Conflicting data vs. TPU reviews

[birdie]

Here's the situation. TPU has reviewed these two GPUs both of which I've personally owned:

https://www.techpowerup.com/review/msi-gtx-1060-armor/ (sold recently)
https://www.techpowerup.com/review/sapphire-radeon-rx-5600-xt-pulse/ (now is my workhorse)

I'd like to know why my data is quite different from to the reviews.

1) Idle temperatures (fan off):

Review:
1060: 45C
5600: 43C

My findings (room temperature 20C, open case):

1060: 42C
5600: 53C, silent BIOS - which is 10C higher than in the review (looking at the temps right now, doing absolutely nothing except typing this text in a Firefox tab, no background applications are running)

amdgpu-pci-0900
Adapter: PCI adapter
vddgfx:       +0.78 V
fan1:           0 RPM  (min =    0 RPM, max = 3200 RPM)
edge:         +53.0°C  (crit = +118.0°C, hyst = -273.1°C)
                       (emerg = +99.0°C)
junction:     +53.0°C  (crit = +99.0°C, hyst = -273.1°C)
                       (emerg = +99.0°C)
mem:          +50.0°C  (crit = +99.0°C, hyst = -273.1°C)
                       (emerg = +99.0°C)
power1:       10.00 W  (cap = 135.00 W)

1) Fan under load:

Review:
1060: 30dB
5600: 28dB

My findings:

1060: is at least twice as quieter as 5600.

I'm not trying to cast doubt on the quality of W1zzard reviews, I just like to know why this could be happening and if anything has changed over the past four years AND if it's possible to re-review these two cards noise levels and idle temperatures just to make sure people can still use the old reviews data as a reference point for their future purchases.

The reason why I'm raising this topic is because I've bought Sapphire Radeon RX 5600 XT Pulse based on the reviews and now I'm really unhappy with what I'm getting. I expected a card which is as quiet as the previous one, with modest temperatures and it's anything but. It's quite hotter and quite louder.

 

[Tatty_Two]

I have been in a similar situation but for different reasons, back in 2009...…….. I bought two new HD4850's for crossfire from the same retailer, got a really good deal, both the same model cards from Asus, can't remember the specific model name but there was a huge difference in temps, both had an identical Bios, so fan curves were the same etc but even putting them in individually and testing them there was 15 - 20 degrees difference in temps at idle and load, having stripped them down one of the cards had thermal paste spread evenly, the other.... well it was like a sparrow had crapped on the GPU, a thick but small pile off centre that was already dry and crusty, less than half the surface covered, suffice to say after a good clean up of both and applying some decent paste temps were fairly consistent, not saying that's what is happening here, just exampling how easy it is for inconsistencies in the assembly process.

 

[birdie]

I've just tried touching the heat pipes right now when the GPU is practically idle and I can keep my finger on them for less than five seconds, so they are really really hot, so it's unlikely that the thermal paste application is to be blamed in my case.

Also, if what you've described was the case I still cannot do anything about it since it will void my warranty and RMA'ing the card right now is near impossible: after all it's working just fine albeit it's a lot hotter and louder than I expected. So, the question remains.

 

[R-T-B]

A few GPU manufactures actually cover a GPU thermal paste swap, unsure what Saphire's policy is but it would not surprise me if they did.

 

[birdie]

A few GPU manufactures actually cover a GPU thermal paste swap, unsure what Saphire's policy is but it would not surprise me if they did.

I live in a country where there's no official Sapphire RMA even available. And that's the largest in terms of surface area country in the world. And it's just like that with a lot of equipment.

 

[RealNeil]

Are all of your system specs the same as the TPU test bench?
Is your airflow the same?

Many things can affect system temps. As someone said, the thermal paste application may be part of the problem.
TBH, I would be surprised to see your temps be the same as the test bench.

 

[birdie]

Are all of your system specs the same as the TPU test bench?
Is your airflow the same?

Many things can affect system temps. As someone said, the thermal paste application may be part of the problem.
TBH, I would be surprised to see your temps be the same as the test bench.

My 1060 GTX temps and noise levels follow closely the TPU review while the RX 5600 XT behaves quite differently, so even if my conditions do not perfectly mimic what W1zzard has in place, it was the case in the past (even by coincidence).

 

[R-T-B]

I live in a country where there's no official Sapphire RMA is even available. And that's the largest in terms of surface area country in the world. And it's just like that with a lot of equipment.

My condolences. They should pay some attention to your market.

 

[FreedomEclipse]

I live in a country where there's no official Sapphire RMA is even available. And that's the largest in terms of surface area country in the world. And it's just like that with a lot of equipment.

Ive delt with Sapphire customer support. They only ever take RMAs through their retail channels. So you'll need to contact the shop you bought it from and start the RMA process through them.

However If you bought a card in a different country to yours then some manufacturers might void the warranty unless you can send the card to a friend or family member with the required documentation and have them deal with the RMA process on your behalf... RMAs tend to take around a month, sometimes it might take longer. id expect yours to take a little longer as you have to ship the defective card back and forth to & from multiple locations.

 

[birdie]

Ive delt with Sapphire customer support. They only ever take RMAs through their retail channels. So you'll need to contact the shop you bought it from and start the RMA process through them.

However If you bought a card in a different country to yours then some manufacturers might void the warranty unless you can send the card to a friend or family member with the required documentation and have them deal with the RMA process on your behalf... RMAs tend to take around a month, sometimes it might take longer. id expect yours to take a little longer as you have to ship the defective card back and forth to & from multiple locations.

But first we need to establish why I would want to RMA it in the first place. Temps alone don't look sufficient.

Here's my GPU idling for a couple of minutes. 47 degrees Celcius with almost no variation.



TPU, AnandTech and all other reviewers paint a very different picture, e.g.

Spoiler: Other reviews

 

[Fouquin]

Looks like the only thing stopping you from checking the thermal paste application is two stickers. Take a razor, slip it under those and just peel them up intact. When you've got the card reassembled you just lay them back down and nobody's the wiser, but your temps could be better.

 

[FreedomEclipse]

But first we need to establish why I would want to RMA it in the first place. Temps alone don't look sufficient.

Here's my GPU idling for a couple of minutes. 47 degrees Celcius with almost no variation.

View attachment 152822

TPU, AnandTech and all other reviewers paint a very different picture, e.g.

Spoiler: Other reviews

It doesnt matter - one time, I contacted XFX to ask if i could remove the coolers off a pair of 6970s to repaste them as they were both hitting about 80-89'c when gaming. Their response was for me to RMA both of the cards.

Warranty varies based on region. some regions are quite relaxed about people removing the coolers off their cards while everywhere else voids warranty.

At the same time, what if it was a temperature probe or sensor that was borked?? I had a 4870 that arrived back from RMA that reported temps 30-40'c lower than a second card of the same make and model. Its quite a rare occurance but it still happens.

 

[moproblems99]

I'll be honest, I doubt they are going to RMA it anyway. I would kill for 53C temps. Those of us who live in hot climates, this is better than our normal. If your temps were in the high 80s, that is different. But to be in the low 50s, not likely.

I was about to ask if your GPU wasn't downclocking but drawing only 10w, it most definitely is.

 

[FordGT90Concept]

1) It's set to Zero RPM so the fan doesn't spin unless the GPU reaches a target temperature (might be 60C) to induce it to spin.

fan1: 0 RPM (min = 0 RPM, max = 3200 RPM)

2) You can adjust the fan curve using Radeon Software.
3) You could have a bad fan if you're hearing a rattle when it does spin.
4) 5600 XT will consume 45w more power (more heat to dissipate, more fan noise) for 40% more performance so it will, overall, be a louder card.

 

[Regeneration]

Sapphire has a hundred variants for each model not listed on their website.

This can also be the result of a different BIOS. There is a new one on their website.

 

[W1zzard]

5600: 53C, silent BIOS - which is 10C higher than in the review (looking at the temps right now, doing absolutely nothing except typing this text in a Firefox tab, no background applications are running)

Here's my GPU idling for a couple of minutes. 47 degrees Celcius with almost no variation.

53°C does seem a bit high, maybe Firefox heats up your GPU a tiny bit?

The 47°C number seems close enough to my review data https://tpucdn.com/review/sapphire-radeon-rx-5600-xt-pulse/images/clocks-and-thermals.png

What RPMs are you getting under load?

 

[mtcn77]

well it was like a sparrow had crapped on the GPU, a thick but small pile off centre that was already dry and crusty, less than half the surface covered, suffice to say after a good clean up of both and applying some decent paste temps were fairly consistent

A similar one: I had an msi 6600gt. Great card with a puny cooler. Great job MSI & Nvidia by the way. Great experience.
Long story short 2 years gone, I buy Arctic Silver 5 and apply a thin coat of the material. The temperatures improve somewhat, it even isn't held by threaded screws, such a funny plugged cooler.
Afterwards, roll on another year, "the gpu is dead."
I try troubleshooting and I see the thermal paste is caked & baked at a corner of the die where the contact patch is most flimsy.
Having to serve regular maintenance intervals for particulate based pastes is real.

I've just tried touching the heat pipes right now when the GPU is practically idle and I can keep my finger on them for less than five seconds, so they are really really hot, so it's unlikely that the thermal paste application is to be blamed in my case.

Also, if what you've described was the case I still cannot do anything about it since it will void my warranty and RMA'ing the card right now is near impossible: after all it's working just fine albeit it's a lot hotter and louder than I expected. So, the question remains.

Could anybody be furtively mining on your rig?

 

[Elysium]

Assuming the data you are providing is correct, you may simply have lost the silicon lottery, albeit badly. The louder noise profile could be due to load temps, although you haven't provided those. Of course, if the card is running hotter at load than the 1060 did, then the fans should be running faster to cool it down. The construction of the fan frame, the fans themselves and the fins of the heatsink, whether there are some kind of minor defects in one or more of them perhaps, could also be responsible for a louder and hotter card.

 

[birdie]

53°C does seem a bit high, maybe Firefox heats up your GPU a tiny bit?

The 47°C number seems close enough to my review data https://tpucdn.com/review/sapphire-radeon-rx-5600-xt-pulse/images/clocks-and-thermals.png

What RPMs are you getting under load?

Again, the test results were taken on a bare Windows 10 desktop while not running anything except Windows Explorer. The systray panel is empty except standard Windows icons and the AMD Radeon Settings icon.

Which test should I run to provide test data? I can run 3DMark/Furmark/certain games (I don't own that many), i.e. The Division 2 or Deus Ex Mankind Divided.

Assuming the data you are providing is correct, you may simply have lost the silicon lottery, albeit badly. The louder noise profile could be due to load temps, although you haven't provided those. Of course, if the card is running hotter at load than the 1060 did, then the fans should be running faster to cool it down. The construction of the fan frame, the fans themselves and the fins of the heatsink, whether there are some kind of minor defects in one or more of them perhaps, could also be responsible for a louder and hotter card.

For a start this GPU has a significantly higher TDP (170W vs 110W) which could explain higher noise levels but according to the TPU review, it's not the case. This is what I'm curious about.

@W1zzard what about TDP? How come in your review the card which has almost twice as high TDP is quieter despite both featuring very similar cooling solutions? And under load Radeon is hotter than NVIDIA GTX 1060. That doesn't make any sense.

 

[W1zzard]

Which test should I run to provide test data? I can run 3DMark/Furmark/certain games (I don't own that many), i.e. The Division 2 or Deus Ex Mankind Divided

any of those should be ok

almost twice as high TDP

I measured 130 W vs 165 W, so roughly +25%. The MSI cooler is definitely worse than Sapphire's.. 2 slot, 2 heatpipe, vs 3 slot 3 heatpipe

 

[EarthDog]

It could be ANYTHING...

Different airflow, ambient temp, idle time to capture data (been on 5 mins or been on for hours?), different bios, borked cooler, borked tim app from the factory. Idle especially can be all over the place.

What are load temps?

Also, forget about comparing the temps of other cards to each other...too many variables. Remember they are different things. A yellow flame on a lighter and yellow flame on a bonfire are the same temperatures... which has more energy and more difficult to cool to the same temperature?

 

[birdie]

When running Furmark stress test (1920x1080, no MSAA).

The noise is like a helicopter taking off nearby.

 

[dorsetknob]

I live in a country where there's no official Sapphire RMA even available. And that's the largest in terms of surface area country in the world. And it's just like that with a lot of equipment.

its Irrelevent that is "Country size"
Country Support would be in relation to population Size /market share
Greenland is largest Island in the world at 839,999 sq miles with a population of 60000 ish
there are villages in europe/america with more people in them.

 

[mtcn77]

160w is normal? What voltage setting? Radeon gpus can and will take undervolting like a great sport, please don't let it overbear on all of us. Just do it and forget it.

 

[Elysium]

When running Furmark stress test (1920x1080, no MSAA).

The noise is like a helicopter taking off nearby.

Wait, 1495RPM sounds like a helicopter taking off? I think that's a fan bearing problem. Even with an open case, that's a pretty distinct description of the sound it's making, which is not normal for under 1500 revs. Moreover though, your GPU is at 99% load but your fans are at less than 50%? There's definitely an RMA-worthy problem with your fans here.