So what has been fixed by cod warzone, to be incl in this driver?
Raytracing performance fixes.
Call of Duty: Modern Warfare may intermittently see a drop in frame rate when ray tracing is enabled
[Pascal] Marvel's Avengers may crash if alt-tabbing while the game is running
[Notebook] Display is not detected when connected to Thunderbolt 3 port on HP Spectre x360 - 15t-df100 Notebook
Minecraft Java Edition may crash when launched with XSplit Broadcaster running in the background
...and back to 451.85. This one and several other newer drivers (almost all of them after 441.xx) give me TDR: Driver crashed and recovered errors, and after playing a game, then right clicking desktop to go to Nvidia Control Panel, I get a long 'hourglass/working cursor' delay, then the right click menu appears with no Nvidia Control Panel in it. I have to reboot to get it working again. Don't have a clue why.
You have unstable system memory.
-Tj-:
So what has been fixed by cod warzone, to be incl in this driver?
sharp framerate dips and raytracing absolutely wrecking MW.
it doesn't really do anything for warzone.
Tried again with new edge version and this driver, did a full DDU uninstall beforehand... chromium edge with hardware accel on still freezes all my monitors randomly at youtube. Never happens with it turned off.
Possible. If so it is barely teetering on the edge of unstable though, because it passes memory tests and exhibits no other issues in any programs or certain graphics card driver versions. Good example is the unoptimized, absolute memory hog Horizon Zero Dawn. I'm one of the few who do not get crashes in that game. So I figure it's pretty "stable".
Hah well that comes up every now and again I suppose, I'd imagine efficiency has a bigger scaling factor though both AMD and NVIDIA have had some issues and possible regressions.
AMD also broke something with low level API performance that still hasn't been resolved for Vega and earlier and that's been going for ~9 months soon with some games losing 30 - 50% compared to earlier drivers.
Driver situations going to be the deciding factor but with a more stable economic situation I would probably have looked at the 3080 as a solid upgrade choice already leaving behind a year of odd driver acknowledgements and somehow even complete obliviousness (They have the Vanguard program too making this even stranger.) it's gotten better but not fully and it's taken way too long. (Vega, Polaris and the Radeon VII user base also affected, R VII in particular how that's been supported is just weird for a supposed flagship GPU model.)
Both have their issues and launch drivers might have some not entirely worked out problems but it's going to be interesting to read up on how this is going to go both for NVIDIA first here and then AMD GPU.
Possible. If so it is barely teetering on the edge of unstable though, because it passes memory tests and exhibits no other issues in any programs or certain graphics card driver versions. Good example is the unoptimized, absolute memory hog Horizon Zero Dawn. I'm one of the few who do not get crashes in that game. So I figure it's pretty "stable".
Maybe, using a lot of memory or absolutely hammering the memory is a bit different though.
Shadow of the Tomb Raider and Battlefield V pops up pretty often for being very quick to crash if there's any memory problems. I would assume the new COD game could be a stability stress test candidate too but the game itself has a few issues to work out and that makes it unreliable for testing against.
Benchmarks might be too fixed and not random enough to trigger some of these borderline stability faults too but some programs a re a bit better about randomizing and alternating the test pattern and conditions for loading up and putting RAM to it's limits.
AMD for their boards also has the infinity fabric and performance on that and stability but the early warnings error 19 or 21 what it is in Windows event log after a day or so of uptime will tell if it's stable or not so if there's no errors of that nature that's a good sign.
(If there's even one however it's not stable.)
EDIT: Argh explanations and what to best call all of this.
Don't just store a ton of data in memory actually send a ton of data back and forth and stress the memory, memory controller and the chipset in full making a good workout of the RAM chips and it's settings from voltage and timings to the infinity fabric on AMD and so on.
Bigger thing for the Navi10 GPU's too due to sending more stuff over the chipset and hitting the memory channel and stability that way confusing some of the GPU crashes with what is actually RAM instability.
(Combined with the other problem around higher transient and peak power draws from the boost algorithm and how the GPU operates though Navi is less extreme than the Vega GPU models here.)
I found out quake champions is the biggest system ram hog atm.
Add that shiity code fix they did on nv+intel and it fcked my haswell+maxell combo, it randomly ctd no matter what, ever since some update back in April 2019? Or was it 2018. Idk anymore I stopped playing it.
I found out quake champions is the biggest system ram hog atm.
Add that shiity code fix they did on nv+intel and it fcked my haswell+maxell combo, it randomly ctd no matter what, ever since some update back in April 2019? Or was it 2018. Idk anymore I stopped playing it.
If so, it would worth a try upgrading the MB firmware further (if available) or downloading the MC and doing it manually (with the winraid UBU tool) if the manufacturer abandoned support.
Tried again with new edge version and this driver, did a full DDU uninstall beforehand... chromium edge with hardware accel on still freezes all my monitors randomly at youtube. Never happens with it turned off.
I used to have an issue with YouTube where after watching one or two 4K videos everything becomes sluggish and really slow. I'd have to for example exit the browser, and go back to YouTube for it to go back to normal speed again.
Anyway the fix I found after MONTHS was to use nVidia Profile Inspector, scroll down to section 5 (common), and find CUDA - Force P2 State and turn it OFF. I do this after every time I update the videocard drivers.
This also fixed any stuttering issues I had with some games that had stuttering every once in a while.
(Combined with the other problem around higher transient and peak power draws from the boost algorithm and how the GPU operates though Navi is less extreme than the Vega GPU models here.)
thanks for your lucid analysis 🙂
btw, I doubt Samsung's quality may compare to TSMC's especially when it comes up to...low power states (forgot exynos' standby flaws?!)
nV said the 12-pin connector had been developed to ensure STABILITY and avoid peaks.
However, AIB will retain 8-pin connectors and 12v madness is expected. (source: [youtube=_e_KDqzlAbk] -> "to cope with fast load changes")
Wonder what the more tuned and fully customized models will be utilizing plus how scalable the cards are although NVIDIA is allowing even the 3090 to be part of it now but the preview image from Reflex and some of the other features also showed that the 3080 at least appears to be capped at 115% power draw extra.
https://www.nvidia.com/content/dam/en-zz/Solutions/geforce/news/reflex-low-latency-platform/geforce-experience-new-performance-tuning-and-monitoring-options.png
If NVIDIA OK's some of the more extreme models for the 3090 GPU that could be a situation where the 12-pin makes a good alternative instead of 3x 8-pin plus a bit extra even since that's 600w and likely more than most of these GPU's would draw.
(A higher power limit on the 3090 would probably still be conservative so maybe 20% or so default and ~420 - 440w over the 350w default power draw specifications.)
Newer power supplies should be able to cope with near idle or close to zero activity GPU cores too and very low power draws plus the short burst and higher transient power draw too but AMD always felt like they had the more problems in this regard and even Navi is sorta matching the 2080 Ti while being slower. (Actually a bit higher and some of the third parties get up to 320 or even 360 I think it was from the bios database on Techpowerup instead of the 280w default.)
Vega64 as the really problematic one at almost double peak transient hitting close to 600w and above 40a every so often. Higher peak draws are a thing but that GPU was really bad with this.
(Navi or rather the 5700 XT defaults "only" hit about 400w though still enough to trip a few PSU's up and that over surge protection safety shutdown.)
I see the 12 pin as bit of both a future possible connector on high-end GPU models and due to lowered power draws and improved efficiency also a bit of a safety feature where the cable can handle up to four times the 8 pin connector and the cable durability and robustness as a result also improves as does things on the PSU end I presume for the 12-pin models coming out though I'm not all that great at how this all works. 🙂
Doubt even the Kingpin would hit that even if NVIDIA were to grant a higher allowed power draw it'd be like 40 maybe 50% and ~500 ish total in terms of wattage so well within spec and if the card is not spiking or fluctuating with these peak power draw transients that shouldn't be a issue but we'll see once the reviews are out sometime mid next week. 🙂
Going by NVIDIA giving certain review sources these monitoring hardware plugins for power draws I'd imagine they are confident and there's probably limits on what the third party models are allowed to do including also high-ends like EVGA's or Asus top end offerings plus what the limits are for the 3070, 3080 and 3090 respectively.
(And how locked this is, Experience might just be a soft cap although I can't imagine NVIDIA letting users lift some of these limits easily either.)
EDIT: And looking it up...
So I was a bit off there the 2080 Ti Kingpin didn't add 40% to the 250w default 2080Ti draw it added closer to 115% from what I can tell and the GPU can hit near 530w but that's everything not just the GPU core.
Hmm but then GFE is still capping it pretty low but the GPU third party designs might be a bit more open to further increasing the limit here and then maybe a bit more through GeForce Experience still from that new default. (Which I think I've missed considering entirely, oops.)
What do I know, might be a possibility of a GPU with more than just the single 12-pin after all.
(Though I do have some doubt on that, has to be a limit somewhere for where it's scaling less with extra power and other thresholds here.)
EDIT: Well it's a user friendly overclock utility same as Wattman it would generally have fairly tight limits and a lower range of causing any actual errors or damage to the hardware.
For AMD's choices here though that's aside from the power limit which instead of 20% can go up to 50% on some GPU models ha ha.
Anyway.
"to cope with fast load changes"
That's probably the big one here, avoiding any fluctuation and issues from rapid changes to the various power states in response to overall GPU load including modern PSU hardware and the cabling too.
(Not just down to near zero load and draw but also quickly up to peak draw from that state.)
Well i installed them and no real difference vs 451.85. At least not in this COD. Will see Desitny2 now. 🙂
But I have one question about COD warzone, are shadows suppose to look like from the 90's in the distance? xD
Well i installed them and no real difference vs 451.85. At least not in this COD. Will see Desitny2 now. 🙂
But I have one question about COD warzone, are shadows suppose to look like from the 90's in the distance? xD
settings:
CoD Warzone Shadows are the worst part of the game. Happens if you don't have 2xSMAA enabled. Even without 'filmic' on, there seem to be issues.
2/3 options really:
a) Live with it. Keep AA off and have the fastest latency response which is critical in FPS games like CoD especially
b) Turn it on to 2xSMAA Filmic and adjust to the blur, but with less responsiveness of having AA enabled.
c) Turn on 2xSMAA Filmic and enable Nvidia sharpening in NVCP. The most visually pleasing, yet still less responsive.
It all depends how sensitive you are to that increased latency that high AA brings to the table. People that say there is no difference are lying or potentially using a controller or something. It is definitely there.
Personally, just keep AA off and try to ignore the terrible shadows as much as you can.
CoD Warzone Shadows are the worst part of the game. Happens if you don't have 2xSMAA enabled. Even without 'filmic' on, there seem to be issues.
2/3 options really:
a) Live with it. Keep AA off and have the fastest latency response which is critical in FPS games like CoD especially
b) Turn it on to 2xSMAA Filmic and adjust to the blur, but with less responsiveness of having AA enabled.
c) Turn on 2xSMAA Filmic and enable Nvidia sharpening in NVCP. The most visually pleasing, yet still less responsive.
It all depends how sensitive you are to that increased latency that high AA brings to the table. People that say there is no difference are lying or potentially using a controller or something. It is definitely there.
Personally, just keep AA off and try to ignore the terrible shadows as much as you can.
Yes I switched to 1xSMAA, couldn't stand that blur, I even sacrificed reflections for it
so only 2xSMAA fixes those shadows, or are they still blocky, just not as noticeable?
I dont mind latency much, play it just for "fun", but I get pissed more then having fun, like usually by cod mp lol
Astyanax:
cascaded shadows are always like that.
Ok, but I can't remember any game atm that looked like that, not in last 4-5years. Maybe Gta5 kinda, but not soooo bad like here.
Would be good if nv and cod team up and came with a better solution, like usually by partnered games. 😉
EDIT: what about this shadow caching system, maybe that's buggy and better off? or that won't affect it, I didnt try turning both off.
settings: