SMP2 introduces new points system

[CJ-5]

My laptop is a core2 duo @ 2.4 GHz with Windows 7 32 bit.

It took about 36 hrs to complete an a1 wu with the older client.

With v 6.29 it will complete an a3 in about 24 hrs.

This is quite an improvement even without bonus points...and just as I was thinking of giving my laptop a rest from folding because it was so slow.

 

[ChasR]

My laptop is a core2 duo @ 2.4 GHz with Windows 7 32 bit.

It took about 36 hrs to complete an a1 wu with the older client.

With v 6.29 it will complete an a3 in about 24 hrs.

This is quite an improvement even without bonus points...and just as I was thinking of giving my laptop a rest from folding because it was so slow.

If you do the math, completing a 1920 point WU in 36 hours yeilds 1280 ppd.
Completing a p6012 in 24 hours yeilds 1680 ppd. So compared to Win SMP, Win SMP2 is a big improvement. This is about the same as your rig would make in a VM running a2 WUs.

 

[ChasR]

i7 920 @ 4.2ghz running -smp 4 and A3 p6012. Credit is only 470 and TPF 3:48 which is reporting a grand ppd of 1780. I upgraded HFM.net to .4.8 and downloaded projects from stanford. I better get one hell of a bonus, people are reporting upwards of 16k PPD.

You'll need -smp 8 if you want to match the folks getting 16,000 ppd.

 

[GIXXERGUY6]

ok now my head is really spinning holy crap.

No wonder Adak told me to shut up and hold on.

the winsmp2 (where do I get it)

this 6.29 where do I get it and which one do I get?

windows 7 64bit here.

I uninstalled my vmware player so I want to know where this stuff is at so I can get back on the block (

 

[torin3]

ok now my head is really spinning holy crap.

No wonder Adak told me to shut up and hold on.

the winsmp2 (where do I get it)

this 6.29 where do I get it and which one do I get?

windows 7 64bit here.

I uninstalled my vmware player so I want to know where this stuff is at so I can get back on the block (

You can find it here: http://folding.stanford.edu/English/DownloadWinOther

The client you want is the "6.29 beta for MPICH (32-bit or 64-bit)"

 

[jonspd]

So compared to Win SMP, Win SMP2 is a big improvement.

very big...


I'm getting 5-6k on my 955 where I was getting 2-3K so that's double in my book

my HFM went from 32-35k to 38-42k

 

[deadlysyn]

I have an A3 WU on my Q6600 right now. It is running at around 5 min frame times, which is about half of what I was getting with the old A1 WU's. HFM is reporting 6700 PPD just from the CPU, but it is choking my GPU2 client like nobody's business. I was getting around 4k-5k out of the GPU2 client, and around 2700 PPD out of the A1 WU's. Now my GPU client is running at around 33 minute frame times, and at about 800 PPD. I don't think it is going to be making deadlines with it running like this. I may switch back to the 6.24 for a while, atleast until we can figure out how to run these new WU's without choking out our GPU's. This is running in Windows 7 x64.

 

[GIXXERGUY6]

now I get this

Note: Please read the license agreement ([email protected] -license). F
urther
use of this software requires that you have read and accepted this agreement.

8 cores detected
If you see this twice, MPI is working
If you see this twice, MPI is working


--- Opening Log file [January 26 20:45:51 UTC]


# Windows SMP Console Edition #################################################
###############################################################################

                       Folding@Home Client Version 6.29

                          http://folding.stanford.edu

###############################################################################
###############################################################################

Launch directory: E:\Folding
Executable: E:\Folding\[email protected]
Arguments: -smp 8 -verbosity 9

[20:45:51] - Ask before connecting: No
[20:45:51] - User name: GIXXERGUY6 (Team 32)
[20:45:51] - User ID: 73F20D5537E229A4
[20:45:51] - Machine ID: 1
[20:45:51]
[20:45:51] Loaded queue successfully.
[20:45:51]
[20:45:51] - Autosending finished units... [January 26 20:45:51 UTC]
[20:45:51] + Processing work unit
[20:45:51] Trying to send all finished work units
[20:45:51] Work type 78 not eligible for variable processors
[20:45:51] + No unsent completed units remaining.
[20:45:51] Core required: FahCore_78.exe
[20:45:51] - Autosend completed
[20:45:51] Core found.
[20:45:51] Working on queue slot 01 [January 26 20:45:51 UTC]
[20:45:51] + Working ...
[20:45:51] - Calling 'mpiexec -np 4 -channel auto -host 127.0.0.1 FahCore_78.exe
 -dir work/ -suffix 01 -checkpoint 15 -verbose -lifeline 4060 -version 629'

[20:45:51]
[20:45:51] *------------------------------*
[20:45:51] Folding@Home Gromacs Core
[20:45:51] Version 1.90 (March 8, 2006)
[20:45:51]
[20:45:51] Preparing to commence simulation
[20:45:51] - Ensuring status. Please wait- Files status OK
[20:45:52] - Expanded 464144 -> 2244013 (decompressed 483.4 percent)
[20:45:52]
[20:45:52] Project: 6316 (Run 372, Clone 15, Gen 21)
[20:45:52]
[20:45:52] Assembly optimizations on if available.
[20:45:52] Entering M.D.
[20:46:08]
[20:46:09] - Expanded 464144 -> 2244013 (decompressed 483.4 percent)
[20:46:09]
[20:46:09] Project: 6316 (Run 372, Clone 15, Gen 21)
[20:46:09]
[20:46:09] Entering M.D.
[20:46:12] s  (92%)
[20:46:12] Extra SSE boost OK.
[20:46:12] Size of work/wudata_01.xtc not what saved.
[20:46:12]
[20:46:12] Folding@home Core Shutdown: FILE_IO_ERROR

did it again and it said client died

do I just need to start over? and if so how can I uninstall this and start over again without deleting my whole folding folder with my gpu stuff in it

 

[ChasR]

It died because you added the -smp flag when running a uniprocessor WU. Restart the client and it should work.

 

[ChasR]

I have an A3 WU on my Q6600 right now. It is running at around 5 min frame times, which is about half of what I was getting with the old A1 WU's. HFM is reporting 6700 PPD just from the CPU, but it is choking my GPU2 client like nobody's business. I was getting around 4k-5k out of the GPU2 client, and around 2700 PPD out of the A1 WU's. Now my GPU client is running at around 33 minute frame times, and at about 800 PPD. I don't think it is going to be making deadlines with it running like this. I may switch back to the 6.24 for a while, atleast until we can figure out how to run these new WU's without choking out our GPU's. This is running in Windows 7 x64.

All I can say is if you have SMP2 priority at idle and gpu priority at low, nothing you do with the smp2 client will affect gpu production. I have a lot of SMP2 clients running now on various OSes (native Linux, VM Linux, XP x32, XP x64 and Win 7 x64) and none of them are suffering a loss of GPU production.

 

[GIXXERGUY6]

It died because you added the -smp flag when running a uniprocessor WU. Restart the client and it should work.

the client isn't working I don't believe

 

[Audioaficionado]

I've got a Q6600 @3GHz running VMP/Linux/SMP A2 + one GPU2/GTX9800+

My main issue with SMP is the Stanford work servers crapping out too frequently and SMP defaulting to less productive WUs. I fold 'em anyway even though it's madining to see only ~50% of my potential processor power being utilized or worse yet, just sitting idle trying to connect to a dead server for hours.

If A3 is on more reliable servers, it would be worth the price of admission even if my PPD potential is the same as A2 WUs.

 

[ChasR]

GIX,
Delete the work dir and queue.dat and restart.

 

[Darur]

Ok, read through this and I'm confused still. I'm running a VM (notfreds) on a C2D (laptop), should I switch to windows SMP or stick with notfred's? I don't know bupkis about A2 this or 6.29 that, just tell me how to help the science

Also, I have an older AM2 Athlon X2 4600 @ 2.4GHz running two instances of the single client along with the GPU client. Should I switch those to SMP2 or just keep with the single clients?

 

[deadlysyn]

All I can say is if you have SMP2 priority at idle and gpu priority at low, nothing you do with the smp2 client will affect gpu production. I have a lot of SMP2 clients running now on various OSes (native Linux, VM Linux, XP x32, XP x64 and Win 7 x64) and none of them are suffering a loss of GPU production.

That may be what my problem is then. I have always had both clients at idle priority, since I never had any problems with production before. When I get a chance, I will reconfigure the GPU2 client with a low priority and see how that works out.

Ok, read through this and I'm confused still. I'm running a VM (notfreds) on a C2D (laptop), should I switch to windows SMP or stick with notfred's? I don't know bupkis about A2 this or 6.29 that, just tell me how to help the science

Also, I have an older AM2 Athlon X2 4600 @ 2.4GHz running two instances of the single client along with the GPU client. Should I switch those to SMP2 or just keep with the single clients?

You may be OK with the regular A1 WU's on that 4600, but it sounds like the A3's will suffer. I would think you are better off with either the 6.24 version of the SMP client, or stick with the dual uniproc clients. I can't say for sure if it will be OK for the standard SMP WU's, but you could always give it a try and see if it will make the deadlines. I think for the C2D, you are probably better off with the notfred's client, as it sounds like the A3 WU's may not make the deadlines with it.

 

[dz_jad]

Thanks for the summary ChasR! On a -oneunit now and going to update to 6.29 as soon as this WU Finishes! Its good to know that I don't have to have the VM now to get some better PPD (rig in sig).

 

[PeddlerOfFlesh]

All I can say is if you have SMP2 priority at idle and gpu priority at low, nothing you do with the smp2 client will affect gpu production. I have a lot of SMP2 clients running now on various OSes (native Linux, VM Linux, XP x32, XP x64 and Win 7 x64) and none of them are suffering a loss of GPU production.

I had the same problem and it was indeed a priority issue. GPU dropped to 400 PPD. Changed priority and GPU went up to 3113 PPD (still slower than it was before). CPU stayed the same.

 

[torin3]

Ok, I was getting about 4800-5200 PPD in my Ubuntu VM.

Switched to WinSMP, and I'm getting about 5400-6000 PPD.

The 2 GPU2 clients on this machine are pulling the same PPD as before.


GPU priority = low
SMP priority = idle

Working great!

 

[deadlysyn]

I had the same problem and it was indeed a priority issue. GPU dropped to 400 PPD. Changed priority and GPU went up to 3113 PPD (still slower than it was before). CPU stayed the same.

Keep in mind that PPD will vary depending on WU. Give it some time, and your GPU is likely to level out. My 9800GT went right back up to somewhere over 4K PPD on a 1888 pointer, while SMP2 PPD may have dropped by a couple hundred. Of course, I waited til the last minute on the passkey, like a moron:, so it's going to take a few more SMP WU's before I can qualify for the bonus points. HFM is reporting somewhere over 10K PPD out of the rig, for the first time.

 

[ChasR]

If you haven't folded with a passkey, don't switch to SMP2 or from SMP a2 WUs untill you've completed 10 a2 WUs with your new passkey. THe a3 WUs without the bonus give truly dismal ppd. At least add the passkey to all your clients to speed up bonus qualification.