GPU Temp: 68°C, Hotspot Temp: 100°C (Sapphire Pulse RX 6700XT)

[oxrufiioxo]

Sapphire said these temperatures are normal for this GPU. I will continue to use it this way. After the warranty period expires, I will replace the paste and pads myself.

I haven't had a chance to try DDU yet. This weekend I will uninstall the driver in safe mode with DDU and do a clean installation. Maybe I can even reinstall Windows from scratch, depending on the situation.

Your temperature values are almost the same as my GPU. Maybe there's really nothing to worry about. Thank you very much for the information you provide.

I've tried. Trixx is a terrible program. When I installed it, only 1-2 menus appeared. When I deleted and reinstalled again, the number of menus increased to 3. When I researched, I saw that there were some that did not even show GPU information, so I deleted the program completely.
For example, after installing it, only this information was available.

View attachment 316367

I was curious and had a friend install it. It looked like this too:

View attachment 316368

I made a slightly more aggressive fan profile and now I'm seeing 65/88C same 20c ambient only other difference is it's in a case now it does have 2 140mm fans blowing air right into it from the bottom of the 011 mini. This was the best of the 3 though but they where all similar and honestly the other 3 RDNA2 GPU's I've worked with where similar with at least a 20C delta. Honestly I wouldn't worry too much about it though they seem fine up to 110C. Run Timespy and see how your system performs compared to others.

Result not found

www.3dmark.com

My Results so you can somewhat compare.

 

[torkema61]

I made a slightly more aggressive fan profile and now I'm seeing 65/88C same 20c ambient only other difference is it's in a case now it does have 2 140mm fans blowing air right into it from the bottom of the 011 mini. This was the best of the 3 though but they where all similar and honestly the other 3 RDNA2 GPU's I've worked with where similar with at least a 20C delta. Honestly I wouldn't worry too much about it though they seem fine up to 110C. Run Timespy and see how your system performs compared to others.

Result not found

www.3dmark.com

My Results so you can somewhat compare.

I did these tests when I first got the GPU. I don't remember the test results, but there was no problem with performance and score. Anyway, thanks for sharing your test results.

I said that when I undervolt, the fans always run at 75%. At 75%, the fan noise was quite annoying for me, but the temperature values were quite low.

GPU Temp: 54°C
GPU Hotspot Temp: 68°C
Fan Speed: 2200-2400RPM

 

[oxrufiioxo]

I did these tests when I first got the GPU. I don't remember the test results, but there was no problem with performance and score. Anyway, thanks for sharing your test results.

I said that when I undervolt, the fans always run at 75%. At 75%, the fan noise was quite annoying for me, but the temperature values were quite low.

GPU Temp: 54°C
GPU Hotspot Temp: 68°C
Fan Speed: 2200-2400RPM

For some reason my Fans where not spinning over 1600RPM at stock so I just raised them to 1850RPM this is a just for fun system so the noise isn't too much of a problem.

 

[torkema61]

For some reason my Fans where not spinning over 1600RPM at stock so I just raised them to 1850RPM this is a just for fun system so the noise isn't too much of a problem.

It is not very disturbing below 2000rpm, but when it goes above it, there is an annoying hum for me.

 

[RJARRRPCGP]

My Pulse 6700XT is about the same 70c edge 95c hotspot swapped it twice and all 3 where about the same margin of error anyways. That was outside of a case in a 20C ambient room.

Those temps *suck* for especially a 20C ambient room! Make sure you don't have the 0-RPM enabled! Then run the fan(s) at least about 2,000 RPM. Ideal would be 1,800-2,500 RPM.

My Nitro+, OTOH gets much better temps!

 

[oxrufiioxo]

Those temps *suck* for especially a 20C ambient room! Make sure you don't have the 0-RPM enabled! Then run the fan(s) at least about 2,000 RPM. Ideal would be 1,800-2,500 RPM.

My Nitro+, OTOH gets much better temps!

I got the card for less than 300 and it's basically a base model so sure something locally that cost 60+ more should get better temps. Although I did try a nitro+ 6800XT and the hot spot also sucked as well as well as a Red Devil 6700XT and pretty much every other RDNA2 gpu I've used only during actual stress test for 4-6 hours though in something like CP2077 after a couple hours it's lower.

I will say it is kinda funny to see cards that uses half the power of my 4090 have a 20C higher hotspot temp though.

 

[kapone32]

It is not very disturbing below 2000rpm, but when it goes above it, there is an annoying hum for me.

That is why I went to Watercooling

 

[physicsjunky]

I would say 100C hotspot is fine my reference 6700 XT was reaching 110C hotspot temp regularly even after a repaste. The highest i've ever seen it was 113C but that was with a heavy OC and benchmarking.

 

[Nike_486DX]

100-68C is a massive 32C delta, while normally hotspot should not be 15+C hotter, which means:

a) the paste pumped out, for a graphics card (or anything with bare die) you SHOULD NOT use stuff like MX4, Noctua NT-H1, Silver 5 or other garbage. these pastes are too liquid, and were made exclusively for benchmarks and positive reviews (easy to apply bla bla). But they really dont last more than 4 months.
b) the heatsink screws haven't been tightened to spec.

Repaste the card with hydronaut (if you are scared of liquid metal), and torque the screws in X pattern to 3Nm (when you feel its tight, but push it a bit more), make sure to use a good quality bit to not strip the screws. Then test and check the temperatures.

I would say 100C hotspot is fine my reference 6700 XT was reaching 110C hotspot temp regularly even after a repaste. The highest i've ever seen it was 113C but that was with a heavy OC and benchmarking.

dude 113C is too much. i had a 5700xt and never saw hotspot higher than 80 (yeah liquid metal)

 

[oxrufiioxo]

dude 113C is too much. i had a 5700xt and never saw hotspot higher than 80 (yeah liquid metal)

Yeah my 5700XT had excellent hot spot temps out of the box Nitro varient but this seems to be a pretty common issue on AMD cards I've never actually had a stock Nvidia card with a higher than 80C hotspot temp although I do tend to buy large models like the FTW3/Strix/GamingX varients. My last two Nvidia cards have like a 5-8c delta my strix 2080ti was a 10c delta but that card rarely broke 65c.

Other than the Reference 7900XTX I don't actually remember any card having performance issues from it though and that was due to a design flaw.

People just have ocd now that it's displayed in software.

I do wonder how bad somthing like a 7970 reference was having an edge temp in the 90C range.

 

[Bernec]

100-68C is a massive 32C delta, while normally hotspot should not be 15+C hotter, which means:

a) the paste pumped out, for a graphics card (or anything with bare die) you SHOULD NOT use stuff like MX4, Noctua NT-H1, Silver 5 or other garbage. these pastes are too liquid, and were made exclusively for benchmarks and positive reviews (easy to apply bla bla). But they really dont last more than 4 months.
b) the heatsink screws haven't been tightened to spec.

Repaste the card with hydronaut (if you are scared of liquid metal), and torque the screws in X pattern to 3Nm (when you feel its tight, but push it a bit more), make sure to use a good quality bit to not strip the screws. Then test and check the temperatures.


dude 113C is too much. i had a 5700xt and never saw hotspot higher than 80 (yeah liquid metal)

I have 6750XT and delta is 30-35 ... Celsius. Core is at 65-75 and hotspot goes to 110-112-108-110 in that region. I repasted it several times. Tried to evenly screw the heatsink ... even lapped the heatsink. Nothing fixes it. Either is a heatpipe broken (from factory Gigabyte ...) or I'm just unlucky. The card rarely shutdowns (but it could be the PSU ?!). In the last few months didnt and it is not boosting as high. Only draw max 205-207W even if its 220W TDP.

 

[RJARRRPCGP]

a) the paste pumped out, for a graphics card (or anything with bare die) you SHOULD NOT use stuff like MX4, Noctua NT-H1, Silver 5 or other garbage. these pastes are too liquid, and were made exclusively for benchmarks and positive reviews (easy to apply bla bla). But they really dont last more than 4 months.

That reminds me of the "white-stuff" that I used on socket 462 Athlons, LOL! The white paste would literally disappear after just some time, LOL. I would be lucky to get 4-5 months out of that!

I regularly found the paste gone after taking the heatsink off!

OTOH, with IHSes on today's CPUs, seems that the Thermaltake TG-50 TIM paste is good. I least I don't recall seeing it disappearing on me when I take the heatsink off.

 

[torkema61]

My friend sent me the temperature image, but only the normal temperature. It also reached 79°C. My graphics card was at 68°C at stock settings. Of course, his fan speed is around 62%, mine is around 75%.

There was no information about hotspot temperature. I asked for the tests again, including the hotspot temperature. Will let you know when available. I am waiting.

Maybe there's really nothing to worry about. If so, I can now easily start the CP2077 Phantom Liberty.

 

[ARF]

Sapphire said these temperatures are normal for this GPU. I will continue to use it this way. After the warranty period expires, I will replace the paste and pads myself.

"Normal" is what they say but doesn't mean you have to accept it. Just don't wait but rather act now - ask them to repaste the card with a better compound. Try ProlimaTech PK-3 Nano Aluminum or Thermal Grizzly Kryonaut.
Then, undervolt following the internet guides. https://www.reddit.com/r/Amd/comments/wqedns

I have 6750XT and delta is 30-35 ... Celsius. Core is at 65-75 and hotspot goes to 110-112-108-110 in that region. I repasted it several times. Tried to evenly screw the heatsink ... even lapped the heatsink. Nothing fixes it. Either is a heatpipe broken (from factory Gigabyte ...) or I'm just unlucky. The card rarely shutdowns (but it could be the PSU ?!). In the last few months didnt and it is not boosting as high. Only draw max 205-207W even if its 220W TDP.

Undervolt it. And try to repair it (RMA) if there is a broken heatpipe.

Got Gigabyte Aorus 6750 XT

Use mine at:Freq - 500/2835 (actual frequency is around 2809 at max)voltage - 1160 (1150 is unstable and will get me artifacts, sometimes even crash. tested on 6500MB donut kombustor + y-cruncher combined load for an hour)Memory - 2262 (actual frequency is around 2250) with Fast timings level 2 enabled. Otherwise can do 2312 (2300) with Fast timings level 1 (wish there was more headroom to overclock)Power limit = +15 (264W)Also have FCLK 2025 (sadly anything over 2070 is definitely unstable), and SOC clk = 1500

Max temps are around 87-89 degrees at hotspot. But i do love my air conditioning, so my room temps are around 22 or may be even bit less



Answer for your questionImagine curve.
When you increase frequency - you open up part to right of it.
When you change voltage - you move it up and down on relative voltage/freq
When you change minimal frequency you tighten up states so ther will be smaller voltage changes when GPU stops render frame and moves to another (if you'll look on oscillograph it downclocks for microseconds if not less)
UPD: For AMD it is bit different than for Nvidia (i just looked in Afterburner). AMD GPU have 2 separate curves, one for frequency scaling and second for voltage scaling depending on load/power/current limiters (actually % of P-state from 0-100, which is written as P-state 0-1)

https://www.reddit.com/r/Amd/comments/wqedns

 

[torkema61]

"Normal" is what they say but doesn't mean you have to accept it. Just don't wait but rather act now - ask them to repaste the card with a better compound. Try ProlimaTech PK-3 Nano Aluminum or Thermal Grizzly Kryonaut.
Then, undervolt following the internet guides. https://www.reddit.com/r/Amd/comments/wqedns



Undervolt it. And try to repair it (RMA) if there is a broken heatpipe.



https://www.reddit.com/r/Amd/comments/wqedns

As I mentioned in the previous post, when I undervolt, the GPU fan is fixed at 75%, even if the temperature is low. Zero rpm up to 50°C, and when it exceeds 50°C, the fan speed becomes 75% directly. When the 23.9.4 update arrives, I will try removing the driver with DDU and doing a clean installation. If it still doesn't work, I'll try fan control with MSI Afterburner.

Since the GPU is still under warranty, I do not plan to open it until the warranty expires. I asked Sapphire company, they said the temperature values were normal. In this case, unfortunately, even if I send it to warranty, I do not think they will change the thermal paste.

 

[ARF]

As I mentioned in the previous post, when I undervolt, the GPU fan is fixed at 75%, even if the temperature is low. Zero rpm up to 50°C, and when it exceeds 50°C, the fan speed becomes 75% directly.

This is a software bug. Try with another app.

When the 23.9.4 update arrives,

There will never be such a driver version. Next is 23.10.1.

Since the GPU is still under warranty, I do not plan to open it until the warranty expires. I asked Sapphire company, they said the temperature values were normal. In this case, unfortunately, even if I send it to warranty, I do not think they will change the thermal paste.

You are wrong. Don't wait. Pay them to change the thermal paste.. Tell them to use or buy yourself and give them ProlimaTech PK-3 Nano Aluminum or Thermal Grizzly Kryonaut..

 

[watzupken]

In my observations with a couple of RDNA2 cards, high hotspot temp is not uncommon. The delta between the GPU vs hotspot temps can go into the 20s degrees. In most cases, I think it is not a problem, but will likely cause some concerns because of the high temps. In your GPU's case, I do think that there may be bad contact between the GPU and heatsink, i.e. could be poor mounting pressure, or some poor thermal compound being used/ applied. Otherwise, you can try and undervolt it, which should help reduce the temps.

This is a software bug. Try with another app.



There will never be such a driver version. Next is 23.10.1.



You are wrong. Don't wait. Pay them to change the thermal paste.. Tell them to use or buy yourself and give them ProlimaTech PK-3 Nano Aluminum or Thermal Grizzly Kryonaut..

In my opinion, for direct die application, the best thermal compounds are the ones that are very hard to apply, i.e. Thermalright TFX. The reason is because it won't be pumped out easily, and therefore, heat transfer will not degrade over time. Kryonaut is very expensive, but it does not work well over time. I even got the batch that scratched the base of my coolers very badly.

 

[Vayra86]

Don't worry about it. I see 100C on my 7900XT with a 68C edge temp as well.
Seems to be normal for these cards, even under water. You can't really fix hotspot temp, that's just what she'll run at, because its sensor is situated where cooling barely gets a chance to do anything yet (on chip). Hotspot temp is directly related to voltage, not cooling.

100-68C is a massive 32C delta, while normally hotspot should not be 15+C hotter, which means:

a) the paste pumped out, for a graphics card (or anything with bare die) you SHOULD NOT use stuff like MX4, Noctua NT-H1, Silver 5 or other garbage. these pastes are too liquid, and were made exclusively for benchmarks and positive reviews (easy to apply bla bla). But they really dont last more than 4 months.
b) the heatsink screws haven't been tightened to spec.

You're telling the user to kill warranty on a perfectly fine operational card that is not exhibiting any issues, apart from the psychology happening where we see 100C and think of boiling water and how not to put our hands in it.

In the meantime all these GPUs run an algorithm for boost clocks that tells them to max boost within spec. The spec is 110C.
Do the math

Every time you fiddle with a GPU, there is a chance you inflict more damage than you fix. In other words, if it ain't broke, don't fix. I still have a GTX 1080 in box here that NEVER had a repaste in 6 years and still works as it does on day one.

 

[torkema61]

I noticed it when I Alt+Tab while playing the game last night. When I Alt+Tab and switch to the desktop, GPU usage increases to 100% and a crackling sound comes from the graphics card. This sound is probably coil whine. How do I fix GPU usage by Alt+Tab?

I regret buying AMD GPU.

Update:

I uninstalled the driver in safe mode with DDU and performed a clean installation, but the fan problem was not resolved. I'll try to install Afterburner and check it with it soon.

Update 2:
Even though I control the fan with Afterburner, the fan speed is fixed at 75%.

Another way I tried is to install version 23.8.2. Many people recommended this version, I tried it and the result did not change. When I write about this problem to Sapphire, they state that the problem is the AMD driver and close the issue.

I was completely undecided about what to do.

 

[RJARRRPCGP]

My friend sent me the temperature image, but only the normal temperature. It also reached 79°C. My graphics card was at 68°C at stock settings. Of course, his fan speed is around 62%, mine is around 75%.

View attachment 316711

There was no information about hotspot temperature. I asked for the tests again, including the hotspot temperature. Will let you know when available. I am waiting.

Maybe there's really nothing to worry about. If so, I can now easily start the CP2077 Phantom Liberty.

Try Unigine Superposition, with the 1080p Extreme preset.

 

[sLowEnd]

I noticed it when I Alt+Tab while playing the game last night. When I Alt+Tab and switch to the desktop, GPU usage increases to 100% and a crackling sound comes from the graphics card. This sound is probably coil whine. How do I fix GPU usage by Alt+Tab?

I regret buying AMD GPU.

Coil whine is not a brand-specific possibility

 

[torkema61]

Try Unigine Superposition, with the 1080p Extreme preset.

I didn't quite understand what you mean. I am sorry. What exactly did you mean?

Coil whine is not a brand-specific possibility

Of course. I don't mind the coil whine sound. The sound can be heard on almost any card and it is very low on my card.

Update:

After my friend sent me the furmark test, he also sent me the hotspot temperatures. There was approximately a 20°C difference between the hotspot temperature and the GPU temperature. I tested it over the weekend and my furmark result is as follows; (10 minutes, 1920x1080 and 8xMSAA) (GPU and hotspot temperatures are available in the upper right corner)

The maximum fan speed is 49%. It would be nice if I could adjust the fan settings, but I cannot control the fan as I want. Sapphire said that the problem was software related, that they also encountered the same problem and reported the problem to AMD.

 

[tugrul_SIMD]

110C ok for card but temperature creeps through pcie and affects mobo and even backside of case where hdmi connector. Anyone knows operating temp range of hdmi connector?

Also my ventus 2x 4070 goes 103C on hotspot while gpu is 90C. But its with silent fan speed and maximum cuda workload. Games rarely go beyon 80C gpu even with silent fan.

 

[torkema61]

110C ok for card but temperature creeps through pcie and affects mobo and even backside of case where hdmi connector. Anyone knows operating temp range of hdmi connector?

Also my ventus 2x 4070 goes 103C on hotspot while gpu is 90C. But its with silent fan speed and maximum cuda workload. Games rarely go beyon 80C gpu even with silent fan.

However, AMD still states that this measured temperature will not be a problem up to 115°C.

 

[tugrul_SIMD]

However, AMD still states that this measured temperature will not be a problem up to 115°C.

Vendors do not care for outside of their own components. They find players responsible for that.