• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Radeon MBA 7900 XTX Junction Temp issue workaround

Associate
Joined
21 Mar 2017
Posts
70
Location
UK
So for those of you like me who have been having this issue, I did the following things to "fix" it:

1. Undervolted to 1120mV
2. Overclock memory to 2714 MHz (Just for performance not the fix)

My Junction Temps went from 112c to 84c max

Also, my Time Spy Extreme score improved from 12,175 to 13,280

Base system for reference is Ryzen 7900X on Asus ROG X670E Hero with 32 GB DDR5 Ram at 6000MHz

Latest drivers including Catalyst 22.12.2

No tilting the case on a corner or any other BS like that :P
 
Last edited:
Then problem with your resolution is that you paid for X performance out of the box, now you are being happy with not getting what you paid for.

You are better off just RMA the card like AMD said you should
True. For me, its a wait and see fix to see if AMD do something about it in the next couple of weeks. If not, RMA.
 
Nope. At stock, my card was hitting 110c after just a few minutes in games and benches, then it would throttle back to ~307w and 2300-2400 core speeds (if not inexplicably start off at those speeds). Benchmarks and games had notably lower results than I've seen in reviews/other cards e.g. <26000 TimeSpy graphics score. I tried the following to resolve to no avail:
  • Similar settings you suggest (1140mv, raised memory clocks to improve performance) - this would net me better scores i.e. right around expected stock performance, and clocks would hover right under 2500, but temps would ultimately ramp up as would fans, and I'd see some instability in games. And as you said, this doesn't fix the issue, just alleviate it.
  • Took side off case - my case is already a mesh type with great ventilation
  • Changed DP cable to HDMI - I know this is akin to replacing a tire to fix a timing belt issue, but someone suggested so I gave it a go.
  • Changed PSU (EVGA GQ 80+ 1000w to an XFX Pro 850w) - both are high-quality power supplies with which I never had an issue, but thought worth a try
  • Reinstalled drivers a few times clean
Then, it dawned on me why the heck am I wasting my time trying to fix a defective card that cost £1000+, and I RMA'd. And I'd agree, I'm not trying to prop my case on its side -- that's ridiculous and completely academic/diagnostic.

If I was to guess, there's either a contact issue between the plate and the GPU die and/or a pressure/filling issue with the vapor chamber where people are seeing this issue (which may explain the better results when tilted). In any case, the fact that these cards passed QA is concerning. There's no driver or bios fix that will resolve this (at least, not one that doesn't hobble performance). My guess is AMD has broader issue as the problem is clear as day.
 
Derbauer thinks the MBA series need recalling.


Derbauer believes all the reference card 7900xtx GPUs should be recalled due to a cooler design flaw

He did a lot of testing to rule out cooler flexing, stand off mount height, thermal paste and the hold down brackets of the cooler itself and mavity pulling on the card and none of them had any effect. In the end, the vapour chamber is the culprit and that means no software update or user fix will be fix the issue - due to a design flaw, the vapour chamber cannot operate as intended when the 7900xtx is mounted horizontally in a case. The 7900xtx's vapour chamber works fine when the GPU is mounted vertically but as soon as you flip it horizontal the temperature will hit as high as 110c and the fans will go crazy, often to max rpm within a minute
 
Derbauer believes all the reference card 7900xtx GPUs should be recalled due to a cooler design flaw

He did a lot of testing to rule out cooler flexing, stand off mount height, thermal paste and the hold down brackets of the cooler itself and mavity pulling on the card and none of them had any effect. In the end, the vapour chamber is the culprit and that means no software update or user fix will be fix the issue - due to a design flaw, the vapour chamber cannot operate as intended when the 7900xtx is mounted horizontally in a case. The 7900xtx's vapour chamber works fine when the GPU is mounted vertically but as soon as you flip it horizontal the temperature will hit as high as 110c and the fans will go crazy, often to max rpm within a minute
its weird that when he moved it back to vertical the issue didnt fix itself, a full recall seems a bit extreme because not everyone has the issue, unless the only ones who arent reporting issues have it vertically mounted :D
 
So for those of you like me who have been having this issue, I did the following things to "fix" it:

1. Undervolted to 1120mV
2. Overclock memory to 2714 MHz (Just for performance not the fix)

My Junction Temps went from 112c to 84c max

Also, my Time Spy Extreme score improved from 12,175 to 13,280

Base system for reference is Ryzen 7900X on Asus ROG X670E Hero with 32 GB DDR5 Ram at 6000MHz

Latest drivers including Catalyst 22.12.2

No tilting the case on a corner or any other BS like that :P

Workaround is to RMA it
 
I think it looks like AIB purchase is the safe bet for these cards, steer clear of reference design, although not all are affected, which leads me to think that some of the vapor chamber coolers are either not filled properly, or badly made with blockages inside so they don't work as they should.
every one who experiences 110c hot spot should just RMA the card.
 
Last edited:
So for those of you like me who have been having this issue, I did the following things to "fix" it:

1. Undervolted to 1120mV
2. Overclock memory to 2714 MHz (Just for performance not the fix)

My Junction Temps went from 112c to 84c max

Also, my Time Spy Extreme score improved from 12,175 to 13,280

Base system for reference is Ryzen 7900X on Asus ROG X670E Hero with 32 GB DDR5 Ram at 6000MHz

Latest drivers including Catalyst 22.12.2

No tilting the case on a corner or any other BS like that :P

That's what it should be doing at stock before undervolting so you already losing out I would never settle for that and would be returning it asap
 
Last edited:
I think it looks like AIB purchase is the safe bet for these cards, steer clear of reference design, although not all are affected, which leads me to think that some of the vapor chamber coolers are either not filled properly, or badly made with blockages inside so they don't work as they should.
every one who experiences 110c hot spot should just RMA the card.
Der8auer just dropped a vid about an hour ago, he bought 4 cards from people who had been using their card in various positions (vertical/horizontal etc.) and ran a whole batch of testing on the cards to repeat the hotspots temps, his conclusion was the vapour chamber on the reference coolers have a design failure.

there was a notable difference in vertical vs horizontal hotspots, but for a bag of sand you should be able to mount this thing however you want and not have to worry about it
 
Derbauer believes all the reference card 7900xtx GPUs should be recalled due to a cooler design flaw

He did a lot of testing to rule out cooler flexing, stand off mount height, thermal paste and the hold down brackets of the cooler itself and mavity pulling on the card and none of them had any effect. In the end, the vapour chamber is the culprit and that means no software update or user fix will be fix the issue - due to a design flaw, the vapour chamber cannot operate as intended when the 7900xtx is mounted horizontally in a case. The 7900xtx's vapour chamber works fine when the GPU is mounted vertically but as soon as you flip it horizontal the temperature will hit as high as 110c and the fans will go crazy, often to max rpm within a minute

Goes to show how much proper testing went into this and how much they cared. They do like to shoot themselves in the foot and then people wonder about market share and AMD image when it comes to their GPU department.
 
Goes to show how much proper testing went into this and how much they cared. They do like to shoot themselves in the foot and then people wonder about market share and AMD image when it comes to their GPU department.
they prob have them all vertically mounted in the testing area so the problem wasnt noticed till people got them home and started using them, even reviewers tend to use test benches instead of actual systems so they get tested vertical too
 
AMD have already acknowledge their looking into this issue and now with Derbauer pointing out a likely culprit I wonder how AMD will react. Most of the time when AMD have had problems it's been due to faulty software and buggy drivers I don't recall a time when they've had such a widespread hardware design fault they could result in a product recall, I wonder how they will respond (I hope they respond better then they did with the 6600XT fault that was never probably addressed besides 'get 800 watt PSU for your 190 watt card' :rolleyes:). I wonder if these were made by PC Partner Group/Zotac?
 
Last edited:
Back
Top Bottom