AMD’s GPU temperatures have always skewed a little higher. In the past, there were perpetual jokes about the company’s GPUs being used as room heaters in colder climes and chilly months. Those who’ve bought the company’s newest cards though, are a little concerned over what they deem to be unnecessarily high temperatures. It’s apparently enough of an issue that AMD’s deemed it necessary to write a blog post about the high temperatures seen in the Radeon RX 5700 cards. In the post, they say that the 110 °C users have experienced when gaming is normal, with those high temps “expected and within spec.”
According to AMD, while these temperatures are higher than you’re probably used to seeing, it’s not necessarily because the cards are running hotter than predecessors, but rather that they now have an array of sensors spread across the cards, and the registered temp is likely the very hottest point on the GPU
“Paired with this array of sensors is the ability to identify the ‘hotspot’ across the GPU die. Instead of setting a conservative, ‘worst case’ throttling temperature for the entire die, the Radeon RX 5700 series GPUs will continue to opportunistically and aggressively ramp clocks until any one of the many available sensors hits the ‘hotspot’ or ‘Junction’ temperature of 110-degrees Celsius. Operating at up to 110C Junction Temperature during typical gaming usage is expected and within spec,” AMD says.
According to AMD, using sensors in this way allows drivers to work with more data, allowing the cards to exhibit “much higher performance and clocks out of the box,” at a lower noise level. It’s got more to do with better measurement than it has to do with higher temperatures, and isn’t anything to worry about, AMD affirms.
Last Updated: August 15, 2019
Yahtzee
August 15, 2019 at 14:08
Chernobyl was also “within spec”. And that was after it blew up.
Magoo マグ
August 15, 2019 at 14:31
I want to understand this reference. But I don’t.
Pofadder
August 16, 2019 at 09:15
They only measured 3.6 Rontgen radiation after the explosion. Which was within spec. The fact that the meters only went up to 3.6 was beside the point.
Viper_ZA
August 15, 2019 at 14:16
No thanks -_-
G8crasha
August 15, 2019 at 15:58
When I watched a review on the new card, the reviewer said the card had insufficient cooling, unlike its bigger sibling, the XT version.
Alien Emperor Trevor
August 15, 2019 at 13:54
So it can make coffee too. Checkmate Nvidia.
Admiral Chief Umbra
August 15, 2019 at 14:02
HAH
Admiral Chief Umbra
August 15, 2019 at 14:02
https://media0.giphy.com/media/8lMQy63W0Cs3EnCDQk/giphy-downsized-medium.gif
Guz
August 15, 2019 at 14:23
Aftermarket cooler companyies are gonna make lekker cash with these cards,
I don’t care if they say it’s in spec, no way I would run a card that hot
Magoo マグ
August 15, 2019 at 14:30
Me neither. I had a 7770 burn out after a few months of not checking temps when I was much younger. Now I have constant anxiety and keep tabs on all of my temps, all the time. 110c would freak me out.
Theo Lubbe
August 15, 2019 at 21:49
It’s typical for VRMs to get over 95C even on well-cooled cards. So if they’re talking about ‘the hottest points’ the way they suggest they are, they could be talking about, well, the MOSFETs of VRMs.
Which are by design typically able to run continuously at temperatures of around 115C for years without a problem in computer applications, and even higher than that when talking about industrial-grade or automotive-use ones.
Magoo マグ
August 15, 2019 at 14:23
My XFX RX580 has a default cooling profile with a maximum of 40% fan usage when the temp gets to 90c+. I’ve always found this to be silly, if not a design flaw.
It resets to default every now and then and I have to manually check that my custom profile is active with every boot.
Theo Lubbe
August 15, 2019 at 21:49
You didn’t buy a used mining card, did you? Because that could be running a custom firmware which was never removed by whoever used it before you, and those often have stupid fan curves on them since the users would employ external cooling instead to cool a bench of the things rather than have their fans inconsistently pump up and down.
Magoo マグ
August 16, 2019 at 11:05
I bought it brand new. It’s not a mining card.
HvR
August 15, 2019 at 14:38
And that kind of thinking is why you get long term use failures; usually those that happen outside of warranties.
Solder joints especially BGA solder joints hate such high temperature fluctuations; causes micro cracks at the top of the joints and when PCB movement either externally or from heating and cooling it goes “pop”
Now for a fact that is what killed my 2nd last AMD card.
Also the major causes of previous generation console failure (ie infamous RROD failures)
So AMD is just confirming that you can now monitor the biggest design fault on their cards especially if plan to use it beyond the 12 month warranty
jimz0r
August 16, 2019 at 08:33
I won’t buy a card without a 3 year warranty. Period.
If you can’t design a product to last little more than a year, I have no interest buying it.