I'm guessing Nelio's work got some nice new shiny servers for him to play with at Christmas!!! Nelio's EOC stats
Well done Nelio. That is a stellar performance you are putting in there. Why not tell the team what you're up to and what you are running to get those results. It would be nice to get a bit more team chit chat going.
Hey guys! Apologies for the delay in response. Just noticed the post now. Coincidentally the original post was on my birthday! Yeh my work got a batch of new kit in for testing HP RGS but we don't have the resource to start the project so we've had 13 new Dell Precision R7610 rack workstations firing away on all cyclinders, they've got dual E5-2687W V2's @ 3.4Ghz so 32 cores, they've got 128GB RAM and a couple SSDs. They have Quadro K4000's as well but I tend not to fire F@H up on both on these systems, they get hella hot! That along with my normal 500-600,000PPD has got me sitting at around 2.6 million PPD at the moment, I was hoping for 3 million but I had to sacrifice a few workstations. Users eh. I think i'll only have them for another 2-3 weeks at most so making the most of it while I have it all! Thanks for the kind words and I'm glad it's boosted the team a bit as well!
ha ha cheers guys. I couldn't let them sit doing nothing while we wait to get the project under way! They're some machines. I'm not sure if anyone is interested but it's interesting to see the difference between the V1 and V2 versions of that chip. The V2's vary quite a bit depending on work unit / project. They're running 7808's and 7809's at the moment, some doing 135,000ppd and some doing up to 146,000ppd, quite a spread for exactly the same chip. I'm wondering if cooling / position in rack is allowing some chips to turbo and others not or just variations in the silicon perhaps. The V1's do between 125,000ppd to 133,000ppd on the same work units so a decent improvement for the V2's. All in all I'm running 24 systems at the moment only three of which don't have some kind of dual Xeon set up! I think I would cry if I saw the electricity usage though!
Ahhh I did wonder what crazy hardware you had running for this sort of PPD, wish we had some new servers right now that needed a 'Burn-in' test All good for the science though mate even though you just overtaken me
hey guys, quickie... We got some new kit, (again) - anyone know how to fold across 48 cores? it seems to stop at 32. I've added a second client on FAHControl to run on the other 16 cores at the moment but would like to nail it out across all 48 in one go if possible. Cheers!
hey Doc! Yeh, got em at full. I did some digging on the f@h forums and another guy was trying the same thing and the threads there never gave a solution. I'm guessing the units/version of software is limited to 32 threads at the moment. I thought there were maybe some secret "extra client options" or "extra core options" under expert mode. I'll keep digging. Keen to see the real PPD of these chips (E5-2697 v2s), they're lower clocked however so think they may just level off with the E5-2687w v2's. p.s. I'm trying my damndest to catch you dude!!!
Strange one because the official blurb says there is no limit to the number of cores. However, looks like the smp client might well be limited to 32 threads, so you're probably better running two CPU slots, each with 24 cores. Did you have to hand the other machines over yet, or are you going to be able to 'double up' for a while with the new batch? PS Been seeing you coming for a long time(!), but there's now't I can do against that sort of output !!!
haha! fortunately I'm good with them for a bit yet so there's another two boxes with 48 cores added for a bit. Can't really say how long but we've had a HP Z820 workstation had to go to a user and two of the Dell T5500's I had both died, PSU in one and not really figured out the other ones problem but they're out of warranty now so down about 200,000ppd from those so was glad we got the two Dell M620 blades lying about for a short while. I'm hoping to get into first place before I lose them all but 3 months may be pushing it a bit! Wish I could get Linux and some bigadv's on them. Would be smashing some serious figures then!
Nelio, run Linux in a VM. VMware 3.0 will allow you to hack a config file so you can run with all cores. Could do it from a USB pen drive. I used to do it with my 2600k, made Linux think it had 12 cores so I could bigadv back in the day.
Good shout dude. Will see if I get a chance to give that a go! Does anyone have a rough guess at the PPD of a 48 core @2.8Ghz bidadv on linux?
Have fun. Remember it has to be version 3.0.0 as the later ones will not allow utilisation of all cores. (unless this has changed in the last year since I done it) As a side note, you have accidentally setup one client and put a space after your username so you have two users, once chucking out 130,000 ppd lol.
Argh no way! Need to check the out asap! Cheers for the heads up! Got three clients on Debian now, the stats they're estimating for PPD are ridiculous!
Got that rogue agent sorted out. Massive thanks! How the 'ell did you spot that one? I should be making over 3 million PPD after the three debian boxes complete a bigadv unit. no idea how long I'm going to enjoy this kit for but good for the science, the competition and of course the CPC/Bit-Tech team! Think some of the rack workstations will be getting Debian on my lunch breaks next week also