Discussion in 'Videocards - NVIDIA GeForce' started by Syranetic, Sep 17, 2020.
Is the monitor 144Hz and OC 165Hz by any chance?
It's Dell S2721DGF.Doesn't really say.Found a couple of links saying its native 165hz
I checked it's 144Hz HDMI and 165Hz DP port I asked because I can set mine to 165 with OC and it gave me only trouble like BSOD and Weird behaving in Games. standard on 144Hz I have no problems with the monitor I have the ASUS VG27A
Yea Im using a display port.
I tried switching to144hz but there wasn't much of a difference so I switched to my old 1080p 144hz monitor and it was noticeably smoother.
Tried the warhammer totalwar 2 benchmark and il let you guess which is the 1080p result.Same thing when starting a campaign on the map.On 1440p id get these occasional frequent micro frametime spikes like shown from the picture before and on 1080p they don't happen.
Its gotta be the extra gpu usage that's being forced on 1440p right?Something aint working right.
How do I go proving there's possibly a fault with my gpu?Got my card from computeruniverse.
Try something with a lower voltage. 0,85-0,9V is enough for any interesting clock speed.
K, tried 1905 at 0,9V and altought the small frametimespike didn't go away they felt less impacting.(if you can make sense of what I mean.The Oomph! wasnt as bad but still there)
Those micro stutters kept coming and going just by simply moving the camera around.
Funnily enough once the 2 armies engaged and gpu usage dropped to around 60% all micro stutters stopped.Again observing the same thing repeatedly.The moment a gpu usage threshold is reached (Around 80-99%) these micro stutters start to happen.
I'm gonna try using my brothers 1440p/144hz monitor just incase it is a monitor issue.
UPDATE:Issue remains on my borthers 1440p 144hz.Again going to my old 1080p 144hz monitor fixes the issue
1080p test Went nuts with the camera and it still smooth.
I'm gonna make a thread on its own just so I don't spam here with troubleshooting stuff.
Any thoughts on using this card for Ethereum mining? Friend of mine talked me into giving it a whirl. EVGA Precision X1 app +1200Mhz, fan 70%, GPU down throttle 300Mhz. Power down 65%. Over 100 on the hash rate.
I ran it for a day or two then stopped. Ran stable but felt bad ripping the GPU 24/7.
Absolutely this ^
2 things with the 3080's:
1) They are not created all equal
2) They are by far one of the worst engineered gpu "software" wise (not talking about optimal performance capability here) I've seen in a long time, all of this most probably for marketing reasons (so that Nvidia could brag having the best performing card vs AMD... clockwise on "maximum boost" on paper and in the best of circumstances in synthetic benchmarks not actually eating as much power as some real world gaming). If you wanna a "plug and play" gpu and enjoy it's full potential out of the box, those are not the GPUs you are looking for...
All 3080s have their stock voltage curve pushed WAY too far (in terms of optimal performance) for their physical capability (maximum power / wattage) vs what those pcb boards are designed to handle (maybe with the exception of the EVGA FTW3 which the board can withstand up to an insane 470W vs Nvidia's reference that can be pushed up to 375W with "power limit" slider at it's max)
My stock Zotac (THE "cheapest" quality build 3080 in the way that they cheaped out even on the hardware, having amongst things less power phases than all other 3080s, having a power limit of only 336W, almost 40W lower than the reference board!!...). That being said, they still left a more or less stock voltage curve implemented in this card's bios, eventhough the card is NOT capable of handling a "stock reference" power draw, so that basically makes the Zotac always hitting the power limit (confirmed with gpu-z) and as a consequence, my card was always throttling in the low 1700's (although capable of hitting 19xx's for a couple of seconds in very light games and drops almost 300mhz right after - and no, it's not a question of overheating, I'm not even hitting the 70's right now on full load - so yeah, this kind of behavior (sudden drop in clocks combined with continual power spikes and drops), can contribute to some additionnal micro-stuttering.
My Zotac's stock voltage curve supposedly allows a max speed of 2070Mhz @ 1162mv (but in reality, never got above late 1800's, early 1900's on stock curve, and in very light games only)
After lots of trials and error (I want rock stable clocks), I ended up with 3 undervolted profiles (if I dont want to hit the power limit, have to choose a different one depending on the game, for example, Metro Exodus, anything above my 850mv and I'm sure to hit my wall and have some throttling). Those profile are ranging from 850mv to a max of 950mv in 50mv increaments, allowing constant and stable load clocks of 1875, 1950 and 2025Mhz respectively
So as you can see, vs stock configuration, I need to drop the voltage anywhere from 18% to 27% in order to get... at least 10% higher clocks AND able to maintain those clocks on load! And by doing so, in my case, it did help to alleviate some microstuttering (but not all of it of course as microstuttering is can be caused by many different factors)
@kman , If you wanna quick test to see if this is actually part of your micro-stuttering, just drop down your max voltage curve to say 825mv with a max clock of 1800mhz
Normally, any 3080's is able to reach constant clock of 1800Mhz at 825mv and technically, you should not hit any power limit at all @825mv, therefore completely avoiding throttling. If you still get the same stuttering with those figures, then your stuttering is coming from somewhere else.
But even then, I think it's still worth it to spend some time for any 3080 owners to find their sweet spot undervolted curve, as you can have massive gains power/watts (for those having a +-650W PSU for example and dont wanna go overboard with their new 3080 and OC CPU's) and temps (It does drop down the temps of the GPU and the whole case due to the amount of heat dissipated by the 3080's high wattage, so specially helpful if your airflow is meh...)
Hope it helps!
Still happens fawk me.What could be the cause tho...It doesn't happen when I play on 1080p (because gpu usage isn't maxed from what I can tell) but happen on 1440p.
If you can backup your OS try an new install to rule out software influence in your case
Due to the fact that it's not happening with your current OS on a different monitor (the 1080p), I dont think it would be OS related in this case (but we never know...)
If your GPU is hitting 99%, but you're not hitting any power limit (you can validate this with GPUz "PerfCapReason" line in sensors), then it might just be that this specific game engine doesn't like hitting the 99% gpu usage wall. The fact that you mention it's not stuttering with the 1080p and that you're not hitting 99% gpu usage with that 1080p tends towards this theory. Try lowering your graphic's settings with your 1440p monitor until your gpu doesn't hit the 99% wall at all and then see if your stuterring is going away or not.
Another weird thing, try (if possible) playing the game with a controller. Dont know why, but some games are buttersmooth with a controller and exhibites some stutters / jerkiness with a mouse... On my end, this happens with Control Ultimate (Steam version) and I have yet to find out why...
I have seen some crazy crap in my 36 years experience with computers trust me it can be very well software related
I totally agree!! 95% pc issues are software (and/or user related) and we'll leave a 5% for the faulty hardware
I just did a clean install of w10 and issue remains.
On my old 1080ti I used to hit 99 gpu usage all the time and NEVER had this issue.
I tried another quick test on with the monitors and If I change to my old 1080p 144hz monitor the issue basically doesn't happen so I tried to downscale my 1440p monitor to 1080p in the ingame settings and altought it was way smoother the gpu usage was still higher then actual 1080p monitor and the occasional 37ms frametime spikes still happened.
As for software being a possible cause.
EDIT:Trying to post in the nvidia forums and I'm getting ghosted https://www.nvidia.com/en-us/geforc...-3080-tuf-ocrandom-micro-stutter-at-1440p-80/ what...
UPDATE:Hmm... another user just posted something earily similar to what I'm getting https://www.reddit.com/r/ASUS/comme...o?utm_source=share&utm_medium=web2x&context=3
The video in question.At the start when the frametime is erratic and he starts to get random micro stutters,the gpu is at around 90% and then its get much smoother when it drops in the 70s.He suggested I go back to driver 456.71 (Guy from video)
Hopefully this is the issue because the similarities are pretty spot on.
Il do this tomorrow tho.Kinda tried of trying to troubleshoot this all day.
Updateidn't work. Setting an fps limit to 100 greatly helps with the issue (probably because it reduces gpu usage to the 70%) but the issue still goes on.
Tried driver 456.71 as a last resort but nothing.Limit fps help obviously because it also reduces gpu usage.I'd have to limit to like a 100 to get signficant smooth results tho :/
165hz with an fps limit of 141 still causes to happen alot https://imgur.com/a/yBE76La you see anything out of the ordinary?
If I was you I would contact the store and see if they can replace the card imo
I'm contacting nvidia support atm and gave them all the info needed.Just gave them a log of gpuz while displaying the issue.
Gonna be a bitch to get a replacement from computeruniverse since they don't seem to stock the card anymore...or they could ask me to send it before they send theirs which would leave me without a pc basically for like 2-3 weeks which is a no go.
Il see what the nvidia support says first.
I've had that happen years ago with an Asus card. Had to RMA ist within a week after purchase, no stock & waited 3 months before I got my replacement.
I'm still not perfectly convinced there's an issue with the card itself. It it doesn't throtte massively? How about monitoring different things, one after the other, looking for an issue. Temperatures, usages, voltages.
Maybe it's something odd like energy saving, the XBox game bar or "threaded optimization".
I'm not that savy in what to look for when it comes to voltages :/ I can check temps/usages and noticed nothing in particular.
I also have everything under gaming disabled and my nvidia global performance is set to maximum just to prevent any "energy saving" (also my windows power plan is set to High performance.
I have the gpu-z log of the issue displayed in the video if you want to take a gander at it https://drive.google.com/file/d/1H0uWRhjdVFRiamwPv_o5Q9vMjWC2N71j/view?usp=sharing its the same file I gave Nvidia.
Have you guys watched this video?
It shows that by using the traditional undervolting method the effective core clock will be much lower than the one actually indicated by msi ab. I tested myself and it's unfortunately true. hwinfo's osd was displaying an effective core clock that was 50-100MHz lower.