Graphics Cards

[whk]

Hello, today I bought a 4k monitor and I started to play doom eternal and cyberpunk 2077 with all the graphic quality to the maximum, with doom eternal I have no problems but with cyberpunk I have a little drop in fps, but the main problem is that During both games the temperature sensor shows 80, 108 and 78 degrees celcius respectively.

I use Ubuntu 22.04 LTS and would like to ask if anyone knows if those temperatures are normal or not. I have read on the internet that 100 degrees of temperature is too much for my graphics card but maybe it is not the sensor that really matters, I don't know, I have used the Psensor to observe the temperature and the GPU has 3 visible sensors.

I have temperature problems or am I misinterpreting the information?

My pc case uses 3 fan corsair water radiator to cool the CPU, besides it has 3 fans in the front and one in the back, the front three draw air in and the rear draws, the top three fans from the radiator also extract.

[EFermi]

What specific model of GPU are you using? Is it reference AMD or some sub-vendor cooling design? What happens to the temps if you ramp up your GPU fan speed to 100% through a third-party app like Afterburner? What happens to the temps when you remove your case side plate (yes, I've heard you have fans in your case but we can't rule out a "what if")? Does it have enough "breathing room" between GPU and PSU (demonstrated by the arrow on picture below)? What are your ambient temps?

Spoiler

Having over 100C on HotSpot sensor is understandable for severely overclocked GPUs that have their clocks ramped up and power consumption increased above manufacturer's standarts via some special tools like MPT (e.g. I have a reference design 6900XT but set the PowerLimit to 350W, full 53 watts over AMD's settings even if +15% PL was applied, so I had 82-83C GPU (didn't check hotspot, but the card didn't throttle so it was below 110) while gaming during summer, when it was 30~33C ambient, with a customized fan curve), but that shouldn't be the case for settings within factory specifications... unless you are for some reason running a "quiet" BIOS instead of "performance" on a dual-BIOS card, or the sub-vendor cooling quality in general is at a "chinese basement manufacturing for aliexpress scam" level.

[whk]

Hello, thanks for your answer, attached image of my setup:

• CPU: Ryzen 7 5800X3D - https://www.amazon.com/-/es/gp/product/B09VCJ2SHD/
• Motherboard: msi x570s edge max wifi - https://www.amazon.com/-/es/gp/product/B09BYXFHLR/
• GPU: Radeon RX 6800 XT - https://www.amazon.com/-/es/gp/product/B08TJ2BHCQ/ - XFX Speedster MERC319 AMD Radeon RX 6800 XT CORE
• Cooler Corsair iCUE H100i RGB Pro - https://www.amazon.com/-/es/gp/product/B08G1Q3GZR/
• Power suply: Pystar 1000w modular gold - https://www.amazon.com/-/es/gp/product/B096KCD9LD/
• Ram: 128GB Kingstore Fury - https://www.amazon.com/-/es/gp/product/B097K3L88K/
• The room temperature in my room is around 18 degrees Celsius.
• Bios settings CPU & Ram with XMP enabled.
• OS: Linux (Ubuntu 22.04 LTS) with amdgpu driver over xorg session.

I have unplugged the second monitor and the temperature has dropped to 90 degrees celcius, but I don't think it is the best solution, during the night I will try opening the side cover to see if the temperature drops with both monitors connected.

[EFermi]

There's no need to drop your entire PC spec for this issue, I'd have asked you otherwise.
XFX Merc GPU series are known to be quite decent performers. 

According to this guru3D temperature graphs, your card is at least 8-10C above what it could be under normal conditions. TechPowerUp also agrees with that, with a couple degree of difference margin. So do Igor'sLab and this lesser-known benchmark resource. Although here legitreviews hit 92C on hotspot, it's still much less than what you have, and the GPU overall was at only 72C.

Even if there is (and obviously always would be) a certain margin of error between GPUs in how they behave temperature-wise under load, you shouldn't be over 100C on hotspot if you didn't severely overclock.

Some off-topic: what do you need so much RAM for? Do you do some specific kind of work that requires lots of memory, like running multiple VMs or whatever? I feel like having 64Gb is overkill already, I barely ever see it half-used.

Also, if all else fails and the issue persists, you can try locking your framerates in games via VSync, RTSS or through the driver, so that your card won't be crunching frames above comfortable threshold, but in Cyberpunk it's not of much help since the game is very demanding and there's no excess performance to cut off in exchange for lower thermals and/or noise. Another option you might try (that I would NOT recommend, but still, if there's no other choice to drive the temps down) is undervolting.

[whk]

---

@EFermi wrote:

Some off-topic: what do you need so much RAM for? Do you do some specific kind of work that requires lots of memory, like running multiple VMs or whatever? I feel like having 64Gb is overkill already, I barely ever see it half-used.

---

My team uses it to work and play, I am a security analyst and developer, sometimes I have to use high-stress tools or sometimes I have to set up different virtual machines with windows to carry out work with clients that require it, especially with somewhat conflicting vpns, in the past I had 12gb of ram but I couldn't keep the virtual machine connected to work and launch a good game at the same time during off hours, I had to close the virtual machines, so now with the ram I have it allows me to have multiple more virtual machines the android emulators when I am performing security tests on mobile apps, in addition to having endless tabs open and debugging applications running and at the same time from time to time launching a good game without affecting the fps in any way, That's why I had to change from ssd to m2 wd black, in fact I had to put two, one exclusively for steam and avoid r io problems while running virtual machines.

I've seen some liquid cooling solutions for gpu, but I don't know if it's worth it or if it gives me the cabinet space.

When I had to choose the parts for my PC I had compatibility and performance on Linux very much in mind, I understand that nvidia has many compatibility problems with linux and to get the best performance out of a radeon gpu an amd ryzen cpu is better, so I decided to go for all the best recommendations from the amd official sites, in fact the motherboard I bought was the one recommended by amd and everything has worked very well for me except for the gpu temperature that worries me.

In the past I had a pc with an overclocked i7 and a nvidia 1060 gpu but the games had compatibility problems and not all the effects or shadows could be seen and it gave a lot of problems when configuring the monitors in gnome, but with the radeon I no longer have those problems, the older games that I like like quake 4 or bioshock infinite look fine.

I have been a linux user since approximately 2009 but I also really like games and I have my work responsibilities, I also like design and music and I think that the only system that accompanies me in all things is gnu/linux now that windows is a system that does not help me in my work as a fullstack. Now with my amd gpu I think I'm finally complete at the hardware level, both in performance and compatibility.

[whk]

Today I tried to remove the side cover of the pc, the temperature now remains stable between 95 and 100 degrees celcius, now it does not exceed 100 and most of the time it remains at 97. There is definitely a ventilation problem in my pc despite having all the fans that the case supports, I will have to find a better alternative.

[kingfish]

Try this....In the power plan (high performance) , advanced settings, change the maximum Processor setting to 99% Max. Reboot your computer play the same game/same settings. .run Hardware Monitor in the background and look at the Maximums.

https://www.cpuid.com/softwares/hwmonitor.html

If it doesn't solve the issue simply return processor to 100%

[Indra881]

Did you read the topic? 

Why are you suggesting this? First, What is the point of disabling turbo boost for GPU temperature?

Second, this is not the correct way to disable turbo boost and OP using ubuntu.

[chiefunplugged]

Depending on how hard, and how long that card has been running. It might be time for the cards thermal pads anew!  and some new paste. would bring those temps down.

[hrpuffnstuff]

Just for comparison using W11 22H2 I run a XFX Speedster MERC 319 RX 6900 XT black and while gaming with it drawing a steady 290-300w I see temps around 180° F while my 5950X is hitting around 155° F.

Just looking at your case you don't have enough air flow over the card.  I run an old school CM HAF 932 with 4 120mm Antec Prism fans blowing directly onto the card from the side panel, another 120mm prism exhaust over it at the rear, 3 exhausting at the top on the radiator and a 230mm at the front blowing down its length.

I also run temp probes in the radiator and the rear exhaust fan so I can see the relationship between my cpu and gpu temps vs the exhaust temps.  At peak I see around 80°F difference between the gpu and rear exhaust temps while my cpu is around 40° F.

[whk]

Thanks, I'll look on amazon for an additional vent just for the gpu. Like as? https://www.aliexpress.com/item/1005003264600722.html

[whk]

It partially solved the problem, while I have the pc lid open and running a heavy game I put my hand inside to feel the airflow and it turns out the cpu radiator was pushing air in instead of blowing it out. I have changed the position of the fans and activate the rage hardware bios mode with the small switch :p, and now everything has lowered temperature, the m2, the gpu and the cpu, now when I play the gpu it remains very stable between 98 and 100 degrees with the lid closed, a big change, but still I'm a bit scared of keeping the temperature constant at ~99.

[Geforsikus_2021]

go to the settings of the amd adrenalin driver, and manually slightly reduce the voltage on the GPU, I have 1075 by default, I have lowered to 1025. and you can try to lower the frequency by 100-200 MHz, and play, check the result, I also recommend manually setting the rotation curve of the cooler, by default, the rotation of the cooler is set to a very quiet mode, and in fact are not particularly cooled even at critical temperatures.

[whk]

Sorry, but there is no official software from AMD to configure the GPU in Linux, Adrenalin does not exist for Linux , i used other free software for change fan to manualy but it is worse. Unfortunately AMD support is focused only on windows, not even the official driver still supports ray tracing natively on linux.