Asus Tuf 3080ti, games crash to desktop... PC power problems? or card?

[darksun9210]

Hey all, hope someone can help shed some light here - or at least offer an opinion?
I have a Gigabyte Vision 3080 from Overclockers way back when they first came out, and that card has been rock solid.
I have also managed to pickup an Asus Tuf 3080ti 12gig card in the last month or so from a project that got wound down at work. i thought it'd be a drop in replacement for my current non-ti 3080, and a 10% uptick in performance for zero cost, so, hey, worth it right?
omg wrong. but i can't figure out where i'm going wrong - i suspect power supply, but hoping to get a 2nd opinion...

so, the Asus card looks like it works fine. i can run all the benchmarks in 3D Mark with it, I had the lightspeed benchmark on stress test for hours with no crashes. Furmark runs perfectly for ages too. GPU-Z is reporting that the Asus 3080ti card is pulling about 360Watts power, versus my original Gigabyte non-ti 3080's 350Watts.
everything looks great. except actual games with the Ti card.
Valheim crashes back to desktop anywhere from 30 seconds to 3 minutes - doesn't matter if you leave it on the initial character selection screen or jump into a game.
Shadow of the Tombraider graphics benchmark never completes. the actual game again crashes back to desktop...
Redout runs fine for a while, then crashes out to desktop.

i'm hoping it's "just" a power supply issue and that the Asus card is good, just that it turns out i've been riding the edge with my current 685W power supply running my non-ti 3080 and not known it.

trouble is, i have nothing else i can test the Asus card in to make sure the card is fine. and as my PC is an ex-office Dell Precision workstation, getting an uprated PSU is a bit of a nightmare.

does this sound like it is a PSU issue, where the benchmarks and stress tests run fine, but games don't? If it was a power problem, wouldn't it wipe out the whole PC with either a bluescreen, or a reset? I know about the "Transient Load" issues the 30 series cards exacerbated from the 20 series. but if i could run a 3080, a 3080ti shouldn't be *that* much heaver to run?

Specs otherwise:-
Dell Precision 5810
Xeon E5-1650 V4
128gig ram (8x16gig DDR4@2400) also tested 64gig with a single quad channel set of 4x16. no change in the symptoms.
Samsung 970 EVO nvme in a pci-e 4x to nvme adapter card
685Watt PSU
Windows 11 latest
Nvidia drivers v511.23

i've had some duff luck buying a higher spec Dell PSU off ebay. basicly dropped nearly £170 on what was supposedly a 1300Watt version - and it arrived DoA, and Dell won't do an out-of-warrantee repair as apparently it's not a supported PSU for this system - unlike what was advertised on the ebay auction, and the listing was "no-returns". ergh. paid all that money to effectively ship some e-waste from china.

anyway, all said and done, after that PSU experience, i don't want to throw money away guessing. now i'm at a place where i want to somehow test this card to make sure the card is known good before i carry on, either 1. get my Dell up to scratch, or 2. build a new machine to host this Ti card.

all thoughts and ideas welcome

 

[Tetras]

does this sound like it is a PSU issue, where the benchmarks and stress tests run fine, but games don't?

CTDs can be a lot of things, unfortunately, both hardware and software.

Does the cooler have a significantly different design? What I mean is: is it likely that the graphics card is dumping a lot of heat into the case in different areas to the previous cooler?

Valheim crashes back to desktop anywhere from 30 seconds to 3 minutes - doesn't matter if you leave it on the initial character selection screen or jump into a game.
Shadow of the Tombraider graphics benchmark never completes. the actual game again crashes back to desktop...
Redout runs fine for a while, then crashes out to desktop.

Do any of these games need to do shader compiles and what not?

You don't have any Gigabyte software installed?

 

[Joxeon]

does this sound like it is a PSU issue

That would be my first thought, is the 685w PSU even a branded make?

 

[G J]

People will be along soon who have more experience its seems from a quick look a 3080ti though it can spike about 50w or more for an non stock one over a 3080 that could be enough to tip your system over the edge. Maybe try lowering the power limit if you can and see if its more stable.

ASUS GeForce RTX 3080 Ti STRIX LC Liquid Cooled Review

The ASUS RTX 3080 Ti STRIX LC is THE overkill RTX 3080 Ti. It comes with a pre-filled watercooling unit, that provides cooling for the GPU and memory chips. Thanks to a factory overclock to 1830 MHz, this is the fastest RTX 3080 Ti we're testing today, but also the most expensive.

www.techpowerup.com

Also the Dell power supply you purchased on ebay see if you can start a return/refund though the cost of shipping it back to china may not be worth it as people cant really state no returns in the listings to absolve themselves of not having to accept returns/refunds when something is not as described.

 

[darksun9210]

ah, good ideas! thanks all

As regards cooling, I've checked the cards and they both have the triple fan setup. the Asus card actually has a more breathable fan shroud, and about twice as much open area on the back plate at the end of the card where the "blow thru" or "flow thru" cooling air goes. but both pretty much exhaust hot air all around the card, into the case and onto the mainboard. there is an IO chip with a heatsink under the card if it's in the top slot. so i'll try the card into a lower slot and try again

I don't know how to force shader re-compilation, i figured that was something that happened at run time?

hah, the RGB control software was one of the first things to go, and since putting a fresh install of win11 on here that's been rock solid, i'm not going to risk it as this is my daily driver work-from-home machine too, so i'm not going to put that RGB stuff back in.

the PSU is what came in the machine from when Dell shipped it. It's their annoying proprietary metal brick, but as it's taken from their server line, i was hoping it had some decent headroom. but i figure this headroom is what i've been running in all this time with my current 3080.

ouch yes that 50W could be critical at this level :/

Thanks for that ebay info! good to know. i'll see if i can get something kicked off. The seller only has a gmail account as their contact details. the more i was looking at this, the more red flags were getting raised. i know i was just that desperate to get something that someone said may fit/work, i just went for it. and have been suitably burnt for my over-excitement... i'll see if there is something i can do besides leaving negative feeback...

 

[Armageus]

That would be my first thought, is the 685w PSU even a branded make?

Normally Dell and HP use reputable OEMs for PSUs (Often Delta or FSP), and are often deliberately underrated (e.g. the 685w might actually be a 700w or 750w)

 

[Tetras]

I don't know how to force shader re-compilation, i figured that was something that happened at run time?

If it didn't recognise the new card automatically, then if you do a major change to the game settings, I think that will usually do the trick.

 

[darksun9210]

Normally Dell and HP use reputable OEMs for PSUs (Often Delta or FSP), and are often deliberately underrated (e.g. the 685w might actually be a 700w or 750w)

yup, that kinda figures. last thing you want as an OEM is to risk your reputation on parts that should just run and run for years.

If it didn't recognise the new card automatically, then if you do a major change to the game settings, I think that will usually do the trick.

oh cool! the games do run though - straight off the bat, it's... just for how long they run... but i get where you're coming from. i'll try that too, see what happens


actually, while i was typing that last post, i was still running the Asus card. my screens went to sleep for no reason. keyboard was still live - caps lock light worked, so i think windows was still running. just couldn't wake the screens. no fan spin up/down to denote a reboot, so powered off anyway. i've put my original 3080 card back in so i'm in a "known good" state.

i've reached out to some family to see if they can help test this card. they have a standard atx build, with a 750W PSU... as a general rule of thumb i'd have thought this would be a bit on the edge for an 80 class card these days, but we'll see how we go.

 

[Tetras]

Does the BIOS allow you to change the PCI-E gen? That could also be worth a try.

 

[darksun9210]

ok, i "think" i may have a bead on the problem.
trying to get my machine to wake up, i could tell my machine was on, and alive, keyboard lights working, but the screens were asleep and wouldn't wake up. unplugged the display port cables and plugged them back in again, and the screens woke back up again, and all fine and dandy. This is with my known good card. odd but, y'know; PC life, right? anyway. it coughed again a few minutes later with the screens going to sleep so did the cable unplug+replug thing, and back up and working. but worryingly, this was one of the symptoms i saw with my other 3080ti card that i can't figure out what the deal is with it not playing games.
so it was with a bit of a sinking feeling i thought i'd give it another go, before i can get the 3080ti tested in another machine. follow up on what's been suggested here that i've not tried.

so, power off, and try another PCIe slot.

unplug one of the 8pin power lines off of my good 3080, and all fine. try to unplug the other, and i can't. it's jammed on there something solid. then one of the ground connectors pops out of the back of the plug... ahhh shoot. this isn't good. more wiggling, more gentle application of force, and the other 8pin power plug finally comes away from the socket on the card.
doesn't take a genius to see that this isn't a good day. the middle 12 volt connector on the bottom row of the 8pin power plug on the cable - the plastic's all burnt away around the metal connector. not to mention one of the ground lines is now loose in the plug. so now i have a load of crud in the plug socket of my good 3080. I now HAVE to use the unknown 3080ti. so it now goes into the other gen3 16x slot i have available. i can't hard limit the PCIe gen in the bios, so this'll have to do.

fortunately, i do have another set of cables for the power supply that was to go into my sons machine to host my old 3080, so i use those new cables on the 3080ti in my machine.

well would you believe it. Cyberpunk2077, works fine. Forza5, works fine. Horizon Zero Dawn, works fine. shadow of the tomb raider and valheim each still crash at the exact same points they did before.
but i can now run 3dmark speedway stress test and benchmark successfully without a crash back to desktop. same for PortRoyal raytracing bench. that wouldn't even get more than about 20-30 seconds in before crashing out. now it completes and gives me a score.

So, conclusions. i recon i AM getting enough power to the cards, just i need to buzz out these cables to double check that all of the 12V and ground wires are connected fully end to end. the problem being, that i'm guessing the power is not spread evenly over the pins it's expected to go over, like that second connector looks like it was taking the full 150Watts it was rated to take, but i suspect over one pin on the connector. Being the "ti", it just exacerbates the issue. what SotTR and Valheim's issues are i can't be sure. but i can probably do a clean windows rebuild on this machine to see if that flushes out that issue. lots of other things that hit the power hard are fine. Valheim just about gets the card over 200 watts, nevermind the 350 - 360Watts i see other games get these cards to eat.

half a mind just to build a new proper PC, not a butchered 2nd hand office workstation, so i'm not dealing with cables that "should" work, off of amazon, but with proper fully wired up power.
just need a set of needle nose pliers to pick burnt plastic out of my original 3080's power socket. *sigh*. anyway, i'm done for now. good nite all

 

[cjgardens]

i've had some duff luck buying a higher spec Dell PSU off ebay. basicly dropped nearly £170 on what was supposedly a 1300Watt version - and it arrived DoA, and Dell won't do an out-of-warrantee repair as apparently it's not a supported PSU for this system - unlike what was advertised on the ebay auction, and the listing was "no-returns". ergh. paid all that money to effectively ship some e-waste from china

You know that no returns is rubbish and you can return it. If that was the case. Everyone would be selling broken items and saying no returns.

 

[Vimes]

Just adding.....

When I briefly had a 3080Ti I had pretty bad problems when running certain games. Everything else seemed fine.
This was at the time when power spikes from such cards was less known.
Even tho monitoring its use for wattage that would not reveal the momentary power spikes that it would cause issues with my PSU.
The game would either crash or the PC reboot.
Two things resolved this....
One, I realised that the power spikes were a possibility, some posts at the time were starting to appear.
Seasonic replaced my 750w Ultra PSU with a brand new model that could accommodate these events.
Two, I refunded the card as I thought that it offered terrible vfm at the time, I bought it new.

So, for me it wasn't the typical wattage used that was the problem but the power spikes that were not able to capture in software, unless you had a scope to monitor, that caused the issues.

I now have a 1000W Corsair PSU, but the 4080 I now own is supposedly much better in terms of a more stable power demand.

 

[darksun9210]

yeah thats what i think is going on here. agreed. the ti's transient load spikes are probably too much.
there is a 1300W psu apparently available for my machine, but its an ebay lottery as to if i get _the_ one that fits/works. and given the price of the unit, i could just start building a new machine. just need CPU, ram, and mainboard as i already have GPU, SSD, and would have a proper brand new PSU for the price of an ebay gamble.
Ebay pricing the 825W psu at £120-140+, so i when all in and dropped £160 for the 1300W, and then another flipping £40 for shipping what turned out to be e-waste half way round the world.
i so feel i should have dropped that money on a new atx PSU as a starting point for a new known good machine...
Talking to Dell support, seeing if there was anything i could do for an "out of warrenty repair", the 1300W psu that i have is not a supported unit for my 5810, so they wouldn't be able to honour a paid repair anyway...

so.. do i drop more money on an 825W psu and hope that it's all good, even with these amazon MacGyver power cables that i'm now worried about going melty because they are causing stress to other points in an already "out of spec" setup?
its £120 for an 825W vs goodness knows what for a whole new setup...

i might see if the guys at a local repair shop can do anything with the e-waste 1300W psu. it fails it's internal self test, so i'm hoping it's just had an internal fuse pop or something, before i try an ebay return/claim...

thanks for your support and ideas so far

 

[Vimes]

IIRC the new Seasonic PSU that I was sent had improved tolerance for the OCP aspect of its design. I suppose that could be relative to the overall capacity of the PSU, but the point I'm making is that the ATX type of PSU that I had, and now have, are bought with a focus on their ability to perform in a gaming environment.
That may or may not be relevant for you and what you want to do with your Frankenbuild, not knowing much about the design or limitations of the hardware. But understanding that it's primary intended use was an office based workstation.
Just be wary of what you are wanting is going to be achievable with what you want to spend.
The 3080ti is an awesome, if spiky, in terms of power spikes, card.

Good luck

EDIT: Do you have to pay returns for the eBay PSU that is faulty, or can that be refunded..?