Overclock.net banner
401 - 420 of 443 Posts

·
Registered
Joined
·
381 Posts
Yes, we have extensively checked to make sure our settings are the same. I have even set my system to run at 2560 x 14440 even though my monitor supports 3440 x 1440 ss we are running the same resolution. Same thing happens in Heaven and 3DMark.

I have 3 x 140mm fans at the front, and 4 x 120mm fans exhausting air through my 360mm rad and through the back of the case. I don't think airflow is an issue here. I have set my system to calibrate the fans and temperatures in BIOS, and the fans hardly ever ramp up their RPM because the case temp never gets very high unless I am benching the CPU and GPU at the same time.

His case is using a 200mm fan at the front, and three 120s in the top and one in the rear. His case is a little smaller and had drive bays which my case doesn't have. So, the airflow should be a little better in my case.

I am going to try the GPU fan curve and see how much that drops temps.

It's not a desperate issue or anything as 66c is fine for gaming, but it would be nice to drop it further.

As I did write, from my guess, even how paste was applied (stock) or how really proper pressure is applied on cooler to chip, can play difference in some degress.

custom fan curve shoud bring temps down, on expense of some noise (for me, it is not much noise).
 

·
Registered
Joined
·
147 Posts
just by moving voltage slider to 100% (need new Beta MSI Afterburner) it can push up to 1.100V. it is not going to induce much more temperature. but also Mhz gains are not any big.

you want to try more than 1.100V. I personally would not go beyond that.

if you can manage to get it cooled little better (as we tested with Kryonaut), it can also help just a little bit
OK thanks, got it. Yeah I saw 1.100v at most, so thats the limit and no need to change any voltage options in afterburner.

Im at +150MHz core, seeing 2100MHz stable in CP2077 at 73 to 77 degrees. Can I expect more?

I didnt OC the memory yet as nvidia doesnt recommend it due to the nature of gddr6x. default, memory temp is max 86 degrees. Is it good? if i want to overclock it, what values should i expect at what max safe temp?

hottest area is 91 degrees btw, reported by gpuz.
 

·
Registered
Joined
·
381 Posts
OK thanks, got it. Yeah I saw 1.100v at most, so thats the limit and no need to change any voltage options in afterburner.

Im at +150MHz core, seeing 2100MHz stable in CP2077 at 73 to 77 degrees. Can I expect more?

I didnt OC the memory yet as nvidia doesnt recommend it due to the nature of gddr6x. default, memory temp is max 86 degrees. Is it good? if i want to overclock it, what values should i expect at what max safe temp?

hottest area is 91 degrees btw, reported by gpuz.


From what I read, allegedly Micron states GDDR6X up to 95C as safe.

from my test, running no overclock on memory, and +1000, increase exactly 4C (but I replace thermal pads for memory, which cut off few degrees, but nothing extreme). so I belive +1000 is ok, or little less, maybe not every memory is exactly same. best test directly in some game, and watch if there is improvement. after +1000 (in my case there is still gains but smaller, so probably correction start to kick in, which you don't want). overall gains from overclocking memory are not big, from what I tested 3-4 fps in GPU demanding game.

we are already discussing here custom fan curve. if you have no problem with it, you can give it a try, it will reduce all temps (including memory), so you should be fine even with overclock on memory.
replacing stock TIM with Kryonaut provide some improvement (but not big one). you can expect there +15Mhz on core if all settle well :)
from what I tested - custom fan curve + Kryonaut + increased voltage slider (so it can push up to 1.100V) you can get +30Mhz, or in best case + 45Mhz
 

·
Registered
Joined
·
147 Posts
From what I read, allegedly Micron states GDDR6X up to 95C as safe.

from my test, running no overclock on memory, and +1000, increase exactly 4C (but I replace thermal pads for memory, which cut off few degrees, but nothing extreme). so I belive +1000 is ok, or little less, maybe not every memory is exactly same. best test directly in some game, and watch if there is improvement. after +1000 (in my case there is still gains but smaller, so probably correction start to kick in, which you don't want). overall gains from overclocking memory are not big, from what I tested 3-4 fps in GPU demanding game.

we are already discussing here custom fan curve. if you have no problem with it, you can give it a try, it will reduce all temps (including memory), so you should be fine even with overclock on memory.
replacing stock TIM with Kryonaut provide some improvement (but not big one). you can expect there +15Mhz on core if all settle well :)
from what I tested - custom fan curve + Kryonaut + increased voltage slider (so it can push up to 1.100V) you can get +30Mhz, or in best case + 45Mhz
Makes sense. So I'll test +500MHz on memory now, which should be reasonably safe regarding people do +1000. My custom fan curve is at max 90% speed from 70-75 degrees, 100% speed makes fans sound like they are struggling.

What's the max. speed you see on core in games?
 

·
Registered
Joined
·
381 Posts
Makes sense. So I'll test +500MHz on memory now, which should be reasonably safe regarding people do +1000. My custom fan curve is at max 90% speed from 70-75 degrees, 100% speed makes fans sound like they are struggling.

What's the max. speed you see on core in games?

2115 with custom fan curve, without added voltage, in most demanding ones. less demanding ones can run up to 2175 with increased voltage. 2190 is not possible in any case.
2115 will never crash
 

·
Registered
Joined
·
1 Posts
Oof, I just checked my temps after playing a round of COD Cold War and I'm actually hitting 78F.
(screenshot below was taken just after exiting the game, but I had it set to only show the highest GPU temp)




I took my side panel off and then tried again but temps and noise were exactly the same. My case fans are from 2012-2013... I have 3x 120mm Corsair AF120 Quiet Edition's in the front at max speed and in the rear a stock fan from a Corsair 550D running at max speed.
I know the cooler is only 2 slot but is something wrong here?
I think you need more exhaust fans. With the lid closed, is it hot to touch at the Evga glow area when closed?
 

·
Registered
Joined
·
147 Posts
2115 with custom fan curve, without added voltage, in most demanding ones. less demanding ones can run up to 2175 with increased voltage. 2190 is not possible in any case.
2115 will never crash
Ok i upped the core to +165 and getting 2115MHz in mid game now.

my power cons. highest is 330W. I wonder those special BIOS cards that can go 380W, actually do more than 2115MHz?
 

·
Registered
Joined
·
381 Posts
Ok i upped the core to +165 and getting 2115MHz in mid game now.

my power cons. highest is 330W. I wonder those special BIOS cards that can go 380W, actually do more than 2115MHz?


that is reason why we opt for Asus Strix. usually Strix model will have very high power limit, but also everything else to cover it - VRM, power phases, cooling, etc. (there is of course also other top models with such thing). and it have 3X 8-pin power connectors.

you can flash BIOS from other card. at least I was doing that on GTX cards many times, did not try on RTX however.
if you are able to flash BIOS with higher power limit, that it is question exactly things I mentioned above, are they going to suffice and handle more power usage (assuming high stress and load). I never killed GPU, not even with XOC BIOS.

Samsung 8nm is not really best what it could be, so higher power consumption on 3XXX.

not every chip is binned same (and not cooled same, due to number of other factors, some of them I mentioned in earlier posts). so what is your stable limit only you can test. if crash, go one step lower. or try to cool it better :)

no matter if some say it doesn't matter much - it is - temperature, for stability. what is stable at 60C may be crashing every now at 70C. it is not bad thing to do some tweaks to improve cooling, or even watercool any card, but in my opinion, only if you intend to keep it for longer time. there is gains, but not big one. with watercooling it is going to be more, but again not something which will make huge difference.
and there is partially answer to your question. mostly how much you are going to hit stable is silicone quality of chip.
and that have nothing with BIOS and power limit.

what cards (bios) with higher power limit is - sustained high clock speed (not hitting power limit).
once you hit power limit, card will downclock

so for example lets look at this scenario.
one user have better binned chip but with lower power limit, so it can run perfect stable 2130Mhz. but if hit power limit, card will downclock
another user have little worse binned chip, so it can run 2100Mhz under same stress. but as power limit is higher, it will not downclock (or will downclock less) in scenarios when you hit power limit. resulting in increased performance, despite worse binned chip.

best of both worlds would be to have good binned chip, higher power limit (and everything else to cover it), and cool card best you can. than you can get little more Mhz and have it sustained under higher load - higher power consumption
 

·
Registered
Joined
·
147 Posts
that is reason why we opt for Asus Strix. usually Strix model will have very high power limit, but also everything else to cover it - VRM, power phases, cooling, etc. (there is of course also other top models with such thing). and it have 3X 8-pin power connectors.

you can flash BIOS from other card. at least I was doing that on GTX cards many times, did not try on RTX however.
if you are able to flash BIOS with higher power limit, that it is question exactly things I mentioned above, are they going to suffice and handle more power usage (assuming high stress and load). I never killed GPU, not even with XOC BIOS.

Samsung 8nm is not really best what it could be, so higher power consumption on 3XXX.

not every chip is binned same (and not cooled same, due to number of other factors, some of them I mentioned in earlier posts). so what is your stable limit only you can test. if crash, go one step lower. or try to cool it better :)

no matter if some say it doesn't matter much - it is - temperature, for stability. what is stable at 60C may be crashing every now at 70C. it is not bad thing to do some tweaks to improve cooling, or even watercool any card, but in my opinion, only if you intend to keep it for longer time. there is gains, but not big one. with watercooling it is going to be more, but again not something which will make huge difference.
and there is partially answer to your question. mostly how much you are going to hit stable is silicone quality of chip.
and that have nothing with BIOS and power limit.

what cards (bios) with higher power limit is - sustained high clock speed (not hitting power limit).
once you hit power limit, card with downclock

so for example lets look at this scenario.
one user have better binned chip but with lower power limit, so it can run perfect stable 2130Mhz. but if hit power limit, card will downclock
another user have little worse binned chip, so it can run 2100Mhz under same stress. but as power limit is higher, it will not downclock (or will downclock less) in scenarios when you hit power limit. resulting in increased performance, despite worse binned chip.

best of both worlds would be to have good binned chip, higher power limit (and everything else to cover it), and cool card best you can. than you can get little more Mhz and have it sustained under higher load - higher power consumption
thanks for the detail. that makes sense, gpuz reports the limit explanations (vrel, power, temp) that why the card doesnt boost etc.

so in my case, i just leave it be. as you mentioned i plan to use this card in coming years, no need for watercooling or bios flash.
 

·
MOD King
Joined
·
263 Posts
I ended up undervolting my Strix 3070 Ti to 1950MHz @ .930mv and it made a huge difference on temps. That's essentially the same factory oc out of the box with ~100w less power draw. Normal temps are 65c and up to 70c tops now, and memory temp maxes at 82c. Memory was maxing at 90c before undervolting. I could probably tweak it a bit more, but I'm actaully leaving the club soon as I came across a 3080 Ti FE at a mostly normal price.
 

·
Registered
Joined
·
289 Posts
I ended up undervolting my Strix 3070 Ti to 1950MHz @ .930mv and it made a huge difference on temps. That's essentially the same factory oc out of the box with ~100w less power draw. Normal temps are 65c and up to 70c tops now, and memory temp maxes at 82c. Memory was maxing at 90c before undervolting. I could probably tweak it a bit more, but I'm actaully leaving the club soon as I came across a 3080 Ti FE at a mostly normal price.
Well done, and congrats on the 3080 Ti
 

·
Registered
Joined
·
3,019 Posts
As I did write, from my guess, even how paste was applied (stock) or how really proper pressure is applied on cooler to chip, can play difference in some degress.

custom fan curve shoud bring temps down, on expense of some noise (for me, it is not much noise).
I was playing with my buddy last weekend. I turned the fans on the card up to 100% and verified they were running at maximum RPM. It made almost no difference in temps. That going from a stock fan curve to running all three fans at 100%.

The card is working very well, and it seems to peak at about 65c when gaming, so I think I won't worry about it until the next time I do a major maintenance. Then I will remove the cooler and apply new paste.
 

·
Registered
Joined
·
1 Posts
Hi,

I've an EVGA RTX 3070 Ti XC3 GAMING with the latest firmware (94.04.5A.00.F9) installed via Precision X1. I'm unable to reduce the gpu voltage below 818mV via MSI Afterburner (4.6.4 beta 3) nor with Precision X1. I'm reading somewhere that one can do 850mV with gpu clock around 1800-1900mhz. Below is the voltage-freq and gpu-z screenshot. I can't seem to force the gpu voltage lower than 818mV even though the gpu clock is already running quite low at 1455mhz.

Anyone has any idea?

Thanks in advance.



2520721



2520720
 

·
Registered
Joined
·
168 Posts
What's the max memory overclock any of you have attained on a 3070 Ti? I'm up to +1500 on my MSI card. I'll add that I'm hashing on it and am not experiencing errors, along with increased hashrate, so I don't think I'm into ECC territory.
 

·
Registered
Joined
·
381 Posts
What's the max memory overclock any of you have attained on a 3070 Ti? I'm up to +1500 on my MSI card. I'll add that I'm hashing on it and am not experiencing errors, along with increased hashrate, so I don't think I'm into ECC territory.

I also get benefits all way up to +1500
but I replaced thermal pads with Thermalright Odyssey 12.8 w/mk
however, after +1200 or maybe +1250 gains are smaller, but still there are some
some users here claimed and showed performance degradation, at some point after +1000
now why I'm not expiriencing same, maybe good temperature are reason, or maybe just luck with memory modules, have no idea
 

·
Registered
Joined
·
381 Posts
now please one question
I'm in confusion with how to turn off RGB lights on my 3070 TI, it is Asus Strix. if you ask me, I personally do not like RGB, and especially not on Strix 3070 TI :)

I installed Armoury Crate, and with it turned off all rgb, including RAM and motherboard (Asus motherboard), and 3070 TI, by setting DARK mode

however, for whatever reason, now it stopped working for GPU, and GPU started to do default rgb (ram and motherboard are still off).
in Armoury Crate GPU is not showed, like it can't recognise card
Device Manager show card properly. MSI Afterburner also. card work normaly. but seems like it is not detected by Armoury Crate anymore.

used DDU to clear drivers, than did new driver install, nothing changes
removed Armoury Crate and all additional stuff with Revo Uninstaller, than installed it again, nothing changes

than I did try LED Visualizer. it doesn't see card
when I replace card with 1080 TI, LED Visualizer recognises 1080 TI

what could be issue here?

and is there any another application which can turn off rgb (but question is, is it going to recognise 3070 TI) ?

I could unplug connector of course, but I would have to open card again I belive
 

·
Registered
Joined
·
61 Posts
hello getting this reading on gpu memory temp it is safe? 102c but i can touch it on my hand.
im using palit 3070 ti gamerock
2523194
 

·
Registered
Joined
·
381 Posts
hello getting this reading on gpu memory temp it is safe? 102c but i can touch it on my hand.
im using palit 3070 ti gamerock
View attachment 2523194


I belive 102C is too much. would degrade faster on such temperature.

maybe to use more aggressive fan curve?

and, unless you have problem with opening card for any reason, I would replace thermal pads
 
401 - 420 of 443 Posts
Top