Overclock.net banner
2,161 - 2,180 of 2,939 Posts
What are your avg clocks for gaming 4K ?
My trash chip can do 1935 effective clock in CP2077 4K maxed out for example with .968mV

I couldnt even do 1980 effective with Galax Bios. Gaming overclock is nightmare.
 
What are your avg clocks for gaming 4K ?
My trash chip can do 1935 effective clock in CP2077 4K maxed out for example with .968mV

I couldnt even do 1980 effective with Galax Bios. Gaming overclock is nightmare.
My lemon of a chip...
Muscle Motor vehicle T-shirt Recreation Event
 
Honestly to maintain above 2010Mhz on any GA102 in chip on the core, at 4K with RTX enabled and without shunt modding or using water cooling, is really REALLY hard lol; on my case (3080, non-Ti) I could only mantain like 2100Mhz-2085Mhz in those cases after shunt modding, before that I was like on the 1905Mhz easily because power limit. (The volts were like 0.95 or even lower)
 
  • Rep+
Reactions: z390e
I have been running my EVGA ftw3 ultra gaming for the last ~9 months with an overclock of around 2130mhz in games - pushing above 2160 seemed to produce some instability in some games but nothing too severe. The gpu and cpu are on a full loop with 6x120mm of rad. The card was on a gen4 pie riser cable for a vertical mount - no issues for the last 9 months or so. The rest of my main system is a 5950x on a crosshair viii formula and a 1000w power supply, so no worries there. After a few months of running the system doing cpu stability tests at night and letting the gpu mine rvn (limited to 80% power in afterburner for better efficiency) I decided to let the system sleep overnight instead of keeping it awake.

When I woke it up, the system ran fine for a minute or two but then all of my monitors went black - after a reboot, this happened again - boot, fine for a minute or so, black screens. I booted into windows and turned off the "plugins" setting on icue having heard that could be an issue with some cards rgb. After that and removing the overclock in afterburner and re-applying it, the system seemed stable for the rest of the night. I speculated that maybe the de-rating of the v/f curve for temperature had caused the card to attempt frequencies that were way too high when waking up from cold. The next morning however, the same crash happened immediately after the system woke up, and now the system did not want to reboot - no video output. The motherboard showed an Asus error boot code 97 - Load VGA Bios. I tried swapping the card to another pie slot, swapping out the riser cable, etc - no good. I tried toggling the switch to the secondary bios - also nothing. I installed a Quadro p600 I had and that allowed me to boot to the bios and into the os - at that point I could see in the os that the 3080ti shows up in the device manager and in gpuz but not in hwinfo64 or afterburner (so if there's any v/f curve changes that got somehow "baked in" to the gpu vbios I couldn't tweak that).

Lacking anything else to try, I swapped cards between systems and put a 3090 in my main pc - that works just fine, so nothing else in the main pc was "the issue". I did skip the vertical mount this time however, and just put the card right into the slot. Problem solved for the main pc, but now I need to see if the 3080 ti still works, or try to figure out what's wrong!

I put the 3080 ti into my secondary system - 9900k, Asus prime z390-A motherboard, again full loop cooled. This system has integrated graphics so that makes debugging a bit easier. If I try to boot with an HDMI monitor plugged into the gpu, I get a VGA light on the board and it doesn't seem to boot. If I boot with the HDMI plugged into the integrated gpu, it boots up just fine and I can see in the bios that the board sees the card in the 16x pcie slot 1 and identifies it as "NVIDIA GPU"
Computer Blue Light Font Line

In windows, the card shows up in gpu-z and in the device manager. I used nvflash to detect it as well, and to backup the bios and then to update the vbios to the latest version. I updated both vbios switch positions just fine and after the update and a reboot they report the correct vbios version numbers. Despite this, the system still does not want to boot or do display output over the gpu. Any thoughts on what to test next? Or shall I just give up and look to do an RMA? It is very confusing that there would suddenly be an issue after the card ran fine for so long - I had assumed the riser cable was a likely suspect but that's no longer in the equation.

Thanks for any help!
 
Honestly to maintain above 2010Mhz on any GA102 in chip on the core, at 4K with RTX enabled and without shunt modding or using water cooling, is really REALLY hard lol; on my case (3080, non-Ti) I could only mantain like 2100Mhz-2085Mhz in those cases after shunt modding, before that I was like on the 1905Mhz easily because power limit. (The volts were like 0.95 or even lower)
I had no problem of feeding power to my GPU, with Galax bios on Strix LC. Problem was the chip cannot handle without crashing above 2025. 3DMark is ease tho.. 4K RTX gaming I better should stick to uV :oops: Also galax bios have glitch for making some of the fan's rpm go like 60k rpm. Not suitable for gaming
 
It puts fans into panic mode.
3000rpm max on GB bios at "100%".
3700rpm max on Galax bios at actual 100%.

As for game clocks ..I let the card do it's thing.

I set 1100mv @ 2250mhz. Then descending the voltage curve, every 3 voltage steps is 1 speed drop until 950mv. From 950mv downwards I go straight line, 1 voltage step 1 speed bin.

I don't expect that will be stable for everyone.

The extra power only helps a lot if you're not getting heavy temperature induced boost reduction (seperate from temp limiter). By 60c GPU temp there is a ton of boost cut occuring already.
 
What are your avg clocks for gaming 4K ?
My trash chip can do 1935 effective clock in CP2077 4K maxed out for example with .968mV

I couldnt even do 1980 effective with Galax Bios. Gaming overclock is nightmare.
I don't run 4K, but at 1440p I've settled at 0.918v @ 1965mhz for my gaming clocks.

I just dropped voltage until Heaven stopped hitting a power limit, but it still occasionally hits it in Cyberpunk :(
 
I got a 2X8 pin GGbyte, 1440p gaming 863v @ 1860mhz, anything in 4K would still drop to 1710 and hangs in 1700s
 
When I woke it up, the system ran fine for a minute or two but then all of my monitors went black - after a reboot, this happened again - boot, fine for a minute or so, black screens. I booted into windows and turned off the "plugins" setting on icue having heard that could be an issue with some cards rgb. After that and removing the overclock in afterburner and re-applying it, the system seemed stable for the rest of the night. I speculated that maybe the de-rating of the v/f curve for temperature had caused the card to attempt frequencies that were way too high when waking up from cold. The next morning however, the same crash happened immediately after the system woke up, and now the system did not want to reboot - no video output. The motherboard showed an Asus error boot code 97 - Load VGA Bios. I tried swapping the card to another pie slot, swapping out the riser cable, etc - no good. I tried toggling the switch to the secondary bios - also nothing. I installed a Quadro p600 I had and that allowed me to boot to the bios and into the os - at that point I could see in the os that the 3080ti shows up in the device manager and in gpuz but not in hwinfo64 or afterburner (so if there's any v/f curve changes that got somehow "baked in" to the gpu vbios I couldn't tweak that).

Did you possibly turn on ReBar?

 
My 2 pin TUF 3080ti can hang above 1800 in 4K most of the time. Only time I think I see a rollercoaster between 1800s and high 1700s is in CP77 4K RT Ultra no DLSS. But I don't play at those settings, so who cares xD
Well, kind of same here. Still jumps to 1800's but mostly stays in 1700's.
I'm still trying to figure out OC or Uv and which bios to keep. I've tried most of the bios available on TechPowerUp from every vendor, found only 2 suitable. This Uv helps to keep it more or less steady, at least @ 1440p.
Colorfulness Font Screenshot Line Technology
 
Well, kind of same here. Still jumps to 1800's but mostly stays in 1700's.
I'm still trying to figure out OC or Uv and which bios to keep. I've tried most of the bios available on TechPowerUp from every vendor, found only 2 suitable. This Uv helps to keep it more or less steady, at least @ 1440p.
View attachment 2554228
For 1440p I get absolutely steady 1920mhz at 900mv, with power to spare in most games, only notable exception being God of War at 1440p. That game hammers the power limit for whatever reason, I get around 1890 there.
 
For 1440p I get absolutely steady 1920mhz at 900mv, with power to spare in most games.
I play Borderlands and now T-Tina's Wonderlands, those games are rollercoasters in terms of performance.
Edit: I'm gonna give your bios a try one more time, it's got higher boost clock @ default,mine is 1710. And it looks like second HDMI and fans work just fine.
Colorfulness Photograph Computer Light Purple
 
I play Borderlands and now T-Tina's Wonderlands, those games are rollercoasters in terms of performance.
Edit: I'm gonna give your bios a try one more time, it's got higher boost clock @ default,mine is 1710. And it looks like second HDMI and fans work just fine.
View attachment 2554239
It's been a hot while since I played BL3

Dusted it off and yeah, it seems to hammer the cores nicely, still my card holds 1920mhz 99% of the time, with an ocassional blip to 1905, never below.
 
It's been a hot while since I played BL3

Dusted it off and yeah, it seems to hammer the cores nicely, still my card holds 1920mhz 99% of the time, with an ocassional blip to 1905, never below.
Meh, stays @1920mhz for most of the time but still fps drops, like I said, Borderlands performance has always been flying all over... Sky Plant Building Automotive tire Vehicle
 
Regarding UV vs OC. Do both.

Find the speed at 950mv that is max stable 2100ish being a fairly safe bet, flatline the curve above 950mv from there. Then step down 2 voltage steps for every 1 speed step from 950 going left on the curve. Tailor as required for your GPU, some will take 3 voltage steps per speed step, others 2 or even 1.

In this way you allow the clock to maintain high speed at low load, 950mv, and throttle voltage to stay below power limit otherwise. The default curve is fairly conservative so UV +OC is very possible.

As mentioned, it really does take 400w+ to hold 2000+ on air for most cards.

Remember too the curve needs to be adjusted always at the same temp or it will self adjust.
 
Regarding UV vs OC. Do both.
As mentioned, it really does take 400w+ to hold 2000+ on air for most cards.
Well, that's my dilemma, to keep this lemon or slap a block on it and forget it. I've got an Aplphacool block laying around for the past 5 months and I'm still considering selling both. It may squeeze another 50mhz on water but will still be a lemon :(, my first non Reference card since the GTX 6 series.
 
  • Rep+
Reactions: z390e
2,161 - 2,180 of 2,939 Posts