Message boards : Graphics cards (GPUs) : New GPU Machine Questions
| Author | Message |
|---|---|
|
I have a XP machine with an Intel quad-core processor that is completely dedicated to WCG. The monitor is driven by the on-board Nvidia video hardware. I would like to stick a GTX 280 in the PCIe slot and put it to work on other things as well. | |
ID: 5115 | Rating: 0 | rate:
 /   Reply
Quote
| |
 K1atOdessaSend message Joined: 25 Feb 08 Posts: 249 Credit: 370,320,941 RAC: 0 Level  Scientific publications      | |
I have a XP machine with an Intel quad-core processor that is completely dedicated to WCG. The monitor is driven by the on-board Nvidia video hardware. I would like to stick a GTX 280 in the PCIe slot and put it to work on other things as well. GPUGrid will use part of a single core, but WCG will continue to run fine on 3 cores plus part of another. However, the 280 is guaranteed not to work with your 300W PSU. The card itself has the potential to use the majority of the power of your 300W'er. See below: Nvidia Specs for 280 from www.nvidia.com: Maximum Graphics Card Power (W) 236 W Minimum System Power Requirement (W) 550 W | |
| ID: 5116 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Sabroe_SMC Send message Joined: 30 Aug 08 Posts: 24 Credit: 500,287,085 RAC: 69,608 Level  Scientific publications  | |
|
>Is the Boinc client smart enough to continue to run WCG projects on the Intel >and ship Seti (or whatever) to the GPU? Or should I be looking at a dedicated >GPUGRID machine? | |
| ID: 5117 | Rating: 0 | rate:
/ Reply
Quote
| |
Kokomiko   Send message Joined: 18 Jul 08 Posts: 190 Credit: 24,093,690 RAC: 0 Level  Scientific publications  | |
That's wrong. The feeding of the GPU use some time of one core, ca. 4%. But in this configuration (we call it's 4+1) the GTX280 can't reach her full potential. I drive them by 3+1 and so the card is up to 30% faster crunching, this give more calculation power than use the fourth core for oher tasks. In this case the dedicated core will be used up to 18% for the feeding of the GPU. ... and about the PSU: No way with your 300 Watt PSU. You need a newer PSU with more than 550 Watt and 2 6 pol PCI-E power connectors for a GTX280. ____________   | |
| ID: 5118 | Rating: 0 | rate:
/ Reply
Quote
| |
Beyond Send message Joined: 23 Nov 08 Posts: 1112 Credit: 6,162,416,256 RAC: 0 Level  Scientific publications | |
Questions: In the current situation you will probably have to do some babysitting to keep the queues full. I would expect that this will be fixed soon though and then running both projects should be pretty much automatic. Will the GPU Grid activities impact WCG processing preformance? With the current v6.55 Windows app (not an issue with Linux) there will be a 15-20% decrease in your WCG production. If they bring back v6.56 the CPU production hit will be less than 1%. Cross your fingers and read this thread: http://www.gpugrid.net/forum_thread.php?id=605 Hopefully the developers will check in with an answer before too long. | |
| ID: 5120 | Rating: 0 | rate:
/ Reply
Quote
| |
With the current v6.55 Windows app (not an issue with Linux) You are suggesting that queuing works better with Linux? I really wanted this machine to be Linux box from the start, but it seemed way slower on WCG, so I went back to XP. I attributed this to a lack of good Linux chipset drivers. Maybe I was wrong. I can always throw another HD in and try again. | |
| ID: 5123 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Beyond Send message Joined: 23 Nov 08 Posts: 1112 Credit: 6,162,416,256 RAC: 0 Level Scientific publications | |
With the current v6.55 Windows app (not an issue with Linux) No, queuing works the same. It's just that with the GPUGRID app v6.55 the Windows client uses a LOT of CPU time needlessly. It's a bug that was fixed with v6.56, but v6.56 was pulled because it didn't work correctly on 64bit XP machines. If you read the thread I referenced above all should become clear. | |
| ID: 5126 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level  Scientific publications | |
|
On the 280 class cards you need either a 6 pin and 8 pin connector, or 3 6 pin and the adapter in the box to power the card ... | |
| ID: 5129 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Ok. | |
| ID: 5130 | Rating: 0 | rate:
/ Reply
Quote
| |
|
ExtraTerrestrial Apes Volunteer moderator Volunteer tester  Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level  Scientific publications | |
Sounds like I should leave XP (32 bit) on my box and wait patiently for 6.5X while I put in a new power supply and a GTX 260. Just a small point to add: if you get the card you can start crunching along.. you don't need to wait for a new GPU-Grid client. Even with the current client the GPU will earn many more credits than the 15 - 20% you'll loose in WCG. And regarding the PS: if you don't plan to go multi-GPU a quality 500W unit is sufficient (e.g. Enermax, Seasonic, Fortron Source). You'd want a quality one anyway to get good efficiency. If you get a 550W unit for the same price it's OK, but going even higher increases cost without any benefit (for single GPUs). MrS ____________ Scanning for our furry friends since Jan 2002 | |
| ID: 5147 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications | |
Ok. As someone that has been trying (for years) to push more projects over the credit earnings I have for SaH, and with WCG on the cusp... I can sympathize ... :) And that is why I compromised to run the GTX 280 on GPU Grid and the 9800 GT "idle" for the moment which, in theory, maximizing my project contributions while waiting. I did not check to see if you have only one system or not (I have 5 so can make these choices) ... AND considering that WCG is going to go through some turmoil over the next couple weeks, well, this might not be a bad time to reduce WCG and let her rip here ... Of course no one can make the choice but you ... and there are no simple answers ... I am about 4 days away from pushing WCG over 388,000 something at which point I will throttle WCG down to its sustaining resource share and push up COsmology (next in line) to see if I can get it over 388,000 something to get it to the point I want ... (16 days if I can get my daily number to 10K per day ... possible if I dedicate the i7 pretty much to Cosmology only). Oh, also to get the "gold" for CEP ... 18 days needed still ... Anyway good luck with the new card and your crunching .... | |
| ID: 5149 | Rating: 0 | rate:
/ Reply
Quote
| |
|
I have two active WCG machines. Both are XP, which seems to be faster than Linux on my particular Gigabyte MBs (both the same). One is a dual core and one is a quad (Intel). My WCG credits are 2,691,789 (rank 4184). I want to keep doing as much WCG work as I can because I really like those projects. It's a shame I can't put a GPU on them. | |
| ID: 5174 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications | |
I have two active WCG machines. Both are XP, which seems to be faster than Linux on my particular Gigabyte MBs (both the same). One is a dual core and one is a quad (Intel). My WCG credits are 2,691,789 (rank 4184). I want to keep doing as much WCG work as I can because I really like those projects. It's a shame I can't put a GPU on them. The only reason WCG would "suffer" on either of those boxes if you started to run the GPU is because of the minor buglet in the 6.55 application. The question is if you want to not allow WCG to suffer at all ... then a new machine would be the best route for the next week or so ... After that, well, you could start all three machines on GPU Grid when the new application comes out. The nub of the question is if you want to not impact on the normal core processing at all ... or you do not mind a minor hit for the nonce... The people here will of course tell you to start here ... :) As soon as possible ... BUT, there are always other considerations for many participants. There are issues here with the servers, now, if you don't mind running dry then it is not an issue. If you hate fiddling and want to maximize the production ... all this comes into play... Oh, and the issues with the BOINC Manager which look to be with us for awhile ... | |
| ID: 5176 | Rating: 0 | rate:
/ Reply
Quote
| |
|
ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications | |
is because of the minor buglet in the 6.55 application. Just a side note, not sure if you're aware of this: it's not really a bug in 6.55. The problem of high CPU usage has been with us from the first day of the open beta. The current state of "few percent under Linux, half a core under win" is already an improvement over the first clients, which needed an entire core. The reason for the difference between Win and Linux is in the way the task schedulers work. So 6.56 didn't fix a bug, it introduced some clever work around or different technique. And back to topic: otherwise I agree with what Paul said! MrS ____________ Scanning for our furry friends since Jan 2002 | |
| ID: 5211 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications | |
is because of the minor buglet in the 6.55 application. Well, I noticed right off that 6.55 is running at a higher priority than other BOINC tasks so that may have something to do with it ... instead of the lowest priority it is running one step up ... this means that it will not "play nice" with the other BOINC tasks running ... And back to topic: otherwise I agree with what Paul said! And I thank you ... :) On my part I did start up the 9800 again ... you talked me into it ... :) In part because I decided that with nearly 14 goals on my list that I would level the playing field and run all 14 projects at 100 Share on all computers and then on two of the 5 I will be narrowing the focus a little more so that one or two projects are at a very high share to direct most of the resources to these two targets ... as the target projects hit the goals I will be setting their shares back down to the planned running share ... As soon as the end of the month I should have hit at least 5 goals with WCG (390K CS and Gold for CEP), Here to 300K plus, and probably Cosmology over 390K too ... Oh, and EaH over 500K (this week?) ... gotta have goals ... or how do you know if you are doing anything? :) | |
| ID: 5219 | Rating: 0 | rate:
/ Reply
Quote
| |
|
ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications | |
Well, I noticed right off that 6.55 is running at a higher priority than other BOINC tasks That is set by the BOINC client when launching CUDA / co-processor apps. With 6.3.21 and 6.4.2 GPU-Grid runs at normal priority, which seems to have been lowered to "lower than normal" in a later version. The idea is that co-processor based tasks need their CPU time right now when they request it and not a second later, otherwise the co-processor will run dry. The same mechanism was already in place when we had pre-6.55 clients. On my part I did start up the 9800 again ... you talked me into it ... :) Cheers mate ;) gotta have goals ... or how do you know if you are doing anything? :) I know that I'm doing fine if my team keeps its world position :) MrS ____________ Scanning for our furry friends since Jan 2002 | |
| ID: 5228 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications | |
I know that I'm doing fine if my team keeps its world position :) Except you are on the wrong team ... :) | |
| ID: 5236 | Rating: 0 | rate:
/ Reply
Quote
| |
|
ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications | |
|
Now that's bold.. and we even have people singing "Join EK now!" | |
| ID: 5242 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications | |
Now that's bold.. and we even have people singing "Join EK now!" Check the water ... must be something in it ... | |
| ID: 5250 | Rating: 0 | rate:
/ Reply
Quote
| |
|
ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications | |
|
Only fermented hops and malted barley, I guess ;) | |
| ID: 5476 | Rating: 0 | rate:
/ Reply
Quote
| |
Message boards : Graphics cards (GPUs) : New GPU Machine Questions
