weird power limit forced cap issue on 1080ti
So after some tuning I got a voltage curve set that prevents driver resets or OS crashes at 1974mhz.
However randomly after some hours of gaming on the profile the power limit will be forced to a lower cap, when this behaviour occurs the usual power limit action doesnt happen so e.g. the clock speed does not reduce, instead the gpu will report 100% usage but at a lower performance level, I hope this makes sense.
When this happens I have to reboot windows, then normal behaviour comes back.
I expect something is happening the GPU isnt liking causing the behaviour. Is it possible its detecting a lack of juice from the PSU so power limit throttling or would this be due to unstable voltage/clock combo?
Usually it restricts to about 70% power usage when it happens.
you mean, your clock speeds gets lower? like in steps of -13mhz? thats normal, cause ur gpu temp goes up, so card clock down a bit, to hold temps in check.
sounds like how boost is designed to work.
minus the needing to restart - unless the restart is long enough to let it cool down enough to clear the downclock.
There are more than one Silent Pro series from Cooler Master so which one do you have?
By far the worst one is the Silent Pro Gold which are god awful so if you have one of those replace ASAP
Edit: if you have a 600 watt unit its either a Silent Pro M or a Silent Pro Gold
The Gold as said is junk and the Silent Pro M is 11 years old and the 12v rail is only rated for 480 watts
In either case you will need a new PSU
I will try and explain it again in as simple terms as possible.
When boost works normally and hit power limit this is what usually happens.
Power limit shown by software is close to power limit set on card so e.g. if set to 120% you would see 110-120%.
Clock speed reduces to reduce power load to stay within limit.
Thats not whats happening.
Instead power usage reported by software shows much much lower than the limit, it doesnt happen immediatly it takes several hours to start happening.
So e.g. if limit set to 120% it will top out at about 70%.
Clock speed will stay at max speed.
But performance will drop.
Basically the 100% utilisation is at a lower level, maybe some of the cuda cores get disabled or something to reduce the power draw, and performance whilst not reducing clock speed.
So e.g. when this occurs I get 50-150fps in unigine heaven, but then I reboot to bring back proper behaviour I get 250fps. Division 2 works flawless 60fps 60-90% load when card works properly, when it starts occurring I get screen tearing as it then cant sustain 60fps even in simple scenes and lower resolution. I have to reboot to restore proper behaviour.
My thoery is the card is detecting lack of capability from the psu possibly (so I am inclined to think yes I need a bigger psu, I was already aware of psu limitation when getting the card which is why I dont set the max 150% limit), and there is some kind of programming on the card that when this is detected it disables a chunk of the cores, as the only rational explanation for what I am seeing is less cores are active, same clock speed, but lower power usage and performance.
I wanted to be absolutely sure tho because changing the PSU is a massive pain in the backside. Its also a fair chunk of change to pay for it as well and if I am wrong its a lot of money to lose on a theory.
This profile is 1.031v 1974mhz 120% power limit, which I think on the card is a 300watt limit.
When I use my low power profile for weaker games I set the limit to 90% so 225 watt limit, and that uses 0.875v for 1846mhz. That profile doesnt get the problem.
The 12v rail has 480 watts capacity so headroom should be there.
I see that newer PSUs at 650watts, even tho only an extra 50 watts allow much higher power on 12v rail like 54amps 648watts. Interesting.
Your PSU is a 11 years old so its ancient and also long out of warranty and it probably cant do its full rated wattage anymore
A Seasonic Focus Plus or a Prime Ultra are both good options
Have you ever heard of this issue before tho with seemingly some cuda cores turning off?
I would expect a crash instead.
do you have by any chance aorus 1080ti extreme? i know one user having exact same issue like you. he had tv connected to the card along with monitor, we never actually pin point whats causing the issue but he suspect it happens cause tv is plugged in and that causes some kind of conflict.
it may also be problem with ur PSU as others point out, never ever take crap PSU and pair it with high end GPU/CPU, u will have only headaches.
if you limit ur boosting to ~1800mhz and 0,8xx mV, no need to lower the power limit, on such low voltages, you card cannot pull more then ~150-180w, try benching with core clock on 1700mhz @ around 0,800mV and see if u get same performance issue then, if it not happens, it may really be PSU issue.
its the non extreme model, and also no tv connected.
New antec HCG 750watt arriving tomorrow (rebranded seasonic focus plus), so I will report if the issue goes away.
guru3d in one of their reviews actually noted this specific problem can occur with a psu incapable of delivering the voltage.
I believe what is happening some cores are getting turned off when a certain failure condition is detected probably as a failsafe measure. Sort of like how the ps4 pro turns off some cores in games that dont have ps4 pro enhanced modes.
|All times are GMT -7. The time now is 08:55 AM.|
Powered by vBulletin® Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.
User Alert System provided by Advanced User Tagging (Pro) - vBulletin Mods & Addons Copyright © 2020 DragonByte Technologies Ltd.
vBulletin Security provided by vBSecurity (Pro) - vBulletin Mods & Addons Copyright © 2020 DragonByte Technologies Ltd.
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2020 DragonByte Technologies Ltd.