I finally got everything working ok on my system,I hope and got my first GPU based workloads from MilkyWay.

What is considered a 'max safe' temp to run your GPU at, I have a control panel which I can adjust the processor and memory speeds with but notice that a little change can make a temp change a lot.

I have a GeForce GTX570 board.

Thank you
Aaron

For a 570 the formal threashold before it downclocks/downvolts is 97 degrees, but stay well well clear of that.

For a fully loaded card in a well ventilated machine, 83 degrees would be the medium temperature on o/c, and the one you should aim for in a fully stressed, fully o/c card, certainly not above 85 as a 7x24 routine.

I would start having at look at whats going on above 86/87, and if it went above 90 seriously think about turning it off until its understood whats happening.

Full 100% stress level with no o/c should be around 77 degrees.

It will protect itself above 97 by downclocking etc .... but I wouldnt like to try and test that :)

Regards
Zy

Thanks for the quick answer. I seen it come up to around 80 degrees and got a bit worried, backed it down a hair and nuked the fan speed to max.

I think I am going to play with the memory speeds and the cpu speeds and bump them up to see if I can squeeze a bit more out of it.

Any suggestions which will have the most affect? Im thinking cpu speed on the card as the memory barely seems to be used at all.

Im completing a task in about 6 minutes? Is that average for a de seperation task?

Thanks
Aaron

Set the memory as low as you can practically get it comensurate with other tasks you use the beast for. MW has a low memory bandwidth need, and even default memory setting of a card is a waste. You will reduce power requirements and heat by lowering memory setting - it has no affect whatever on MW as the dataset it sends from the CPU to the GPU is small. Collatz for example is the opposite, it sends very large tables and the bigger the memory bandwidth the better (circa 1200 to 1300Mhz).

The higher the Core GPU speed the better.

CPU speed has a very small affect on any GPU WU in any project, with a single 570 you will not become CPU bound, and the datasets MW handles are relatively small. You will notice a difference in times if you run a CPU WU concurrently due to the way BOINC handles requests from the GPU for a CPU assist. Horses for courses, if you are crunching for pure credits, leave the CPU WU alone, its likely you will lose more credit by the slowdown of the CPU WU (from any Project) than you would gain by running MW GPU only.

Re speed of 570, cant nail it further as I do not run one, but I would have thought you could reduce that time a fair bit - treat the latter with caution, I dont run a 570, best wait for someone to come in who runs one.

There is an excellent review on the GTX570 if you have not seen it - well worth a read in detail Guru3d GTX570 Review

Bare in mind that MW is a Double Precision Project, and NVIDIA nerf the DP to a far greater degree than ATI does. The upshot of that is you will find your performance relative to ATI cards to be approaching poor.

If you crunch purely for Credits, Prime Grid (which is single precision only) is probably your best bet at present with their new NVIDIA apps which pay very high compared to other NVIDIA based Project WUs.

The next best would be GPUGRID, and the choice for "crunching to make a difference". Its a "non-mathematics" one that crunches molecular structures for medical research (nice bunch, the Project Scientists and Admins are good people, friendly etc, they are based on Barcelona University, and are a well respected research crowd).

Regards
Zy

Thank you for the in depth reply Zydor. I have been playing with the card a bit and am slowly learning it's quirks.

You said Nvidia nerfed their DP, hmm, that's a shame, I was under the impression it was one of the better cards out there, it certainly cost enough to be!!

I like crunching for the sake of crunching, why waste the processor cycles when this computer is pretty much on 24/7 anyways. The credits, yah they are nice but that's not solely why I am doing this, if I all I wanted was credits I could just write an app that says presto Aaron, you have a bajillion credits, wow you are sooo super :D

These projects are for a good cause and it's nice to be able to say, yes, I helped with that. Especially the medical ones, as I am sure pretty much everyone of us here has had someone we know / loved suffer from and / or die of cancer. In this case yes, your time is definately going towards something that will better mankind one day.

Aaron

In that case definitely head for GPUGRID - they do genuine cutting edge medical research that has been cited in many respected journals.

I will join you there, hopefully in the not too distant future, as they are close now to putting out an ATI app which is hoped for Q1 next year. I was there up to April this year when I went purely ATI for other work related reasons, and am itching to return for many reasons, some the genric work they do, some personal reasons, hopefully will not be long now.

Regards
Zy

In that case definitely head for GPUGRID - they do genuine cutting edge medical research that has been cited in many respected journals.

Especially if you enjoy constant frustration and arrogant admins who don't care about the people doing their work for them. Check out the forum before joining.

In that case definitely head for GPUGRID - they do genuine cutting edge medical research that has been cited in many respected journals.

Especially if you enjoy constant frustration and arrogant admins who don't care about the people doing their work for them. Check out the forum before joining.

Unfortunately this is a bit true I have experienced. I crunch space physics prior, Einstein and Milkyway. But some medicine at times when both other have problems. My pc with nVidia can only crunch for them and when I was not getting new work they told to update the driver. I got the latest, also good for Milkyway, and then it was too new. They asked for more info but don't read the post. So if it works, okay but otherwise then I will not put any effort in it again. Then I went to Rosetta which is also for medicine (i.e. cancer and Alzheimer). So indeed with GPUGRID you are a bit on your owne. The best experince I have is with Einsten@home where admins and other crunchers are very helpfull and within a few minutes to hours after posting an issue.

---

Greetings from,
TJ

Especially if you enjoy constant frustration and arrogant admins who don't care about the people doing their work for them. Check out the forum before joining.

I'm going to have to call you on that cheap shot. The guys at GPUGRID have been great in the forum. I've been running GPUGRID plus MW for a few months in the same machine and have had very few problems. Normally, most project issues are the fault of the driver or BOINC, but there's nothing a rollback can't do.

The science they are doing is great, and it is a good community to be with.

Perhaps you can be more constructive by actually stating what kind of frustrations you were having? Following their instructions verbatim is the best way to go, including the swan_sync = 0 switch if you're in windows.

Anyway just my 2c

---

Goal: 1 BILLION cobblestones

Progress: 2%

Especially if you enjoy constant frustration and arrogant admins who don't care about the people doing their work for them. Check out the forum before joining.

I'm going to have to call you on that cheap shot. The guys at GPUGRID have been great in the forum. I've been running GPUGRID plus MW for a few months in the same machine and have had very few problems. Normally, most project issues are the fault of the driver or BOINC, but there's nothing a rollback can't do.

The science they are doing is great, and it is a good community to be with.

Perhaps you can be more constructive by actually stating what kind of frustrations you were having? Following their instructions verbatim is the best way to go, including the swan_sync = 0 switch if you're in windows.

Anyway just my 2c

yeah, ok, right. HUDSON!!! ;) I am with beyond regarding their support, but my experience is now dated back when a ps3 was a nice cruncher...WAY back then. not once did they [do 'guys' = admin?] answer a question of mine and I can assure you it not about drivers and boinc--their flaky linux clients. after a very frustrating time over several months, I gave them every opportunity, I left thinking they were the worst admin in the community. when they did sing for whatever topic, it was so often a blow off. there were non-admin trying to help, of course, many helpful cruchers out there, but rarely the admin. however, based on your view, maybe they turned over a leaf and now up to minimum support. good for them! regardless, I do not consider beyond's view a cheap shot--the thought to check the forums is reasonable.

I am happy for their crunchers if the admin got with it, but the lasting impression, for me, is poor.

The best project I was on was Nano-Hive. The other projects have been hit or miss.

---

Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

Especially if you enjoy constant frustration and arrogant admins who don't care about the people doing their work for them. Check out the forum before joining.

I'm going to have to call you on that cheap shot. The guys at GPUGRID have been great in the forum. I've been running GPUGRID plus MW for a few months in the same machine and have had very few problems. Normally, most project issues are the fault of the driver or BOINC, but there's nothing a rollback can't do.

The science they are doing is great, and it is a good community to be with.

Perhaps you can be more constructive by actually stating what kind of frustrations you were having? Following their instructions verbatim is the best way to go, including the swan_sync = 0 switch if you're in windows.

Anyway just my 2c

The single GPU you're running on GPUGRID is one of the least problematic. Consider yourself one of the lucky ones. But your 2 posts and 3,238,524 credits don't add up to being an expert on the project. That said I sure do defend your right to your opinion even if it doesn't have much experience behind it.

Go to their forum and read my 399 posts, and yes I ran 37,217,581 credits worth of work for them. Spent a lot of time (with others) trying to help with debugging their problems (usually at their request). When we found an answer it would almost always be ignored, or covered up, or denied if it wasn't what they wanted to hear. Now they're complaining that there's been a big decrease in people crunching. Why do you think that is? It's because so many became frustrated and left. For every one posting their frustrations there's probably dozens more that said nothing and walked away. I hope they turn things around. If they do I'll go back again, but to improve they need to have an attitude of cooperation with the guys doing their work and their testing. So far that's been missing.

Guys, please. I understand your frustrations with the lack of support but can we please not turn this thread into a GPU grid bash a thon?

I have never experienced anything like you said however wish to say this. I have worked in a pharmaceutical company before and the researchers, ie the doctors with the PhD's, were good at what they did, but not really good with 'human interaction'. We'd get called for a problem and getting one of them to explain it to you... WHAT? you can't see that the nanofluxer is sticking on the knibbly pin and that is causing temporal rifts in the wobulator which is causing the doofulator valve to overload the kanuten flapper? Followed by a look of disbelief, an eyeroll that my subspecies is even breathing, and a disgust look of why am I not on my knees bowing in their exhalted presence.

They may come off as arrogant but I honestly don't think they are trying to be, that's just the nature of that beast. On the not answering your questions, maybe they didnt know the answer. They may be good at gene crunching but are not Unix / SQL guru's. so are not sure really. Just a thought on my part, take it or leave it.

Aaron

It's often cultural. Although there are many exceptions if a project is based in Spain or the USA, in my experience there is a greater chance of an abrasive and seemingly uncaring attitude to their contributors.

Some projects have a great sense of self worth and do not believe in allocating any time or resources to addressing the concerns of their contributors. This breaks the idea of a partnership between project and contributor but they will never be able to see it. Focussed so strongly on their research goals they tend to treat contributors as a resource to help them achieve their aims rather then as partners.

When choosing which projects to donate computer time to, some are prepared to overlook such arrogance because they believe strongly in the particular research that is being done. Others' choices are influenced greatly by the level and quality of support offered by the project staff and whether or not they treat contributors in a reasonable and adult manner or prefer to rely solely on paternalistic forum administrators to run interference and fob off contributors' genuine concerns as annoying distractions.

I can understand what you mean Kashi, some folks may want to act like they are doing you a favor by allowing you to crunch their numbers.

Sometimes projects have very little money as it is and they can't afford to spare anyone to be a permanent or semi permanent 'liason' to essentially babysit message boards and appease. However these projects will also find themselves constantly looking for people to do their bidding as folks get tired of the attitude and walk away.

Then again, there are plenty of projects out there too and one can also, simply goto an other if they feel undertreated at one project.

It was mentioned more than once that many folks don't feel like their work is appreciated over at other sites. Has anyone brought this up to them? I don't mean a flame war or ridicule laced post but has anyone genuinely tried to let them know that the crunchers need a little love once in a while?

Problems can't be solved if nobody brings them up to the folks that can solve them.

Aaron

Yes quite right, differences in funding are relevant, but even that differs in how it affects projects in different countries. The cultural differences are not just in personalities, but extend to the research institutions themselves. Where there is not security of tenure and funding, competitive pressure can lead to a sense of urgency and insecurity that is not conducive to a reasonable and well thought out approach. Some of these people do not have the correct frame of mind or cannot afford the time to be polite even if they wished to be. As I implied before, often they are not even aware of it or consider it quite normal behaviour. Occasionally you may be able to notice changes in the mood and tone of researchers who were previously based in one country who then take up a position in another country.

Another generalisation, but regardless of the culture/country the problem is more often seen in biology/medical types of projects rather than those involved in physics or mathematics. The physics and mathematics people often go into incredible detail explaining the work their projects are dealing with. They do not find this a bothersome chore but are keen and excited to share with their contributors. Even if I cannot always understand all the details I am still interested to read about what I am actually crunching and feel privileged that the researchers have taken the time and effort to share their enthusiastic spirit of discovery.

The biology/medical projects however sometimes have the concept of the layperson who should be told little of the detail because they would not be able to understand it. They tend to more of a closed shop. There is sometimes a sense of entitlement that because the projects are researching cures for diseases then any explanations are just unnecessarily wasting time and money that would be better used finding cures quicker.

I suppose this is partly traditional in medicine, kind of like doctors writing some things in Latin. Us and them, treat them like children approach. Us medicine men have great status in the tribe, you shall not question or upset us lest we cast a spell on you. We know what is best for you, we will not deign to listen to your uninformed mutterings until you are "peer reviewed". Our great wisdom and knowledge are wasted being freely shared with the masses so sometimes we ensure the details of our research are available only to those with professional qualifications who pay to subscribe.

A trifle exaggerated and unkind perhaps, but just painting a picture.

Wow Kashi, it sounds like you worked for Vistakon too at one time :D

Oh yes, I understand exactly where you are coming from there and the point about the sci/math geeks being very willing to 'show you' how intelligent they are .vs. the medical geeks being frustrated that you can't automatically see how intelligent they believe they are.

Yes they do tend to surround themselves with a caste system at times too. However as you have also stated, there is a LOT of empire building going on, that's their job security. If they are the only ones who know something important, then they are indispensable right? Many places the backstabbing and peer knifing is very prevalent and thats the only way one can stand tall is when standing on the back of a colleague. I do believe though that they are also taught a bit of this mentality in their schools too, they are told that they are somehow better than their fellow man and to act that way too. Shame really.

Yes I remember those days and not very fondly either, you would start the first few hours of the morning wading through the ego's and listening to each re iteration to the Nth +1 degree, on how they are going to save the world and you should be privileged to even be on the same planet as them. But then, for every 5 arrogant doctors you'd have to put up with, you'd have one that was really genuinely a great person once you figured out that it wasn't that they were intentionally being rude, but that their brains are wired up 'scientifically' and they never really learned .. for lack of a better description, how to interact with people properly, they were always busy mentally working out their next problem and socializing to them, was more of a waste of time when they could be doing something 'constructive' . Maybe that's what makes them so good at what they do? But hey, it takes all kinds to make the world go round right?

Aaron

i just got my system back up from a harddive crash/death. i upgraded with zotac gtx570. they are running hot and would like to adjust them...
clock speed: 723Mhz
memory clock 3800Mhz
shader clock 1464 Mhrs
Fan speed to Auto

My temp hits 88-91 c 9 night.

q 1 what is the Shader clock can i lower it to make the system run cooler but not lose RAC

q 2 how much can i lower the memory clock and still keep a good RAC?

i have a cool setting for the daytime
632 core, 3280 memory, 1264 Shader.

now this works okay but when summer hits south AZ i need a plan to keep it cooler

thanks for any help

any body out there?

Cliff

Memory Clock is not relevant at Milkyway, the data sets MW processes are relatively small, and you can reduce your memory clock way down. To give a relative example, AMD cards with a default memory clock of about 1000 depending on the card, can easily run MW on 300 or less with no effect on RAC (memory setting refers to bandwidth not speed). Reducing that will reduce your heat output and power needs.

Your fan is on auto, switch to manual and crank it up a bit - I am aware its noisy on those, but manual control allows a higher setting and is usually better for o/c with the attendent flexibility.

Another thing you can do is pop over to look at Guru3d's 570 review. Read that in detail, Guru3d is an excellent Site, well worth bookmarking, and has an excellent Forum populated by very knowledgeable folk in the hardware world - always helpful provided the requestor remembers their please & thank you's. They saved by butt a few times over the years.

Edit:

Forgot an obvious one, but often overlooked - and one expensive one.

Clean out the case, cpu and cards if not done for a while, dust builds quickly. Maybe possible to fit another fan in your case, or go up a fan size? Fans are cheap these days, and anything you can do to increase airflow in the case will have a marked effect.

The relatively expensive, but really effective solution is the Arctic GPU cooler. They claim a 40 degree c reduction - and they are right, I use them on my twin 5970s, superb, use those and you will not care about summer, no matter where you are, winter or summer ... Arctic Cooling GPUs. Its a chunky price, and you woulds be wise to get it fitted by a local IT Shop, its a fiddly install not to be taken lightly, but hugely Hugely worth the time, cost and hassle. Fantastic bit of kit.

Regards
Zy