Hi there
I have in issue with the Powermizer of the GTX970. Although I set it on “Prefer Maximum Performance”, the graphics freq is still varying and never really running on max stable. We used GTX 480,580,680 and TITAN and never had a similar issue. In our application we need FULL performance stable, no freq variation. Is this a known issue? It is more than one month ago since the last (and first) driver. When are updates available?
Bumping this thread…
Why is the graphics clock varying, although I set to prefer Maximum Performance?
I noticed this too on MSI 970 Gaming
Switching between auto and performance does change the memory clock but the graphics clock for some reason only go to 1113 instead of maximum ~1500.
Ahhh… then finally I’m not alone with my worries. Thanks for posting! I hope somebody at NVIDIA is listening to this…
Hello?
I think this is expected. The max. performance option locks it into the highest performance level (i.e. locking the memory clock to the highest available), but the GPU Boost dynamic clocking stuff still runs in that mode.
Actually reading about Boost 2 think it is like you say. Looking at Hardcoop review of 970:
MSI GeForce GTX 970 GAMING - Base: 1114MHz / Boost: 1253MHz / Actual: 1366MHz. Default voltage 1.206v on this video card.
Maximum performance stops at 1113/1114 and Graphics Cloc Max does indeed show the maximum current possible boost state. Bit unclear presentation maybe in the UI =)
Also while we are talking about it, is there a way to modify the Power Limit with the nvidia drives like on Windows with MSI Afterburner etc? If not, pretty pretty please =)?
Yes, you can unlock clock range adjustments by setting bit 3 (i.e. 8) in the Coolbits option. Caveat user; please see the giant warning in that section.
Not exactly that.
I have coolbits 12 in Xorg so I got fan control and offset for Core and Memory but the function I ask about is the Powerlimit that limits the boost frequency TDP limit. On Windows with MSI afterburner you can change that between a range that could also help OP here if he does not want to have a boost frequency but always stable. (shown in the left picture) Some cards can raise it but to go further you mostly need to modify BIOS of it. But then it allows you to overclock a bit more.
I know that you cannot change voltage through the drivers, are there any new plans on that btw?
Also thanks for the fast repsonse!
I also had a look into the Coolbits option for setting clock offsets. But this is actually not what I need.
I would actually want to be able to DISABLE GPU Boost, and just have a stable running clock.
And I don’t understand why setting the card to the highest performance level only increases the max clock level. It effectively happens that even on max performance level, my clock sometimes goes down (see screenshot) and my application starts to stutter and cannot hold the vsync of 60Hz anymore.
I consider this more a BUG than a feature for now…
Um what card and what is the temperature of the GPU when it drops? Also are there any XID errors?
I wonder if this could be a downclock because of TDP limit when you stress the card.
I think the temperature was around 90°C . I have no clue what average temperature is normal for a GPU to work happily. Could it be that the fan is not working strong enough?
That seems a bit high, I cannot comment on when these cards starts to down-clock but its “close” I would say.
My 970 reaches 63°C on Load, I know that before it was about 100-110°C when cards started to throttle. But you cannot really compare one generation or vendor to another. There is probably or should be some documentation somewhere.
I checked on some other Cards here (GTX TITAN, GTX 680), the temperature is normally between 60 and 70 degrees even in high Performance.
But my GTX 970 reaches to 90 and more degrees.
There is one important Detail that I noticed. The fan RPM Shows always 0 (Zero). After opening the case, I see that the fan starts Spinning and some seconds later stops again, and this continues over and over.
Is there a cooling issue with this Card?
Danoo
can you check the fan behavior in windows or even a different linux nvidia driver version? your fan should NOT be doing that under any circustance IMO. My GTX 760 running 346.16 fan idles at around 1200RPM. It’s no wonder your temps are as high as they are. Did you check anything related to fan control within the nvidia-settings, did you maybe set a fan variable with some script?
CommanderAlchemy
You can adjust the voltage with coolbits “16” (Bit 4) but can only adjust the voltage thru the nvidia-settings command line interface
The MSI 970 Gaming if that’s the one Danno has, got a passive fan with a target temp ~60c. Until it reaches that temp the fan will stay turned off and the card is cooled passively.
@Danoo, you have to check the temperature when the fan goes to 0, if that is on load with temps around 60 or above i would say that something is not right with the card BIOS or fan.
Also found max temp for this card.
http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-970/specifications
So around 98 the card should throttle down.
@ubuntuaddicted i know that, but what does bit 4 mean?
Coolbits 12 gives you fan-control and overclocking, and 16 gives only voltage control, how do you combine those, just two coolbit option lines in Xorg or?
Edit:
Nvm found a updated wiki about this in Arch:
To enable multiple features, add the Coolbits values together, so 28 would be the value for coolbits to enable fan control (4), overclocking (8), voltage-control (16).
glad you found it, i was the one who made that wiki change. ;) well, not the way it’s formatted but the adding of the coolbits values to enable all the things you want.
Blasted powermizer and other silliness. Here’s the thing Nvidia/AMD/Intel/others, i really didn’t want to pay a substantial amount of money for a gpu to run like crap most of the time, with very little if any real control over these GREEN technologies. My computer ends up using 99 percent of computing power figuring out which devices should be at what performance level and fan, like a control freak…but really out of control. Performance is woeful. Why after paying so much money for a GPU i have to spend months trying to figure out how to turn OFF green features so that it runs almost acceptably on a 2003 opengl game? Still get horrible tearing on a 144hz monitor with all that power and all. It’s crap is what it is. Without special xconfig and quickly changing powermizer setting upon login, xorg gets funny colors and crashes. How much would it cost for a 2003 computer that can still run properly without re-programming the space shuttle to turn off and tweak every part of it to stop the stupid non-predictive power saving features which are all buggy and end up causing more problems than they’re trying to fix. Just make my gtx 770 run 700MHz without throttling/boost/boost2/bla bla bla technologies that sound fantastic on marketing hype but don’t pan out well at all for users.
Before all these crappy technologies we used to have a $100 dollar gpu that worked fantastic day in day out. I bought a $300+ gpu just so it: crashes on login, stutters, tears, runs like shite…and I need a degree in Nvidia university to make it run.
When will i have time to actually use the card?