Discussion in 'Videocards - NVIDIA GeForce' started by Syranetic, Sep 17, 2020.
Prices are crazy but I paid for my RTX 3080 with Crypto money
Anyone with a 3080tuf OC noticing a similar issue with how the card boosts?It seems the cards seems to downclock slowly down once temps hit like 62-65C where it was dropping to the1800s (Card's default is 1785mhz and anything under 61C the card stays at 1950mhz)
I'm asking because I I'm investing an issue with a game and micro stuttering.Noticed the moment I was using skill ingame (and I guess a spike in temps because of graphic effects) the clock was flactuating a bit so I'm wondering if there's a connection
Wierdly my old 1080ti used to hold 1924mhz up to 70C+ yet this 3080 starts downclocking waaay earlier.
I'm using 2 seperate cables with an rmx 750 PSU
From what you posted i see at least too much voltage on GPU (1080mV), try using curve with 1025mV max on 1950Mhz
Welp.That's default voltage :/ I would rather not have any 3rd party software to mess with the core voltage.(Mostly because I have never oc'd my gpu)
I guess its fine to leave it alone too?I wonder if default could be the cause of random micro stutters.I think I remember Jayz saying something similar on an early review of the 3080s where he noticed micro stutters that where resolved by lowering the voltage/clock (It's probably a game issue tho in my case) I've also seen a couple of threads on reddit saying something similar about getting smoother frametime by lowering voltage/clock on their 3080s
Even that is a lot, default voltage is way too high for the chip to run efficiently.
I'd recommend to test out a setting with something around 0,85V. Then max out your power target to prevent stuttering (should be what you're experiencing there).
With 0,85-0,862V, my card runs at 1.900MHz GPU and 10GHz memory. My power target is set to 450W even though the card doesn't exceed 280-300W on a regular basis. This way I'm safe for the odd situation where it may need a lot more juice or jump between clocks and throttling.
So it is a possibility the voltage being too high (by default) to be a cause of micro stutters?
Shouldn't they test these cards for stability/efficiency?(even if it didn't cause micro stutters)
You’re simply maxing out the power target and/or reaching the thermal target.
The undervolting capabilities are second to none, I haven’t seen anything alike in the past two decades.
Very little overclocking headroom but huge gains through undervolting.
Know what these dips might be?had them happen to be while playing.Wasn't a loading screen when it happened either.
UPDATE: NVM it's happening occasionally when I alt-tab out and do something else.If I play continuesly without doing anything else the strange dips don't happen.
BDO still stutters tho so they where unrelated. https://drive.google.com/file/d/1IRCFbeoWK7pl3YOELLMkxcfzvIVAijJa/view?usp=sharing gpu z-log I don't see anything wrong so it has to be the game.
I think he lost silicon lottery here, it won't working on 825-850Mv, so i suppose 1000-1025Mv is target, but he must try all variants IMO
My card uses all the voltage it can get too, so I don't see that as a point. My old 1080Ti did the same.
Keep it cold enough and it'll raise the voltage, eg. go further to the right on the pre-set curve.
That's why I'd test a lower voltage. It certainly helps against the micro-stutter being induced by hitting the power target and/or by the constantly changing frequencies.
With a proper curve, the card won't ever hit the power target, will have a stable frequency under load and usually show lower temperatures and power consumption.
Hmm I've been looking around and most threads I'm finding are saying hitting the power limit shouldn't cause any micro stutters.Changing clocks?possibly?When I did my timespy benchmark https://www.3dmark.com/3dm/59313958 the clock was averaging 1,830mhz with a 1950mhz max and it was smooth (I watched it all xD)
Maybe it is indeed the game/s in question thats the cause.
Just for comparison's sake: https://www.3dmark.com/compare/spy/18908343/spy/18921466#
Even though your card boosts higher, the average speed is lower and you're losing performance while burning at least 100W more for nothing in return.
Why not give a custom curve a try?
I'm trying to do as this guy but for some reason the line is not straight as his video.Goes up a bit on the 1100 mark.
Just play with undervolting. Start with -270-290 on core and just do some testing. Spend some time get it right. Don't go chasin' waterfalls. Please stick to the rivers and the rocket lakes that you're used to.
I left my RTX 3080 OC Tuf untouched for so far except bios version no problems at all
The bigger cards will get along without undervolting, sure.
Still, running the card on full voltage simply isn’t efficient. You don’t gain anything by running it like that.
Tried doing 1900mhz at 0.975 and tried timespy benchmark,The benchmark was going on fine but the clock kept going down at 1875mhz. The voltage wat at 0.975 while it was going and now that I had msi afterburner frametime graph up I noticed it wasn't quite as smooth as I tought.
Think il just give up.I just wanted to try and stop the micro stutters from happening.
There could be something going on with the card just saying
playing Apex.I get those tiny frametimespikes pretty much in every game randomly.Noticed they happen mostly when the gpu usage pushes past 70%+. The only game I haven't seen it happen is Wow(probably because it doesn't push the gpu hard enough as other games.Ever since I got my new 1440p 165 hz monitor those tiny spikes seem much more obvious.(could it be the higher res is pushing the card more and its making the issue come earlier?)