AMD RX 5700 XT GPUs reaching 110°C “expected and within spec” for gaming

auntjemima

auntjemima

[H]ard DCOTM x2
Joined
Mar 1, 2014
Messages
12,144
https://www.pcgamesn.com/amd/amd-rx-5700-gpu-110c

AMD says that temperatures of up to 110°C with its RX 5700-series graphics cards are “expected and within spec” for typical gaming usage. The company’s latest GPUs, utilising the RDNA architecture, are known to crank out some heat, but according to AMD’s latest community update that’s within the safe operating range of the graphics card and nothing to worry about.
 
So AMD is now redefining what safe operating voltages and temperatures are. Should we assume they created some type of super silicon that prevents degradation over time?
 
Remember when Ryzen X series used to have that 20C offset? I wonder if the hot spot temp is actually 110c... or if its some throttle point they introduced to protect it.


AMD designed it, so they would know best. The 290x ran all day at 94c and no catastrophic failures. If anything, they are the kings of heat.
 
ZeroBarrier said:
So AMD is now redefining what safe operating voltages and temperatures are. Should we assume they created some type of super silicon that prevents degradation over time?
Click to expand...

The theoretical maximum for long-term operation is between 125 and 150°C, though there are exotic designs that can survive much higher temperatures (though I don't know how long). At my job we generally test ICs up to 125°C.

I remember 'back in the day' in the Athlon XP era when 90°C was the maximum safe temp for consumer CPUs. Better hardware and temperature monitoring pushed that up to 100°C nowadays.

As the article says they now have more accurate temperature monitoring across the die, which allows for even more aggressive clocks.
 
I noticed when running FurMark on burn-in test for my 5700 XT, the temp reading on the app was up to 99C after 5 minutes, while the AMD overlay showed significantly lower, at 82C.

I'm not sure what either measurement is referring to (junction, hotspot, etc.) or which is correct. However, the card seems to be working fine.
 
If it is within spec, it's within spec. If it fails running it within spec isn't going to void the warranty so get it replaced. If it lasts 3+ months and especially if it hits a year, it's not likely to just up and die at that point heat or not. Same story with Ryzen processors and voltage, it's just how PBO, CPPC2, and XFR work, if the voltages weren't safe AMD wouldn't be allowing it in their designs and default out of the box configurations. Just because peope "feel" it's too hot or too much voltage doesn't actually mean that's the case regardless of what reddit thinks lol. The fact that AMD had to make a chipset driver available that tones down the auto overclocking just so that people didn't have to see 'scary' "idle" (though modern chips are rarely ever actually idling with all the background tasks they are running) voltages in their software monitoring apps is crazy.

Stay [H]ard people, or should I say stay [H]ot while gaming on 5700XT ;-).
 
Yep they are talkign about the hotspot temp. That is not average across GPU. So the title is kinda misleading. Honestly I think AMD is doing a better job here and being honest. Nvidia doesn't do hotspot temp or atleast they don't reveal it publicly.
 
NKD said:
Yep they are talkign about the hotspot temp. That is not average across GPU. So the title is kinda misleading. Honestly I think AMD is doing a better job here and being honest. Nvidia doesn't do hotspot temp or atleast they don't reveal it publicly.
Click to expand...

Yeah, plus the fact that with the new cards, that temp does not appear to be even reaching above 80C. On my Vanilla 5700, even when overclocked, the hot spot temp is no more than 90C or so and if the load on the GPU is removed, the temp drops quickly and immediately.
 
wolfofone said:
If it is within spec, it's within spec. If it fails running it within spec isn't going to void the warranty so get it replaced. If it lasts 3+ months and especially if it hits a year, it's not likely to just up and die at that point heat or not. Same story with Ryzen processors and voltage, it's just how PBO, CPPC2, and XFR work, if the voltages weren't safe AMD wouldn't be allowing it in their designs and default out of the box configurations. Just because peope "feel" it's too hot or too much voltage doesn't actually mean that's the case regardless of what reddit thinks lol. The fact that AMD had to make a chipset driver available that tones down the auto overclocking just so that people didn't have to see 'scary' "idle" (though modern chips are rarely ever actually idling with all the background tasks they are running) voltages in their software monitoring apps is crazy.

Stay [H]ard people, or should I say stay [H]ot while gaming on 5700XT ;-).
Click to expand...

Pretty much this. That was the position I took when I bought my P4 Prescott and it lived its full lifetime. If it's built for it, it's nothing to worry about.
 
I thought we were led to believe that die shrinks result in significantly less power requirements, super-fast linear performance increases and eventually passive cooling, this is a crock of BS.

If AMD/ NVIDIA was presenting a thesis defense/ professional conference with their latest advertising slides, we would still be there separating the BS from reality. Anytime I was on a committee I would tell the presenter to be prepared to backup the slide with real science or remove it.

Anyone care to tell me why the new custom AIB cards have a larger cooling system at 7 nm than my 390 X has at 28 nm ? I mean really a 3 slot cooler at 1/4 the scale?
 
owned a 290x that just wasn't happy at stock clocks or temps so I upped the voltage and that gpu chilled at 104 for 6 months mining away and never did die :p
 
drutman said:
I thought we were led to believe that die shrinks result in significantly less power requirements, super-fast linear performance increases and eventually passive cooling, this is a crock of BS.

If AMD/ NVIDIA was presenting a thesis defense/ professional conference with their latest advertising slides, we would still be there separating the BS from reality. Anytime I was on a committee I would tell the presenter to be prepared to backup the slide with real science or remove it.

Anyone care to tell me why the new custom AIB cards have a larger cooling system at 7 nm than my 390 X has at 28 nm ? I mean really a 3 slot cooler at 1/4 the scale?
Click to expand...

The heat is condensed to a smaller area so you have higher temperature hot spots. The trade off is you pack in more transistors per nm but I don't think any of them made a claim it would run cooler with a smaller process. As long as the gpu is designed to work at those temperatures for 5 years its fine.
 
cybereality said:
I noticed when running FurMark on burn-in test for my 5700 XT, the temp reading on the app was up to 99C after 5 minutes, while the AMD overlay showed significantly lower, at 82C.

I'm not sure what either measurement is referring to (junction, hotspot, etc.) or which is correct. However, the card seems to be working fine.
Click to expand...


so up until vega every gpu including nvidia's 2000 series uses gpu edge temp to give the gpu temp reading. with vega and navi you now get edge temp + highest temp sensor reading within the gpu it's self.. so the 82C is the edge temp, the 99C is the hottest point inside the silicon it's self.

drutman said:
I thought we were led to believe that die shrinks result in significantly less power requirements, super-fast linear performance increases and eventually passive cooling, this is a crock of BS.

If AMD/ NVIDIA was presenting a thesis defense/ professional conference with their latest advertising slides, we would still be there separating the BS from reality. Anytime I was on a committee I would tell the presenter to be prepared to backup the slide with real science or remove it.

Anyone care to tell me why the new custom AIB cards have a larger cooling system at 7 nm than my 390 X has at 28 nm ? I mean really a 3 slot cooler at 1/4 the scale?
Click to expand...



for the same reason why gpu's while getting smaller are now running almost 1000mhz faster than your 390x.. if you took the same 390x gpu at the exact same clock speeds and put it on a 7nm process then yeah it'll run cooler because it won't need any where near the same voltage to hit those clock speeds.. what the process shrink allows is the ability to add more to the gpu, higher clock speeds at lower voltages, more CU's, etc, etc within the same power envelope of a previous process but there is always a trade off and in this case it's heat.
 
Last edited:
drutman said:
I thought we were led to believe that die shrinks result in significantly less power requirements, super-fast linear performance increases and eventually passive cooling, this is a crock of BS.

If AMD/ NVIDIA was presenting a thesis defense/ professional conference with their latest advertising slides, we would still be there separating the BS from reality. Anytime I was on a committee I would tell the presenter to be prepared to backup the slide with real science or remove it.

Anyone care to tell me why the new custom AIB cards have a larger cooling system at 7 nm than my 390 X has at 28 nm ? I mean really a 3 slot cooler at 1/4 the scale?
Click to expand...

who on earth were you listening to that was making wild claims about a passively-cooled future?

a 5700XT has 66% more transistors than a 390X, runs at nearly double the clocks, and is rated at a 50W lower TDP. sounds like process shrinks are working to me (y)

p.s. there were 390X cards with rediculously massive custom coolers too, bad comparison is bad.
 
110C hotspot temp. Now you actually have insight as to what the hottest temp is (well Vega has it as well) whereas before you didn't .... Shit was running hotter than it displayed before in every other design, you just didn't know. It's not like AMD is redefining what it safe for silicon and 110C is just fine, it's hot, sure but it's safe. Can't believe everyone is freaking out over this it's not even news.
 
Doh...

My 5700xt hit 100c and I thought my fan failed and I was going to have to RMA it. Well a custom fan curve in wattman and now it games at 78c all day long albeit the blower is obnoxious.
 
I wonder if anybody is building a case mod that lets these things double as an air popper. Sweet gaming sesh and Popcorn.... how much better could it get.
 
Lakados said:
I wonder if anybody is building a case mod that lets these things double as an air popper. Sweet gaming sesh and Popcorn.... how much better could it get.
Click to expand...
Did you know:
Popcorn doesn't effectively pop until 150C, with optimum popping temperature being in the 200-235C range.

upload_2019-8-13_13-42-50.png
 
Lakados said:
I wonder if anybody is building a case mod that lets these things double as an air popper. Sweet gaming sesh and Popcorn.... how much better could it get.
Click to expand...

I'd prefer a George Foreman 5700XT-G, the lean mean fat reducing Crypto-mining machine for your laptop.

Something to go with my 290x Toaster:

2019-08-13 13.24.02.jpg


Works great and cools better than the stock XFX blower, too.
 
Last edited:
I heard that the next Navi (Big Navi) will need to be current limited, there is concern that the 7nm matrix could cause a miniature black hole singularity due to the high energy concentration in a small space in the (relatively) low gravity of Earth, versus say the gravity at the surface of the Sun. No word on how that will affect the performance though, but sucks for AMD fans.
 
Not sure why this is a big deal...AMD giving you the actual temp of the hottest part of the die is a GREAT thing in my opinion. I wish Nvidia and soon Intel would offer the same thing.

I wish the Reddit crowd would actually take the time to learn something instead of just being paranoid about things they shouldn't be.

GoodBoy , I LoL'd. Thanks for that!
 
I guess the RX 5700 is not as hot as the XT model .. RD 2 is stock fan profile and SB ( 11:00 min mark action starts) is one I am working on for my card .



 
DeathFromBelow said:
I'd prefer a George Foreman 5700XT-G, the lean mean fat reducing Crypto-mining machine for your laptop.

Something to go with my 290x Toaster:

View attachment 180650

Works great and cools better than the stock XFX blower, too.
Click to expand...

:ROFLMAO: I almost burned myself on my Vega exhaust b4 undervolt and WC and definitely burned myself on 290X exhaust back in the day... go Team Red!
 
DeathFromBelow said:
I'd prefer a George Foreman 5700XT-G, the lean mean fat reducing Crypto-mining machine for your laptop.

Something to go with my 290x Toaster:

View attachment 180650

Works great and cools better than the stock XFX blower, too.
Click to expand...

At one time I had 4x 290x in crossfire. Man was it pulling the powah! Tripped a 15 amp breaker like it was a zit on a teenagers face.
 
crazycrave said:
I guess the RX 5700 is not as hot as the XT model .. RD 2 is stock fan profile and SB ( 11:00 min mark action starts) is one I am working on for my card .



Click to expand...


Meh Crap Poopgade was a crappy game hah but it sure runs good on that 5700xt
 
Dayaks said:
The 290x actually had super high failure rates: https://hardforum.com/threads/amds-failure-rates-part-two-2015-edition.1899524/
Click to expand...

It wasn't the heat that was causing the failures, the cards were designed to run at 95 degrees. Most of the RMA's were due to the black screen issues that those cards had. Issues that were caused by the Elpida ram. The first reference and AIB cards from Sapphire had them. And I would not be surprised if a lot of people returned them when they saw they were running at 95 degrees and thought they had a problem.

Also the first real mining boom happened back then and ended after three months, it wasn't like the more recent boom, It was the first time that normal joes were getting involved with trying to mine. I wonder how many cards were bricked by idiots trying to mine for the first time. Remember how RMA procedures tightened up a lot after that.
 
DeathFromBelow said:
I'd prefer a George Foreman 5700XT-G, the lean mean fat reducing Crypto-mining machine for your laptop.

Something to go with my 290x Toaster:

View attachment 180650

Works great and cools better than the stock XFX blower, too.
Click to expand...

Bet that works pretty well.

I used to have a GeForce FX 5700. Was great in the winter at keeping me warm... (Still got it somewhere). Can't find the video where the nVidia execs made a parody video about it. "We'll be the Harley Davidson of Graphics Cards!" lol, it had a loud ass blower, maybe the first one ever used on a video card. Was loud as hell. Now AMD makes a video card with the same model number, and they used a blower too. We've come full circle.

Edit: It was the FX 5800, and found the video!
 
Last edited:
GoodBoy said:
Bet that works pretty well.

I used to have a GeForce FX 5700. Was great in the winter at keeping me warm... (Still got it somewhere). Can't find the video where the nVidia execs made a parody video about it. "We'll be the Harley Davidson of Graphics Cards!" lol, it had a loud ass blower, maybe the first one ever used on a video card. Was loud as hell. Now AMD makes a video card with the same model number, and they used a blower too. We've come full circle.
Click to expand...

maybe that was the hidden joke all along from AMD, lol.
 
The AMD 5700 XT has now been listed as a climate change catalyst (pun) and now has restrictions under the Paris Agreement.
 
Yeah, I think all of the Crypto-mining farms were a measurable factor in greenhouse gas emissions, due to the high energy usage. All them GPU's running at 100% for months on end..

Myself on the other hand, buy 100% wind power from my electric company :) Adds like $2 to $5 per bill. But 10 years-ish ago when economy took took a big shit, I was actually saving money by $20 or more per bill, which went on for several years. That was why I moved it to 100%.
 
Last edited:
reaper12 said:
It wasn't the heat that was causing the failures, the cards were designed to run at 95 degrees. Most of the RMA's were due to the black screen issues that those cards had. Issues that were caused by the Elpida ram. The first reference and AIB cards from Sapphire had them. And I would not be surprised if a lot of people returned them when they saw they were running at 95 degrees and thought they had a problem.

Also the first real mining boom happened back then and ended after three months, it wasn't like the more recent boom, It was the first time that normal joes were getting involved with trying to mine. I wonder how many cards were bricked by idiots trying to mine for the first time. Remember how RMA procedures tightened up a lot after that.
Click to expand...

You are correct in the fact that many had faulty Elipda memory and then tons of cards were bricked during the flashing daus...


AMD allowed a custom Bios to be signed that disabled 1/2 of the ROPs on the cards and changed memory straps/timings. The reduction in ROPs allowed the card to run about 45% less power with virtually no mining performance hit ....
 
So, should I wait for a good blower, or pull the trigger on the cheapest sucker I can find?

rx 5700 without the xt
 
You must log in or register to reply here.
Back
Top