Jul 09, 2016 I already know the issue happens when using multiple monitors with different refresh rates. My question is since this is a 60Hz Freesync panel, can there be a driver update to make it so that the lower refresh Freesync capable monitors can be used with multiple monitors at the default Idle Memory Clock of 300Mhz? Jun 18, 2018 Memory clock in MHz or GHz, is an indicator of how high of throughput it can sustain but full calculation also needs total bit number too. Higher frequency means higher throughput on same memory. Core clock of a GPU indicates throughput too. But this thime it needs number of pipelines and some other complex equations to fully guess real throughput.
Greetings,I'm having issue with high GPU temperature and clock speed on idle. I don't understand why. I turned off my computer and after 3 or 4 hours after turned on again. But it's starting with high GPU temp and clock/memory speed. It's like a game is running on background but no. My CPU temp and useage is normal.Now I'm downloading NVIDIA driver and will uninstall current one, then install the new driver that I downloaded.But if it's not neccesary, what should I do?
I added a screenshot below.GPU-Z. 58 on idle is high. I do not know if it's considered 'normal' for your specific model (that's the MSI one, right?), if it's you altering the fan curves, or if it's just time taking it's toll; but for idle, it's high.You should experiment with fan RPM and/or have a look at the actual fans, see their condition. Since you mentioned it, i doubt this has anything to do with Nvidia's drivers.
Next time you post, do state whether this is a recent occurence/you just noticed/card is new, etc?One general bit of advice however.Everybody, their mom, da, granda and this site included will tell you what a wonderful thing software that turns fans off is; it's not.The fastest way to kill or at best 'tire' your fans is to do exactly that, have them going on and off a couple of dozen times a day just. Because.Find a nice sweetspot between noise and heat and manually leave them there. Fixed.They will last you a hell of a lot longer.They will not (assuming they're of any decent quality) eventually start rattling.They will not (assuming they're of any decent quality) 'degrade'. At least not before it's time for a new GPU/CPU/whatever.And of course, cleaning. 58 on idle is high. I do not know if it's considered 'normal' for your specific model (that's the MSI one, right?), if it's you altering the fan curves, or if it's just time taking it's toll; but for idle, it's high.You should experiment with fan RPM and/or have a look at the actual fans, see their condition. Since you mentioned it, i doubt this has anything to do with Nvidia's drivers.

Next time you post, do state whether this is a recent occurence/you just noticed/card is new, etc?One general bit of advice however.Everybody, their mom, da, granda and this site included will tell you what a wonderful thing software that turns fans off is; it's not.The fastest way to kill or at best 'tire' your fans is to do exactly that, have them going on and off a couple of dozen times a day just. Because.Find a nice sweetspot between noise and heat and manually leave them there.
Fixed.They will last you a hell of a lot longer.They will not (assuming they're of any decent quality) eventually start rattling.They will not (assuming they're of any decent quality) 'degrade'. At least not before it's time for a new GPU/CPU/whatever.And of course, cleaning. ProcessorIntel Core i7 5960X - 4.5 GhzMotherboardAsus X99-E WSCooling360 Monsta + 360 ST30 9 White EK VardarMemory32GB Corsair LPX 3000MhzVideo Card(s)EVGA GTX 1080 Ti FTW3StorageIntel 750 400GB, Intel 750 1.2TBDisplay(s)Asus PG279Q 2 x Dell U2412MCasePhanteks Enthoo Primo SE Black/WhiteAudio Device(s)Schiit Jotunheim - Xonar Essence One Fostex TH-900, Fidelio X2 - KRK Rokit RP-6 G3Power SupplyCorsair AX1200i - Black/White Cablemod SleevesMouseLogitech G303 Daedalus ApexKeyboardDucky Shine 1 - MX BrownSoftwareWindows 10.
Gpu Memory Clock Too High School

Gpu Memory Clock Too High Top
New Here?Please keep in mind that we are here to help you build a computer, not to build it for you. Rules.
Be respectful to others. Follow the and of posting. No piracy or grey-market software keys. Limit posts to one every 24 hours. Do not use all-caps post titles.
Link shorteners are not permitted. No intentionally harmful or joke advice. No offering rewards to solicit help.ResourcesChoosing Parts/Components:.The Build:.After Building:.Filters.Related Reddits. I have the Asus Strix GTX 960 2GB. After boosting up my once anemic CPU with OC'ing, I learned GPU's are less resilient to adding voltage and OC'ing in general in terms of lifespan.So I am more or less following the setup of these guys, who overclocked my card 100mHz core and 1000mHz memory without overvolting!What really worries me is the 1000mHz memory overclock, as everywhere on the internet for other cards, the memory oc is much lower even after raising voltage. And also people say that you can't tell if your memory is running too fast b/c it doesn't give the same visual errors as an unstable core clock, causing people to fry their memories even with stable cores and temps.
Core Clock Gpu
And my temps are very stable, around 60C when stress testing with MSI Kombuster.So is there a way to see that you are turning your memory up too fast? As I don't want to blow up my card for it, especially since it supposedly doesn't improve performance as much as the core clock does. Causing people to fry their memories even with stable cores and temps.You basically can't kill GDDR5 by overclocking it. Power consumption scales linearly with clockspeed, so a boost like that shouldn't hurt the memory VRMs either.What can be dangerous is overvolting memory. Which you generally shouldn't do unless you're trying to get every last drop of performance out of your card. It's tricky b/c you have to watch memory VRM temperatures with something like HWInfo.The press (aka legitreviews) tends to get cherry-picked GPUs that overclock better than your average off-the-shelf GPU. So don't expect to get that same crazy memory OC:POCCT has an error-checking test that should show bad memory overclocks very quickly.

But generally speaking, if you get artifacts/crashes in your games, just bump the memory clock down 100mhz and try again.