cancel
Showing results for 
Search instead for 
Did you mean: 

Graphics Cards

danpets
Adept I

AMD Pro w6800 Temperature

I just bought an AMD Pro w6800, and while working with my AutoCAD, GIS Mapping and a few Games with ray tracing on my GPU temps are around 86c. To me that is very hot for a GPU. The cooler is spinning and it is throwing a ton of hot air out the back of the pc. What is the safe high temp for the w6800 and what can be done to fix this (ie an aftermarket cooler)?

My PC is an Epyc 7543 with 256gb Ram, 2Tb 980 Pro. The ambit room temperature is 20c and I'm only running one 5k x 2k monitor. Any ideas would be appreciated. 

18 Replies

I don't know how accurate this Maximum Temperature Website is but it shows the Maximum Operating Temperature for the W6800 is 99c:

Screenshot 2022-04-09 210010.png

If the above is accurate then your Professional GPU card is running below its maximum operating temperature of 99c.

This Tech website mentions that the author believes the W6800 needs a slightly better cooling solution than the one it has now: https://www.igorslab.de/en/amd-radeon-pro-w6800-32-gb-in-test-beats-the-workstation-card-a-nvidia-qu...

The above link shows the performance using AutoCad 2d & 3D.

0 Likes

Wow, 99c is extremely hot. Even at 86c I see the card throttling down, I wonder how bad it would be a 99c? I guess I will just use it as it is and hope that this card does not burn the house down. Also if any one knows of a better cooling solution I would be interested in trying it as I was hoping to keep the temps down to the point were this card would not throttle the clock down. 

Thank you for the help on this card, as finding information is not readily available. 

0 Likes

Well like I mentioned I can't verify if the Maximum Temperature is 99c from that website image I posted.

AMD Moderator of Professional GPU card can probably verify if that is truly the correct information from that website and possible GPU cooling solutions for you @fsadough 

0 Likes

Thank you for your time and in-site elstaci, I hope an AMD Moderator will be able to answer my questions. 

Oh and I am still impressed with what this card can do. I just wish I didn't have this issue to deal with.

0 Likes

Thermal throttling (throttling clocks and power reduction) will occur if either junction or GDDR6 memory temperature exceeds their limit of 110 and 100 degrees respectively.

I understand the reason for throttling. I wanted to know if my card doing this at 86c was by design or is there an issue with my card? Or should I look for an aftermarket cooling solution for my w6800 to keep is cooler?

 

0 Likes

You will lose your warranty by replacing the GPU fan. Are there any problems with the GPU?

0 Likes

I would consider it a problem when the GPU throttles down doing workloads that it was built for. So Yes it is not working as designed, but getting AMD to also except it like that is another story. My thought is if I'm paying north of $2K it should work like intended out of the box under the designed work loads. I have had problems like this in the past with AMD cards (gaming GPUs) and thought that getting a Professional workstation card I wouldn't have these types of issues. I do think that AMD should look at the cooling solution as if they can fix that the w6800 would be a great card for more people.

0 Likes

If you think there is a problem with the GPU. I need detailed info to your system. Preferably an AMDZ-Report included.

AMDZ Report
- Please extract the amdz-v311.zip available from https://we.tl/t-xKeE2sH9ua
- Run amdz.exe file as an Administrator
- Select “Save All“ and “TXT“ as the output format
- Click on the blue button to save the report
- The .txt file will be saved in the same folder where you extracted the zipped file

0 Likes

Here is my problem, I don't think the GPU is bad, I do think that the cooling solution is not quite right for this GPU configuration. I will get you the information your asking for, but to me it's an assembly/cooling thing not a GPU thing.

The software I use is called Fan Controller https://github.com/Rem0o/FanControl.Releases. Its free software from GitHub and this thing works great. You can watch a video about it here https://youtu.be/uDPKVKBMQU8 this is a YouTube channel I watch a lot very informative and helpful guy. Not really a fix or anything but very useful software for PC.

0 Likes

Thanks for your comment It’s a cool solution, However, this card launched last year with a 3 year warranty anyone who has it is still under warranty terms and conditions so no external tool or cooling solution can be used (I asked support) besides the ones provided by AMD and even then you cant tweak without voiding the warranty. And let me be clear, enterprise and Adrenalin drivers have the option to set target and minimum fan speeds BUT if you use them you are voiding the warranty. AMD made a mistake I intend to make them correct that mistake maybe recall all those cards so called PRO but that are unable to move even the “certified” unreal engine or retract those flashy comments promoting this card as the second coming of JC ..Anyone reading please reach out All your replies are VERY IMPORTANT FOR ME ..this WILL get interesting and drama will spawn … I will travel to AMDs headquarters and chain myself to its doors if needed but I won’t loose a year of my savings in this PRO shenanigans 

29D903F2-F0DF-4DAE-9364-D11A43F6A592.jpeg

 

0 Likes
Alexa22
Adept I

i had the same issue I bought few months ago for my website designing and graphical work but i faced the same issue

0 Likes
Jabyer
Adept I

Hey .. did you get any solution for the temperature problem?? I got my w6800 last week and I'm pretty concerned about that problem I noticed after a huge render that the junction temp was 110c and GPU temp was 96c. I don't think you're wrong about the card being a potential fire hazard . Please update your experience if you can.. Thanks for taking the time

0 Likes

I've been reporting the same issues for months now. To me it seems as if the fan doesn't spin up as much as it should, and I've resorted to manually raising the minimum fan speed in the Radeon utility. Same issues whether using Enterprise or Adrenalin drivers. For extended periods of rendering (hours) of 3D fractals, I'll typically set the minimum fan speed to 2,500rpm (max fan speed slider (Target) set all the way to the right at 4,950. I generally cannot run games on anything like high - ultra without crashing the whole machine. GTA5, Forza 4 Horizon I set fan to 2,500. Cyberpunk 2077 I set fan to 3,000. Ray tracing in Cyberpunk is a no-no, causing GPU to get seriously hot to the touch pretty quickly. It might be helpful if AMD could give us the ability to properly set fan curves, instead of just two sliders. I realise this is not a gaming card, but I would expect to be able to run a game like GTA5 at max settings without issue. 

ASUSTeK COMPUTER INC. PRIME B450M-A (AM4), AMD Ryzen 7 2700X, 32.0GB Dual-Channel GSkill, 

AMD Radeon PRO W6800, Samsung SSD 970 EVO 500GB, Windows 11 Pro 64-bit.

0 Likes

Hi, Have you tried to ask for a refund or exchange>?. Today my machine just shut down reaching 105c (in the picture) and 110c when it went off. I was running an after-fx animation using element 3d with raytracing ambient occlusion, first, it went up to 99 then, the fan started to spin really fast, render time started increasing, and then it just shut down.

Later today, I opened the car's sample scene in Unreal Engine, and the same behavior, a spike in temperature before shutting down. I didn't even set ANYTHING to ultra or raytracing it took 1 minute NOT EVEN MOVING THE CAMERA and it hit 100+c and shut down.

In my opinion sentences like

" this GPU will politely laugh at projects that before made your previous GPU nervously sweat and shake. Leap into those projects knowing that this GPU is here to support, not let you down.

"As your projects get bigger and more demanding, your GPU can keep pushing workloads towards that impending deadline."

"Built from the ground up, with the established graphics foundation of the leading games consoles. AMD RDNA™ 2 Hardware Raytracing is now available for professional graphics."

"Modern workstation users demand performance, stability, image quality and software certifications for increased productivity, while using more graphically intense applications than ever before. Discover how the entire AMD Radeon PRO graphics range answers this challenge."

"Our GPUs are designed for 24/7 use in modern professional environments. This extends to the physical design of the GPU where the predictable heat generated when under full load, is carefully exhausted out the back. And not into your powerhouse of a machine. We keep cool (and quiet), so you can keep your cool when things get hot."

ALL THOSE are false advertising!! and I feel we should get compensated for our trouble. In my use case those applications, I mean after fx or Unreal Engine don't get more professional than they are at the moment, I don't know how other "PRO" hardware works, I come from using a GTX1070 but after READING the description AMD gives of the pro W6800 you expect AT LEAST be able to open ANY SAMPLE in unreal, AT LEAST a 1080p render with raytracing in after fx, so I feel we deserve better than "you're operating within parameters"

In my case, I imported this hardware to my country paying a 50% extra for VAT and import FOB taxes, so I'm ready to give the fight CONTENT CREATOR style. I invite you, and anyone reading to join and share your experience. Let's make some noise! Ill upload my experience very soon to you tube

 

 

 

 

 

IMG_5436.JPEG

IMG_5464.JPEG

0 Likes

Hi there, So here are some pointers I found and tested,  

0. AMD pro software is a MUST if you want to tame this girl. And let me be crystal clear, all your problems will go away as soon as you understand how this piece of software works.

 1. AMD Pro software is more than just the driver, In the performance controllers (the ones you can use without the warranty warning) you will find a frame rate limiter. The card will work within parameters all the time but its max parameters are kind of wild if you let it take control it will try to perform its max that's why it goes fan and temp crazy and then it crashes when performing some demanding apps.  This is the way you also tune games like cyberpunk or in my case Eve-online which was also hitting the system BTW 

2. The viewport utilities will let you work ultra-fast so make sure you take into account what are you going to work on and how that function impacts your performance, for example in unreal engine that will lower the resolution during movement and increase it in a kind of flash render it is really useful (there is an unreal plugin btw)

3. Aperture of the PCI-E and the way raytracing works are still a mystery to me but I will post later

4. To avoid tuning and retuning you can save performance profiles so you can switch the way you work on the go

5. Remember the PRO part is like an O.S. the more flexible the more you have to set and configure, BUT the multitasking is the key at least for me/ that is really impressive

To close, first sorry if I bother everyone with my posts but I was really lost, and to be FAIR AMD should walk the extra mile and put together more documentation and "how-to" step by step. Most of my problems came because there is no clear directions on how to work with the tool in an intended way and let's be real this one brings A LOT of o things to configure, there are even render SDKs for specific tasks,,, At this point, I feel a little better, I did want a plug and play solution but I guess I'll have to learn on the go....

0 Likes

I've also reached out to AMD support but they unfortunately couldn't help me. In the end, I did manage to fix the problem myself, so I figured I'd describe it here.

The problem lies in the power play table that resides in the card's VBIOS. The firmware running on the GPU uses this table to determine power limits, fan speed, etc. The table that's programmed into the W6800 has a hard limit on the fan speed of 2150 RPM. With good, unobstructed airflow into the GPU and at normal ambient temperatures, this is just about sufficient to cool the GPU and keep it at around 100°C (which is the target temperature set in the VBIOS). I suspect AMD put in this limit to prevent the GPU from getting excessively loud under load.

Unfortunately, if the computer's airflow isn't perfect, the card overheats. I've got another card plugged into the slot below the W6800, which I can't move, so my card overheated because the airflow is slightly obstructed.

It turns out that you can modify these VBIOS-provided firmware parameters and increase the fan speed limit, but you have to do a bit of a dance to make the GPU's firmware actually reload them.

I'm using the W6800 on Linux (for compute with ROCm), so I'm not sure if this is possible on Windows too, but it should be doable somehow.

 

You can generate a new power play table with a higher fan speed limit and lower temperature target (3200 RPM and 90°C, respectively) like this:

1. Copy the power play table into a file: cp /sys/class/drm/card0/device/pp_table ./pptable.bin

2. Install a power play table editor ("upp") with pip: https://github.com/sibradzic/upp

3. Run upp to modify the power play table: upp -p pptable.bin set smc_pptable/FanAcousticLimitRpm=3200 smc_pptable/FanThrottlingRpm=3200 smc_pptable/FanTargetTemperature=90 --write

4. Optionally, verify the new pptable with upp dump and compare it to the old one

 

Once you have the new pptable (which you of course only have to create once), you have to perform the following little dance to load it and make the GPU firmware actually use the new parameters:

cat pptable.bin | sudo tee /sys/class/drm/card0/device/pp_table > /dev/null
echo 1 | sudo tee /sys/class/drm/card0/device/hwmon/hwmon6/pwm1_enable
echo 2 | sudo tee /sys/class/drm/card0/device/hwmon/hwmon6/pwm1_enable

These three commands have to be run every time you boot the system. The paths might be different on your system, so you might have to change them slightly (i.e. different card number, different hwmon number). You could put the commands into /etc/rc.local or a systemd unit file to automate this process. It's not necessary to modify ppfeaturemask; this works without GPU overdrive.

The first command loads the new pp_table, which should be pretty obvious. The other two less obvious commands put the card into manual fan control mode temporarily (mode 1), and then back into automatic firmware-based fan control (mode 2, the default). It turned out to be necessary to poke the card like this to make it actually read in and use the new parameters.

I hope this is at all useful to anybody. Of course, if you decide to follow these instructions, you do so at your own risk.

It's likely also possible to write this pptable into a VBIOS image and flash that to the card, but doing that will likely also void the warranty (and it's quite risky). I've asked AMD support if they could provide a VBIOS with these changes, but so far they unfortunately haven't.

Maybe a mod could poke AMD staff about it? Modifying the VBIOS pptable fan speed limits as described above would make the W6800 much less prone to overheating, at the expense of a bit of noise.

With the new pptable loaded, as described above, my W6800 stays well below 90°C - problem fixed. It is louder, though.