Hello

I want to have the software to monitor the performance of the card and to reduce the voltage as you suggested me in this topic.
I already lost the previous graphics card and I want to reduce the risk of losing this card.


Thanks for the reply.

I have already downloaded the EVGA Precision X and I have it installed, I have successfully changed the skin so everything works fine.

Can anyone tell me how to lower the voltage at 90% and "save" the changes to the card always work well?
I've set the fan speed to "automatic by software."


Tanks.

Setting the "Power Target" value to a number below 100%, will instruct the GPU to not ramp its clock all the way, when being utilized. So, maybe try 75% Power Target, and hit Apply, if you are that worried?

I do not know how much risk involves processing data GPUGrid project or other projects using the graphics card and I do not know how often is that a graphics card is damaged by processing data from Boinc.
What do the users of this forum to reduce the risk of damaging your graphics card?


Thank you.

In the time I've been crunching for BOINC, I've used 7 graphics cards in two machines and have never had any of them damaged. As long as you're operating them within safe limits and keeping them as cool as you can, you should be fine. As a general rule for temperature, try keeping them below 75C if they will be running for long periods of time but ideally below 70C. The lower the better. Do some research on your new card to find out all you can about its capabilities and limits. Google is your best friend for that.

I've been running BOINC also for several years, using 5 different GPUs. I prefer to push them as hard as they will go, without adjusting any of their factory-preset voltages. (So, for instance, I will set the Power Target % to 140%, so it'll up-clock as much as it can within voltage tolerance, regardless of Power usage).

The only GPU problem I had was when I bent a fan blade, while cleaning a GPU fan. It made a high-pitch whirring sound from that point.

But I've personally not had any problems running the cards as hard as they will go [without adjusting voltages, and without adjusting clock rates].

Keeping the temps below 70*C is best. I set a custom fan curve in Precision-X, so that it will set the fan at full-speed by the time it gets to 69*C.

Keeping the temps below 70*C is best. I set a custom fan curve in Precision-X, so that it will set the fan at full-speed by the time it gets to 69*C.

How long does it stay at 100%? I've always tried to keep the fan no higher than 80% to save wear on those parts. Am I being overly cautious?

My eVGA GTX 660 Ti 3GB FTW... has a fan that only allows settings of 30% to 80%, as witnessed by the yellow dashed lines on the fan curve. My fan curve is set so that it reaches max fan (80%) right before thermal limiting temp (70*C), so... 80% at 69*C. To answer your question, it's usually at 80% 24/7, because I run a hot computer. And I've not experienced any harm running my fans like that.

In fact, even with those settings, I was having trouble keeping it below 70*C. So, I've recently written a program to have my system fan ramp up in accordance to my 660 Ti's temperature. It works nicely (I take a lot of pride in it), and now my GPU stays below 70*C (meaning it'll stay at Max Boost 1241 MHz, without dipping down 13 MHz to 1228 MHz)

For reference, I have 2 other GPUs in the system, a GTX 460, and a GTS 240. The GTX 460 can do fan settings of 30% to 100%, and the GTS 240 can do fan settings of 35% to 100%. Neither of these 2 GPUs support "boosting" or "thermal down-throttling", so they will run at full Mhz regardless of my fan setting. Nevertheless, my fan curve does not simply stop at 69*C. I have another point on the curve set at 85*C at 100% fan, so that those other 2 GPUs can ramp up beyond 80% fan, up to an all-out fan setting at 85*C (what I consider to be the beginning of the danger zone for a GPU). Temps are usually between 70*C and 77*C on them, so fan speeds are usually around 80%-90%. 24/7. No problems with fan wear and tear, just a bit noisy.

I don't know if you're being over-cautious. But I would feel pretty comfortable pushing the components as hard as they will go without overclocking or overvolting. (So, maybe you ARE being overly cautious)

Hope this helps!

Thanks, Jacob. That's a lot of good info.

My EVGA GTX 780Ti Superclocked cards will go to 100% on the fan curve. Currently I'm running the new NATHAN_RPS and SDOERR_BARNA WUs on these. One card is holding 70C at 80% fan and the other 71C at 81% fan. About 85% GPU utilization each. The room temp is 14C. The 780Ti seems to run a bit warmer under higher loads than my two EVGA GTX 680 FTWs did in the same box before I swapped them out.

I'll try adjusting the curve more to see if I can get them to stay below 70.

I'm trying to think, in my head, what would happen if you set your fan curve to be at 100% at 69*C. I guess it would mean that the fans would work harder, and your end result would be cooler GPUs, which would be overworking the fans unnecessarily. Hmm... Not sure what I'd do in your scenario. You might want to just try that setting, to see what the end result is -- I'm very curious. Or maybe, since they "usually do okay at around 80% fan", you could try setting a point at 67*C 75%, then another point at 69*C 100%.

But here's a tip that'll help performance:

There's a separate thread on maintaining Max Boost, if you want to keep it clocked at Max Boost even when the stupid drivers stupidly think the utilization is "not high enough". In it are instructions to create a .bat file, that you can set to run at Windows startup, to ensure maximum performance. If you're going to run the GPU(s) 24/7 anyway, it makes sense to ensure they are at Max Boost, for maximum performance.

Original Forum post:
http://www.gpugrid.net/forum_thread.php?id=3647
Post summarizing the .bat file:
http://www.gpugrid.net/forum_thread.php?id=3647#35562
Post detailing exact procedures for the .bat file:
http://www.gpugrid.net/forum_thread.php?id=3647#36320

EMYArg: I apologize for sounding so rude earlier. If Precision-X doesn't properly set your custom Power Target % every Windows startup, you could read through the instructions in those 3 links, to create a .bat file that does.

On these cards the base clock is 980MHz. GPU0 at full boost goes to 1124MHz and GPU1 goes to 1137MHz at full boost. (My 680s did the same thing: 1 boosted about 1 13MHz "step" more than the other.) GPU1 also seems to run consistently 2 - 3 degrees cooler as well.

Currently I'm running on 335.23 with a Power Target in Precision X at 106% (the max for me) and a temp target of 72C. Under these conditions I've not had any trouble with downclocking below full boost when under the Temp Target. I had the "under-utilization" problem with earlier drivers but since I've gone with my current driver I haven't seen the problem come back.

I read those threads and was set to implement the suggestions but then, as I said above, the problem went away. Thanks for putting in the effort on that, though. It seems to have helped a number of people already.

Currently a fan speed of 94% is holding GPU0 at 70C at full boost. GPU1 is at 68C with 88% fan and also at full boost. Both are still averaging around 85% utilization with NATHAN_RPS and SDOERR_BARNA WUs.

I'm jealous of your GPUs :) I would love to have a 700-series, not only for the performance, but to learn more about Boost 2.0 (with the Temperature Target). From what I understand, the user gets to choose one of the target types to be active, and the other target type is completely ignored. But maybe I'm wrong. NVIDIA's webpage is horribly lacking in terms of describing the functionality. http://www.geforce.com/hardware/technology/gpu-boost-2

I'm told that, with the Titan's at least, they actually don't start thermal downclocking from Max Boost, until the temp reaches 80*C. So, that's another thing to keep in mind. You could very easily test this, by unsetting "Auto Fan" in Precision X, and then manually setting the fan to lower speeds to watch the temp go up, and then monitor the temperature where you first see the card downlock. Usually it's about 2*C above the thermal drop, so... it'll start downlocking around either 72*C or 82*C. But then, it stores a "history" of temps, and won't upclock unless it thinks it won't get tripped again. So that's why I recommend max fan at 1 degree below the thermal limiting temp.

So... If you're up for testing it, I'd just be curious to know if a 780 Ti starts thermal-downclocking at 70-72*C, or if it's 80*C-82*C instead. You don't have to do the test if you'd prefer not to, but I generally think anything up to 85*C is safe.

Regarding that... One time, I had suspended BOINC and unset Auto on Precision-X, and manually set a low fan %, so I could sleep in the room. When I woke up, I found that I had used BOINC snooze by accident (so BOINC kicked on an hour later), and my GPUs were COOKING. They were actually at the "critical" thermal limit -- Precision-X was reporting 100*C on the GTX 660 Ti, and instead of the 1241Mhz max boost or 1024Mhz normal 3D, it was clocked at something ridulous like 365Mhz. I'm glad that extra safety net was there; that was a true scare. I'm never ever ever ever going to uncheck the "Auto" fan curve again, unless I'm doing a test and will remember to put it back on.

I'm almost positive the thermal dropoff for a 780Ti is 80*C, not 70*C. So, in terms of performance, you could get away with a curve that has 100% fan at 79*C, and it wouldn't thermal downclock. But, if you wanted to keep the GPUs even cooler (for less risk of work unit errors maybe?), you could put the 100% point even lower (like at 74*C, 70*C, or 69*C). That's my opinion at least.

Hello

I could not complete the download of any GPU GPUGrid unit, even if I have come GPUGrid CPU (CPU Only App 1.05 ) units , I did not know existed in these units to process CPU. The units I try to download so far have been only four , all of them have been in " Download: retry in HH : MM : SS " but never finishes downloading , at this time there is a acemd.815 -42 unit. exe in that state but never finishes downloading .
When units of several projects started to download Avast Free began to block them , I accept them one to one but I think one of them mistakenly select the option that prevents downloading the file, maybe that's the problem , Avast may be blocking the download and for that reason never ends.
As I can not get GPUGrid units , I downloaded some Nvidia Prime Grid units at this time are being processed but I do not know what percentage of the capacity of the graphics card is being used because I still do not understand the operation of the interface monitoring software, EVGA Precision , but the temperature of the graphics card is maintained in 47 degrees Celsius and the speed of fans of the graphics card is kept in 48% .

Edit: Now, in the section of transfers Boinc Manager, download the GPU unit slope is marked as "Stoped Project".


Thank you all for the answers.

EMYArg: I don't use Avast so I'm not sure what to do in order to successfully download and run GPUGrid tasks. I've never had any trouble with AVG and BOINC projects.

Precision-X won't give you a good idea of the actual utilization of your card by various projects and tasks you're crunching for them. For real-time monitoring I would recommend Nvidia Inspector. It's also a very powerful tweaking tool in its own right once you know more about your hardware.

Jacob: Yes, I can independently set separate Power and Temp Targets and tell Precision-X which to prioritize. The default Temp Target is 82C, so I believe you are right about the raised threshold of GK110.

I'll run those tests you suggested sometime in the next day or two probably and let you know what I find.

Avast may be blocking the download and for that reason never ends.

Thank you all for the answers.

Go into your Avast program and exclude the Boinc folders to keep Avast from checking them. IF there is a virus then it WILL try to get out of the Boinc folders and get caught by Avast, if it is a false positive, as is most likely, you will be fine. Either way the only thing directly affected will be Boinc and if it gets a virus as long as the only one it connects to is the project servers it isn't your problem.

As far as gpu crunching I too have been doing it a LONG time and have never adjusted the software beyond the defaults and have never burnt up a video card. I HAVE had some get so gunky the fan slowed down or stopped and it overheated, but the machine shut itself down as opposed to burning up the gpu. I was able to free the fan and get back to crunching with no problems. You can also buy aftermarket fans for your gpu's, but I have not done that either, although the last few gpu's I have bought do have multiple fans on them instead of just a single one. I use both Nvidia and AMD gpu's and on some projects crunch more then one unit at a time, while at other projects one unit at a time keeps the gpu busy.

My eVGA GTX 660 Ti 3GB FTW... has a fan that only allows settings of 30% to 80%, as witnessed by the yellow dashed lines on the fan curve. My fan curve is set so that it reaches max fan (80%) right before thermal limiting temp (70*C), so... 80% at 69*C. To answer your question, it's usually at 80% 24/7, because I run a hot computer. And I've not experienced any harm running my fans like that.

Right and we are talking here about an EVGA GTX660, I have two of those and the maximum of the fan is 75% while it actually runs at 74% max. No other program like Precision X, MSI afterburner or Asus GPUTweak can get the fan at 75%. With more rigs in the room the ambient temperature will rise quite quickly when outside becomes warmer and thus is my first EVGA 660 often running at 74°C and 76-78° with an SDOER_BARNA. However after a year 24/7 no issues with the cards, besides a lot of error with SANTI's, but I will replace soon with one GTX780Ti. As it can do little more in the same time as 2 660's

---

Greetings from TJ

Hello

Avast Free was blocking the download of the GPUGrid units but i already solved the problem, I have a unit "Long Runs" running and according to the EVGA Precision and Tthrottle the temperature of the graphics card is 50 degrees Celsius, and the fans of the graphic card work at 50 %.
I think that 50 degrees celcuis of temperature are perfect, what do you think?


Tanks to all.

I think that Temperature, GPU Usage, and Power Usage, are all 3 factors that you should care about, if you are worried about wear and tear on your GPU.

If you want to somehow LIMIT how hard your GPU works, you should do it using the POWER TARGET % value, setting it to something below 100%. This value determines "how hard" the GPU is allowed to use its resources to do work while maintaining the fans. The real thing it is determining is how many watts of energy are allowed to be used by the GPU, to do the work and run the fans. But it's the best thing to manipulate.

Whatever you do (limit or no limit)... if you keep the temperature below 70*C (by setting a custom fan curve in Precision-X), and you stay away from any options regarding over-clocking or over-volting... then your GPU should be perfectly fine.

Regards,
Jacob

I am processing without having changed anything related with voltages or other values, the graphics card is processing with the settings that it had when it came out of its box, so I am surprised that this is processing at 50.3 degrees Celsius when the graphics card Asus Nvidia Geforce GTX 650 TI Boost that i had before was never below 70 degrees celcuis.


Tanks.