Potential thermal throttle on my one of my 3080's, pls help :-(

[elontusk69]

I recently built two rigs, one with an evga 3080 ftw3, (got lucky at microcenter), & in my other rig I have a Gigabyte RTX 3080 Xtreme, the one with an LCD.  (payed scalper price, but was bnib, with receipt from microcenter)  Before I built my second rig, I put the new Gigabyte 3080 in, just to make sure it's working, didn't notice anything weird and everything checked out, didn't have it in my rig for too long. Moving forward to now, I build my second rig and I got a riser /cable for the card, I got it so I can mount the gpu vertically, and I didn't notice any issues. At first glance, both systems work fine. I would say for most idle, and web browsing, light usage, they are both identical. But from gaming on for anything longer than a few minutes, I've noticed a significant difference between the performance of the two cards, of about 10-30% fps drop on the rig with the Gigabyte card. It's consistent across all games(except really basic games, that run identical on both), and upon looking further in the situation, I've noticed that the EVGA card runs at 300w+ when under a load, and the gigabyte card is only running at around 170w when playing/it being under load. I swapped the cards out to see if maybe it was the riser, or a difference between the systems, and the EVGA card was pulling 300w+ no problem on the riser. Upon looking further in to comparable stats to try to troubleshoot, it looks like the Gigabyte card is running almost 10-20c cooler than the EVGA card (under load)  but the VRAM temp on the gigabyte card is much higher, despite the cooler overall temperature. My guess is that the gigabyte card is thermal throttling the performance due to the high vram temp upwards of 110c+, but is there nothing that can be done? despite the card running cooler than the EVGA one, I am so confused. If anyone knows anything that would help, I would appreciate it. (Even though the card temp was cooler, I took 3 120mm fans and pointed them at the card, as well as put it on an open bench to try and make sure it's as cool as possible, still reaching those high vram temps.. 

[Maticks]

There are a few threads about riser cards causing issues with FPS. with Riser Cards there are many types of quality with them and also the 3080 sucks a decent amount of power over the PCI Express port.

I would probably try switching your EVGA and Gigabyte cards around so the Gigabyte card was on the motherboard directly.

See if the FPS pick up and the EVGA card starts dropping frames.

 

I did see some people setting the PCI Express slot to 3.0 Forced when they have a 3.0 Riser Cable. It's likely one of these problems that has several fixes.

is it this type of riser cable?

 

If it's one of these ones then thats likely your issue.

[Nathanpete]

Your GPU's heatsink may not be making contact with the VRAM. There is a supposed to be a layer of thermal pads in between, but it is possible that they are missing or misplaced, an error that likely occurred during manufacturing in China. You can disassemble the card, and as long as your have some isopropyl alchohol, a rag, and some more thermal paste to fix the dried paste that results when removing the cooler, you can confirm for yourself if the VRAM is making contact with the memory dies. You can probably find a tear down guide for your specific GPU on yt, or a more generic tear down for 3080s might also help, especially if the GB card uses the reference board. 

 

Contact between the memory and heatsink can be confirmed with impressions marks being made on both side of the thermal pads that lay atop of the VRAM. Take a pair of tweezers or a very thin wedge, and peel up. If there are clear marks outlining memory modules on one side, and marks where the GPUI cold plate made contact with the thermal pads, then your problem is stranger. 

 

Better yet, if you open up your graphics card and there aren't thermal pads on some or all of the VRAM dies (they can stick to either the dies on the board, or the metal on the heatsink, so look at both), then that is your problem: An error in the manufacturing of your specific card has resulted in a lack of contact between the VRAM and the heat sink. 

 

Please post back if you decide to open up your card and peek at the VRAM cooling. It could also be possible, depending on your GPU shroud, then you can peak in the side and looking closely under the heatsink it may be possible to find if there is a thermal pad atop the PCB between the relatively short VRAM modules, but you can only confirm like 2 or 3 of the modules with this method, not all of them. Also don't get them confused with the much taller VRMs beside it, which also require thermal pads. 

[elontusk69]

On 4/26/2021 at 5:16 PM, Maticks said:

There are a few threads about riser cards causing issues with FPS. with Riser Cards there are many types of quality with them and also the 3080 sucks a decent amount of power over the PCI Express port.

I would probably try switching your EVGA and Gigabyte cards around so the Gigabyte card was on the motherboard directly.

See if the FPS pick up and the EVGA card starts dropping frames.

 

I did see some people setting the PCI Express slot to 3.0 Forced when they have a 3.0 Riser Cable. It's likely one of these problems that has several fixes.

is it this type of riser cable?

 

If it's one of these ones then thats likely your issue.

I have that same riser actually, but it works fine with my other card, is the weird thing. I took your advice and put it directly on the motherboard, and I got excited because the card was operating as normal, but then after more than like 20 minutes it began to lose performance. I actually put it on my main/original rig, and had the EVGA card running on the riser, the EVGA card did not lose performance, and the Gigabyte began to throttle after a while. I am starting to think I am going to have to tear apart the card to fix this, I'm thinking now it's something to do with the thermal pads. 

[Maticks]

you've switched out the rigs they are running on so that rules out drivers as well as other factors but you've also plugged it onto a motherboard directly.

Based on that it's clearly something physical with the Gigabyte Card, there was a decent amount of talk about the VRM Thermal pads on the Gigabyte 3080 people have been replacing them with better quality ones with a lot better results.

 

There have been a few memory overheating and throttling posts but mostly related to mining not gaming.

But i would imagine stressing a GPU with gaming at high utilization for 20 minutes will result in something similar.

It does sound like a heat soak situation where the card heats up and then its throttling. 

[Nathanpete]

On 4/29/2021 at 12:44 AM, elontusk69 said:

I have that same riser actually, but it works fine with my other card, is the weird thing. I took your advice and put it directly on the motherboard, and I got excited because the card was operating as normal, but then after more than like 20 minutes it began to lose performance. I actually put it on my main/original rig, and had the EVGA card running on the riser, the EVGA card did not lose performance, and the Gigabyte began to throttle after a while. I am starting to think I am going to have to tear apart the card to fix this, I'm thinking now it's something to do with the thermal pads. 

Can you reply with photos of profiles shots showing what both GPUs look like inside your case when vertically mounted. I need to know how close the GPU is to the side panel. If the gigabyte card is wider than the EVGA card, or there is too much shroud, you card could be what is called "choked for air." This is usually a problem imposed by inconsiderate case designs, or a case that was made before cards significantly wider than 2 slots were popular. 

[elontusk69]

On 4/30/2021 at 9:08 AM, Nathanpete said:

Can you reply with photos of profiles shots showing what both GPUs look like inside your case when vertically mounted. I need to know how close the GPU is to the side panel. If the gigabyte card is wider than the EVGA card, or there is too much shroud, you card could be what is called "choked for air." This is usually a problem imposed by inconsiderate case designs, or a case that was made before cards significantly wider than 2 slots were popular. 

Yeah I have a helios strix case on where it will be vertically, but I never put the glass panel on over the case over the main components unless I'm done tinkering/adding fans etc. I wasn't even done with the whole build before I noticed the issue  But I also tried putting it outside and adding fans, the gpu gets cooler but the vram doesn't. So far I've only found one video on my specific card, and it was around mining. Said the back thermal pads are not covering all modules, and on the front they're low quality or something. I guess I'm gonna try to replace the pads, kinda hoping for a decent waterblock to be released for it lol I will definitely be posting pics when I open it, I'm really annoyed that only because I thought I would water cool the evga one first, just because it already has multiple blocks available, but oh well lol 

[elontusk69]

On 4/29/2021 at 5:06 AM, Maticks said:

you've switched out the rigs they are running on so that rules out drivers as well as other factors but you've also plugged it onto a motherboard directly.

Based on that it's clearly something physical with the Gigabyte Card, there was a decent amount of talk about the VRM Thermal pads on the Gigabyte 3080 people have been replacing them with better quality ones with a lot better results.

 

There have been a few memory overheating and throttling posts but mostly related to mining not gaming.

But i would imagine stressing a GPU with gaming at high utilization for 20 minutes will result in something similar.

It does sound like a heat soak situation where the card heats up and then its throttling. 

yes, I found a video about the exact card, and it was from a miner saying the hashrate went down, and it was due to the thermal pads on the back not covering all modules. I think it's normal to replace the pads after a long time, but I think it's weird considering how I haven't used the card too much. But looks like I'm gonna replace the pads and hopefully that fixes it lol