| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Conquer Cancer
Thread: Will HCC publish list of approved GPU cards for Windows? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 21
|
|
| Author |
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: 
|
Someone stated that a desktop would need a 500watt power supply...that seems hard to believe...most desktops only come with 300-400 unless it's a gaming system. Can anyone else confirm that the GPU cards even list that as a Hardware Requirement? Usually the requirements are simply OS and slot type. I have been Folding on a GTX 560 Ti for over a year with a Seasonic 330 watt supply with no problem (and running WCG on all the CPU cores, originally a Core2 Quad and now an i5 3550). It draws no more than about 220 watts, and usually less than 200 watts now. It is duck soup, though I did get an error above 280 watts when I added another card, and so pulled it out. NOTE: these powers are measured at the AC input to the power supply. But it depends on the quality of the power supply a lot. [Edit 3 times, last edit by Jim1348 at Sep 4, 2012 9:55:11 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
That's not to say that Nvidia cards won't work well. They will. It's just that ATI will tend to be more efficient and faster for cards of equal gaming performance. (Emphasis mine).I'll settle for the more efficient part and give nVidia fans the faster part. After all, one can not and should not have them all.  It would be extra nice if the list could be sortable by speed/performance. Hmm... I don't know. Would that sound like asking for trouble of the kinds that hardware-review sites have in their forum? But it depends on the quality of the power supply a lot As for the matter of power-requirements, I think that is best left for crunchers to do some research on before deciding. Many factors come into play and they are all influenced by current, projected, and future usage profiles for a cruncher.; |
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
I'll settle for the more efficient part and give nVidia fans the faster part. After all, one can not and should not have them all.
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Ok, nanoprobe. Got you. I will also have to strike off the wink-smiley in the new version below.
Thus: I'll settle for the more efficient part for OpenCL computing in AMD cards, and give nVidia fans the faster part in many, but not all, games -- after all, one can not and should not have them all. ; |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
They just did a beta (GPU 6.49) this morning; blink, and you missed it. The main purpose was to determine why some cards could not be identified properly, so I don't know if the work units were a standard size or not.
But at any rate, I picked up 25 of them on my two Radeon HD 7770s, and 23 completed properly on my end (but not validated yet). These had an average elapsed time (as measured by BoincTasks) of just under 3 minutes, including both GPU and CPU time (I have a Core2 Duo at 3 GHz). The other two errored out early, in only 2 seconds, but they were part of the first group sent, and did not identify the card properly. However, they must have fixed that quickly since all the subsequent work units properly identified the cards and ran fine. |
||
|
|
mikey
Veteran Cruncher Joined: May 10, 2009 Post Count: 821 Status: Offline Project Badges: 
|
BOINC will only crunch one at a time unless the project provides a mechanism for multiple instances. This is mistaken info...you can use an app_info.xml file and force the gpu to run as many workunits on a single gpu as your heart desires. The relevant line is here: <coproc> <type>NVIDIA</type> <count>0.5</count> </coproc> The default is 1.0 which runs 1 unit at a time, but putting 0.5 runs 2 units at once, putting 0.333 would run 3 units at once, putting 0.25 would run 4 units at once, putting 0.125 would run 8 units at once. Now doing that is NOT always helpful, to be FAST and efficient a unit MUST fit TOTALLY into the gpu cards onboard memory. Once you go beyond that it is swapping to either ram or the harddrive and ALL of your efficiency is lost! I use the above at MilkyWay for a 1gig gpu and it churns out units with only a very slight time loss, seconds, compared to running a single unit, which is more than made up for by the fact that I am running 2 units at once. It doesn't matter if the type is Nvidia or AMD it works for both. IF you do this you NEED to be careful and monitor the heat temps of your gpu though, it can easily reach the max recommended and beyond if your case is not designed for good cooling. I recycle my cases and leave the side off on most of them. By recycle I mean I replace the inards and reuse the case to build a new pc out of parts. Most pc's are like chicken mcnuggets, 'parts is parts'.   |
||
|
|
JCMarsh [U.S. Army]
Cruncher Joined: Feb 8, 2012 Post Count: 5 Status: Offline |
@mikey159b Yes, parts is parts, and I've done plenty of that, but my latest cruncher is much happier with the side cover on. There's about a 4 degree C change in temp, which may not sound like much but every bit makes a difference. If you're running cooler without the side cover on your pc case, perhaps you'd benefit from adding or upgrading a fan. Or just do what the serious geeks do and go liquid. I did, and I absolutely love it. Temps (cpu, ps, mb) are down across the board and really easy on the silicon.
----------------------------------------As for GPU, I'm using a 550ti on GPUGrid and it practically never has errored out on a work unit. When it has, it's been due to user error (shutdown now, without at least snoozing or exiting BOINC client first). If HCC is having problems with CUDA computing, it's a problem with code, not hardware or architecture. Isn't there any interop/cross-talk between the CS geeks who implement the science applications? Saying CUDA is less efficient or more error prone is like saying you can't convey a certain message in another language. Of course it's possible, you just don't know how yet!  |
||
|
|
sean0118
Cruncher Australia Joined: Feb 7, 2010 Post Count: 33 Status: Offline Project Badges: 
|
I have been Folding on a GTX 560 Ti for over a year with a Seasonic 330 watt supply with no problem (and running WCG on all the CPU cores, originally a Core2 Quad and now an i5 3550). It draws no more than about 220 watts, and usually less than 200 watts now. It is duck soup, though I did get an error above 280 watts when I added another card, and so pulled it out. NOTE: these powers are measured at the AC input to the power supply. But it depends on the quality of the power supply a lot. Well if it works it works. However the majority of power supplies are most efficient when running around 50% load. Like you said Seasonic are a good quality brand, some of the cheap ones will not have much power output on the 12V rail. For example a 1000W psu for $20 might only supply 200W on the 12V rail and the rest on the 3.3V and 5V, which is no good for GPUs. |
||
|
|
sean0118
Cruncher Australia Joined: Feb 7, 2010 Post Count: 33 Status: Offline Project Badges:
|
Sorry if this is not the right place, but it seemed a waste to start a new topic.
Can anyone recommend a suitable GPU that's <$50US (new or used). I know that's not really enough to get a decent card but it is all I have in PayPal at the moment and I would like to test this project out. thanks ;) |
||
|
|
KWSN - A Shrubbery
Master Cruncher Joined: Jan 8, 2006 Post Count: 1585 Status: Offline |
Can't really make a recommendation but I can tell you for $50 you will not find a card that will work on this project. None of the older and therefore affordable cards have sufficient capacity to be allowed work.
---------------------------------------- Distributed computing volunteer since September 27, 2000 |
||
|
|
|