Asus Rog Strix RX 5700 XT thermals

[AusWolf]

Hi,

As I got tired of waiting for a decent midrange offering (or available stock) from either AMD or nvidia, I've just bought an Asus Rog Strix RX 5700 XT. I read the initial bad reviews (except for TPU which was fairly positive), but 1. this was the cheapest model in stock, 2. Asus supposedly corrected their thermal/voltage issues with the 5700 series around January, so I thought I'd give it a go - not to mention Aura sync compatibility with my motherboard which is a secondary factor, but still nice.

Anyway, long story short, my temperatures don't look too great at first glance. I might be just worrying too much, as I haven't owned an AMD card in the last 6-7 years, but better safe than sorry. So, after 30+ minutes of gaming, I get GPU temps of 78-79 °C, hotspot (junction) temp between 90-95 °C and VRAM temp of 94-96 °C. I've read AMD's statement that junction temps of up to 110 degrees are tolerable, but I'm not sure about the VRAM. Part of me tells me that it's the same "junction temp" thing so it's OK, but another part of me thinks something's not right. I get no artefacts or other issues whatsoever. I've also tried undervolting and lowering the power target on the GPU, but that also relaxes the fan curve so temperatures remain the same. I basically have to run a (pretty loud) manual fan curve to keep temps at fairly decent levels.

But as I said - I haven't owned an AMD graphics card in years, so I don't really know what's acceptable nowadays and what isn't. What do you guys think?

(hardware specs are in my profile)

 

[WarTherapy1195]

Hi,

As I got tired of waiting for a decent offering (or available stock) from either AMD or nvidia, I've just bought an Asus Rog Strix RX 5700 XT. I read the initial bad reviews (except for TPU which was fairly positive), but 1. this was the cheapest model in stock, 2. Asus supposedly corrected their thermal/voltage issues with the 5700 series around January, so I thought I'd give it a go - not to mention Aura sync compatibility with my motherboard which is a secondary factor, but still nice.

Anyway, long story short, my temperatures don't look too great at first glance. I might be just worrying too much, as I haven't owned an AMD card in the last 6-7 years, but better safe than sorry. So, after 30+ minutes of gaming, I get GPU temps of 78-79 °C, hotspot (junction) temp between 90-95 °C and VRAM temp of 94-96 °C. I've read AMD's statement that junction temps of up to 110 degrees are tolerable, but I'm not sure about the VRAM. Part of me tells me that it's the same "junction temp" thing so it's OK, but another part says something's not right. I've also tried undervolting and lowering the power target on the GPU, but that also relaxes the fan curve so temperatures remain the same. I basically have to run a (pretty loud) manual fan curve to keep temps at fairly decent levels.

But as I said - I haven't owned an AMD graphics card in years, so I don't really know what's acceptable nowadays and what isn't. What do you guys think?

(hardware specs are in my profile)

wasn't there a "first revision" that had mounting issues with the wrong screws or something?
that lead to insane memory temps.

 

[AusWolf]

wasn't there a "first revision" that had mounting issues with the wrong screws or something?
that lead to insane memory temps.

There was, but I thought Asus got rid of the issue with batches after January. Link

 

[WarTherapy1195]

but to get to the temperatures.

the GPU is "fine" and quite normal for a 5700XT (not the best but there are worse cards)

the memory is unnecessarily hot but still a few degrees away from dangerous.

 

[AusWolf]

but to get to the temperatures.

the GPU is "fine" and quite normal for a 5700XT (not the best but there are worse cards)

the memory is unnecessarily hot but still a few degrees away from dangerous.

Wow! I remember the times when running your Bulldozer CPU at 61 °C was considered bad.

It's funny that I've never worried about VRAM temperatures before - not because they were always OK, but because I've never owned a graphics card that reported memory temps. Weird how a single piece of data in GPU-Z can change one's perspective on hardware monitoring.

Thanks for the reply btw. This might be worth a shot, I guess.

 

[WarTherapy1195]

Wow! I remember the times when running your Bulldozer CPU at 61 °C was considered bad.

It's funny that I've never worried about VRAM temperatures before - not because they were always OK, but because I've never owned a graphics card that reported memory temps. Weird how a single piece of data in GPU-Z can change one's perspective on hardware monitoring.

Thanks for the reply btw. This might be worth a shot, I guess.

i'd rather keep the card untouched and have the warranty TBH..

better have worse temps but the ability to RMA instead of better temps and with bad luck a dead card in a month.

 

[AusWolf]

i'd rather keep the card untouched and have the warranty TBH..

better have worse temps but the ability to RMA instead of better temps and with bad luck a dead card in a month.

True. I might do it by the end of the warranty period - if I end up keeping it that long.

 

[Grom777]

Thanks for the reply btw. This might be worth a shot, I guess.

oh, you would know how I suffered with this video card, I endured black screens, blue screens, restarting the computer, artifacts in games for a whole year, plus I was lucky with the bonus marriage of the first batch because of this very high memory temperatures, hot spots, RMA I could not use it because I bought a video card in another country and sending it was very expensive, I tried on the YouTube link Husky put nylon washers, it did not help me, the temperatures were still high, I tried undervolting, it also did not help, I am sad and do not know what to do , technical support refuses to tell me the sizes of the necessary screws, they also refuse to send me the necessary screws

 

[AusWolf]

oh, you would know how I suffered with this video card, I endured black screens, blue screens, restarting the computer, artifacts in games for a whole year, plus I was lucky with the bonus marriage of the first batch because of this very high memory temperatures, hot spots, RMA I could not use it because I bought a video card in another country and sending it was very expensive, I tried on the YouTube link Husky put nylon washers, it did not help me, the temperatures were still high, I tried undervolting, it also did not help, I am sad and do not know what to do , technical support refuses to tell me the sizes of the necessary screws, they also refuse to send me the necessary screws
View attachment 178884

Wow! Did this happen before or after you tried the nylon washer fix? Your temps should never spike straight up to max if there is sufficient contact between the heatsink and the GPU die.

Sometimes applying more pressure on the screws can undo the GPU standoffs on the heatsink, eliminating all contact between the heatsink and the GPU. I'd check the standoffs right away, making sure I see no thread between the heatsink and the PCB - that is, the standoffs are in place and holding the heatsink onto the GPU die and not above it.

It also seems like you applied a pretty heavy overclock on both the GPU and VRAM, which I would totally not recommend with any modern graphics card, this one especially. If you check some recent graphics card reviews on youtube from Hardware Unboxed or Gamer's Nexus, you'll see why. Put all your Wattman settings back to default and check your temps again.

 

[kapone32]

i'd rather keep the card untouched and have the warranty TBH..

better have worse temps but the ability to RMA instead of better temps and with bad luck a dead card in a month.

It's really not that difficult. There were Gigabyte RX470 cards that got terrible retailer reviews. I found the issue was that Gigabyte was not applying enough thermal paste for the fin array they had gone with for that cooler so most of the thermal paste sat in the ribs of the fins. This made the card that least expensive RX470 cards to buy but doomed it to poor sales. Indeed it is one of the only RX470 or Vega 56 cards that you can still buy new retail. Being that it is so inexpensive I use them in my customer builds as they used to be a consistent $149.99 Canadian which is academic for a ultra budget Gaming rig as anyhting decent (Faster than a Vega 56 or 1070) is at least double that with a premium. I run a MSI Kombustor run to see if the card needs it first though.

 

[AusWolf]

It's really not that difficult. There were Gigabyte RX470 cards that got terrible retailer reviews. I found the issue was that Gigabyte was not applying enough thermal paste for the fin array they had gone with for that cooler so most of the thermal paste sat in the ribs of the fins. This made the card that least expensive RX470 cards to buy but doomed it to poor sales. Indeed it is one of the only RX470 or Vega 56 cards that you can still buy new retail. Being that it is so inexpensive I use them in my customer builds as they used to be a consistent $149.99 Canadian which is academic for a ultra budget Gaming rig as anyhting decent (Faster than a Vega 56 or 1070) is at least double that with a premium. I run a MSI Kombustor run to see if the card needs it first though.

It's weird how these tiny mistakes can doom a line of product. It's also weird how these mistakes do not appear to show during QA (or maybe some QA people don't care and just pass anything).

Update on my ROG Strix 5700 XT: I built up the courage to apply 1 mm plastic washers to the two VRM heatsink retention screws and swapped my 3 front case fans to more powerful units (not by much, as I'm a silence freak). Now my GPU temp hovers around 72-74 °C (junction between 90-95 °C) and VRAM between 90-92 °C. Still not great, but acceptable, I guess.

 

[Grom777]

Wow! Did this happen before or after you tried the nylon washer fix? Your temps should never spike straight up to max if there is sufficient contact between the heatsink and the GPU die.

Sometimes applying more pressure on the screws can undo the GPU standoffs on the heatsink, eliminating all contact between the heatsink and the GPU. I'd check the standoffs right away, making sure I see no thread between the heatsink and the PCB - that is, the standoffs are in place and holding the heatsink onto the GPU die and not above it.

It also seems like you applied a pretty heavy overclock on both the GPU and VRAM, which I would totally not recommend with any modern graphics card, this one especially. If you check some recent graphics card reviews on youtube from Hardware Unboxed or Gamer's Nexus, you'll see why. Put all your Wattman settings back to default and check your temps again.

the temperatures did not change before and after the washers, I did not notice the difference, when I pressed the board and the radiator, nothing moved, everything was tight, on the screen I only raised the memory frequency, here is another screen where everything is at the stock value. temperatures are also not very good

but with power the limit is 175w and low frequency

 

[AusWolf]

the temperatures did not change before and after the washers, I did not notice the difference, when I pressed the board and the radiator, nothing moved, everything was tight, on the screen I only raised the memory frequency, here is another screen where everything is at the stock value. temperatures are also not very goodView attachment 178984
but with power the limit is 175w and low frequency
View attachment 178985

Hmm... there's a huge gap between your GPU temp and hotspot temp, which makes me believe that the cooler is not seated properly (which can be the case even if nothing is moving - thanks to the standoffs I mentioned), or there's bits of thermal paste missing. Or maybe your card's cooler is simply defective.

 

[Grom777]

Hmm... there's a huge gap between your GPU temp and hotspot temp, which makes me believe that the cooler is not seated properly, or there's bits of thermal paste missing. Or maybe your card's cooler is simply defective.

show your tests 5-10min in FurMark against the background of GPU-Z sensors

 

[AusWolf]

show your tests 5-10min in FurMark against the background of GPU-Z sensors

5 minutes of Furmark 1280x720 default, all AMD driver settings on default, 100% power target. Everything is stable, my only problem is the memory temp - which hovers around 90-92 °C in The Witcher 3, so not a big deal I guess.

Edit: I see your GPU gets quite a lot of voltage on stock. I wonder why that is. What is the BIOS version on your card?

 

[Grom777]

5 minutes of Furmark 1280x720 default, all AMD driver settings on default, 100% power target. Everything is stable, my only problem is the memory temp - which hovers around 90-92 °C in The Witcher 3, so not a big deal I guess.

made 1280x720 by default we have almost the same temperature difference, the last BIOS costs
017.001.000.049.000000
Device Id: 1002 731F
115-D182PI0-100

we need to get rid of this video card or look for the necessary screws

 

[AusWolf]

made 1280x720 by default we have almost the same temperature difference, the last BIOS costs
017.001.000.049.000000
Device Id: 1002 731F
115-D182PI0-100
View attachment 178998

we need to get rid of this video card or look for the necessary screws

Well, your peak GPU temperature looks similar, but your hotspot spikes straight up to 105 C, which is just a hair away from the limit. As for me, it's a steady increase to about 92-94 C which is still fine. Not to mention your fans spinning at 2700 rpm, while mine barely touches 2000 (that's probably why my VRAM gets so hot).

I agree that this cooler design from ASUS is flawed, but your sudden temperature spike (compared to my steady increase) makes me believe that there is some contact issue between your GPU and the cooler, which could be improved (maybe with the shorter screws, or seating the standoffs properly, or better thermal paste, or something).

I bought mine 3 weeks ago. It would be a waste to get rid of it so soon. Besides, there's literally nothing to buy instead.

 

[Grom777]

I bought mine 3 weeks ago. It would be a waste to get rid of it so soon. Besides, there's literally nothing to buy instead.

can you tell you bought this new card in the store? is she not from the first batch?

 

[AusWolf]

can you tell you bought this new card in the store? is she not from the first batch?

I bought it from a store brand new, but I have no clue whether it's from one of the fixed batches or not.

 

[Athlonite]

take it back tell em it's crap and get a refund go buy a sapphire pulse or nitro 5700XT instead

 

[Grom777]

I bought it from a store brand new, but I have no clue whether it's from one of the fixed batches or not.

check the batch number on the box and compare with this https://rog.asus.com/articles/gamin...og-strix-radeon-rx-5700-series-graphics-cards

if your card is from a new batch, then it's very sad, what kind of trust do you have in Asus...

 

[AusWolf]

check the batch number on the box and compare with this https://rog.asus.com/articles/gamin...og-strix-radeon-rx-5700-series-graphics-cards

if your card is from a new batch, then it's very sad, what kind of trust do you have in Asus...

Yep, it's listed. I guess I've got an old one after lying around the warehouse for quite some time.

Edit: I've just sent an email to ASUS UK customer support regarding the upgraded screws. We'll see what they come up with.

 

[Grom777]

Yep, it's listed. I guess I've got an old one after lying around the warehouse for quite some time.

my part No: 90YV0D90-M0NA00 on the box and it gets into the first batch

 

[Grom777]

only 6 screws are needed, so a person changes them, how would you know the dimensions of these screws?

 

[AusWolf]

only 6 screws are needed, so a person changes them, how would you know the dimensions of these screws?

I've read (or heard) it somewhere that regular m.2 screws do the job, though I wouldn't try that.

Besides, I've received an answer from ASUS UK Tech Support. They said that the retailer is responsible for all returns, replacements and repairs, they (ASUS) do not deal with such matters... so back at stage 1.