Dell XPS 9520 / 3050 Ti Issue

[witblitz]

Hi all,

Hoping someone has a clue here as I'm stumped.

I occasionally game on my XPS 9520 w/ 3050 Ti... The CPU is usually hitting between 85-100 celsius and the 3050 Ti is hitting its peak of 74 celsius pretty much all game until back in the menus... I really don't see a way to bring these temps down apart from re-pasting (a second time) and/or replacing thermal pads.

My issue is that all of a sudden, 10-30 minutes into a game of COD MW2 my framerate literally halves from 70/80 down to 30/40. Nothing else changes... I can change GPU settings in-game yet nothing "resets" this behaviour. You have to restart the game completely.

The CPU performance seems constant so I really think its either a bug in a recent update to the game (that nobody else is reporting), a bug in the Nvidia drivers OR the GPU is hitting some kind of thermal limit... Is this a thing and is there any way to check this? I've just downloaded GPU-z so may get some logs there but I'm really not sure what is causing this out of the blue after almost a year of problem-free gaming. I'm sure most of you will say "driver updates" which may very well be the case.

Anyways, any ideas are welcome.

Thanks in advance,
WB

 

[Kursah]

I have this laptop, but use it for work and not gaming. It still runs hot under some workloads compared to my personal gaming laptop with similar specs (Lenovo Legion 5i Pro 2022), which has better cooling.

Could be a bug, have you ran GPU-z or HWInfo or anything like that to keep track of your min/max/current temps?

Could also be that you need to clean your laptop, take the bottom shell off, remove the dust from the fans, maybe even reseat the heatsinks to make sure the TIM is in good shape.

If still under warranty, you could contact Dell Support, they may report a bug exists or otherwise. Have you ran the Dell Update software recently? That looks for firmware updates too.

Are you aware of anything else that might have changed around the time this started happening? Was it recent enough you could go through Windows Event Viewer to see what kind of Critical or Errors you find near the same timeframe and consistent since then?

 

[witblitz]

Thanks for the reply. I'm not sure what to look for, but I do know that the GPU runs at 74 consistently for a solid year now. Temperatures haven't changed - I get a solid 80 FPS without a single drop but then all of a sudden it tanks to 30 out the blue. Its not a matter of poor performance - its perfectly good performance but then a huge drop which remains. There is a single event that occurs that precipitates this and I just dont know what it is... and its driving me nuts

Im thinking it must be a recent driver update or something... there have been driver, BIOS, Windows and GPU updates in the past week or so... just cant prove it and rolling back all of them is a hassle.

Just wondering if theres a clue to look for when this happens - GPU memory is full, some kind of throttle warning etc...

As you can tell Im not too clued up on this stuff

 

[Kursah]

I'd say run GPU-z and HWInfo in the background while you're gaming, make sure there's not a thermal spike or something that's causing throttling. Though once cool enough the thermal throttle should cease. Which is interesting if it isn't.

I'd suspect drivers, BIOS, etc. Did you try to install GPU drivers directly from Nvidia.com?

Do you use a laptop cooler or any kind of riser? I use a case for my 9520 to allow it to have more airflow underneath, and it did help with my CPU temps. At work I use a lot of programs, windows, tons of browser tabs, etc. all happening at once, so mostly the iGPU and CPU get bogged down.

Here's the case I'm talking about, cheap but effective: https://www.amazon.com/mCover-Shell-Precision-Laptop-Computer/dp/B08HKN5HYQ

I have the black colored case, has held up great over the past year. My boss just broke a leg on his after about 14 months of abuse for reference.

I'd track temps with logging (both GPU-z and HWInfo I believe support logging), install NVidia drivers if you haven't. Any recent updates to MW2? I don't own the remastered version, not sure if there's a spot where this occurs, but sounds more like just after so long the frames cut down as-if a thermal throttle occurred.

Have you considered using Lossless Scaling or something similar to upscale MW2? Maybe try a lower resolution that still maintains the 16:10 ratio. Say 1680x1050, upscale that to 1920x1200. Any changes? Lossless Scaling is cheap on Steam, there's Magpie which is free. Not sure if or how well they work with a game like MW2, I primarily use them for classic games that I want to scale up to full screen that themselves don't scale well. But maybe it'll help with your gaming experience here?

 

[tabascosauz]

Thanks for the reply. I'm not sure what to look for, but I do know that the GPU runs at 74 consistently for a solid year now. Temperatures haven't changed - I get a solid 80 FPS without a single drop but then all of a sudden it tanks to 30 out the blue. Its not a matter of poor performance - its perfectly good performance but then a huge drop which remains. There is a single event that occurs that precipitates this and I just dont know what it is... and its driving me nuts

Im thinking it must be a recent driver update or something... there have been driver, BIOS, Windows and GPU updates in the past week or so... just cant prove it and rolling back all of them is a hassle.

Just wondering if theres a clue to look for when this happens - GPU memory is full, some kind of throttle warning etc...

As you can tell Im not too clued up on this stuff

iirc Windows shouldn't automatically update your GPU drivers on the day-to-day? At least, with the auto update feature disabled under Advanced System Settings>Hardware>Device Installation Settings.

Dell is a vendor that pushes capsule UEFI updates (essentially BIOS updates but pushed through Windows update instead), on my 9370 it is enabled by default and almost caught me flatfooted. I disabled it in BIOS as later updates surreptitiously disabled undervolting. Seeing as you've had the laptop a while, if you download CPU-Z and find that the BIOS build date in Mainboard tab is newer than expected, there might have been a capsule BIOS update.

Hard to say for sure what it is unless you either have a way to actively monitor HWInfo right when it happens, or start logging data (ie. HWInfo, CapframeX) when you play and analyze the data after. The disparity in performance seems about right for a 3050 Ti switching over to the Xe iGPU, since it's not a gaming laptop and isn't MUX'd - Optimus handles the iGPU/dGPU switching on demand, instead of having the ability to specifically disable one or the other.

Also, MW2022 or OG MW2? MW19 and MW22 are incredibly buggy games and sometimes bugs have extremely unlikely and illogical solutions.

If you suspect the BIOS you could check if you are still allowed to rollback (iirc it's an option in the BIOS if it exists, rollback protection or something). If you suspect Optimus is to blame you could DDU and then use Nvcleanstall for latestt drivers.

 

[witblitz]

Thanks for the replies... I do indeed have a laptop stand. Its a must for gaming on these terribly designed XPS's

I just kept GPU-z open while gaming and this happened after like 10 minutes - maybe less:

I see the "Perfcap" reason is thermal meaning (I assume) that I'm running into a performance cap because of thermals.

Also noticed that GPU temp is only 63.8 degrees whereas "Hot Spot" is almost 72 degrees...

Still - weird this never happened before. not even once in a year....

Guess I'm going to have to re-paste :/

 

[tabascosauz]

If you haven't repasted before I'd hold off on that until you're sure of what's going on. XPS serviceability is not particularly good and I think they're still using that hard thick Shin-etsu type stuff that is hard to replace properly with typical liquidlike pastes due to the thickness. I would try a piece of Honeywell PTM7950 instead to fill the gap.

The logging doesn't really show much. The Perfcap reason is constant the whole time. Can you alt+tab out immediately when it happens and get some before data in the graphs? It looks like this GPU-Z data is all just well after the fact.

9520 didn't have great thermals in the reviews and as far as I can tell nothing's out of the ordinary? The difference between edge temp and hotspot temp isn't anything special, you should see what desktop GPUs look like (up to 25C difference).

Perfcap isn't always definitive.

 

[DarkDreams]

Thanks for the replies... I do indeed have a laptop stand. Its a must for gaming on these terribly designed XPS's

I just kept GPU-z open while gaming and this happened after like 10 minutes - maybe less:

I see the "Perfcap" reason is thermal meaning (I assume) that I'm running into a performance cap because of thermals.

Also noticed that GPU temp is only 63.8 degrees whereas "Hot Spot" is almost 72 degrees...

Still - weird this never happened before. not even once in a year....

Guess I'm going to have to re-paste :/

A few things I noticed from this log:
- Hot spot temperature just disappears for some time (maybe a faulty sensor; could cause weird throttling issues maybe?)
- Power readings seem a little low for a 40W chip although it is only at 26% utilization, so maybe accurate
- Voltage for PWR-SRC is off the charts and definitely wrong (maybe because GPU-Z doesn't know the laptop)
- System RAM is stuck at 16 GB, if you have only 16 GB total ram in your laptop you might be running into constant page file access which can kill performance as well

edit:
one more thing: is your laptop plugged in when gaming and what are battery levels like? I've heard about some laptops that don't get enough power while gaming when they start charging the battery.

 

[witblitz]

OK thanks guys will run GPU-Z before and after this happens to see the differences... i do notice that the "perfcap" shows "idle" shortly after closing the game down and then re-appears as soon as im in the game but this was just a quick in and out...

I have re-pasted this laptop before about 6 months ago but I think Im gonna go PTM7950 either way. Im in a very hot climate and I need all the performance I can get

still not 100% convinced that this perfcap is the culprit but I will investigate

thanks again for your help it is really appreicated

 

[Kursah]

Are you using a laptop cooling pad?

Also, consider that laptop case I suggested above. Using the legs allows a lot more space for the intakes to breathe and take more air in, it did help my work 9520 cool down a bit compared to just the stock feet.

Keep us posted!

 

[eidairaman1]

Thanks for the replies... I do indeed have a laptop stand. Its a must for gaming on these terribly designed XPS's

I just kept GPU-z open while gaming and this happened after like 10 minutes - maybe less:

View attachment 302502

I see the "Perfcap" reason is thermal meaning (I assume) that I'm running into a performance cap because of thermals.

Also noticed that GPU temp is only 63.8 degrees whereas "Hot Spot" is almost 72 degrees...

Still - weird this never happened before. not even once in a year....

Guess I'm going to have to re-paste :/

So you know all mobiles are not laptops but notebooks due to lack of cooling all of them have today, they all thermal or power throttle.

The only way they be good anymore is if their chassis was that of the Dell XPS Gen 1/Inspiron 9100 (2004)

 

[witblitz]

RIght so it takes about 6 minutes for this to occur now. Here are 3 screenshots - before playing, start of play and immediately after the FPS drop occurs

1. Idle - before I even start a game. No GPU usage

2. A few seconds into a Multiplayer game with a solid 80 FPS

3. Right as the FPS drop happens - only a couple of minutes into the game... Drops to 30 FPS and stays there

Is there anything here that would indicate why this happens? I see the PerfCap appears almost instantly when I open the game but I still dont really understand what this means/does

i.e. is it just capping performance to remain at 74 degrees? If so this is normal. OR - is it responsible for this sudden FPS halving after a few minutes of play?

Anyways - as always, any clues welcome... Appreciate all your input!

EDIT:

i've noticed that as soon as the performance drops, the PerfCap reason in GPU-z changes from just "Thrm" to "Pwr Thrm" indicating that there is a power draw issue (I think). It seems to happen at the exact time of the FPS drop... still no idea why! Im plugged in with my standard Dell 130W charger and there is very little else sucking power. So there are no anomalies there....

 

[tabascosauz]

@witblitz I see the simultaneous Pwr and Therm flags but it doesn't immediately scream problematic. I think Therm is just being triggered from hotspot at 82C or higher, which is normal even for desktop Nvidia as GPU Boost just starts very gradually reducing freq after that point. Temps are still good.

What I can see however is VRAM clock drops to 200MHz, which would certainly be a huge problem for game performance. I'm not sure if that's normal behaviour because of the alt tab.

In the past few generations on desktop Nvidia likes to err on the safe side and keep VRAM clock high whenever it thinks there might be a 3D load. Works to prevent unpredictable stuttering due to fluctuating VRAM clock.

Are you able to run some sort of OSD in-game that shows clocks? I think that's the only way to tell what's going on - if you add a second screen or alt-tab both will interfere with GPU behaviour.

What profile are you in, Ultra Performance?

 

[witblitz]

Thanks @tabascosauz

I use Greenshot for screenshots which immediately freezes both screens so the alt-tab isnt an issue

Regardless - I completely uninstalled all Nvidia drivers and software and installed an older version of my GPU driver. Issue now seems resolved - far less PerfCap going on in GPU-z which is odd. the GPU is behaving differently

i.e. drawing less power / getting less hot

I only tested with one game but will try do a longer session later.

Anyways thanks again for your help and sorry to have wasted your time.

Peace!
WB

 

[tabascosauz]

Thanks @tabascosauz

I use Greenshot for screenshots which immediately freezes both screens so the alt-tab isnt an issue

Regardless - I completely uninstalled all Nvidia drivers and software and installed an older version of my GPU driver. Issue now seems resolved - far less PerfCap going on in GPU-z which is odd. the GPU is behaving differently

i.e. drawing less power / getting less hot

I only tested with one game but will try do a longer session later.

Anyways thanks again for your help and sorry to have wasted your time.

Peace!
WB

Interesting. I still can't find an explanation for why it's suddenly doing that after x time. Happy to help, let us know how it goes.

 

[witblitz]

Sigh... looks like I spoke too soon. Same thing happened but took a lot longer than normal...

Anyways, here's a screenshot a few minutes after it happens. I'll leave this here just in case anyone has any other ideas

 

[DarkDreams]

Sigh... looks like I spoke too soon. Same thing happened but took a lot longer than normal...

Anyways, here's a screenshot a few minutes after it happens. I'll leave this here just in case anyone has any other ideas

The memory clock drops sharply at the same time as memory controller load goes up. I assume that is the time the problem starts. Looks like you might be running out of VRAM. You can try setting texture quality to lowest and lowering resolution to see if it still happens with less load on the video memory. I also noticed your task manager says your system memory is shared with the GPU, maybe try turning that off and see how it behaves.