-bigadv, Set yourself up BIG, a users guide

With a microcenter run, you should be able to build a i7 920 system for about 600.
 
mrmylanman said:
Thanks for the heads up Apollo. I priced out a Phenom II system using this processor

http://www.newegg.com/Product/Product.aspx?Item=N82E16819103727

And it cost me about $600. Which is a little more than what it would cost for one of those Sun servers with dual core processors put in (maybe less, who knows?). Plus being brand new components, I'm sort of feeling that route a little better, only problem is that wouldn't really be up to -bigadv folding.
Click to expand...
The Phenom won't be able to run -bigadv even if you can magically OC it well over 4GHz. It only has four cores and unless things have changed recently, the IPC of a Phenom isn't as high as an equivalent Intel CPU. At any rate, no quad core can manage it without an unrealistic OC, and very high frequencies (even with i7) is not something I would aim for with -bigadv because of the nature of the work, which requires a stable system that can turn in WUs quickly and more importantly, reliably. The one system running these WUs in my farm has a moderate OC, and I still get the occasional lockup with a native Linux install. It's a server setup without enthusiast board OC features, so no surprise there. I'm lucky I can OC at all and it's convoluted. This is something I'm trying to overcome but it's extremely difficult since these WUs are highly sensitive as they are. If anything, I recommend people to back off the OCs a little with -bigadv just to improve the likelihood of completion within an appreciable time frame.

Having said the above, if it's -bigadv you want to run and you have a limited budget, I still think the best setup would probably be an i7 processor with a moderate OC to frequencies which can accomplish the bonus deadline and still be very stable. I have no idea what the balance of performance and price is with S-1366 to give safe recommendations at this juncture since I haven't looked into the new architecture for prospective purchase scenarios yet.

Alternatively, in the second-hand/EOL market your best bet by far IMO is a dual S-771 bare bones system or S-771 parts purchased separately, not a quad-socket server and there are various reasons why I say this which I won't get into here. Used S-771 boards can be found for well under $200, but then you'll need the processors and FB-DIMM memory in sufficient quantity to run -bigadv (4GB+). It all depends on your budget and what other things you wish to do with the system besides folding.
 
APOLLO said:
Having said the above, if it's -bigadv you want to run and you have a limited budget, I still think the best setup would probably be an i7 processor with a moderate OC to frequencies which can accomplish the bonus deadline and still be very stable.
Click to expand...

ASRock X58 Extreme - $169.99 + Free Shipping
Core i7 920 - $199.99 - Microcenter
OCZ Gold 6GB Kit - $149.99 + Free Shipping ($120 after MIR)
Corsair 750TX PSU - $109.99 + Free Shipping ($89.99 after MIR)

Total: $630 - just add your favorite case, an HD, and an optical. OC her to 3.6 GHz and sit back and watch it crunch. :cool:

ASRock has come a long way from their shaky past to produce some really fine X58 platform boards. I haven't used one myself but they have been getting very nice reviews and there are several big time folders running them with 4xGTX-295's on the 4-slot PCI-E version of the board (ASRock X58 SuperComputer). I believe the stability is there with both models. If one is on a real tight budget, they are definitely worth considering.
 
I have a spare i7 920. If I wanted to get another one and put in a Dual 1366 MoBo, which board is recommended? How much memory is suggested for this setup? How much PSU do I need to keep it happy? If I went Corsair H50 (dual), is this a good way to go?

This would serve as a workstation for work and also a Big WU SMP client.
 
Well right now I have an i7 920 and it's at stock clock speed (heat is an issue... I need to mod my PC-A05). I'm still getting it within the timeframe to get the bonus points. I'm getting ~48K a work unit, and it's taking 3 days and change to get the work units done.
 
mrmylanman said:
Well right now I have an i7 920 and it's at stock clock speed (heat is an issue... I need to mod my PC-A05). I'm still getting it within the timeframe to get the bonus points. I'm getting ~48K a work unit, and it's taking 3 days and change to get the work units done.
Click to expand...

Do you not have a top fan? That's pretty much a must have for the A05. I installed a 2x120 exhaust fan and it improved my temps dramatically, and it looks SICK.
 
Thanks all for the help. Since I am working from home (yes, it is nice not having to get on an airplane on a Monday morning), I've had some time to review the OC on my sig rig.

I'v got it 48 hour stable @ 4.00. I can probably squeeze 4.2 out of it, but will do so when I have a little more time to figure out what vcore will make it happy. I found another much more simple guide on the eVGA forums that made OC'ing the i7 much simpler and frankly temps are great (68F @ 4.0 @ full load).

I am running 3 x GTX285's (GPU2) and I'm getting nearly 50K (18K of out VM and 32K out of GPU2) total out of it. I was running all of the GPU2 clients while I was stress testing to ensure that it could handle both CPU load and GPU load.

My frame times seem high given the OC, but it could be the VM I am using. I repurposed the VM that was running on this box. OS is Ubuntu 8.04. In checking other i7 920's @ 4.0, they are getting 25K PpD. I assume the OS of choice will make some difference. This rig is NOT a dedicated folder so I am hesitant to use the eVGA VM as I have read about lost work as there isn't a backup. My VM even if shut down ungracefully, still bounces back without losing work.

I also understand that GPU2 will also steal some cycles, but did not believe it would make a significant difference (per previous posts regarding BigAdv).

I am using WinAFC (yes, I am a believer and thus the reason it was included in the VM guide I wrote earlier in the year). GPU2 set to high and VM set to normal.

Any thoughts/suggestions on what to tweak?
 
capreppy said:
My frame times seem high given the OC, but it could be the VM I am using.
Click to expand...
Most likely it's due to the three GPU clients you're running. A single client is just about negligible, but three clients certainly are not as I can tell you from personal experience.
 
I've bumped up the VM via WinAFC to Normal and it is up to 22K PpD. I can live with that as the GPUs are at 30 to 32K PpD.
 
Woot Woot!!!

Thanks all for the help. Just finished my first BIG WU
 
Just a quick heads up ..........
Mine tend to take over 4 hour from upload till when they show up in my stats.
So don't panic if they don't show up in the first stats update from when you uploaded them.
Mine tend to show up on the second stats update from upload.

Luck ............ :D
 
Hey guys linuxrouter has posted a new version of his VM on the EVGA page linked in the OP.

The 0.6 image has the Langouste Decoupler installed so you can upload and download at the same time.
This should really help those that have slow upload speeds

Hope this helps :)
 
nitrobass24 said:
Hey guys linuxrouter has posted a new version of his VM on the EVGA page linked in the OP.

The 0.6 image has the Langouste Decoupler installed so you can upload and download at the same time.
This should really help those that have slow upload speeds

Hope this helps :)
Click to expand...

I saw that the other day, this VM just keeps getting better and better. The setup web page has improved quite a bit too.
 
tjmagneto said:
I saw that the other day, this VM just keeps getting better and better. The setup web page has improved quite a bit too.
Click to expand...

I'll grab it.... my current WU should finish at 6AM or so.
 
linuxfah has just now made some more changes as well, I will quote him below.

I made a few more changes to the image and posted an updated image in the guide. The core Linux system is still the same. The changes are mainly related to the web interface and init scripts.

MtM suggested that I add NTP support in the event there is still some clock drift. I added a checkbox to enable that. This starts up ntpd on boot. I used the following NTP servers in /etc/ntp.conf:

0.pool.ntp.org , 1.pool.ntp.org , 2.pool.ntp.org , 3.pool.ntp.org

Feel free to change them to something else if you want.

I also added an option to specify Ramdisk size and to use tmpfs in instead of a block device in RAM. The last option I added was for enabling -oneunit with the fah6 client.

I setup the shutdown script (rc.local_shutdown) to backup the data before shutting down the fah6 client. Then on boot, the init script does a restore from that backup. This only happens when either auto-startup or Ramdisk is enabled. Hopefully this will help with failed work units due to shutting down the client before completion. The backup script keeps a maximum of three backups in directories backup, backup.1, and backup.2 where backup.2 is the oldest. Also, before doing the restore, the restore script copies the existing files (queue.dat, unitinfo.txt, work) to a directory called old in case something is wrong with the backup of the files being restored.
Click to expand...
 
FrEaKy: Rather than quote the linuxfah post from the EVGA forums, can you please just link it instead. It is developing rapidly and you are now at least two versions behind.
 
Sorry about that. I noticed that too.

*edit* Information now updated on primary thread to show new information on eVGA page. This guy updates alot quicker than I thought (couple hours ago since last update)
 
Last edited:
FrEaKy said:
Sorry about that. I noticed that too.

*edit* Information now updated on primary thread to show new information on eVGA page. This guy updates alot quicker than I thought (couple hours ago since last update)
Click to expand...

He does some real strong work. I wish I had the time to do stuff like that. Damn job gets in the way. (does pay for the toys though)
 
Yeah thats my biggest problem. I am at work 90% of the time, so I do what I can with the time I have, he has alot more time and is capable of getting that information updated alot. I will try to update it at least once a week to catch up to him.
 
Finally got around to settling on an i7 oc last night (3.67 @ stock volts), and got the VM going (using the eVGA build). I tried getting one up using ubuntu x64 and felt pretty lost most of the way and then couldn't get the client to run anyway. :eek:

Question about the -bigadv...noticed my cpu use hovers around the midpoint (50%) each core; is this right? I'm used to the VM of notfred or the Win SMP clients putting the cpu use at 100%... just wanted to check.

TIA :)
 
zero2dash said:
Question about the -bigadv...noticed my cpu use hovers around the midpoint (50%) each core; is this right? I'm used to the VM of notfred or the Win SMP clients putting the cpu use at 100%... just wanted to check.
Click to expand...
It should be utilizing 100% or very close to full. How much memory do you have the VM using? I tried a -bigadv client with less than the 3600MB pre-configured in the EVGA VM, and it behaved in a similar fashion to what you described. It would process very erratically and sometimes just halt.
 
zero2dash said:
Finally got around to settling on an i7 oc last night (3.67 @ stock volts), and got the VM going (using the eVGA build). I tried getting one up using ubuntu x64 and felt pretty lost most of the way and then couldn't get the client to run anyway. :eek:

Question about the -bigadv...noticed my cpu use hovers around the midpoint (50%) each core; is this right? I'm used to the VM of notfred or the Win SMP clients putting the cpu use at 100%... just wanted to check.

TIA :)
Click to expand...

make sure in the .VMX file that CPUs is set to 8
And make sure you start the client with -smp 8
 
Well the VM settings had it on 8 cores but checking the VM config through the IP/web interface, that was (for some reason) set to "cores: 4" so I changed that to 8 and I'm rebooting it now (thank god for logmein).
[crosses fingers]
I did have it set to take 4600mb for ram so that shouldn't be the issue there; I'm sure it's this core thing in the VM config. I have it set to autostart bigadv

[update]
Yep it was the config file.
8 cores @ 100% now, checked the log and it does say
Arguments: -bigadv -smp 8
Click to expand...
so I'm good to go now. :)
I'll post some hardware pr0n pics later :D

As always - thanks for the help fellas ;)
Can't believe I missed that last night...guess I was too excited to finally get my stuff running and get the bigadv going that I didn't even notice. :eek:
 
Last edited:
Does this mean you'll finally show up on my threat list?? :D

Seriously, great to see the configuration all ironed out.
 
I'll pop up eventually, I need to get my quad back together (basically migrating the board to the living room/e8400's current case)...that'll add another couple thousand PPD as well.

I'm *hoping* once all's said and done (and I get everything moved around and running again) - I should get similar PPD to you so I'm sure we'll be clotheslining (old school wrasslin' style) everyone ahead of us. :D
i7 920 @ 3.67, 2x 260s, q9450 @ 3.6 (probably use the evga client with it since it's more updated now than notfred), 8800gt and (eventually) e8400 will go in my work computer and the e2200 will be thrown in the closet or sold. :)
 
Since just I finished and uploaded another bigadv this morning I decided to download and fire up the 0.7 version of the EVGA VM.

Here's a screenshot of the latest web configuration page. oops it went bye-bye


I enabled the following stuff: FAH auto startup, enable bigadv, use 8 cores, and enable Langouste Decoupler. I left everything else unchecked for now.

What wierded me out at first was that with the autostart I didn't see any progress messages in the VM window. HFM.NET was displaying the log so I felt happier. I went to work right after the WU downloaded and started so I don't have much to report until later.
 
Last edited:
Everything looks fine so far, now I just need to see how long it takes from finishing one unit to starting the next. I'll have that answer late tomorrow.
 
Note: Do NOT play MW2 when you have a Big WU about to drop!!!

I was playing MW2 (was supposed to be working, but needed a break). A big WU completed and started uploading in the middle of a team death match. I was hosting :D

Needless to say, EVERYONE started experiencing really bad lag because all of my upstream bandwidth immediately went to crap because my VM was sending the data (pretty damn close to 100MB).

Sorry to the other 11 people that were playing with me. I have learned my lesson and will keep tabs on my Big WU uploads so I don't host and then screw up everyone's game :D :D
 
One advantage of bigadv over standard I've found is that you only upload once every three days and not twice a day.
Yes it is a 100Mb upload but the standard is around 50Mb.
So I end up uploading less with bigadv vs standard.

Luck .............. :D
 
The -bidadv WU seem to be very delicate. On a machine I consider stable, be running for 16 days without a hiccup, crashed today. I fire it back up, tells me basically corrupt WU
you are boned, get a new work unit. UGHHH, does qfix works with these WUs?, or any other advice on recovering a borked WU.....:confused:
 
You must log in or register to reply here.
Back
Top