could be wrong tho.

 

[El Fiendo]

Will I have to delete and resinstall my current GPU Folding clients to input the passkey?

I don't think the GPU side is affected by the bonus point system, so adding in a passkey to them wouldn't gain anything. So far as I can tell this is CPU and CPU only.

ok so i ran configonly and put in the passkey. now do i execute with the -smp flag and -bigadv ?

Apparently the reason these aren't bigadv capable is because Windows still only has access to A1 and A3 coress, while Linux has A1-A3 capability. This is supposed to change over the coming weeks, along with the phasing out of the A1 cores. I'm currently strip mining and trolling around a few forums trying to piece together the puzzle, but apparently if your rig is bigadv capable you should keep the VMs but update their image (in a few weeks time they will no longer get work units on the outdated client version). If your CPU is not -bigadv capable but still folds, then you can switch to this one to try it out. You have a chance to get A1 cores if no A3 cores are available which would mean a hit in your PPD, but A3 should yield about the same as you were getting before due to the bonus system.

 

[Easy Rhino]

alright so let me get this straight, according to the instructions after entering the passkey i need to complete 10 work units with an equal to or greater than 80% success rate. then i get the advanced units or something which get points bonus? do i have to have -bigadv set for this to occur?