Is RX 9070 VRAM temperature regular value or hotspot?

[Vayra86]

The fact that I don't want to live in it, doesn't mean I'm foolish enough to ignore the fact that I do live in it.

It is what you make it out to be mate.

 

[rusty caterpillar]

Those are useless, definitely. But those are a couple of cards with a bad design from an AIB. The thing is, VRAM runs hot on almost every 9070 XT. If someone wants to tell me that every single AIB botched every single 9070 XT, and they're all gonna fail soon, go ahead.

Botched no, they have done the minimal, and minimal is always problematic. Minimal is for light usage.

 

[AVATARAT]

I understand Marketing but, hurting the product on the long run is not Marketing rather, planned obsolescence.

One of the jobs of marketing is to make money, right?
So planned obsolescence is also under their hat

 

[rusty caterpillar]

I guess they're stuck in 2012 when your 32 nm CPU had a 65 °C tJmax

More like 2019, when VRAM start frying more often.

Micron chips starting with 8 manufactured in 2018. Replacing mem modules on Turing cards EVGA and Gigabyte >>>> GDDR6

Replacing mem modules on 1080 Ti >>>> GDDR5X

Replacing mem modules and title goes like "Common problem with most GPUs" >>>>> GDDR6X 3080 Ti


Your card has GDDR6, hopefully a better revision. Mine has GDDR5X. Not playing the Hot VRAM lottery but, you my friend is entirely up to you.

I really don't think they reach the threshold where suppose to throttle down, are just frying before that.

One of the jobs of marketing is to make money, right?
So planned obsolescence is also under their hat

You could say that, it seems is just a thing in our heads , all is well otherwise

When you look at something expensive that you bought, you are supposed to take pleasure in the little details, they are supposed to give warm and fuzzies

My point exactly "warm" VRAM and "fuzzies" FSR4 on FSR2 games.

 

[AusWolf]

Botched no, they have done the minimal, and minimal is always problematic. Minimal is for light usage.

Some of these cards are dual slot, 2.5 slot, 3 slot, maybe even 4 slot. Some of them have 5 heatpipes, some 6, some even 7 or 8. All of these run their VRAM hot regardless of cooler design. So how is that "minimal"?

 

[rusty caterpillar]

Some of these cards are dual slot, 2.5 slot, 3 slot, maybe even 4 slot. Some of them have 5 heatpipes, some 6, some even 7 or 8. All of these run their VRAM hot regardless of cooler design. So how is that "minimal"?

You can put 8 heatpipes if you cover the exhaust of the heatsink with plastic shrouds or the top gap you left for the third fan to blow trough.
You can use 8 heatpipes if the fin soldering is poor
You can have heavy sinks if thermal pads are poor and don't take heat fast enough or fin soldering poor
You can have big heatsinks if the thickness of the fans is too little , blade counts too low, or simply rotors gave up to pressure created by air turbulence hitting the fins.
You can have vapor chamber for GPU core if your design is detrimental for VRAM
You can have good cooling for VRAM if DRV mosfets are not cooled properly, VRAM will remain hot as result or get even hotter
You can have big nice cooler design if you didn't calculate the PCB density before doing the actual design and reused old design from lower density PCB

I have some more but, I feel I'm wasting my time.

Some of these cards are dual slot, 2.5 slot, 3 slot, maybe even 4 slot. Some of them have 5 heatpipes, some 6, some even 7 or 8. All of these run their VRAM hot regardless of cooler design. So how is that "minimal"?

If you think all 9070XT are VRAM hot, maybe is just a failed PCB design or GDDR6 is not up the GPU core tasks, fast enough, therefore GDDR6 is continuous under operational high stress also under high voltages.

 

[AusWolf]

You can put 8 heatpipes if you cover the exhaust of the heatsink with plastic shrouds or the top gap you left for the third fan to blow trough.
You can use 8 heatpipes if the fin soldering is poor
You can have heavy sinks if thermal pads are poor and don't take heat fast enough or fin soldering poor
You can have big heatsinks if the thickness of the fans is too little , blade counts too low, or simply rotors gave up to pressure created by air turbulence hitting the fins.
You can have vapor chamber for GPU core if your design is detrimental for VRAM
You can have good cooling for VRAM if DRV mosfets are not cooled properly, VRAM will remain hot as result or get even hotter
You can have big nice cooler design if you didn't calculate the PCB density before doing the actual design and reused old design from lower density PCB

I have some more but, I feel I'm wasting my time.


If you think all 9070XT are VRAM hot, maybe is just a failed PCB design or GDDR6 is not up the GPU core tasks, fast enough, therefore GDDR6 is continuous under operational high stress also under high voltages.

Or maybe 90 °C is normal operation on these chips? Next to that list of yours, is it a completely unimaginable option?

 

[Vayra86]

Or maybe 90 °C is normal operation on these chips? Next to that list of yours, is it a completely unimaginable option?

Cant we just find the part number and check what the max temp is? If it is 108C, we are good right.

 

[AusWolf]

Cant we just find the part number and check what the max temp is? If it is 108C, we are good right.

Micron lists 95 °C I think, but that's edge temp, while the 9070 (XT) measures hotspot temp.

 

[Vayra86]

Micron lists 95 °C I think, but that's edge temp, while the 9070 (XT) measures hotspot temp.

If its edge temp I wouldnt be so happy about a 90C hotspot tbf. Its running on the brink of destruction. Add some age and dust and you are in the danger zone. Im sure there is some increase in heat density with newer more dense chips but still, temps be going up. We see similar things in CPUs; but silicon doesnt magically become more temp resistant...

This could be viewed in similar ways as creative accounting. THAT is why I would really appreciate AMD stepping in with an explanation.

 

[rusty caterpillar]

Or maybe 90 °C is normal operation on these chips? Next to that list of yours, is it a completely unimaginable option?

Hopefully it is one, and fixable....

If its edge temp I wouldnt be so happy about a 90C hotspot tbf. Its running on the brink of destruction. Add some age and dust and you are in the danger zone. Im sure there is some increase in heat density with newer more dense chips but still, temps be going up. We see similar things in CPUs; but silicon doesnt magically become more temp resistant...

This could be viewed in similar ways as creative accounting. THAT is why I would really appreciate AMD stepping in with an explanation.

Clearly, factors like dust maybe and leaking pads will add to this already high temps.

Before AMD jumps in maybe someone on YT will change the thermal pads so we can start somewhere, if is the case removing plastic film which might act as insulator. Some AIB back plate have that plastic thick foil glued to the backplate forcing the heat to circulate only trough thermal pads areas.
Don't know what kind of plastic is that foil, couldn't find any info.

 

[AusWolf]

If its edge temp I wouldnt be so happy about a 90C hotspot tbf. Its running on the brink of destruction. Add some age and dust and you are in the danger zone. Im sure there is some increase in heat density with newer more dense chips but still, temps be going up. We see similar things in CPUs; but silicon doesnt magically become more temp resistant...

This could be viewed in similar ways as creative accounting. THAT is why I would really appreciate AMD stepping in with an explanation.

Why would it be on the brink of destruction? Every card seems to have a 108 ˚C limit. I assume that's the limit on the hotspot. With a 90 ˚C hotspot, your edge temp is way lower.

For example, HWinfo says I'm using 83% of the available thermal headroom with 90 ˚C VRAM.

Hopefully it is one, and fixable....

Again, why would it need to be fixed if it's part of its normal operation?

Why would you automatically assume that something is wrong? Even your CPU has a 95 (AMD) or 100 (Intel) ˚C max operating temp. It surely isn't some manufacturing error, right?

Don't know what kind of plastic is that foil, couldn't find any info.

Probably the film that's only used for transportation, and you remove before first use.

 

[rusty caterpillar]

Again, why would it need to be fixed if it's part of its normal operation?

Why would you automatically assume that something is wrong? Even your CPU has a 95 (AMD) or 100 (Intel) ˚C max operating temp. It surely isn't some manufacturing error, right?


Probably the film that's only used for transportation, and you remove before first use.

Don't think is normal operation not on 92-94 C, new cards.

I seen too many cut corners to hardware over the years.


You don't understand which film...

You see it now, the lighter gray. This is on Sapphire Nitro

 

[Vayra86]

Why would it be on the brink of destruction? Every card seems to have a 108 ˚C limit. I assume that's the limit on the hotspot. With a 90 ˚C hotspot, your edge temp is way lower.

For example, HWinfo says I'm using 83% of the available thermal headroom with 90 ˚C VRAM.


Again, why would it need to be fixed if it's part of its normal operation?

Why would you automatically assume that something is wrong? Even your CPU has a 95 (AMD) or 100 (Intel) ˚C max operating temp. It surely isn't some manufacturing error, right?


Probably the film that's only used for transportation, and you remove before first use.

Myeah I don't buy that CPUs running at 95C-100C 24/7 are going to survive much longer than the warranty period either. Seen too much over the years, anything running at those temps is on the way out sooner or later. I'm sure a lot of companies say its fine. Buying a replacement is fine too for their bottom line. Laptops are squarely in that territory. 3-4 years at best and they start exhibiting problems.

We've seen plenty of mass market products fail spectacularly. Need I remind you of a degrading 14th gen... RRODs on consoles, blu ray players on PS3s needing replacement after X years... And as always, radio silence around these things never spells something good.

All we can do now is 'assume'. That's bad. Assumptions are the mother of all fuckups.

 

[AusWolf]

Don't think is normal operation not on 92-94 C, new cards.

What do you base that on?

I seen too many cut corners to hardware over the years.

Same here, but that doesn't make me automatically assume that what we see here is a bad thing. If it was, then I'm sure at least one AIB would have gone "hang on a minute, this isn't OK, I'll make a hefty cooler for the VRAM modules on my premium card".

You don't understand which film...

View attachment 391219

You see it now, the lighter gray. This is on Sapphire Nitro

Interesting. I wonder why that's there. But I'm also wondering if it restricts anything. It seems to have cutouts for the VRAM pads.

 

[tugrul_LordOfDrinks]

Its weird. You are able to r&d to create graphics cards, able to write fan control systems in software, able to design something in 3D space, but still can't think of the cooler to touch memory.

Perhaps it is time for in-GPU memory tech. Because they wont touch memory for some reason ever but still remember to touch GP. So put it inside gpu.

I have 5070 and its vram heats to 80 for extended use. Even with +2000 oc, it heats only 2-4 Celcius more. Ambient is 20Celcius.

 

[JustBenching]

Myeah I don't buy that CPUs running at 95C-100C 24/7 are going to survive much longer than the warranty period either. Seen too much over the years, anything running at those temps is on the way out sooner or later. I'm sure a lot of companies say its fine. Buying a replacement is fine too for their bottom line. Laptops are squarely in that territory. 3-4 years at best and they start exhibiting problems.

We've seen plenty of mass market products fail spectacularly. Need I remind you of a degrading 14th gen... RRODs on consoles, blu ray players on PS3s needing replacement after X years... And as always, radio silence around these things never spells something good.

All we can do now is 'assume'. That's bad. Assumptions are the mother of all fuckups.

It's 101 of electronics. Lower temps = more longevity. Im fine with anything running at 75-80 since I prefer lower noise levels but stuff approaching 90 is a nogo for me.

 

[AusWolf]

Its weird. You are able to r&d to create graphics cards, able to write fan control systems in software, able to design something in 3D space, but still can't think of the cooler to touch memory.

More assumptions that no cooling engineer knows what they're doing at any of the AIBs. Well-funded indeed!

 

[Vayra86]

More assumptions that no cooling engineer knows what they're doing at any of the AIBs. Well-funded indeed!

Its really of no higher value than your assumption they got it right and there is no pressure from anyone to reduce cost and perhaps reduce it a tad too much.

We've seen AIBs deliver mismatched coolers on many occasions because 'cost' and we also know AMD historically doesn't have a tight leash on its AIB partners. Naaah, sorry. The faith in this whole thing is too low to ride on assumptions.

No less than one gen ago I believe we had this

AMD Radeon RX 7900 XTX Overheating Tests Point to Vapor Chamber Issues

Over the holiday period, AMD has been embroiled in a bit of controversy. Its all-new flagship GPU is overheating for some users, which has spurred the company to investigate. With most of the tech media on holiday, though, there hasn’t been much new information. However, overclocking wizard...

community.amd.com

So fuck assumptions. AMD has a job to do here.

 

[JustBenching]

More assumptions that no cooling engineer knows what they're doing at any of the AIBs. Well-funded indeed!

Don't you also believe that all engineers didn't know what they were doing with the design of the 12vh?

 

[tugrul_LordOfDrinks]

Is 80Celcius on gddr7 Samsung ok? 82-84 with +2000 oc.

 

[Vayra86]

Don't you also believe that all engineers didn't know what they were doing with the design of the 12vh?

Precisely. Its honestly RAINING engineering screwups the past few years.

I assume nothing anymore honestly. Similar things apply to politics. All bets are off past 2024.

 

[JustBenching]

Precisely. Its honestly RAINING engineering screwups the past few years.

I assume nothing anymore honestly

I don't think they screwed up with the 12vh, but he does, so the question is why the double standards

 

[Vayra86]

I don't think they screwed up with the 12vh, but he does, so the question is why the double standards

Well they don't revise a connector because its perfect, to begin with. But let's not go there Gotcha, and agreed either way. You've got a good antenna for those double standards

 

[tugrul_LordOfDrinks]

Which is worse:

- heating from 40C to 95C and staying 95C for 10 hours of gaming
- heating from 40C to 95C and cooling for 200 times a day in a CUDA coding session, the CUDA code also heats from 90 to 95 for 5 times per second due to high power requirment and 200milliseconds of code running.