Posts by Emanuel

21) Message boards : Number crunching : Boinc Manager Regression? (Message 39338) Posted 1 May 2010 by Emanuel Post: Although I don't know his particular card, judging by the name it's a 7-series card (duh I know, just being careful), which preceded the unified shaders required for all GPGPU calculations. I think he knows this though, and meant that since he won't be affected anyway, he's sticking with 6.10.43 for the moment.
22) Message boards : Number crunching : v 0.26 and 0.03: Less than 20 Cr./hour on Nehalem! (Message 39318) Posted 30 Apr 2010 by Emanuel Post: Those use a different background model for the Milkyway, but not the new search method that was mentioned. The differences are important, of course, but they don't require a full rewrite.
23) Message boards : Number crunching : v 0.26 and 0.03: Less than 20 Cr./hour on Nehalem! (Message 39303) Posted 30 Apr 2010 by Emanuel Post: It's also been mentioned that they're working on a new type of search that might not be as GPU-friendly (though that remains to be seen). With GPUs working so well here at the moment it's probably best to focus your CPU on other projects until they get that search underway. I realize that's not technically what the topic is about, but ...
24) Message boards : Number crunching : Is there an easy way to get the memclock below 490 ? (Message 39139) Posted 25 Apr 2010 by Emanuel Post: ATI Tray Tools is third-party software, its not included in the drivers; it's very well known though. Try Googling it.
25) Message boards : News : testing new application (milkyway3) (Message 38960) Posted 21 Apr 2010 by Emanuel Post: If it can't do double precision math you can't use it. Sorry, but we need more precision than single precision :( The Fermi cards can all do double precision, even if Nvidia -have- artificially capped them. The problem, as I recall, is that the application checks for CUDA compute capability 1.3, and the Fermi cards have compute capability 2.0. So it should test for >= 1.3 instead.
26) Message boards : Number crunching : GPU Requirements [OLD] (Message 38944) Posted 21 Apr 2010 by Emanuel Post: Actually for Mw you can downclock. Oc doesn't speed up the wus. You can downclock the memory, anyway. I do believe MW benefits from OCing the core and shading units.
27) Message boards : Number crunching : Still getting 10% invalid results (Message 38824) Posted 18 Apr 2010 by Emanuel Post: Maybe the authors of Furmark should talk to Milkyway@Home to see how they're stressing the cards this thoroughly XD
28) Message boards : Number crunching : server initiated delete of outdated stars* files (Message 38807) Posted 17 Apr 2010 by Emanuel Post: Einstein@Home has this sort of system in place; might be a good idea to ask them how they manage it so you don't have to figure it out on your own.
29) Message boards : Number crunching : Waiting for validation... (Message 38682) Posted 13 Apr 2010 by Emanuel Post: They should validate (or not) eventually. If you've ever crunched for Einstein@Home, you should be familiar with pending credit. They validate every WU, and since their deadlines are pretty long you can get unlucky and end up with WUs still pending credit from a month ago. It's just the way their system works, and no one complains about it over there; the situation should always be better at Milkyway@Home because the deadlines are much shorter, and they won't validate more WUs than they need to be safe (of course things are still in flux at the moment, but they should settle in the coming weeks).
30) Message boards : Number crunching : Marked as Invalid? (Part 2) (Message 38534) Posted 10 Apr 2010 by Emanuel Post: Huh, I thought they were supposed to show the reported fitness in the stderr.txt log ... but I don't see it in there. Otherwise you would be able to compare the fitness you got to the fitness the others got. The only way the situation -should- occur is if their fitnesses were very close together, and yours was somehow off. That would still be strange, but would be correct as far as the validator is concerned.
31) Message boards : Number crunching : Marked as Invalid? (Part 2) (Message 38486) Posted 9 Apr 2010 by Emanuel Post: Speaking of Detaching/Reattaching, I think a project Reset should do just as well if you're using BOINC client version 6.10.45 due to the following bug fix: David 4 Apr 2010 - client: clean out project dir on reset. fixes #978 Could be wrong though.
32) Message boards : Number crunching : IMPORTANT! Nvidia's 400 series crippled by Nvidia (Message 38459) Posted 9 Apr 2010 by Emanuel Post: For as long as there has been Quadro cards, they have been using firmware to only enable a number of OpenGL hardware acceleration paths on Quadro cards used with Quadro specific drivers, despite Quadro products being based off identical silicon to their GeForce counterparts. Then let's hope they did the same thing here so we can crossflash our cards to the Fermi equivalents.
33) Message boards : Number crunching : IMPORTANT! Nvidia's 400 series crippled by Nvidia (Message 38441) Posted 9 Apr 2010 by Emanuel Post: As I posted on Einstein@Home: "I'm very disappointed, and really torn about getting one of these cards now. Even if I admit to myself I won't be getting it just for its science computing capabilities, this artificial limiting leaves a very bad taste in my mouth. There's simply no way any consumer is going to buy a Tesla card, so I really think they're shooting themselves in the foot with this. Tesla should simply imply direct, lifetime support and maybe have all the shader clusters enabled. They may think this is a good business decision, but I've seen a little of just how much money some of the more competitive crunchers have - I think doing this is just going to lose them a percentage of their sales without actually helping out Tesla in the slightest." If this is something that was done at the hardware level there's not much we can do about it; if it's purely software, maybe Nvidia will change their minds and give us the full potential of these cards, or maybe someone will figure out how to hack their drivers to do the same thing. Only time will tell I suppose - but let's hope they at least get some seriously bad PR for this. If they're able to keep this mostly under wraps, they have nothing to worry about.
34) Message boards : Number crunching : Odd Validation Result (Message 38374) Posted 8 Apr 2010 by Emanuel Post: Can you really call that a problem with the validator? It seems to be doing its job correctly - it's the older applications that mess things up. Pushing out the new, not backwards compatible applications should fix this, but that doesn't mean the validator is broken. Unless I'm misunderstanding you.
35) Message boards : News : stock ATI 58x0 apps updated (Message 38311) Posted 7 Apr 2010 by Emanuel Post: I would think that the implementation of the new model takes precedence. If all applications run well with the new code (was just released as a beta version, it is not final yet), one can think about getting the results of CPU/ATI/CUDA versions even closer together. What I've done is basically only a different method for the final summation of the integration steps (using two double values instead of a single one to hold the values, limits the precision loss for this step and greatly reduces the effect of the different summation order of the values between CPUs and GPUs) and to use a simple trick in the likelihood/fitness calculation which buys you one or two digits of precision there. As long as the individual mathematical operations used are precise to all but the very last bit (which is true for all GPUs), minor differences in the very last bit doesn't change the final likelihood result, as the limit for the precision is the summation of the individual values from all the integration points. Sounds like it should be relatively simple to implement when the time comes. Thanks for the explanation :)
36) Message boards : News : stock ATI 58x0 apps updated (Message 38289) Posted 7 Apr 2010 by Emanuel Post: [double post, please delete]
37) Message boards : News : stock ATI 58x0 apps updated (Message 38288) Posted 7 Apr 2010 by Emanuel Post: Now that the ATI applications have been updated, I'm curious - Cluster Physik mentioned that with a few changes, the CUDA applications could give exactly the same results as the ATI and CPU applications. They are currently close (well within tolerances) but not the same. Do you intend to update the CUDA applications based on his advice, or does the current work updating them with the new science and server communication code prevent this from being a priority?
38) Message boards : Number crunching : Marked as Invalid? (Part 2) (Message 38218) Posted 6 Apr 2010 by Emanuel Post: It's just that, well, this is a forum. Forums tend to have turnaround times of -at least- 6 hours, and that's if you're lucky. There really isn't any point in posting again and again because you're frustrated - it only serves to piss off the people who might actually be able to answer you. At the moment, I think problems are to be expected. Work on the validator is ongoing, the AMD/ATI CAL applications need to be updated to get accurate results from the HD5800 series (which yes, that means that all results they've returned up until now have been inaccurate and invalid to some degree), -and- work on new science models is being finalized. The first is mostly up to Travis, the second is being worked on by Cluster Physik and will have to be checked by Travis, and the third is being worked on by the other project scientists and will also have to be passed on to Travis. I wouldn't be surprised if the guy is currently a bit too overworked to communicate much, but he has been giving some status updates in the News section of the forum; I hope you've been keeping up with the threads there.
39) Message boards : Number crunching : ATi 58x0/59x0 DP problem and science (Message 38193) Posted 6 Apr 2010 by Emanuel Post: I think they will probably opt to move on to the new Milkyway background models they've been setting up. But I agree that this problem makes it painfully clear that we always have to be on the look-out for structural issues of this kind; in a way we can almost be thankful to the scammers for causing the validator to be revisited :s
40) Message boards : Number crunching : Marked as Invalid? (Part 2) (Message 38184) Posted 6 Apr 2010 by Emanuel Post: http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=90334482 Some get credit and some don't for the same WU???? Why would they get credit? Their results were marked as invalid. Mind you, in this case it looks like the wrong hosts got the credits, as it's the HD5800 cards that are giving the wrong results (this will hopefully be fixed soon, perhaps even today). There is no way for the server to tell which results are the valid ones except by majority vote, and unfortunately the HD5800 cards were in the majority on this one.

Previous 20 · Next 20