Windows BIGADV is Here

amdgamer

Supreme [H]ardness
Joined
Oct 27, 2004
Messages
4,880
Check this out! Look at what my Windows SMP client just downloaded! Lets hope it is more stable running natively in Windows than it was in Linux.


[15:07:07] Completed 490000 out of 500000 steps (98%)
[15:10:26] Completed 495000 out of 500000 steps (99%)
[15:13:45] Completed 500000 out of 500000 steps (100%)
[15:13:46] DynamicWrapper: Finished Work Unit: sleep=10000
[15:13:56]
[15:13:56] Finished Work Unit:
[15:13:56] - Reading up to 20302920 from "work/wudata_03.trr": Read 20302920
[15:13:56] trr file hash check passed.
[15:13:56] edr file hash check passed.
[15:13:56] logfile size: 55430
[15:13:56] Leaving Run
[15:14:01] - Writing 20392626 bytes of core data to disk...
[15:14:01] ... Done.
[15:14:03] - Shutting down core
[15:14:03]
[15:14:03] Folding@home Core Shutdown: FINISHED_UNIT
[15:14:07] CoreStatus = 64 (100)
[15:14:07] Unit 3 finished with 96 percent of time to deadline remaining.
[15:14:07] Updated performance fraction: 0.950619
[15:14:07] Sending work to server
[15:14:07] Project: 6020 (Run 0, Clone 51, Gen 206)


[15:14:07] + Attempting to send results [June 29 15:14:07 UTC]
[15:14:07] - Reading file work/wuresults_03.dat from core
[15:14:07] (Read 20392626 bytes from disk)
[15:14:07] Connecting to http://171.64.65.54:8080/
[15:19:54] Posted data.
[15:19:54] Initial: 0000; - Uploaded at ~57 kB/s
[15:19:55] - Averaged speed for that direction ~57 kB/s
[15:19:55] + Results successfully sent
[15:19:55] Thank you for your contribution to Folding@Home.
[15:19:55] + Number of Units Completed: 23

[15:20:00] Trying to send all finished work units
[15:20:00] + No unsent completed units remaining.
[15:20:00] - Preparing to get new work unit...
[15:20:00] Cleaning up work directory
[15:20:00] + Attempting to get work packet
[15:20:00] Passkey found
[15:20:00] - Will indicate memory of 12286 MB
[15:20:00] - Connecting to assignment server
[15:20:00] Connecting to http://assign.stanford.edu:8080/
[15:20:01] Posted data.
[15:20:01] Initial: 43AB; - Successful: assigned to (171.67.108.22).
[15:20:01] + News From Folding@Home: Welcome to Folding@Home
[15:20:01] Loaded queue successfully.
[15:20:01] Connecting to http://171.67.108.22:8080/
[15:20:09] Posted data.
[15:20:09] Initial: 0000; - Receiving payload (expected size: 24840475)
[15:20:54] - Downloaded at ~539 kB/s
[15:20:54] - Averaged speed for that direction ~517 kB/s
[15:20:54] + Received work.
[15:20:54] Trying to send all finished work units
[15:20:54] + No unsent completed units remaining.
[15:20:54] + Closed connections
[15:20:54]
[15:20:54] + Processing work unit
[15:20:54] Core required: FahCore_a3.exe
[15:20:54] Core found.
[15:20:54] Working on queue slot 04 [June 29 15:20:54 UTC]
[15:20:54] + Working ...
[15:20:54] - Calling '.\FahCore_a3.exe -dir work/ -nice 19 -suffix 04 -np 8 -checkpoint 15 -verbose -lifeline 2944 -version 629'

[15:20:54]
[15:20:54] *------------------------------*
[15:20:54] Folding@Home Gromacs SMP Core
[15:20:54] Version 2.22 (Mar 12, 2010)
[15:20:54]
[15:20:54] Preparing to commence simulation
[15:20:54] - Looking at optimizations...
[15:20:54] - Created dyn
[15:20:54] - Files status OK
[15:20:59] - Expanded 24839963 -> 30791309 (decompressed 123.9 percent)
[15:20:59] Called DecompressByteArray: compressed_data_size=24839963 data_size=30791309, decompressed_data_size=30791309 diff=0
[15:20:59] - Digital signature verified
[15:20:59]
[15:20:59] Project: 2684 (Run 2, Clone 13, Gen 5)
[15:20:59]
 
Interesting. Let us know how those 2684s go. ;)
 
Interesting. Let us know how those 2684s go. ;)

I'm actually more curious about how our fellow Gulftowners are going to do since they will now be able to throw a full 12 threads at it without that VM. Or how our new SR2 guys will do with their ability to throw a full 16 threads in Windows!

Part of me is scared because we know the 2684's wern't too stable.
 
What flags if any did you have?

We need info on how you grabbed this WU.... MEOW.
 
What flags if any did you have?

We need info on how you grabbed this WU.... MEOW.

This.

I put -bigadv in my SMP2 clients a few weeks ago...is that all you did?
(Mine are set to advmethods always in the config, and the flags I use are -bigadv -smp)

Sounds good. Both of my VM's will finish their 2684's around 4am CST tomorrow (Wed.), looks like I can switch 'em off permanently. :)

Fuck it, post EVERYTHING. WE WANT DEETS.
TPF, flags, credit, the whole 9. :D
Control+C it, see if it flushes. :D nah j/k
 
Control+C it, see if it flushes. :D nah j/k
I second this, seriously. Stop and restart the client to see if these WUs are more stable than what we've seen so far. You don't have anything to lose, the WU has just begun. That alone will determine it for me whether I run -bigadv at all. No sense processing -bigadv regardless the OS it's available in if there's no guarantee of completion. /2cents

Just an FYI for those that are running a VM and aren't getting any bigadv WU's...apparently they've been turned off.
After nearly 48 hrs, I figured as much. Well, I would imagine at some point they'll make a comeback to Linux for those who prefer running an alternative OS.
 
I'm actually more curious about how our fellow Gulftowners are going to do since they will now be able to throw a full 12 threads at it without that VM. Or how our new SR2 guys will do with their ability to throw a full 16 threads in Windows!

Part of me is scared because we know the 2684's wern't too stable.


I'll let you know how my Gulftown handles it once we finish the BOINC run this month.
 
As much as I wanna say that's awesome, AMDGAMER, I suppose now you're gonna get the mower out again :(...

Is it safe to say there is still no way around the 8 core requirement?

 
Just loaded up a 2684 on my i7 930 (slipped by home on the way home from a client) it's runnign good and here is the best part....

The system is only using 1.77GB of memory with Win 7 Ult 64bit. Now that is something. Now I can harvest the rest of the memory from the system and make Nexus 12GB trip channel :cool:

Nexus gets switched over when I get home (I have to do a HDD swap and it be nice for this WU to finnish
 
Ever since I abandoned the Linux client in the VM due to stability issues with the 2684 A3 bigadv, i've been running the Windows SMP2 client with advmethods turned on by default, and using these switches at the command line

- smp 8 -verbosity 9 -bigadv

This morning after it completed a regular A3 work unit, I noticed that my wireless router was showing activity for a very long time. It just got done uploading a 21mb finished work unit, so it couldn't still be uploading. I ran downstairs to the storage room where the folding rig is, and it sure as hell downloaded a 2684 A3 bigadv in Windows.

I'm am hoping for stability, although part of me is scared that the 2684 A3's may not be any more stable running natively in Windows. If this 2684 crashes and flushes on me, i'll just remove the -bigadv switch from the command line.
 
The system is only using 1.77GB of memory with Win 7 Ult 64bit. Now that is something. Now I can harvest the rest of the memory from the system and make Nexus 12GB trip channel :cool:
Now that's something I want to hear. I could eventually get rid of half the modules running in my systems, and cherry pick the best out of the lot for a bit higher OCing, or just build another budget system with the removed parts. Good stuff. :cool:
 
If this 2684 crashes and flushes on me, i'll just remove the -bigadv switch from the command line.

Seriously, control+c the window, then bring it back up again...that was a guaranteed crash and flush with 2684's in linuxrouter VM for me. You're early on in the WU, might as well test it before it does a ton of work.

.....and post your rig stats, TPF, PPD and credit! Inquiring minds wanna know! :D TIA
 
Confirmed! 'Bout frickin' time!

This is an i7 920 running at 203x19. TPF is about on par with Linux VM, but that's just the first frame, and it usually improves after frame 3 or 4.

Right now it's using approximately 1GB of RAM.

HFM.NET says about 19.5K PPD. Hopefully that will improve. Can't wait until my 4050 (203*20) i7 920 gets a hold of this one! :cool:

Here is my client.cfg (without Passkey of course!) :)

Code:
[settings]
username=10eeee
team=37412
passkey= BA A3 windows FTW
asknet=no
machineid=1
bigpackets=big
extra_parms=-bigadv -smp 8 -verbosity 9
local=245

[http]
active=no
host=localhost
port=8080
usereg=no

[core]
priority=96
cpuusage=100
checkpoint=9
nocpulock=1
addr=

[clienttype]
memory=4096
type=3

Code:
[15:45:33] 
[15:45:33] *------------------------------*
[15:45:33] Folding@Home Gromacs SMP Core
[15:45:33] Version 2.22 (Mar 12, 2010)
[15:45:33] 
[15:45:33] Preparing to commence simulation
[15:45:33] - Looking at optimizations...
[15:45:33] - Created dyn
[15:45:33] - Files status OK
[15:45:37] - Expanded 24831552 -> 30791309 (decompressed 124.0 percent)
[15:45:37] Called DecompressByteArray: compressed_data_size=24831552 data_size=30791309, decompressed_data_size=30791309 diff=0
[15:45:38] - Digital signature verified
[15:45:38] 
[15:45:38] Project: 2684 (Run 3, Clone 9, Gen 4)
[15:45:38] 
[15:45:38] Assembly optimizations on if available.
[15:45:38] Entering M.D.
[15:45:48] Completed 0 out of 250000 steps  (0%)
[16:31:39] Completed 2500 out of 250000 steps  (1%)
 
As much as I wanna say that's awesome, AMDGAMER, I suppose now you're gonna get the mower out again :(...

Is it safe to say there is still no way around the 8 core requirement?


I'm afraid there is no way around the 8 core requirement. The Nehalem based Core i7's more or less got lucky that they are able to run 8 simultaneous threads, and the folding client apparantly is fooled into thinking it is an 8 core processor when it is only a quad core. The problem with us Core i7 folks is that running the 2684's are almost a nailbiting experience as it runs incredibly close to the 4 day deadline. Bigadv from what i've been told, was never intended to be run on anything less than 8 cores which shows you how incredibly powerful the Nehalem architecture is.
 
Yes, so anything less than 3.4 Ghz is pretty much on the line depending on some other system variables.

I am noticing as someone else mentioned (Tigerbiten I think) that bus speed and memory speed can help in those situations.

The above config I have is 203x19, 1624mhz RAM speed at 8-8-8-24, 2T, so it's by no means a super box, but a tiny bit better than mainstream in terms of RAM speed.

I'm afraid there is no way around the 8 core requirement. The Nehalem based Core i7's more or less got lucky that they are able to run 8 simultaneous threads, and the folding client apparantly is fooled into thinking it is an 8 core processor when it is only a quad core. The problem with us Core i7 folks is that running the 2684's are almost a nailbiting experience as it runs incredibly close to the 4 day deadline. Bigadv from what i've been told, was never intended to be run on anything less than 8 cores which shows you how incredibly powerful the Nehalem architecture is.
 
Guess now's as good a time as any to start re-tweaking my oc's that were stable in everything (incl. 20 passes of LinX) except bigadv VM. :)
 
All you have to do is add -bigadv to your SMP2 client.

2684's are much much more stable in windows than on linux. In fact i got better TPFs in windows than on linux. When i get home ill post all my numbers.
 
I tried it on a Dell dual E5506 workstation and got 56:54 for the first frame. I'll let it run a few more frames to see if it will be able to complete on time when the machine is in use. I'm thinking no, since it's like 40 seconds away from the bonus cutoff. The machine has Windows XP, with only 4GB of RAM.
 
Soo... Just shut down, add big adv to my flags, restart? Ooooo soooo excited!! Headed off to do this to Skynet right now!
Posted via [H] Mobile Device
 
i7 930 @ 3.8 is doing 49:34 TPF or 17600ppd.

That is a few k more than what the average SMP client was puling.
 
Soo... Just shut down, add big adv to my flags, restart? Ooooo soooo excited!! Headed off to do this to Skynet right now!
Posted via [H] Mobile Device

Yep and make sure the work size is set for big...I made that mistake when I first tried it.
 
when I went thru setup the first time I put in big, nothing to change now? If thats the case I'll drop 7-64 on my i7 build tonight!
 
I'm excited to try this as well but my i7 920 is only running at 3.4 atm...whats my TPF going to be with that? Any idea?
 
when I went thru setup the first time I put in big, nothing to change now? If thats the case I'll drop 7-64 on my i7 build tonight!

Nope, just make sure advmethods is on by default in the config and make sure you add the -bigadv command line switch. Hurry up, I want to see numbers from you SR2 folks!
 
when I went thru setup the first time I put in big, nothing to change now? If thats the case I'll drop 7-64 on my i7 build tonight!

I think that's it. I just needed to add the bigadv flag and I was good to go.
 
Just grabbed one on skynet :)
Altho I will say this, is there utilization issues with this?? with A3 SMP's I had a steady power draw, maybe a 2 watt vary in each direction. Now, my draw is jumping around between 380 and 417 (417 where I was with A3's) but task manager shows me at constant 100% CPU use. ideas??
 
I just pulled down a -bigadv. I'll let you know in a bit how my Gulftown is doing.
 
I'm excited to try this as well but my i7 920 is only running at 3.4 atm...whats my TPF going to be with that? Any idea?

52-54:ish range. My work computer is at 3.3GHz, but I forgot the number.
 
Back
Top