Is RX 9070 VRAM temperature regular value or hotspot?

[AusWolf]

I've slow-cooked VRAM on a video card before. It can take a while but it's irreversible. Once it's cooked, it's either scrap, or underclock and pray, or use for 2D only. The 2 year warranties aren't long enough unless that's your upgrade cycle (and that's assuming they honour the warranty).

Did you overclock it by any chance? I don't overclock, and I've never had VRAM die on me that wasn't already dead.

Did you check the links in my post?

 

[Event Horizon]

Did you overclock it by any chance? I don't overclock, and I've never had VRAM die on me that wasn't already dead.

Did you check the links in my post?

I never overclocked it either. The card was factory OC'd. I shoved it into a microATX case and didn't increase its fan speed to compensate. It ran fine for a couple of years, and then started artifacting in 3D loads with increasing regularity.

Yes I saw the link, and that's great news. For my part, I've turned off zero fan and slightly bumped the fan curve. The noise difference is imperceptible to my ears, but the temps are a nice deal lower.

We can check in again after 5 years, and hopefully everyone's 9070(XT) is as good as the day they bought it (assuming you keep hardware for as long as I do).

 

[rusty caterpillar]

I've slow-cooked VRAM on a video card before. It can take a while but it's irreversible. Once it's cooked, it's either scrap, or underclock and pray, or use for 2D only. The 2 year warranties aren't long enough unless that's your upgrade cycle (and that's assuming they honour the warranty).

You are not alone, 2080, 3080, 3070, 3060Ti users which believed too much in the official VRAM specs given by AIB, than it goes artifacts following by code 43, black screen with no detect and so on.
I feel for you being not aware but, They, deserve what comes to them, I warned them and they come with same old text "is in the official specs".

When I open the GPUs was to late:
1. T pads, didn't had any drv mosfets, too much oil and dust from VRAM pads
2. some of pads very dry(like some wall filler) even on 3 months old GPU
3. Fan drips oil - too much heat under rotor

Some ppl doesn't want your help is no written on your forehead Asus

 

[Code3]

:bastone:

 

[AVATARAT]

:bastone:View attachment 392829

I'm not sure what you're trying to show that in 60 seconds your graph would be lower?
How about 2 hours of gameplay?

 

[freeagent]

I run my GPU OC a few ticks away from crashing. It has thousands of hours of heavy compute, and thousands of hours of games, this summer will be its 3rd summer with me, and it has never run at stock intentionally

I would hope your AMD cards can do the same, because if not, then its just a waste of money imo.

 

[AVATARAT]

If the memory on any of the cards fails, I don't think it will be AMD's failure, but the partners who made poor coolers.

 

[freeagent]

poor coolers.

I don't understand why it is so hard for them to build a decent cooler. There may be exceptions to the rule, but it seems like 80-90c is ok to run. I dunno bro.

I am skeptical.

 

[rusty caterpillar]

I don't understand why it is so hard for them to build a decent cooler. There may be exceptions to the rule, but it seems like 80-90c is ok to run. I dunno bro.

I am skeptical.

Who wouldn't be, I am

:bastone:View attachment 392829

Furmark use very little VRAM as I hope you noticed. Just go for Superposition 4k optimized 20 min, cinematic loop.



If your VRAM is 66 C in 2:27 min Furmark imagine in 20 min Superposition 4k optimized cinematic loop

 

[freeagent]

I would beat it like it owed me money for a couple of months, just like I did with this one lol..

 

[AusWolf]

I run my GPU OC a few ticks away from crashing. It has thousands of hours of heavy compute, and thousands of hours of games, this summer will be its 3rd summer with me, and it has never run at stock intentionally

I would hope your AMD cards can do the same, because if not, then its just a waste of money imo.

Are you saying that if you can't / don't run your card overclocked, then it's a waste of money? If so, that's weird thinking.

 

[LittleBro]

I've slow-cooked VRAM on a video card before. It can take a while but it's irreversible. Once it's cooked, it's either scrap, or underclock and pray, or use for 2D only. The 2 year warranties aren't long enough unless that's your upgrade cycle (and that's assuming they honour the warranty).

Exactly. It's always better with lower temps, as those increase semiconductor lifespan significantly.
I'd like to see my GPU not degrade even in year or two past the warranty period.

 

[Assimilator]

I've slow-cooked VRAM on a video card before. It can take a while but it's irreversible. Once it's cooked, it's either scrap, or underclock and pray, or use for 2D only.

And how did you know it was the memory chips themselves that were the problem? As opposed to the circuitry powering them, or the core's memory controller failing? How do you know that it was temperature to blame, versus defective components? You don't know, you're just theorising, and anecdotes are not compelling evidence.

The 2 year warranties aren't long enough unless that's your upgrade cycle (and that's assuming they honour the warranty).

The millions of people with GPUs older than 2 years make this claim complete bulls**t. Also, warranties are law.

 

[Event Horizon]

And how did you know it was the memory chips themselves that were the problem? As opposed to the circuitry powering them, or the core's memory controller failing? How do you know that it was temperature to blame, versus defective components? You don't know, you're just theorising, and anecdotes are not compelling evidence.


The millions of people with GPUs older than 2 years make this claim complete bulls**t. Also, warranties are law.

Clarify the point you're trying to make. You can feel free to disagree about VRAM degradation and failing outside of warranty being your concern. That's pretty much it. Seems like your sole goal is being argumentative and rude and none of it is in good faith.

 

[rusty caterpillar]

And how did you know it was the memory chips themselves that were the problem? As opposed to the circuitry powering them, or the core's memory controller failing? How do you know that it was temperature to blame, versus defective components? You don't know, you're just theorising, and anecdotes are not compelling evidence.


The millions of people with GPUs older than 2 years make this claim complete bulls**t. Also, warranties are law.

I'll try to explain what happens outside VRAM own operation and their load temperatures.

Power delivery(drv mosfets, inductors capacitors) too hot will mess up voltages which are fed to VRAM and G core, oscillating voltages is enough to keep VRAM hot and jittery FPS.

Hot and too hot Drv mosfets will deliver bad voltages to inductors- those can come to core saturation or not but, you can have coil whine, we go forward to the capacitors which if are hot by proximity heat, the electrolyte inside will dry.
Capacitors will not blown up as the capacitors crisis 2002-2005( when Dell has to pay 300 millions $ blown capacitors Optiplex to mainstream and corporate users) but just being dry will not hold enough electricity, therefore oscillations again.

Memory controller chips, too hot(no T pads in place see MSI and Giggle bite) will not handle properly VRAM clocks resulting in oscillating FPS.

My sources : my own experience in working on several systems.
NorthWest repair channel on YT I believe I watched 90 % of his videos, IMO not an experience to ignore.
Yes has some of his videos where he repairs:
1. knock off components by the users,
2.cracked PCBs user and shipping _ but why are PCBs so weak and brittle to start with > this started around 2018.
3.Ripped pads under the GPU core or VRAM(due to the fact that they use weak PCB and soldering without lead) Check yourself when AIB start to replace lead soldering

You can ignore 1, 2, 3 and focus on cooling videos.

2 examples I have at hand : Micron 2018 Look at the title

9070 and 9070 XT much more condensed PCB compared to a 3080, so we have more heat by proximity.

Electronics mostly capacitors bellow.

The capacitor crisis and the new capacitors electrolyte drying up causes

Consider all above "anecdotes" but, AIB will be paid in if your GPU bricks not me.

Hope it helps

Clarify the point you're trying to make. You can feel free to disagree about VRAM degradation and failing outside of warranty being your concern. That's pretty much it. Seems like your sole goal is being argumentative and rude and none of it is in good faith.

I believe he is just confused how other components on GPU PCB is influencing VRAM temps also ignores heat by proximity, specially on condensed PCB like 9070 and 9070XT.

 

[Assimilator]

Clarify the point you're trying to make. You can feel free to disagree about VRAM degradation and failing outside of warranty being your concern. That's pretty much it. Seems like your sole goal is being argumentative and rude and none of it is in good faith.

You're the one who claimed that you "slow-cooked VRAM" without providing any evidence that it was actually the VRAM that failed, or that said failure was caused by temperature. Making claims without providing evidence is bad faith.

I'll try to explain what happens outside VRAM own operation and their load temperatures.

Power delivery(drv mosfets, inductors capacitors) too hot will mess up voltages which are fed to VRAM and G core, oscillating voltages is enough to keep VRAM hot and jittery FPS.

Hot and too hot Drv mosfets will deliver bad voltages to inductors- those can come to core saturation or not but, you can have coil whine, we go forward to the capacitors which if are hot by proximity heat, the electrolyte inside will dry.
Capacitors will not blown up as the capacitors crisis 2002-2005( when Dell has to pay 300 millions $ blown capacitors Optiplex to mainstream and corporate users) but just being dry will not hold enough electricity, therefore oscillations again.

Memory controller chips, too hot(no T pads in place see MSI and Giggle bite) will not handle properly VRAM clocks resulting in oscillating FPS.

My sources : my own experience in working on several systems.
NorthWest repair channel on YT I believe I watched 90 % of his videos, IMO not an experience to ignore.
Yes has some of his videos where he repairs:
1. knock off components by the users,
2.cracked PCBs user and shipping _ but why are PCBs so weak and brittle to start with > this started around 2018.
3.Ripped pads under the GPU core or VRAM(due to the fact that they use weak PCB and soldering without lead) Check yourself when AIB start to replace lead soldering

You can ignore 1, 2, 3 and focus on cooling videos.

2 examples I have at hand : Micron 2018 Look at the title

9070 and 9070 XT much more condensed PCB compared to a 3080, so we have more heat by proximity.

Electronics mostly capacitors bellow.

The capacitor crisis and the new capacitors electrolyte drying up causes

Consider all above "anecdotes" but, AIB will be paid in if your GPU bricks not me.

Hope it helps


I believe he is just confused how other components on GPU PCB is influencing VRAM temps also ignores heat by proximity, specially on condensed PCB like 9070 and 9070XT.

None of those videos have anything to do with overheating VRAM. The capacitor plague was over 2 decades ago now.

 

[rusty caterpillar]

None of those videos have anything to do with overheating VRAM. The capacitor plague was over 2 decades ago now.

1. You'll find plenty of evidence. Early micron failed because of heat, the official threshold didn't hold ground. With few clicks and 5 min watching time you not gone learn anything, even you the "Assimilator"
2. Yes, but that doesn't mean the new capacitors doesn't have dried out electrolyte over exposure to heat by proximity. Again, not gonna pop or explode but, will not hold properly the voltages resulting in chain reactions.

 

[Assimilator]

Early micron failed because of heat

Once again, zero evidence provided.

Yes, but that doesn't mean the new capacitors doesn't have dried out electrolyte over exposure to heat by proximity.

Which is completely irrelevant because MODERN GPUS DON'T USE ELECTROLYTIC CAPACITORS you ABSOLUTE SPANNER. Because solid-state caps are more heat-resistant. Hey it's almost like engineers know what they're doing and you don't.

 

[rusty caterpillar]

I'm not gonna make a collage out of his videos every time he mention overheating or lack of pads on drv mosfets or Memory controllers and so on. Is up to you.

Once again, zero evidence provided.


Which is completely irrelevant because MODERN GPUS DON'T USE ELECTROLYTIC CAPACITORS you ABSOLUTE SPANNER. Because solid-state caps are more heat-resistant. Hey it's almost like engineers know what they're doing and you don't.

You chose to ignore something called respect as much as the information around you, yet you, are very opinionated without yourself coming with evidence, refusing to spend time and analyze correctly what was given to you.

Zero evidence in your opinion.

Assimilate the following and next time pls be respectful:

Modern GPU use not only SP caps but, also Ceramic capacitors for WHICH excessive heat can cause changes in the ceramic material's properties, leading to a reduction in capacitance or increased losses. Over time, this can degrade the performance of the capacitor.
"Drying out" term is still used in the electronics jargon because has same effect over the new ceramic capacitors as it has for electrolytic capacitors when they dried out: reduction in capacitance or increased losses.

SP caps or solid polymer capacitor used widely in GPUs

The polymer electrolyte in solid polymer capacitors can deteriorate over time due to thermal degradation. This leads to a decrease in electrical conductivity, which can increase the ESR and reduce the capacitor's performance.
The life of a solid polymer capacitor increases by a factor of 10 for each 20°C reduction in operating temperature.

Because solid-state caps

Solid-state caps is just a marketing term which again, proves where your "knowledge" comes from... Same as the one that tells you VRAM is fine at 90C. Right?

Solid Polymer caps or SP caps, is not easier to write SP caps instead of Solid-State caps?

 

[cinemaware]

Just wrote an update about my up-to-102C 9070 XT here.

 

[AusWolf]

Chill out, lads. There's still only one conclusion, and it's this:

1. Micron specifies GDDR6 to be safe to run up to 95 °C package temp.

2. Your 9070 (XT) only reports junction temp.

3. There is a considerable delta between package and junction temp that you don't see.

4. Considering the delta, your 90 °C junction temp means a considerably lower package temp, therefore, it's fine. The hard 108 °C junction temp limit coded into the VBIOS isn't an engineering mistake that a whole bunch of engineers made across several AIBs.

 

[LittleBro]

Burn baby, burn!

I mean high temps problems affects Nvidia, too. AMD fucked up VRAM temps to stay competitive with Nvidia temp-wise. Collegionalism. ol'Jensy and ol'Su are cousins, after all.

 

[JustBenching]

Chill out, lads. There's still only one conclusion, and it's this:

1. Micron specifies GDDR6 to be safe to run up to 95 °C package temp.

2. Your 9070 (XT) only reports junction temp.

3. There is a considerable delta between package and junction temp that you don't see.

4. Considering the delta, your 90 °C junction temp means a considerably lower package temp, therefore, it's fine. The hard 108 °C junction temp limit coded into the VBIOS isn't an engineering mistake that a whole bunch of engineers made across several AIBs.

What does "safe" mean though?

 

[rusty caterpillar]

Burn baby, burn!

I mean high temps problems affects Nvidia, too. AMD fucked up VRAM temps to stay competitive with Nvidia temp-wise. Collegionalism. ol'Jensy and ol'Su are cousins, after all.

Jensy lend some ROPs to Su so she can sell some cards, he asked in return to increase clocks on VRAM, so 9000 will die faster than Nvidia. Sneaky bastard

What does "safe" mean though?

No AIB stated that VRAM can run at those temperatures for 6 months or 2 years. They prolly just said 24/7 but didn't specified for how long.

 

[AVATARAT]

Burn baby, burn!

I mean high temps problems affects Nvidia, too. AMD fucked up VRAM temps to stay competitive with Nvidia temp-wise. Collegionalism. ol'Jensy and ol'Su are cousins, after all.

I just put an extra fan on the back panel and the VRAM temp dropped to 88 -> 78-82C, GPU 85C -> 76C (+undervolt), so purely a design issue with the cooler.

Nvidia models have much more massive coolers and temperatures are 10C+ lower (depend of model).

If we water cooled them, all temperatures would drop significantly, so AIB are just making crap decisions. Same with the Igorslab card - low quality control.

But hey, maybe Nvidia/AMD/AIB just fired their quality control teams and replaced them with AI AI AI and this is the result