No. The EL WU's require at minimum 8 threads.Quote:
Originally Posted by GTSRboy
Printable View
No. The EL WU's require at minimum 8 threads.Quote:
Originally Posted by GTSRboy
reading through some threads and they use points as an incentive to get people to use certain clients. that would lead me to believe bigadv is very important if it has bonus points.
Exactly, that is why I tried these and am still doing them, double or better PPD also...:up:Quote:
Originally Posted by Chumbucket843
Any word on a windows client for these? I'd love to try my i7 on here for a bit.
no. just linux. they will have smp2 coming out soon, maybe then. virual machines become very unstable when running this. at least virtualbox does.Quote:
Originally Posted by [XC] flat-four
Last time I checked they were doing good to get the A1 core working on windows, and A2 they were having no luck with which is needed... :shrug:
i posted this a week or so ago in this section. http://folding.typepad.com/news/2009...d-clients.html
the desmond core scales much better with more atoms in the sim so it should be faster than A2 anyways.
For windows they definitely need to do something to speed it up compared to linux, I could take the same rig with a quad and put linux on it and get 800 more ppd than windows. It says "The main goal here was to make it MUCH easier to use" hopefully there is more to it than that. There is also a heck of boost in ppd in linux from A1 to A2 core...Quote:
Originally Posted by Chumbucket843
When you're mixing in big WUs with your "regular" WUs, is there anywhere (read: stats site) to check how much your individual big WU was worth?
I finally got my dual Harpertown up under water so I've been stress testing it with Ubuntu + bigadv. Running ~32:35/frame @ stock before overclocking.
Not that I know of, I PMed Bruce at the folding forum and give him the info on the work unit and ask him if he would kindly tell me what points it got. The 920 @ 4.5Ghz got 62186.5 points took 48 hours to complete or a couple minutes shy of 48 hrs...Quote:
Originally Posted by kevikev
Yeah, I also noticed a big of time lag on the reporting, my big WU was finished @ 10:50AM but didn't get included in the intra-daily stats aggregation until 6PM.
Yes I noticed that also... :shrug:Quote:
Originally Posted by kevikev
31,000 PPD on a single cpu! :shocked:Quote:
Originally Posted by road-runner
RR, have you put one of your i7 rigs under a Kill-A-Watt to measure what the kW/H comes out to?
I'm seriously considering dumping my GX2 rigs for i7s if the kW/H is much, much more economical.
Figuring on about $800 in sunk cost per quad GX2 rig, each rig ends up costing $80/month in electricity, or an additional $320/month. That's also an additional $1k/quad-gx2-rig over the course of a year.
I saw a calculation of about 200W output for a dual QC E5335 IBM server producing 17k PPD. Extrapolating that for a 25k PPD, let's say the output is 300W (vs 900W for a quad GX2 rig) which works out to about $30/rig/month in electricity.
If bonus points continue to be awarded, then 4 25K PPD rigs costing $120/month in electricity would start to pay off (when considering the new i7 investment - electricity savings) in only 3-4 months.
Does my math work out?
Not to mention the cooling on an i7 is much, much quieter than 4 GX2 fans set to 100% ;)
I have.
At 3.8GHz w/ just mobo, 6GB RAM, one HDD, and a basic PCI video card I'm pulling about 250W from the wall. WAY more eco friendly than the dual GX2s were (600-650W with the i7).
my points still arent registering. this is getting annoying. i would post logs of the upload but im in windows right now. everything is entered correctly. i didnt get any credit for the WU.
Overclocking is scaling linearly in terms of time savings. 10% overclock from 3.16GHz to 3.47GHz is also producing 10% better frame times. 32:35 down to 29:41. Hopefully the last 10% bump to 9.5x400 will be stable, and I'll finish with 27:00 frame times.
Kev, you should buy a Kill-a-Watt or similar to see at what point the power use moves off the linear too
Yeah, I do have one ... it's around here ... somewhere ...Quote:
Originally Posted by OldChap
For some reason, the 2nd WU folding @ 3.8 is producing frame times of under 27:00, averaging about 26:55 the last 38 frames
Still on the same pace?
So 8 threads are required to even run bigadv? ....or just needed to get the bonus. If it is the latter, maybe a Core i5 750 @ 4+ Ghz would be fast enough for the bonus points?Quote:
Originally Posted by ColonelCain
Required (as in, not optional).
i5 has zero chance of completing them in time to get bonus (and probably would not even make the deadline).
Stanford was adamant that they did not want anything less than 8 threads attempting these (and they even said they frown on virtual cores but the i7 when OCed would be sufficient).
so you would have to OC in order to hit deadlines or get bonus?Quote:
Originally Posted by mstp2009
I don't think hitting the deadline would be all too difficult, but to get bonus you'll have to make it in 3 days I think. As a comparison, my i7@4.2GHz takes everywhere from 2 days 2hrs - 8 hrs depending on workunit and what I'm doing on that rig. I think you need at least 3.8GHz for a single i7, and you need it with HT. I assume memory speed also plays it's part (you have around 420MB memory usage per thread), but there's no solid numbers on that. :shrug:Quote:
Originally Posted by coo-coo-clocker
Scheduler performance is a very important topic on these units, your frame times with the same kernel can differ a good 3-5 minutes per frame depending on what scheduler you're using (all the more important if you're running GPU's at the same time).
Anyways, for everyone willing to attempt this, do not format the work partition as ext4, use ext3!