About a 2.8GHz i7 will make the deadline if folding exclusively, but no bonus. Everything faster than that is gravy.
Printable View
About a 2.8GHz i7 will make the deadline if folding exclusively, but no bonus. Everything faster than that is gravy.
if you have fast memory then yes. communication across 1 million atoms is intensive. i have been able to run bigadv with 4 threads but you still could get the WU's in time.Quote:
Originally Posted by DrJay
Average frame time is dipping to 26:29 now, but the rig is -very- hot.Quote:
Originally Posted by coo-coo-clocker
I had to top off the reservoir when I got back from 5-days away and for some reason, it was half empty this morning :shrug:
you might be able to get frametime lower with the new kernel.
http://en.fah-addict.net/news/news-0...ux-folding.php
I figured this out ... the reservoir shroud was so hot that I couldn't touch it for more than a few seconds.Quote:
Originally Posted by kevikev
I had a kink in my loop, once the kink was solved, temps are back to "normal".
Isn't ready for the Big WUs yet:Quote:
Originally Posted by Chumbucket843
Quote:
This kernel has some scaling problems on BigAdv WUs and more than 8 cores ... we'll try to build a new one using a different basis (2.6.31 for instance).
Man, I hate when that happens.Quote:
Originally Posted by kevikev
bigadv can run in windows now.:D
http://www.evga.com/forums/tm.asp?m=100994828
http://img245.imageshack.us/img245/5499/bigadvwin.png
Awesome news! Now all our fellow folders with Windows on their boxes and a fast enough i7 will be able to push team PPD.
In that thread they advice to use -smp 7, which really is an ideal solution. You can use WinAFC to move all GPUs to thread 8 and have them run at full production.... makes me almost want to go back to Vista and try it. Full GPU production + bigadv + the ability to play SC2 and D3 when it's out = unbeatable.
Wonder if I can try that on my X3440@4ghz :shrug:
X3440 can perform on par with a 920@4ghz on WCG.
Question I have though is; will linux still do better than windoze as it usually does? :shrug:
I do not know anything about linux
i am getting more ppd in windows actually. make sure you have enough ram. i am using 5.10GB our of 6GB. your x3440 will do fine just make sure you have HT on.
1st 8-thread i7 rig has come up. The save-electricity-by-switching-from-GX2s-to-i7 mission has commenced!
Had to bump up VCore to 1.500V (1.476V in BIOS) in order to get the 200 BCLK to work on the D0. Bumped RAM down to 6x ratio.
We'll see how frame times are before I do further tweaking.
Do please keep us updated. I'm seriously contemplating a 920 D0
1.5v:eek: you should be able to hit 4ghz on a D0 with 1.3v if not less.Quote:
Originally Posted by kevikev
I have a D0 and a W3520 4ghz@1.248v with no issues (knock on wood), this is on air (True/Megahalem)
Also I thought they said you can only run 7 cores, the 8th core creates errors :shrug:
I am thinking the same thing. I have a p55 rig I can pull from WCG to test on. I need to get my sheet figured out. I do not know if I should build another p55 rig when my mobo comes back from RAM, or just go with another cheap X58 rig.Quote:
Originally Posted by coo-coo-clocker
The only thing with my X3440 is that it seems like I have to run it @ 1.4v to get 4ghz stable, but I am an amateur when it comes to overclocking past the easy level. :shrug:
I already have a cheap P55 mobo thats getting replaced and a X3440 is approx $215 after cb. or get a UD3P x58 for approx $130 or so after rebate/cb and either find a used D0 for $175 or pay $215 for a new one from microcenter.:confused: :shrug:
Could you be more specific please? GHz, how many minutes for one frame on Linux and in VMWare? Which Kernel on Linux, which in VMWare? :confused:Quote:
Originally Posted by Chumbucket843
3.7GHz with 38 min per frame. on ubuntu 9.04 it was ~41 minutes and its only using 7 threads in windows. its linux 2.6 64 bit and vmware 3.0. everything else you need to know is in the guide i linked.Quote:
Originally Posted by p2501
Thanks man, I went back to Vista since my overclock got flaky and I need to reevaluate it. Did my own VM install of Ubuntu 9.10 Server and running advmethods now since I'm back at stock clocks for now. But wow, you really can run 8 cores now in VMWare, you just need to edit the .vmx file to set this since the VMWare setup lets you only choose up to 4. This didn't work in VMWare 2.x versions.Quote:
Originally Posted by Chumbucket843
WinAFC fun, first 7 threads SMP, last one two GPUs:
http://img200.imageshack.us/img200/1931/7threads.png
I thought Linux was doing better than windows :shrug:
I will be back soon for your guy's help. I need to get a cheap HDD to put Linux on, unless Win7 is doing just as well. I will still need your help. Either way I will need another HDD as I have plans for 1 more rig for the moment.
I honestly do not know what I am doing, especially with plans. It is very hard to keep up with F@H, GpuGrid, & WCG; especially when funds are limited and I have my wife always breathing down my neck to quit spending money on computers :rofl: :shrug:
Everyone folding bigadv and Linux SMP, if you're not already running it you might be interested in the langouste project. It's very useful if you have limited upload bandwidth like me (like 45kb max, on DSL), it's easy to setup and not noticeable that it's even there. On bigadv the client idles until it finished uploading the unit, which can take 50 minutes and more for me (standard a2 is not so bad but still takes a while).
What it does is just this:
- client finishes one unit, tries to upload it
- langouste is setup as proxy for the client, but it denies upload until the client gives up on it (takes maybe 5 minutes)
- client downloads new unit to work on, saves not uploaded unit for later
- client starts working again, while langouste forks a new client in the background which just uploads the finished unit and the exits
- client deletes said unit from queue, cycle done.
Time saved for me: at least 45 minutes.
Check it out. :up:
Got a weird?; If I have all my systems on win7 and want to try linux for the big wu's, would I be able to network the win7 machins with the linux machine?
Yes you can. You'll want to install a ubuntu app called SAMBA. It's available from the menu on the top of the screen at the left hand side. They've got a new name for the menu,
something like "get free software" or somesuch. This will set up file sharing with Windows machines.
Thanks,Quote:
Originally Posted by don_xvi
I already got some info from RR as well:up:
Now the question is to run linux/big wu's or win7/big wu's?
Also they are saying you run win7/big wu's plus GPU? Does this work on linux too?:shrug:
I want to try the big wu's and I also want to finally learn linux, but I do not want to compromise my production. Maybe I will try win7/vmware, plus gpu's; Then I can try to get another cheap system together for linux/big wu's :shrug:
I've been wondering about the 250W - 300W draw of the dedicated i7 folders ... thinking if my 85% efficient Antec Phantom 350W could power a barebones i7 rig.
Sure why not :shrug: I would imagine it would be fine.Quote:
Originally Posted by kevikev
On a side note I am building a new 920 system next week and think I might give this a shot.
Can anyone clarify if you can still run this with bonus points and still run a gpu on folding as well? Last I saw on the evga forum they said if you run on 7 threads you can use the last one for gpus.:shrug:
I ran bigadv -smp 8 on Fedora together with two GPU processes with no problem at all. On pure Linux you can just let the scheduler do it's work and throw these two different folding processes together. Just remember to set different nice values on startup (like nice -n 13 for bigadv and -n 10 for the GPUs). On Windows with VMWare it's better to sort processes, like -smp 7 on the first seven threads, leaving the last one for GPUs. Otherwise the VM will kill your GPU output.