Is RX 9070 VRAM temperature regular value or hotspot?

[Eugene0]

Я упомянул понижение напряжения как тест, потому что вы утверждали, что карта не деградировала за 3 года (частоты остались прежними). Понизьте напряжение и выясните, насколько она деградировала за 3 года. Она деградирует (любой кремний деградирует), но запас прочности от производителя не позволяет вам это увидеть... в основном.

you mean, got damaged by composition around

 

[AusWolf]

I was mentioning undervolting as a test because you claimed that the card did not degrade in 3 years (clocks remained the same). Undervolt it and actually find out how much it degraded in 3 years. It is degrading (all silicon degrades) but the safety margin from the manufacturer is preventing you from seeing it....mostly.

And how would I know that if I never undervolted the card before?

Yeap, exactly. Was trying to explain it before but didn't manage and gave up

You had a point which I disagreed with. It's not the same as "failing to explain".

 

[TechBuyingHavoc]

And how would I know that if I never undervolted the card before?

Then you can't make solid claims on if a card is degraded or not (from high VRAM temps or anything else) as you did earlier. You didn't know what the limits of the card were at the start and what they are now after the degradation period.

 

[AusWolf]

Then you can't make solid claims on if a card is degraded or not (from high VRAM temps or anything else) as you did earlier. You didn't know what the limits of the card were at the start and what they are now after the degradation period.

I wouldn't say a card has degraded if it can't do something that it is not supposed to do by its normal operation in the first place. Or at least not degraded enough to be of concern.

 

[JustBenching]

I wouldn't say a card has degraded if it can't do something that it is not supposed to do by its normal operation in the first place. Or at least not degraded enough to be of concern.

It can't do something that it was doing due to electromigration. That is literally the definition of degradation.

 

[AusWolf]

It can't do something that it was doing

Read my post above - I never undervolted, so I have no idea whether it was ever doing "that" or not. It ran stock clocks at stock voltages 3 years ago, and it does now as well. I'm not interested in pushing limits.

Even if it has degraded,
1. I have no means to obtain any proof of it, and
2. It hasn't degraded enough to affect its normal operation.

Everything else is an empty claim, so why worry?

 

[pulcsi]

1. What is the idle power (TBP) consumption of the card?
2. What is the idle speed of VRAM?
3. What is the idle speed of GPU?

1. --About 13W (if i open chrome it goes up to 40W and then goes back to 13W)
2. --57Mhz
3. - Jumps between 54 - 105 Mhz
At these the vram sits at 60-62°C. On full load it goes to 84°C.

What I found out is that no matter what, even on the base 9070, the Vram goes up to 84°C, which is better than the XT BUT it should be lower because my 9070 uses the XT's cooler solution (9070 Sapphire Pure).
Somewhere I read that the vram is too close to the gpu so it gets hot.

 

[Eugene0]

1. --Около 13 Вт (если я открою Chrome, то он поднимется до 40 Вт, а затем снова опустится до 13 Вт)
2. --57МГц
3. - Скачки между 54 - 105 МГц
При этом температура видеопамяти составляет 60-62°C. При полной нагрузке она достигает 84°C.

Я обнаружил, что независимо от обстоятельств, даже на базовой модели 9070, температура видеопамяти достигает 84°C, что лучше, чем у XT, НО она должна быть ниже, поскольку в моей модели 9070 используется более охлаждающее решение XT (9070 Sapphire Pure).
Где-то я читал, что видеопамять расположена слишком близко к графическому процессору, поэтому она нагревается.

this is a temperature of seriously stress testing level ?, great temperature

 

[pulcsi]

this is a temperature of seriously stress testing level ?, great temperature

Metro Exodus Enhanced Edition, Extreme preset, Raytracing max + everything maxed.

Lighter load vram likes to sit on 80°C

 

[Eugene0]

Metro Exodus Enhanced Edition, настройки Extreme, трассировка лучей на максимум + все на максимум.

Более легкая загрузка vram любит сидеть при 80°C

Это хорошая температура

аnd at full load your temperature too is good

 

[QuietBob]

I've been watching this thread, and I've seen similar VRAM temps on my MBA 7900XTX. I am of the opinion that this is by design, rather than a sign of inadequate cooling, incomplete die contact or poor workmanship on the part of the AIB.

The highest VRAM temp I've observed on my card was 94 C. That was in a sort of a worst case scenario: gaming in a 28 C room with the GPU pumping out constant 355 W. And even then the fans went up to only 68%.

I mean, if this is considered a safe temp on a card made precisely to AMD design guidelines, I see no reason for alarm. What's interesting, maximum VRAM temp for me does not really depend on the ambient -- it'll approach 90 C even in a 20 C room.

For reference, these are idle values with Zero RPM off in a 23 C room:

 

[cinemaware]

I've been watching this thread, and I've seen similar VRAM temps on my MBA 7900XTX. I am of the opinion that this is by design, rather than a sign of inadequate cooling, incomplete die contact or poor workmanship on the part of the AIB.

The highest VRAM temp I've observed on my card was 94 C. That was in a sort of a worst case scenario: gaming in a 28 C room with the GPU pumping out constant 355 W. And even then the fans went up to only 68%.

I mean, if this is considered a safe temp on a card made precisely to AMD design guidelines, I see no reason for alarm. What's interesting, maximum VRAM temp for me does not really depend on the ambient -- it'll approach 90 C even in a 20 C room.

I don't mind the high VRAM temps, even if they reach 100C (as it is in my case, with default fan curve) as long as I can read somewhere official that "this is fine" (but non-ironically of course)

 

[rusty caterpillar]

I don't mind the high VRAM temps, even if they reach 100C (as it is in my case, with default fan curve) as long as I can read somewhere official that "this is fine" (but non-ironically of course)

Others do mind high VRAM temps as order for cooling priority on a card is VRAM, pwr delivery, specially drv mosfets, GPU core.

 

[AusWolf]

Others do mind high VRAM temps as order for cooling priority on a card is VRAM, pwr delivery, specially drv mosfets, GPU core.

It's all been accounted for by your card's maker. All you need is good case airflow.

 

[TechBuyingHavoc]

It's all been accounted for by your card's maker. All you need is good case airflow.

Has it been accounted for? I would like to see an official statement before I just believe these companies on an implicit assumption that all of their products are working as designed.

 

[AusWolf]

Has it been accounted for? I would like to see an official statement before I just believe these companies on an implicit assumption that all of their products are working as designed.

You've got a warranty on it.

 

[Zach_01]

I've been watching this thread, and I've seen similar VRAM temps on my MBA 7900XTX. I am of the opinion that this is by design, rather than a sign of inadequate cooling, incomplete die contact or poor workmanship on the part of the AIB.

The highest VRAM temp I've observed on my card was 94 C. That was in a sort of a worst case scenario: gaming in a 28 C room with the GPU pumping out constant 355 W. And even then the fans went up to only 68%.

I mean, if this is considered a safe temp on a card made precisely to AMD design guidelines, I see no reason for alarm. What's interesting, maximum VRAM temp for me does not really depend on the ambient -- it'll approach 90 C even in a 20 C room.

For reference, these are idle values with Zero RPM off in a 23 C room:

View attachment 392197

With zero rpm after 15min idling

Nitro+

 

[Knight47]

How's the VRAM temps with multiple high refresh rate monitors with Zero RPM?

 

[rusty caterpillar]

9070XT @ 1440 P in CP 2077 from min 4:33 to 6:56, fans at approximative 3500 RPM which means noise, yet VRAM at 81-82 C.

 

[Code3]

Old its better lol

9800x3d >push pull aio 480 radiator
9 fans reverse pwr

 

[Assimilator]

Wow this trainwreck of a thread is still going.

 

[AusWolf]

View attachment 392544
Old its better lol

9800x3d >push pull aio 480 radiator
9 fans reverse pwr

What's better? Where is your VRAM temperature? Have you read the thread title by any chance?

 

[Code3]

YES, but mine doesn't have sensors for the vram and anyway it was just for fun given the little sense of this thread. How annoying, brother drink less coffee

 

[AusWolf]

Wow this trainwreck of a thread is still going.

It's self-made expert engineers discussing expert engineering stuff, because the professional ones working for AMD, Nvidia and all their AIBs surely don't know what they're doing. Also, the Earth is flat.

You've got to love how everybody knows everything better than everybody else these days.

YES, but mine doesn't have sensors for the vram and anyway it was just for fun given the little sense of this thread. How annoying, brother drink less coffee

Ah, so if AMD didn't put VRAM sensors onto the 9070 (XT), then threads like this wouldn't exist, and everybody would live in peace. Kind of like what Nvidia did with the hotspot sensor on Blackwell. I get 'ya.

---

Lads, 95 ˚C is the max case temp for GDDR6, not tJmax.

Sapphire Radeon RX 9070 XT Pulse

Are you talking about the very very gradual decline in clocks from 2800 to 2779 after the initial spike, or the initial spike? That brief spike to 2910MHz isn't representative of the sustained load, it's not a valid part of the sustained load test, but a necessary part as the test obviously...

www.techpowerup.com

(in case someone is confused: "case temp" is the temperature measured at the casing of a chip, junction temp or tJ is the hottest temp measured among any of the actual semiconductor within the chip)

Junction temperature - Wikipedia

en.wikipedia.org

 

[Event Horizon]

It's self-made expert engineers discussing expert engineering stuff, because the professional ones working for AMD, Nvidia and all their AIBs surely don't know what they're doing. Also, the Earth is flat.

You've got to love how everybody knows everything better than everybody else these days.


Ah, so if AMD didn't put VRAM sensors onto the 9070 (XT), then threads like this wouldn't exist, and everybody would live in peace. Kind of like what Nvidia did with the hotspot sensor on Blackwell. I get 'ya.

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).