Texture pop, poor draw distance, AA, LOD and tessellation, shadow flashing

slim01

slim01

slim01

Associate
Joined
1 Sep 2010
Posts
1,252
Location
Bristol, South West
Above sound familiar at all? don't even know where to start with this one...

-Went from a 3070 (FHR)+r5 3060
to
(updated mobo bios)
-3080ti fe +r7 5800x
-Installed crucial p5 ssd (nvme)
to
-3070ti fe.

Something has happened in that upgrade path, as now I've been getting horrible texture pop ins, draw distances and shadow glitches across multiple games (below). All 6 games tested have glitches.
I only played horizon on the 3080ti for 30+ hrs, and only noticed the odd bit of bad texture pop once or twice. Now I find it every 10mins. Could 3070ti be bad?
**edit also rarely get a massive fps drop in horizon. 60fps constant drops to like 25-35 for 3 secs then back to normal. Happens every hour or so.


Details:
Poor draw distances and texture pop in whilst moving ...
-Horizon zero dawn : individual bush's, textures on rocks like moss and snow patterns pop in/out in some areas, grass AA looks very fuzzy and also grass is very thinly built after about 20-30 metres, at the point where its detail draws/pops in).
-Mechwarrior 5: about 5% of the ground rocks pop in at 20-30 metres also.
-Comanche: basically every tree pops into view at around 400metres.

Shakey low fps like mini map/compass in...
-All 3 games with mini-map/compass
(was the same on 3080ti in horizon),+ Comanche, Mech 5.

Flashing Shadows...
-Commanche: REALLY bad shadows, almost all of them constantly pop in and out whilst moving, some shimmer without moving, tree shadows constantly flash.
-Horizon: shadows around the stalactites (could be textures) are a very odd colour and pop in and out of normal/blue at around 5 metres in one cave.
-Man Eater: some shadows (not all shimmer as they move and are very made of massive pixels.



Potential issues...
(yes, **worst build I've done:))
-spilt a bit of kyronaut paste into cpu socket but thought cleaned up nicely.
-I did have to YANK a 3080ti out of the pcie socket (snapping the retention clip of mobo). There was some bending on the mobo and also dropped a screwdriver on mobo trying to open clip.

Also...
-something hot in pc (46-53c idle), mobo fan idles around 2000rpm and often goes up to +3000rpm.


**what I've tried so far as of 12/11/2021**
-RE-installing newest 3070ti drivers in safe mode after using DDU.
-tried 10+ games across different SSD's. All inhibit some issues.
-"trc" changed "bank cycle time" from 75 to 58 as per jdec specs (all others to xmp).
-tried 3070ti in both pcie slots.
-updated chipset drivers (I was on stock).
-replaced with a gigabyte 3080.
-loaded optimised defaults in BIOS, note: left ram on stock from here.
-fiddled with most display settings in nvidia control panel.
Then...
-disconnected everything (USB 3.0, sound card, all fans ALL old SSD's), left memory in as can't access easy.
-fresh copy of Windows 10, on a brand new ssd.
-did win10 updates,
-installed newest gpu driver (ddu+safe mode).
-loaded Horizon, still looks bad.
-Tried last know working gpu driver (same one I would have used on rtx3070 (June 2021) using DDU + safe mode.
-tried all psu cables in different ports.
-different sata and pcie power cables.
-tried powering pc from different wall socket, room, and also a mains filtering extension.
-plugged gpu into a work monitor with DisplayPort (instead of usual hdmi).
-cleared CMOS.
-48 passes (all) of memtest (default clocks)
-tried praying to the PC gods....

Next steps?
-Try ram in diff slots.
-Try 3080 and ram in new system when it arrives(3700x, a used replacement x570-p, seasonic 850w and the ssd I got before).
-Re-inspect cpu/socket for excess TIM.

Specs...
R7 5800x
Asus x570-p
32gb Vengeance 3200mhz
Evga p2 850w
TV - Panasonic 120hz 4k hdr TV


Any ideas guys? Is it likely hardware or software?
 
Last edited:
Is the screen ok in other uses? Ruling out a screen issue

5800x runs warm but nothing to be concerned about. Well within spec.

I'm hazarding a guess the PCIe slot is damaged and trace connections aren't right. Could be wrong though.
 
subbytna said:
Is the screen ok in other uses? Ruling out a screen issue

5800x runs warm but nothing to be concerned about. Well within spec.

I'm hazarding a guess the PCIe slot is damaged and trace connections aren't right. Could be wrong though.
Click to expand...

Screen seems fine. No other issues.

Slot is a good shout, it was a hell of a tug (was very scary with a £1400 value gpu for sure I was sweating!).
I'll try a diff pcie slot. Annoying as that'll be right up against my psu and will get no air from above but may rule the mobo as the issue good shout.

Yeah those temps aren't the cpu, it's something else. I'm guessing it's whatever is under the mobo fan (northbridge?). Only noticed the noise of the fan since I updated mobo bios to accommodate the 5800 series.
 
slim01 said:
I'm guessing it's whatever is under the mobo fan (northbridge?). Only noticed the noise of the fan since I updated mobo bios to accommodate the 5800 series.
Click to expand...

It is the chipset under the fan on X570 mobos. Current AMD chips have the memory controller on the die so does not have what we would consider a seperate Northbridge. I have Asus Prime X570-pro mobo and I would guess it has identical chipset fan to yours and I never really hear it but I do have old ears.

Just some clarification. Did you have any of the issues when you were using the 3080ti or is it just an issue since you put the 3070ti in the pc ? Like said above , trying in a different pci slot is a good place to start....or test the gpu in another system if you have one or use a different gpu in that system if you have one.

Edit- I would say do not manually mess around with the memory setting at this time. Either run at stock 2133mhz while you are testing or just use the DOCP settings. In BIOS look under AI Tweaker and in the DOCP drop down just choose DOCP.
 
Last edited:
Haz123 said:
It is the chipset under the fan on X570 mobos. Current AMD chips have the memory controller on the die so does not have what we would consider a seperate Northbridge. I have Asus Prime X570-pro mobo and I would guess it has identical chipset fan to yours and I never really hear it but I do have old ears.

Just some clarification. Did you have any of the issues when you were using the 3080ti or is it just an issue since you put the 3070ti in the pc ? Like said above , trying in a different pci slot is a good place to start....or test the gpu in another system if you have one or use a different gpu in that system if you have one.

Edit- I would say do not manually mess around with the memory setting at this time. Either run at stock 2133mhz while you are testing or just use the DOCP settings. In BIOS look under AI Tweaker and in the DOCP drop down just choose DOCP.
Click to expand...
Yeah using DOCP settings, just weird I never notice that higher trc timing being off (don't even know what it is tbh).

Well on the 3080ti I played horizon for around 30-40hrs on the 3080ti, its the only game I played on it and I don't recall seeing hardly any glitches... Maybe one or two similar issues.
However now I've gone back to the start of the game and seeing these issues pretty much every 10-20mins of running around.
I'm sure I would have noticed it before, not something I have to even look for it's quite distracting when textures pop in and out even if its not in the exact area you're looking at.

Will try in diff pcie slot later. Don't have another gpu unfortunately.
Will also try memory at 2444mhz.
And remove the eco setting on ryzen master (although that was on when I had 3080ti as cpu was reaching like 80c in windows doing basically nothing before.

As mentioned the mobo did take a beating (screw driver drop, and the bending from pulling out the gpu.


You rekkon thermal paste in the socket could cause the issue? Not sure how better I could clean it... I wiped with alcohol swap wipes.

Any nvidia/windows settings you'd check?
 
My tRC is also sitting at 75 when I am on the desktop. I am not concerned about it tbh.

With you having issues in 3 seperate games it is unlikey to be individual games. Is best to test with benchmarks and not games , Heaven is old but free so you can test with that and see if you get any issues and whether it occurs over time. Or if you have any of the 3Dmark tests use those as well and see if you are getting the types of scores you should be.

The best way to test is to put the gpu in another pc and see if it works as expected.
 
Texture pop is definitely hinting to GPU/driver issue.
Those are used from graphics card's memory and not constantly accessed over PCIe.

Draw distance would similarly fit to at least driver not working.
Same for that "shadow flashing".


If there were broken traces into slot, that should result in number of used lanes dropping down from full 16 to say 8.
(which wouldn't show in any way in performance with PCIe v4)
And if communication was tried over malfunctioning lanes, those errors would propably cause major issues for rest of the PC.
 
Ok so I've tried gpu in different slot, issues remain. Memory tried at 2444mhz as well.

EsaT said:
Texture pop is definitely hinting to GPU/driver issue.
Those are used from graphics card's memory and not constantly accessed over PCIe.

Draw distance would similarly fit to at least driver not working.
Same for that "shadow flashing".


If there were broken traces into slot, that should result in number of used lanes dropping down from full 16 to say 8.
(which wouldn't show in any way in performance with PCIe v4)
And if communication was tried over malfunctioning lanes, those errors would propably cause major issues for rest of the PC.
Click to expand...

So what you're saying is its likely a driver issue... Not likely hardware related?

I'm deciding whether I should be trying to arrange a return to nvidia, as I unfortunately don't have another pc to try it in or another gpu (nor can I afford one).
Hopefully I can even do that, as got off a social network marketplace. Id have to lie to nvidia and give the sellers address if he even gives me proof of purchase... Then hope the guy sends to me... Last time I buy privately.
 
Ok weird one... Not sure if this means anything to anyone,

Just turned my TV on (pc was already on) signed in to windows and resolution was at like 1024x768 for some reason... Now I can't change it (only 2 res showing in display settings, that one and 800x600)

Nvidia panel says

Connector:
Dvi - display port

But the cable is just hdmi to a TV

**edit - fixed by turning TV off at the wall, and plugging hdmi back into 3070ti**
 
Last edited:
Ok have tried
-reseating gpu,
-running memory at stock
-put cpu back to stock (no eco mode)

Issue persists. Anything else you guys would try before selling gpu as faulty?
Maybe a windows os refresh or new mobo?

Anyone want to donate me a gpu to test on?

I guess I could just buy like a gtx970 or something but even they are like over £100 atm. So annoying we can't just buy gpus at retail price.
 
slim01 said:
So what you're saying is its likely a driver issue... Not likely hardware related?
Click to expand...
Lots of graphics glitches without crashes, shutdowns or Windows giving errors hints strongly toward graphics card's internal functionality/its drivers.

Live Linux with some GPU test would be one way to exclude software issues.
https://www.maketecheasier.com/stress-test-graphics-card-linux/
https://www.geeks3d.com/gputest/
https://www.howtoforge.com/tutorial/linux-gpu-benchmark/


slim01 said:
Hopefully I can even do that, as got off a social network marketplace.
Click to expand...
Sounds like you were sold broken card.
 
EsaT said:
Lots of graphics glitches without crashes, shutdowns or Windows giving errors hints strongly toward graphics card's internal functionality/its drivers.

Live Linux with some GPU test would be one way to exclude software issues.
https://www.maketecheasier.com/stress-test-graphics-card-linux/
https://www.geeks3d.com/gputest/
https://www.howtoforge.com/tutorial/linux-gpu-benchmark/


Sounds like you were sold broken card.
Click to expand...

Honestly I'm not sure it's the gpu... Multiple people have now told me its not and also looking back at my benchmark notes, I did actually write "texture pop and bad draw distance" but was only one or two areas... But It happened. Might not have been looking though.

I found a vid to show exactly what's happening with texture pop in, in pretty much all my games (which same as video poster):
https://youtu.be/zjgoCGRmi98

I'm starting to think this happened when I updated mobo bios and new cpu.
If this issue was there before, it must be the mobo bios/drivers or a bad cpu, and I can rule out the damage from the screwdriver drop or paste spill. You don't often get bad cpus though right? They often work or they don't?


I'm picking up a 3080 soon anyway of mm here, so can test gpu, but I haven't got my hopes up.

Anyone got an am4 mono to donate? 3080 will clear me clean Haha
 
It's like there's an invisible circle around my character, about 10 meter diameter in most games.
When moving forward, shadows (especially grads shadows and little grass textures) and some objects only render once inside that circle.

I had a similar issues with gta v but that was 2 years ago... But on this same mobo. Not sure what I did to fix it... Maybe a Windows refresh.
 
Everything points toward issue in graphics pipeline and rendering. (either physical graphics card or software)


CPU itself simply doesn't have any idea about function of bits it handles.
It simply runs code, like OS and usual software, or game code/DirectX/GPU drivers etc, which are the ones controlling what's done.

Motherboard has even less idea.
It's basically just physical platform into which different things are attached and connects them into each others.

Fault in either really couldn't just cause such picky symptoms and errors should happen in every kind data causing crashing of programs and very likely also OS errors.
Along with possible corruption of data handled/modified and written to disk...


Neither could possibly damaged trace on GPU slot cause such picky symptoms of only showing in whole objects blinking and far away textures not working.
Errors in that physical layer/signaling would randomly corrupt every kind data transferred over link.

And possibly crashing whole PC before even getting into OS, unless broken wires using lanes were correctly recognized and disabled in hardware initializing.
PCIe communication/link always starts from single lane x1 config and anything faster is result of negotiating and testing between host and client.
(dead GPU can even block whole PC from starting)
 
New gpu, exactly the same issues. In fact maybe even a new one... Don't remember having so many jaggies.

-the gpu in the other pcie slot does the same thing so the slots not damaged
-now using pcie 3.0 riser
-Also changed ram timings to what they should be

I'm sure it's not the ssds themselves as its doing this across multiple games on different SSD's.



I've heard it can be psu / dirty electrical current... Find that hard to believe though as these glitches can be repeated in the exact same areas, it's not intermittent, its constant. The same texture of snow over the same rock will glitch in the same way.
 
EsaT said:
Everything points toward issue in graphics pipeline and rendering. (either physical graphics card or software)
Click to expand...

Well we know now it's not the gpu itself as replaced, could it be something is not reacting well with the drivers? I mean... I did notice that my 3080ti and 3070ti used same drivers, and now my 3080... I downloaded from nvidia site, and I have the same number (2) as its a 3rd copy.

Are all 3000 series really on the same drivers? Is that the issue?
 
I would probably suggest against Windows 11, it has its fair share of performance issues so you are better sticking with a mature OS. Even the fixed Ryzen issues apparently come back if you swap CPU.

What BIOS version are you using?

Memory might be a good shout, especially with the timing changes. Have you run Memtest on the memory? Memtest86+ via a USB stick for at least 2 or more passes with no errors should be enough. Might take a fair few hours with 32GB though so keep that in mind. If you get any errors you'll want to ideally tweak your settings and retest rather than leave it running.

Have you tried a clean install on a spare empty SSD? If not you'll want it to be as bare bones as possible. Latest version of Windows 10 from the Media Creation tool installed via a known good USB stick, latest X570 motherboard chipset drivers from the AMD site, GPU drivers from the Nvidia and just the game client for the games are you testing.
 
You must log in or register to reply here.
Back
Top Bottom