Personal computing discussed

Moderators: renee, morphine, SecretSquirrel

 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

NVIDIA P-states bug

Tue Dec 03, 2013 2:01 pm

Is anyone else familiar with this?

I used to get this bug on my GTX 460(s) where sometimes they'd get stuck in P1 state, which is the second-fastest state. On my GTX 460s, which were at 822Mhz, it was 411Mhz. Nothing I would do could raise or lower the P-state, until I rebooted.

I assumed they'd fixed it in drivers because it hadn't come up in a long time, but I've just recently noticed my GTX Titan is doing it! And HAS BEEN doing it for awhile now!

What the heck is the deal? Has anyone else had this problem?
 
Deanjo
Graphmaster Gerbil
Posts: 1212
Joined: Tue Mar 03, 2009 11:31 am

Re: NVIDIA P-states bug

Tue Dec 03, 2013 2:08 pm

auxy wrote:
Is anyone else familiar with this?

I used to get this bug on my GTX 460(s) where sometimes they'd get stuck in P1 state, which is the second-fastest state. On my GTX 460s, which were at 822Mhz, it was 411Mhz. Nothing I would do could raise or lower the P-state, until I rebooted.

I assumed they'd fixed it in drivers because it hadn't come up in a long time, but I've just recently noticed my GTX Titan is doing it! And HAS BEEN doing it for awhile now!

What the heck is the deal? Has anyone else had this problem?


I haven't noticed anything on my Titans behaving like this (I do run them in linux however). Are you sure they are actually only running in P1 or is it just what is displayed on the control panel? I would try a 3rd party monitoring program to verify that it is actually being "stuck" in P1 and not just the panel not updating.
 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

Re: NVIDIA P-states bug

Tue Dec 03, 2013 2:25 pm

Deanjo wrote:
I haven't noticed anything on my Titans behaving like this (I do run them in linux however). Are you sure they are actually only running in P1 or is it just what is displayed on the control panel? I would try a 3rd party monitoring program to verify that it is actually being "stuck" in P1 and not just the panel not updating.

Yes, yes, of course. When it happens, my GPU shows 601Mhz clock rate steady in MSI Afterburner/EVGA PrecisionX and Nvidia Inspector, regardless of applications open or active.
 
Arclight
Gerbil Elite
Posts: 768
Joined: Tue Feb 01, 2011 3:50 am

Re: NVIDIA P-states bug

Tue Dec 03, 2013 2:45 pm

What driver version? I would recommend trying 331.93 and report back.
 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

Re: NVIDIA P-states bug

Tue Dec 03, 2013 2:56 pm

Arclight wrote:
What driver version? I would recommend trying the driver version 331.93.
I've used every driver, Beta and WHQL, since 290 series, but I've only confirmed this since 314. I have confirmed it on all since 314, though. It's intermittent, but it seems guaranteed to happen after ~60-70 hours of uptime.

I tend to leave a game running all the time; maybe that's related?

I'm also using my Titan alongside two displays connected to my Intel HD; however, this also happened on my GTX 460s when I wasn't using any other adapters (besides those), and a 260 before that. It was REALLY bad on my 260 where I first noticed it, but they did fix something that was causing it in a later driver (that is, it was in the patch notes.)
 
Arclight
Gerbil Elite
Posts: 768
Joined: Tue Feb 01, 2011 3:50 am

Re: NVIDIA P-states bug

Tue Dec 03, 2013 3:01 pm

auxy wrote:
Arclight wrote:
What driver version? I would recommend trying the driver version 331.93.
I've used every driver, Beta and WHQL, since 290 series, but I've only confirmed this since 314. I have confirmed it on all since 314, though. It's intermittent, but it seems guaranteed to happen after ~60-70 hours of uptime.

I tend to leave a game running all the time; maybe that's related?

I'm also using my Titan alongside two displays connected to my Intel HD; however, this also happened on my GTX 460s when I wasn't using any other adapters (besides those), and a 260 before that. It was REALLY bad on my 260 where I first noticed it, but they did fix something that was causing it in a later driver (that is, it was in the patch notes.)


Have you or have you not tried 331.93? If you have and the issue persists then do one last test by installing the card in another system. If the issue can be reproduced then RMA the card.
nVidia video drivers FAIL, click for more info
Disclaimer: All answers and suggestions are provided by an enthusiastic amateur and are therefore without warranty either explicit or implicit. Basically you use my suggestions at your own risk.
 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

Re: NVIDIA P-states bug

Tue Dec 03, 2013 3:06 pm

Arclight wrote:
Have you or have you not tried 331.93? If you have and the issue persists then do one last test by installing the card in another system. If the issue can be reproduced then RMA the card.

I am using 331.93 now.

Why would I RMA the card? Are you suggesting that it can't possibly be a persistent driver bug and that somehow my card has failed/is failing? Isn't it awfully coincidental that a Tesla arch card, a Fermi arch card (and then a pair of them), and a Kepler arch card have all had the same problem?

I was going through release notes and I found this:
NVIDIA 331.93 release notes wrote:
NVIDIA GPU PhysX acceleration is not available if there is a non-NVIDIA graphics processor in the system, even if it is not used for rendering.
I use GPU-accelerated PhysX all the time despite having Intel graphics enabled. How amusing.
 
ClickClick5
Gerbil Elite
Posts: 581
Joined: Fri Aug 15, 2008 8:28 pm
Location: Alaska
Contact:

Re: NVIDIA P-states bug

Tue Dec 03, 2013 3:34 pm

Arclight wrote:
auxy wrote:
Arclight wrote:
What driver version? I would recommend trying the driver version 331.93.
I've used every driver, Beta and WHQL, since 290 series, but I've only confirmed this since 314. I have confirmed it on all since 314, though. It's intermittent, but it seems guaranteed to happen after ~60-70 hours of uptime.

I tend to leave a game running all the time; maybe that's related?

I'm also using my Titan alongside two displays connected to my Intel HD; however, this also happened on my GTX 460s when I wasn't using any other adapters (besides those), and a 260 before that. It was REALLY bad on my 260 where I first noticed it, but they did fix something that was causing it in a later driver (that is, it was in the patch notes.)


Have you or have you not tried 331.93? If you have and the issue persists then do one last test by installing the card in another system. If the issue can be reproduced then RMA the card.


60-70 hours? How about a restart? Problem solved.
i7-5930k (4.1GHz), 32GB DDR4-2400, 2080Ti, GA-X99-UD3, Samsung 860 Pro 256GB (os/programs), Toshiba 5TB (games), WD VelociRaptor 500GB (games/scratch disk), Win 10 Pro x64
Check out my site for trance sets and other goodies! clickclick5.com!
 
JohnC
Gerbil Jedi
Posts: 1924
Joined: Fri Jan 28, 2011 2:08 pm
Location: NY/NJ/FL

Re: NVIDIA P-states bug

Tue Dec 03, 2013 4:08 pm

There is no "bug". My Titan SC is always idling at 324MHz, using any recently released driver version, beta or WHQL. You either have something (game, program or some OS service) running in background which uses "GPU acceleration" or there is an issue with some setting in your modified BIOS (if you flashed your Titan with it) which may also cause the card to get stuck at certain P-state. Most likely the former, considering that the problem disappears after you reboot.
Gifter of Nvidia Titans and countless Twitch donation extraordinaire, nothing makes me more happy in life than randomly helping random people
 
Arclight
Gerbil Elite
Posts: 768
Joined: Tue Feb 01, 2011 3:50 am

Re: NVIDIA P-states bug

Tue Dec 03, 2013 4:42 pm

auxy wrote:
Are you suggesting that it can't possibly be a persistent driver bug and that somehow my card has failed/is failing? Isn't it awfully coincidental that a Tesla arch card, a Fermi arch card (and then a pair of them), and a Kepler arch card have all had the same problem?[...]


Unless all GTX Titan cards out there behave the same as yours with the same driver then it's a defective card. There is no guess work. The card is not working as advertised therefore the only permanent fix is to RMA it.

ClickClick5 wrote:
60-70 hours? How about a restart? Problem solved


That's not an insane amount of time for someone that uses their desktop daily. That's a week or a week and a half at most of casual daily use.

Nevermind, i thought you were refering to the testing period...Even so when you have a top of the line card like the GTX Titan you expect it to work perfectly all the time. Only 60-70 hours of normal operation is not acceptable for a consumer card that costs $1000.
nVidia video drivers FAIL, click for more info
Disclaimer: All answers and suggestions are provided by an enthusiastic amateur and are therefore without warranty either explicit or implicit. Basically you use my suggestions at your own risk.
 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

Re: NVIDIA P-states bug

Tue Dec 03, 2013 7:52 pm

JohnC wrote:
There is no "bug". My Titan SC is always idling at 324MHz, using any recently released driver version, beta or WHQL. You either have something (game, program or some OS service) running in background which uses "GPU acceleration" or there is an issue with some setting in your modified BIOS (if you flashed your Titan with it) which may also cause the card to get stuck at certain P-state. Most likely the former, considering that the problem disappears after you reboot.

Hey John! Wanna try reading the thread and posting again?
 
JohnC
Gerbil Jedi
Posts: 1924
Joined: Fri Jan 28, 2011 2:08 pm
Location: NY/NJ/FL

Re: NVIDIA P-states bug

Tue Dec 03, 2013 10:42 pm

I did read your thread, did you read mine? :roll: Once again, my Titan always switched "states" properly, using stock BIOS and any driver version or compatible OS (Win7, Win8 or Win8.1). And I don't know anyone else who had such issues with their Titan cards. Unless they modded their card's settings, like the person from my previous link did. So once again, if your current card is getting "stuck" at certain state, there are only 2 possible causes:
1) You may have some background software which causes the GPU to get "stuck" at that state, this may also include multi-monitor adjustment programs or various redundant overclocking/monitoring crap, which may interfere with each other (or override each other's settings) in some weird way.
2) Your non-stock adjustments to clock speed or power/voltage limiter through BIOS modifications (you said it before that you flashed a custom BIOS to it) may cause this, just like they did to other Titan owners (once again, see the link in my previous post).

Arclight wrote:
If the issue can be reproduced then RMA the card.

Oh, it can be reproduced very easily. For example, if I will intentionally enable "K-Boost" setting in EVGA Precision I can get my GPU "stuck" at about 575MHz clock speed, even if it is idling or running any other program. Of course, such adjustment (intentional or accidental, using overclocking apps or even a BIOS reflashing) is perfectly reversible and does not warrant the RMA. :wink:
Gifter of Nvidia Titans and countless Twitch donation extraordinaire, nothing makes me more happy in life than randomly helping random people
 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

Re: NVIDIA P-states bug

Wed Dec 04, 2013 12:03 am

JohnC wrote:
I did read your thread, did you read mine? :roll:
I'm not talking about lower p-states -- that is, lower clocks and power-saving modes. I'm talking about p0 state. My problem is that I can't get the card to go into p0 state (to raise clock above ~600Mhz) without rebooting. It also doesn't go into the lower p-states (100Mhz etc) but I'm not that worried about that.

How exactly do my modifications to my Titan affect my older GPUs that had this same problem?
auxy wrote:
Isn't it awfully coincidental that a Tesla arch card, a Fermi arch card (and then a pair of them), and a Kepler arch card have all had the same problem?
The only overclocking tool I use is MSI Afterburner, and in any case, background apps shouldn't keep my GPU from going up to P0 state when I am running a fullscreen game. It would make sense if they kept it from dropping to a lower p-state, but that's not my problem.

That's why I said you didn't read the thread -- you didn't understand the problem (when I feel I explained it perfectly well already.)
ClickClick5 wrote:
60-70 hours? How about a restart? Problem solved.
Do you restart your desktop every few days? I sure don't. I prefer to leave it up constantly. When I reboot, I have to restart a bunch of apps and put all my windows exactly the way I like them, so rebooting is an inconvenience for me, and I'd prefer to avoid it. Aside from this stupid NVIDIA ... thing, there's no reason for me to reboot.
 
Convert
Grand Gerbil Poohbah
Posts: 3452
Joined: Fri Nov 14, 2003 6:47 am

Re: NVIDIA P-states bug

Wed Dec 04, 2013 12:30 am

I'm not familiar with the flash process on newer cards, is it a hassle to put it back to a stock firmware just to rule that out? I know it used to be a PITA.

I never had that problem with my 460's, the problem I had is that the fans wouldn't ramp past 60% under some workloads so the system would BSOD after hitting 80?C (somewhere around there, maybe it was 90C). That was pretty annoying!

Doing some research I saw some people posting similar issues where it won't ramp up clock speed (other nvidia cards) and once they put them back to factory clocks it started working normally. Something else to try.

Who reboots their computer anymore? Mine only gets rebooted for patch Tuesday.
Tachyonic Karma: Future decisions traveling backwards in time to smite you now.
 
JohnC
Gerbil Jedi
Posts: 1924
Joined: Fri Jan 28, 2011 2:08 pm
Location: NY/NJ/FL

Re: NVIDIA P-states bug

Wed Dec 04, 2013 12:47 am

auxy wrote:
How exactly do my modifications to my Titan affect my older GPUs that had this same problem?

Simple - you used same combination of crappy tools (Afterburner/Precision/Inspector/whatever) which may interfere with each other and/or cause certain settings to be permanently stuck. Read up this thread for an example (it's not exact but close enough):
http://www.overclock.net/t/1363440/nvid ... club/10140
Post #10145 and especially the post #10148.

You CAN get a GPU to "stick" to certain fixed frequency, such as 600MHz and stay there (meaning it will NEVER switch to any other P-state, "high" or "low"), regardless if GPU is idling OR under the FULL LOAD, just by screwing up with overclocking tools (and possibly other similar tools). I can easily reproduce it using my own Titan by intentionally "screwing it up" - it will stay at exactly 575MHz even when running FurMark or playing BF4. I can still undo it, though, and such thing NEVER happens at stock settings with no crappy utilities running in background.
Gifter of Nvidia Titans and countless Twitch donation extraordinaire, nothing makes me more happy in life than randomly helping random people
 
ClickClick5
Gerbil Elite
Posts: 581
Joined: Fri Aug 15, 2008 8:28 pm
Location: Alaska
Contact:

Re: NVIDIA P-states bug

Wed Dec 04, 2013 8:10 am

auxy wrote:
ClickClick5 wrote:
60-70 hours? How about a restart? Problem solved.
Do you restart your desktop every few days? I sure don't. I prefer to leave it up constantly. When I reboot, I have to restart a bunch of apps and put all my windows exactly the way I like them, so rebooting is an inconvenience for me, and I'd prefer to avoid it. Aside from this stupid NVIDIA ... thing, there's no reason for me to reboot.

:lol: :lol: :lol: The amount of first world issues here is just amazing. You running a SSD? So about a 7 second reboot? Oh lordy and having to drag that app to screen two the way I like it, heavens! Just unacceptable! Satan himself could not do worse! OH THE HUMANITY!
i7-5930k (4.1GHz), 32GB DDR4-2400, 2080Ti, GA-X99-UD3, Samsung 860 Pro 256GB (os/programs), Toshiba 5TB (games), WD VelociRaptor 500GB (games/scratch disk), Win 10 Pro x64
Check out my site for trance sets and other goodies! clickclick5.com!
 
jihadjoe
Gerbil Elite
Posts: 835
Joined: Mon Dec 06, 2010 11:34 am

Re: NVIDIA P-states bug

Wed Dec 04, 2013 9:20 pm

To be fair to auxy, I wouldn't see it as a simple first world issue thing, but more of "something is wrong with my computer!" I dunno about everyone else, but when things aren't exactly the way they should be on my own desktop it drives me up the wall.

@auxy
I'd consider what JohnC said. Have you tried killing all of the OC tools and using just GPU-Z to monitor your card?
Failing that maybe consider killing the game in the background.
 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

Re: NVIDIA P-states bug

Thu Dec 05, 2013 7:28 am

Okay, well, anyway, I can't be bothered to keep restating things if people won't read. Thanks for your replies everyone. This thread served well enough to answer the question that "no, nobody on Techreport knows what I'm talking about".
 
Arclight
Gerbil Elite
Posts: 768
Joined: Tue Feb 01, 2011 3:50 am

Re: NVIDIA P-states bug

Thu Dec 05, 2013 7:40 am

auxy wrote:
Okay, well, anyway, I can't be bothered to keep restating things if people won't read. Thanks for your replies everyone. This thread served well enough to answer the question that "no, nobody on Techreport knows what I'm talking about".


Shots fired.

It's not our fault your card has issues. You should receive way better support on nvidia's forums, try to post there.

https://forums.geforce.com/
nVidia video drivers FAIL, click for more info
Disclaimer: All answers and suggestions are provided by an enthusiastic amateur and are therefore without warranty either explicit or implicit. Basically you use my suggestions at your own risk.
 
JohnC
Gerbil Jedi
Posts: 1924
Joined: Fri Jan 28, 2011 2:08 pm
Location: NY/NJ/FL

Re: NVIDIA P-states bug

Thu Dec 05, 2013 8:38 am

Good luck reading the same answers at different places.
Gifter of Nvidia Titans and countless Twitch donation extraordinaire, nothing makes me more happy in life than randomly helping random people
 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

Re: NVIDIA P-states bug

Thu Dec 05, 2013 11:40 am

Arclight wrote:
It's not our fault your card has issues.
You are dumb. It's not my card. Please stop posting. Or breathing.
JohnC wrote:
Good luck reading the same answers at different places.
I'm just going to deal with it. It's not THAT big a deal restarting every few days. It's just annoying. As someone else said above, it's not "first world problems", it's just that things aren't working as they should and that frustrates and irritates me.
 
ClickClick5
Gerbil Elite
Posts: 581
Joined: Fri Aug 15, 2008 8:28 pm
Location: Alaska
Contact:

Re: NVIDIA P-states bug

Thu Dec 05, 2013 1:05 pm

auxy wrote:
Arclight wrote:
It's not our fault your card has issues.
You are dumb. It's not my card. Please stop posting. Or breathing.


auxy wrote:
Is anyone else familiar with this?

I used to get this bug on my GTX 460(s) where sometimes they'd get stuck in P1 state, which is the second-fastest state. On my GTX 460s, which were at 822Mhz, it was 411Mhz. Nothing I would do could raise or lower the P-state, until I rebooted.

I assumed they'd fixed it in drivers because it hadn't come up in a long time, but I've just recently noticed my GTX Titan is doing it! And HAS BEEN doing it for awhile now!

What the heck is the deal? Has anyone else had this problem?


But..but you say it right there... I even made it bold and underlined it.
i7-5930k (4.1GHz), 32GB DDR4-2400, 2080Ti, GA-X99-UD3, Samsung 860 Pro 256GB (os/programs), Toshiba 5TB (games), WD VelociRaptor 500GB (games/scratch disk), Win 10 Pro x64
Check out my site for trance sets and other goodies! clickclick5.com!
 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

Re: NVIDIA P-states bug

Thu Dec 05, 2013 1:17 pm

If the card is the problem, why did it happen on three other cards, across two more (three total) generations of cards...? (in different motherboards, different processors, different operating systems...)
 
ClickClick5
Gerbil Elite
Posts: 581
Joined: Fri Aug 15, 2008 8:28 pm
Location: Alaska
Contact:

Re: NVIDIA P-states bug

Thu Dec 05, 2013 1:28 pm

auxy wrote:
If the card is the problem, why did it happen on three other cards, across two more (three total) generations of cards...? (in different motherboards, different processors, different operating systems...)


I could make an Nvidia joke but I will leave with this: No matter the company, drivers are a pain to get flawless. Nvidia or AMD, there ARE BUGS in both, and always will. A $40 card or $999 card, there will always be bugs. With the cards working with heat throttling these days, it could be a firmware issue? You might can try finding a firmware mod tool for the Titan and tweak the thermal cutoff? I modified the thermal range on a Radeon 4870 back in the day to solve the fan ramp issue it had. If it IS a driver issue, you just have to wait it out with Nvidia sir.

Also, I remember Nvidia having some other time based bug long ago that ran from the 2xx to the 5xx series I think...
i7-5930k (4.1GHz), 32GB DDR4-2400, 2080Ti, GA-X99-UD3, Samsung 860 Pro 256GB (os/programs), Toshiba 5TB (games), WD VelociRaptor 500GB (games/scratch disk), Win 10 Pro x64
Check out my site for trance sets and other goodies! clickclick5.com!
 
auxy
Graphmaster Gerbil
Topic Author
Posts: 1300
Joined: Sat Jan 19, 2013 4:25 pm
Location: the armpit of Texas

Re: NVIDIA P-states bug

Thu Dec 05, 2013 1:45 pm

It's a firmware bug that's persisted through three generations of graphics architectures?
 
Ryu Connor
Global Moderator
Posts: 4369
Joined: Thu Dec 27, 2001 7:00 pm
Location: Marietta, GA
Contact:

Re: NVIDIA P-states bug

Thu Dec 05, 2013 2:01 pm

auxy wrote:
You are dumb. It's not my card. Please stop posting. Or breathing.


Telling people to die is grounds for a timeout. This thread is done.
All of my written content here on TR does not represent or reflect the views of my employer or any reasonable human being. All content and actions are my own.

Who is online

Users browsing this forum: No registered users and 1 guest
GZIP: On