32 cores on Debian - Is this right?

[ekiro]

I have numerous servers sitting at the data center, powered on with lots of disk activity, but the CPUs are 99% idle. Worst thing is, these servers have 4x 8-core CPUs (32 cores total).

I installed the available debian package and set the load to 'full'. Here is what I see now. Can someone confirm I am pushing as much power out of this thing as I possibly can?

[Zagen30]

Whatever you linked to didn't show up, probably because there are restrictions on new users to combat spammers.

If possible, paste the contents of the log file in code tags, which you can find in the Full Editor. Make sure to include the System section at the top of the log.

[k1wi]

ekiro - personal servers or servers (work or otherwise) that you have access/control of? It's a common first question, as if they aren't personal servers then as per the EULA you need to get permission of the owners/business to run FAH on them (ideally in writing).

It would probably be useful to keep an eye on power usage, as depending on hosting (i.e. in-house or off-site), they an add up quite quickly in a datacentre environment...

[ekiro]

ekiro - personal servers or servers (work or otherwise) that you have access/control of? It's a common first question, as if they aren't personal servers then as per the EULA you need to get permission of the owners/business to run FAH on them (ideally in writing).

It would probably be useful to keep an eye on power usage, as depending on hosting (i.e. in-house or off-site), they an add up quite quickly in a datacentre environment...

It's OK. This is all my equipment. My company. My datacenter. And power isn't an issue. There is plenty of that too. Any recommendations to fine tune my system for max output so I can apply it to the rest of the boxes?

Whatever you linked to didn't show up, probably because there are restrictions on new users to combat spammers.

If possible, paste the contents of the log file in code tags, which you can find in the Full Editor. Make sure to include the System section at the top of the log.

Also, you can see the attached image by visiting: http://i.imgur.com/nyRXej7.png

AND here is the entire log so far:

Code: Select all

*********************** Log Started 2013-10-28T21:38:45Z ***********************
21:38:45:************************* Folding@home Client *************************
21:38:45:    Website: http://folding.stanford.edu/
21:38:45:  Copyright: (c) 2009-2013 Stanford University
21:38:45:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
21:38:45:       Args: --child --lifeline 17868 /etc/fahclient/config.xml --run-as
21:38:45:             fahclient --pid-file=/var/run/fahclient.pid --daemon
21:38:45:     Config: /etc/fahclient/config.xml
21:38:45:******************************** Build ********************************
21:38:45:    Version: 7.3.6
21:38:45:       Date: Feb 18 2013
21:38:45:       Time: 07:24:08
21:38:45:    SVN Rev: 3923
21:38:45:     Branch: fah/trunk/client
21:38:45:   Compiler: GNU 4.4.7
21:38:45:    Options: -std=gnu++98 -O3 -funroll-loops -mfpmath=sse -ffast-math
21:38:45:             -fno-unsafe-math-optimizations -msse2
21:38:45:   Platform: linux2 3.2.0-1-amd64
21:38:45:       Bits: 64
21:38:45:       Mode: Release
21:38:45:******************************* System ********************************
21:38:45:        CPU: Intel(R) Xeon(R) CPU E5-2650 0 @ 2.00GHz
21:38:45:     CPU ID: GenuineIntel Family 6 Model 45 Stepping 7
21:38:45:       CPUs: 32
21:38:45:     Memory: 63.02GiB
21:38:45:Free Memory: 250.01MiB
21:38:45:    Threads: POSIX_THREADS
21:38:45:Has Battery: false
21:38:45: On Battery: false
21:38:45: UTC offset: -7
21:38:45:        PID: 17874
21:38:45:        CWD: /var/lib/fahclient
21:38:45:         OS: Linux 3.2.0-4-amd64 x86_64
21:38:45:    OS Arch: AMD64
21:38:45:       GPUs: 0
21:38:45:       CUDA: Not detected
21:38:45:***********************************************************************
21:38:45:<config>
21:38:45:  <!-- Client Control -->
21:38:45:  <fold-anon v='true'/>
21:38:45:
21:38:45:  <!-- Folding Slot Configuration -->
21:38:45:  <power v='full'/>
21:38:45:
21:38:45:  <!-- User Information -->
21:38:45:  <passkey v='********************************'/>
21:38:45:  <user v='Ekiro_Starkana'/>
21:38:45:
21:38:45:  <!-- Folding Slots -->
21:38:45:</config>
21:38:45:Switching to user fahclient
21:38:45:Trying to access database...
21:38:45:Successfully acquired database lock
21:38:45:FS00:Set client configured
21:38:45:Enabled folding slot 00: READY cpu:32
21:38:45:WU00:FS00:Connecting to assign3.stanford.edu:8080
21:38:46:WU00:FS00:Connecting to assign3.stanford.edu:8080
21:38:46:WU00:FS00:News: Welcome to Folding@Home
21:38:46:WU00:FS00:Assigned to work server 128.143.231.202
21:38:46:WU00:FS00:Requesting new work unit for slot 00: READY cpu:32 from 128.143.231.202
21:38:46:WU00:FS00:Connecting to 128.143.231.202:8080
21:38:48:WU00:FS00:Downloading 3.67MiB
21:38:49:WU00:FS00:Download complete
21:38:49:WU00:FS00:Received Unit: id:00 state:DOWNLOAD error:NO_ERROR project:8585 run:1 clone:5 gen:71 core:0xa3 unit:0x0000025b0a3b1e5952288a23cefb225d
21:38:49:WU00:FS00:Downloading core from http://www.stanford.edu/~pande/Linux/AMD64/Core_a3.fah
21:38:49:WU00:FS00:Connecting to www.stanford.edu:80
21:38:49:WU00:FS00:FahCore a3: Downloading 2.56MiB
21:38:54:WU00:FS00:FahCore a3: Download complete
21:38:54:WU00:FS00:Valid core signature
21:38:54:WU00:FS00:Unpacked 5.98MiB to cores/www.stanford.edu/~pande/Linux/AMD64/Core_a3.fah/FahCore_a3
21:38:54:WU00:FS00:Starting
21:38:54:WU00:FS00:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/www.stanford.edu/~pande/Linux/AMD64/Core_a3.fah/FahCore_a3 -dir 00 -suffi$
21:38:54:WU00:FS00:Started FahCore on PID 17891
21:38:54:WU00:FS00:Core PID:17895
21:38:54:WU00:FS00:FahCore 0xa3 started
21:38:54:WU00:FS00:0xa3:
21:38:54:WU00:FS00:0xa3:*------------------------------*
21:38:54:WU00:FS00:0xa3:Folding@Home Gromacs SMP Core
21:38:54:WU00:FS00:0xa3:Version 2.27 (Dec. 15, 2010)
21:38:54:WU00:FS00:0xa3:
21:38:54:WU00:FS00:0xa3:Preparing to commence simulation
21:38:54:WU00:FS00:0xa3:- Looking at optimizations...
21:38:54:WU00:FS00:0xa3:- Created dyn
21:38:54:WU00:FS00:0xa3:- Files status OK
21:38:54:WU00:FS00:0xa3:- Expanded 3851666 -> 4394668 (decompressed 114.0 percent)
21:38:54:WU00:FS00:0xa3:Called DecompressByteArray: compressed_data_size=3851666 data_size=4394668, decompressed_data_size=4394668 diff=0
21:38:54:WU00:FS00:0xa3:- Digital signature verified
21:38:54:WU00:FS00:0xa3:
21:38:54:WU00:FS00:0xa3:Project: 8585 (Run 1, Clone 5, Gen 71)
21:38:54:WU00:FS00:0xa3:
21:38:54:WU00:FS00:0xa3:Assembly optimizations on if available.
21:38:54:WU00:FS00:0xa3:Entering M.D.
21:39:00:WU00:FS00:0xa3:Mapping NT from 32 to 32
21:39:01:WU00:FS00:0xa3:Completed 0 out of 500000 steps  (0%)
21:43:19:WU00:FS00:0xa3:Completed 5000 out of 500000 steps  (1%)
21:47:44:WU00:FS00:0xa3:Completed 10000 out of 500000 steps  (2%)
21:51:51:WU00:FS00:0xa3:Completed 15000 out of 500000 steps  (3%)
21:56:02:WU00:FS00:0xa3:Completed 20000 out of 500000 steps  (4%)
22:00:00:WU00:FS00:0xa3:Completed 25000 out of 500000 steps  (5%)
22:04:06:WU00:FS00:0xa3:Completed 30000 out of 500000 steps  (6%)
22:08:04:WU00:FS00:0xa3:Completed 35000 out of 500000 steps  (7%)
22:12:00:WU00:FS00:0xa3:Completed 40000 out of 500000 steps  (8%)
22:15:56:WU00:FS00:0xa3:Completed 45000 out of 500000 steps  (9%)
22:19:50:WU00:FS00:0xa3:Completed 50000 out of 500000 steps  (10%)
22:23:47:WU00:FS00:0xa3:Completed 55000 out of 500000 steps  (11%)
22:27:41:WU00:FS00:0xa3:Completed 60000 out of 500000 steps  (12%)
22:31:41:WU00:FS00:0xa3:Completed 65000 out of 500000 steps  (13%)
22:35:37:WU00:FS00:0xa3:Completed 70000 out of 500000 steps  (14%)
22:39:35:WU00:FS00:0xa3:Completed 75000 out of 500000 steps  (15%)
22:43:35:WU00:FS00:0xa3:Completed 80000 out of 500000 steps  (16%)

Mod Edit: Fixed URL Link - PantherX

[7im]

Passkey? Check.
Linux? Check.
client-type = bigadv?
Edit CPU slot to specify 64 cores? (Assuming HT is enabled...)

[ekiro]

How?

[7im]

FAH Configuration Guide: Customize One Slot

client-type = advanced is the example. replace advanced with bigadv

You can edit the CPU core count from -1 to 32 or 64 on that same tab.

Then wait for next WU. Check your PPD after 4-5 frames are completed.

There's more info about Big work units on that page.

[k1wi]

I'm pretty sure that the computer cited is dual socket - the 2650 is dual socket max, for a total max of 16 physical cores and 32 threads...

[Joe_H]

First, did the folding client detect the processors correctly as Xeon E5-2650 processors? Those are rated for dual processor installations, not quad processor. Two of those in a dual processor server with HT active would show as 32 cores.

If this is quad processor setup, not sure what issues you would run into if you have dual rated processors installed. In any case, if HT is disabled, that gets enabled in the BIOS.

Finally, since you just started your passkey would not be qualified yet for the Quick Return Bonus. Your username and passkey need to complete 10 WU's first, after that if you maintain a better than 80% successful return rate you will get the QRB. So you may want to stay folding non-bigadv WU's until you finish and turn in 10 before switching to bigadv.

[ekiro]

What is a WU?

[Joe_H]

WU is the abbreviation for Work Unit. Each one gets downloaded, processed and then returned for points credited to your username. The V7 Introduction includes descriptions of some of the common terms used.

[k1wi]

As Joe_H mentioned, get the first 10 work units out of the way and then change the client type to bigadv. That should boost your ppd once it kicks in.

Better yet, get all your servers up and running, using the same username & passkey combination and then switch them over to bigadv once they pass 10 work units combined as they'll all contribute to your account.

Your username won't show up until you've completed your first work unit (and the stats files are updated every hour), but once it has you can keep track of all the WUs you have completed using the F@H stats links at the top of the forum pages. It must be pretty close to completing though?

p.s. I love seeing that many cpus in top

[ekiro]

My apologize. I didn't exactly know what I have. Looks like they are all 2x E5-2950s. 8-cores each. 16 physical cores. 32 /w HT.

Anyway, is there anything else I can do with the config to maximize computational output?

ALSO, what GPUs are best for this? In the case of nVIDIA, GeForce or Quadro?

[k1wi]

In terms of value for money, you're going to get better bang for your buck going with 'regular' GeForce GPUs - the benefits associated with Quadros (i.e. DP compute) don't really apply for F@H at the present time.

The quick return bonus (QRB) rewards faster GPUs in an exponential manner, but this is often offset by the increasing cost of GPUs. I'm not up to date with the 700 series, so someone else might be able to let you know where the sweet spot currently is.

[ekiro]

In terms of value for money, you're going to get better bang for your buck going with 'regular' GeForce GPUs - the benefits associated with Quadros (i.e. DP compute) don't really apply for F@H at the present time.

The quick return bonus (QRB) rewards faster GPUs in an exponential manner, but this is often offset by the increasing cost of GPUs. I'm not up to date with the 700 series, so someone else might be able to let you know where the sweet spot currently is.

What you think of NVIDIA's Tesla line (K20?). I see we have 4 of those in storage.