I saw a video claiming that this 5700xt is one of amd's hottest cards available and slamming it for poor heatsink etc. what mods can you make to a card that overheats easily like this cool down? will just lowering the temp of the case, adding extra exhaust fans help?
TechPowerUp compiled a nice chart of the 5700 and 5700XT cards they tested. Aftermarket cards do do a better job than AMD's reference cooler, but mid to upper 70s on the GPU under gaming loads isn't bad, much better than the 90s of the reference edition. Not really anything you can do to improve it except spin the fans faster.
According to this Tom's Hardware article, AMD says the Maximum Operating Temperature of the RX5700 is 110C: AMD Explains Why 110-Degree Operating Temps Are 'in Spec' for RX 5700 | Tom's Hardware
Excerpts from above link:
AMD said that it previously relied on "a single sensor that was placed in the vicinity of the legacy thermal diode" to measure a GPU's core temperature. That changed with the release of the AMD Radeon VII in February. The graphics card has "enhanced thermal monitoring," according to AMD, and laid the foundation for the RX 5700's "extensive network of thermal sensors distributed across the entire GPU die." Where once there was one sensor, now there are many.
"Instead of setting a conservative, ‘worst case’ throttling temperature for the entire die, the Radeon RX 5700 series GPUs will continue to opportunistically and aggressively ramp clocks until any one of the many available sensors hits the ‘hotspot’ or ‘Junction’ temperature of 110 degrees Celsius. Operating at up to 110C Junction Temperature during typical gaming usage is expected and within spec. This enables the Radeon RX 5700 series GPUs to offer much higher performance and clocks out of the box, while maintaining acoustic and reliability targets."
You bought one of the worst AIB RX5700XT according to this review:
Worst 5700 XT Period, Asus TUF Gaming X3 RX 5700 XT OC, DON'T BUY! - YouTube
It is operating with GDDR6 temps above safe operating temperature according to them.
I would recommend you just RMA the card and refuse the same one as a replacement and point to that review.
There are some good AIB RX5700XT cards you could buy as alternative.
As far as I am aware, Just because you bought a GPU Card that has bad reviews isn't a reason to be RMAed.
You have heard of that adage "Buyer's beware". Well that applies here. If the buyer didn't do any research and bought the GPU card and later on finds out the buyer purchased a subpar GPU card, in respect to other similar GPU cards, that isn't the manufacturer's fault but rather the Buyer fault. As long as the Manufacturer didn't advertised the GPU card fraudulently.
I doubt the OP can RMA the card and request a different model GPU card because of one bad review. Asus can probably come up with just as many good reviews to counter any bad reviews.
Besides, RMA involves replacing or repairing the same model. Unless the Manufacturer doesn't have any of the same model to replace, then I guess they can replace it with a similar or upgrade model.
Thing is though, I found this video of the KitGuru, a reputable third party reviewer, of the 5700 TUF model, and they point to a lack of memory cooling as well. Given the only difference between the 5700 and 5700XT models, likely, are the GPU, it's entirely possible ASUS messed up as badly as AMD did when they designed their reference HD 5970 which lacked cooling for an entire bank of VRMs.
I am playing Devil's Advocate in my reply:
Yes, I understand what you and Colesdav are saying and I also agree. But as long as Asus doesn't publicly acknowledge and announce that their RX5700 model is badly designed, As far as I am concerned, Asus won't do anything about it. This would be an very expensive recall to replace everyone that purchased that particular model or to modify it to work better.
In that case, IMHO, then the only recourse that the Asus Users have is if Asus has a Product Recall or the Users start a Class Action lawsuit against Asus.
Without a Product Recall, I don't believe Asus will replace the badly engineered GPU card for another model under RMA terms.
Unless Asus publicly announces that the RX5700 is to be returned so that Asus can modified or replace the defective engineered GPU Card under RMA terms.
I imagine Asus would send either an Email or letter to everyone that purchased that model if there is an official recall.
Put yourself in Asus place. A User RMAs the RX5700 saying reviews indicates it is badly designed and the User wants the RX5700 model to be replaced with a different Model under Warranty. I bet Asus will reply that is not possible and that the GPU card works as advertised. Then sends the RX5700 back to the User. Now Asus could secretly modify the RX5700 before returning it to the User when it is RMAed.
Telling Asus about the badly designed engineered GPU model is something that Asus probably is well aware of but won't admit it unless they are planning on a product recall. The moment Asus admits to one User that the GPU card was badly designed then they will be forced to make a Product Recall.
You mentioned about the HD5970, yet I don't recall AMD making any Product Recalls to either replace the GPU card or to modify it. Did AMD ever admit that the GPU card was badly designed?
I hope you guys understand my logic behind my argument. Unless there some legal thing I am not aware of that you both might be.
Sadly that's a pretty common occurrence lately, design defects that hinder performance, probably the most widely known currently is the fact that 5G phones overheat so badly they drop to 4G, and don't work in ambient temperatures above 85F. But of course through the years there's a host of badly designed laptops which overheat and throttle (even the new expensive Surface Pro laptops), CPUs which ship with a cooler barely sufficient and cause it to throttle, GPUs which require manually setting the fans very high to keep them from throttling...Sadly bad/cheap engineering does not mean a product is defective.
In this case, however, it seems as if ASUS is citing AMD as the reason for their badly designed cards, as this seems to be their copy paste response to negative Newegg reviews about this very card. Also in their response they ask about which fan curve they are using, which goes to another well known issue that effective coolers for high wattage cards are loud as a heck unless they're properly designed and made with proper conductive materials, so manufacturers have been spinning the fans slowly until temps rise to near or at throttling speeds.
But it is one of the least expensive models, and there's a tradeoff that comes with cheapness, and that's usually undesired operation.
And on higher end products, those problems are reduced to non existent.
The REAL problem, however, is the fact that when someone buys one of these cards, which may be more expensive than the nVidia counterpart, and has a very negative experience, there's a high degree of likelihood that they will choose the competitor product the next time, or even return it and get the competitor product, in which case AMD lost a customer though the fault of a board partner, and it should be AMD's standard procedure to require all board partners to send them products to certify it meets certain standards and does not cause negative experience, press, and revenue loss due to faulty or cheap engineering.
Yes, if a manufacturer puts out a badly designed product and the majority of reviews are negative this will hurt the company in the long run. Since present and future buyers will equate the company with bad products.
I know many Users have said they will never purchase a Asus or MSI or Gigabyte etc. products because of the negative experience they had.
This is probably the main reason why manufacturers will never admit they put out a badly designed product. I imagine the last thing a manufacturer wants is to admit they put out a badly designed product.
I am in full agreement. That would be the ideal solution.
But as always the Manufacturer looks at the bottom line when making that type of decision - will it hurt our profits.
NOTE: Personally I would be very upset if I purchased an expensive computer hardware and later on found out it was proven to be badly designed and I can't get a full refund or compatible replacement. To me, they just lost a future customer.
I was looking at purchasing one of these AIB Vega 64 last over past few years, as there is no good choice for me w.r.t. AIB Vega 64's it seemed the least worst one, from my point of view:
ROG-STRIX-RXVEGA64-O8G-GAMING | Graphics Cards | ASUS USA
I contacted ASUS about the VRM Thermal Pad issue, and told them I just wanted to buy a working GPU and not have to fix it.
I pointed them to the above video detailing the problem.
I asked if the problem had been fixed.
I asked how could I identify a fixed card versus cards that had been released with the problem.
The response from ASUS Support was useless, and from memory the summary was:
"Buy it, and if a problem RMA and we will fix it. But do not attempt to fix it yourself as you will void the warranty"
After that I just will not buy ASUS AMD GPU.
One of my best GPU owner experience is with an Nvidia Asus GTX 780TI Direct CU II OC card.
I will not consider ASUS AMD GPUs at all any more as they seem to have problems or be overpriced compared to alternatives.
If it's Vega 64 16GB then you are likely looking for a Vega Frontier Edition GPU.
In that case the watercooled version is the version to go for but question is why?
Why not just go for an RX Vega 64 Liquid if you are after one of those Frontier Edition GPUs?
They can be bought for ~ 270-300 each.
They run Radeon Pro Enterprise Drivers.
They perform much better in compute than the RX5700XT.
They are ~ as fast in gaming if the HBM2 is run at ~ 1120MHz (DX12) or up to 1150MHz (DX11) from what I have seen so far.
They do use lots of power though. BFV 4K Ultra (no DX11 Crossfire or DX12 MultiGPU) a single RX Vega 64 Liquid reports ~ 360-380 Watts of power.
A pair of them running DX12 MultiGPU at 4K Ultra they are pulling about 270-300 Watts each.
Good cards to run in the Winter.
Then there is always the Radeon VII ... which are starting to sell for ~ 400.
Correction - A pair of them running BF1 DX12 MultiGPU at 4K Ultra they are pulling about 270-300 Watts each.
now you see why I use the HX1000i so I can indulge in a pair of power pigs and have no fear
Yes it is bad if a board partner card puts people off purchasing.
However I would not go near an RX5700XT reference or especially an Anniversary Edition either for the same reason - temps are too high / cooler not enough.
I think these RX5700XT's are running at GPU and VRAM clock speeds that are on the edge of power efficiency, even on 7nm, in an attempt to beat Nvidia cards.
Yes I saw that one as well. No VRAM Cooling.
Asus also did a bad job of cooling the ROG Strix Vega 64 thermal pads according to other reviews and user feedback.
According to this reply in General Discussion concerning the article I previously posted. Asus is using a Carbon Sheet instead of Thermal Paste which is why the vRAM doesn't overheat and get damaged: https://community.amd.com/thread/247082
There is no cooling at all on the VRAM of that Asus TUF RX5700.
See : ASUS RX 5700 TUF Gaming X3 Review - NO memory cooling?! - YouTube
No thermal paste.
No carbon cheet.
The VRAM Chips are sitting on the PCB and rely on air blowing down on them, from somewhere.
If you fit your own watercooler to a GPU, you are normally supplied with a set of small thermal pads and heatsinks for the VRAM.
There is a good reason for that.
High speed VRAM Chips generate their own heat, and you need to get rid of it.
according to the Youtube you linked, The reviewer mentions the sole cooling comes from the Air flow from the three fans on the GPU card.
From the Youtube link you posted it shows a heatsink for the vRAM:
So the Asus card does have a heatsink. may not be the most efficient but with the 3 fans blowing air looks like it is enough from causing the vRAM to overheat.
Still, if you pay an inflated premium for a midrange card, even the cheapest model of such, you shouldn't have to deal with any issues such as these. Even temperatures which aren't actually damaging can result in lower than expected performance due to throttling compared to competitor models.
I'd suggest returning it as well if they don't charge a "restocking fee" and you don't have to pay return shipping. If either apply, then you should just go ahead and use it.
I agree when you mentioned about returning the GPU card and getting a refund.
That is not the same thing Colesdav was suggesting. Colesdav was advising to RMA the GPU card and demanding that Asus replace it with a different model due to the bad reviews. That will be wasting the OP and Asus time. Because Asus isn't going to replace a GPU card that is working according to their design and because the User is saying Reviews is saying that it is a badly designed GPU card and demands that Asus replace it with a different model.
This is what Colesdav wrote originally:
"I would recommend you just RMA the card and refuse the same one as a replacement and point to that review."
(1). Regarding the Asus TUF RX5700XT review.
No that is the point of the review. The heat spreader over the VRAM on that RX5700XT is NOT enough to prevent the VRAM Modules from overheating. They are hitting 104'C whilst the Spec for the VRAM on the PCB is max operating temp of 95'C.
I believe the whole AMD saying GPU Tjmax = 110'C is o.k. refers to the AMD GPU, not the VRAM on the PCB.
Both the Radeon VII and the RX5700XT introduced more thermal sensors on the GPU dies and those cards report more hotspot temps, not just an average GPU die edge temp which is used on older AMD GPUs.
(2). Regarding the new temperature sensors. It is detailed here:
AMD Radeon RX 5700 XT Review: Thermals, Noise, Gaming, & Broken Drivers - YouTube
The testing was done on an open air bench PC with Ambient temp of ~ 20-21'C.
Put that GPU in a PC case and you will easily get higher temps.in the PC Case.
(2). Regarding the ASUS TUF RX5700 review.
The issue is still the VRAM Cooling.
Those results were taken using an open air test bench again.
I have not determined what Ambient Temperature was for their testing.
Even if I make an assumption that it was 20-21'C, on an open air test bench, and assuming they report absolute temps:
There is only 95'-84' = 11'C of thermal headroom below maximum operating temperature specification for the VRAM on the PCB in an open air test bench at an assumed 20-21'C (I will try again to find out what Ambient temps were but that's normally what is used).
So what happens when you put that GPU in a PC Case where Ambient is ~ 10'C higher?
You will be operating on the edge of or beyond the VRAM operating spec.
I do not think 11'C of thermal headroom is enough.
"You have heard of that adage "Buyer's beware". Well that applies here. If the buyer didn't do any research and bought the GPU card and later on finds out the buyer purchased a subpar GPU card, in respect to other similar GPU cards, that isn't the manufacturer's fault but rather the Buyer fault. As long as the Manufacturer didn't advertised the GPU card fraudulently."
That sounds like an excuse to let manufacturers get away with releasing a bad product.
The GPU is supposed to work and be cooled properly.
If the GPU is missing thermal pads on VRM then something is wrong in the design of the GPU and it should be RMA'ed.
Yes, it also the Manufacturer to decide if the bad product should be recalled or not.
Before I purchased my Asus GPU card, I did a lot of research comparing various different Manufacturer's GPU cards within my price range. I read many reviews from tech sites and from Retail stores Customers reviews. Finally I settled on Asus GPU due to good reviews from both Tech sites and Customers Reviews.
Now if I would have bought the Asus RX5700 with hardly checking any tech or Customer's reviews. Who fault is that? The manufacturer or the Buyer?
Unless there is enough bad publicity or it becomes a safety issue or Users start a Class Action suit against the Manufacturer than the manufacturer might be forced to do a Product Recall under RMA terms.
But if the GPU is working even though it is running hot, but it still stays under the 110C AMD Specs for the GPU card then there is nothing the User can complain about. Plus when it reaches 110C it starts to throttle. Unless the badly designed GPU card is a safety issue like it will catch on fire due to extreme overheating or some other safety problems then Asus would be immediately forced to make a Product Recall since the Government will get involved.
Also since you said the vRAM has no thermal padding, then Asus will expect a huge amount of RMAs from Users. Since the GPU vRAM will become defective from excessive heat in a short time of use. Obviously that hasn't happened as far as I am aware. Seems like the system Asus designed is just enough to prevent the vRAM from being damaged from overheating.
I am not excusing Asus for putting out a badly designed GPU card. What I am saying you can't tell someone to RMA a GPU Card because of bad reviews and expect Asus or any manufacturer to replace it with a different model under Warranty when the GPU card is working as designed. The User can inquire to see if the Manufacturer will replace it with a different model or modify it by mentioning the bad reviews before deciding whether to try and RMA it under Warranty.
Like I mentioned in my previous reply. I am just playing the Devil's Advocate in seeing Asus reaction to that type of RMA when the card was turned in by the User only because the User saw some bad reviews that mentions the GPU card is badly designed and wants it replaced with a different GPU model under Warranty RMA Terms.
Guess I have been lucky, my video cards all work fine but I have had to refurbish a few.
Video cards should not be as hot as hell aside the fact that I play Doom.
This is the evidence you need to send your GPU back, assuming I have the correct model and you can repeat these results:
Just click on this link for more explanation of VRAM temperatures in the review: Worst 5700 XT Period, Asus TUF Gaming X3 RX 5700 XT OC, DON'T BUY! - YouTube
The ASUS TUF Gaming X3 RX 5700 XT OC in that review is showing a GDDR6 Temperature of 104'C which is outside the maximum Operating Spec of the VRAM memory.
If you can repeat that test or show VRAM temps above 95'C in gaming or other tests then the card has a problem.
From the Creators of the YouTube that Black Zion linked: https://www.kitguru.net/components/graphic-cards/dominic-moass/asus-rx-5700-tuf-gaming-x3-oc-edition...
For our temperature testing, we measure the peak GPU core temperature under load, as well as the GPU temperature with the card idling on the desktop. A reading under load comes from running the 3DMark Fire Strike Ultra stress test for 30 minutes. An idle reading comes after leaving the system on the Windows desktop for 30 minutes.
Looking closer, however, we can see that all is not well. Above we have five RX 5700/5700 XT cards on their own chart, with those cards also including built-in memory temperature sensors where AMD reports the maximum temperature of the hottest memory module through Wattman (and GPU-Z).
As you can see, the TUF Gaming X3’s memory peaked a whole 10C hotter than the reference RX 5700. We mentioned at the start of this review how the lack of any kind of VRAM cooling was surprising, and it certainly shows based on this testing.
It’s not just that it runs hotter under load, but as we will see on the next page, the card isn’t as quiet as it could be due to the fact that the VRAM relies entirely on the fans’ airflow for cooling. That also means the fans don’t have a zero rpm mode, as otherwise there’d simply be no cooling for the memory chips.
To be clear, edge and junction temperatures are great, the heatsink and its 4 heatpipes are more than sufficient to cool the GPU core – it’s the memory which has been entirely neglected here. Do bear in mind though, that GPU temperatures will always be higher if there is no VRAM cooling, as the heatsink isn’t having to deal with the heat from the memory as well, so we would expect better GPU-only temperatures considering the lack of VRAM cooling.
Obviously if you stress test the Memory is going to get hotter than during normal computer use such as playing games such as the results your posted from Crystal.
According to Kitguru, the Asus GPU card has one of the best thermal performance equivalent to the Red Devil RX5700.
NOTE: The above Kitguru Review is a very detail review of the Asus RX5700.
This whole thread is going off on a tangent. So I won't be replying anymore.
I believe the OP original post has been answered and the OP has plenty of information on whatever he decides to do in the future.
The OP has the RX5700XT version correct?
That is the one that has the heat spreader and VRAM temps of 104'C which is above maximum operating spec of 95'C.
So if Asus have supplied a GPU that runs with VRAM temps of 104'C and the the OP can repeat those temps that is enough to prove there is something wrong.
Those thermal tests were run at 21'C room temp.
I run most of my PC Builds with the side panel removed now, in an attempt to keep the internal of the case as cool as possible.
An RX Vega 64 Liquid or R9 Fury X Watercooler Radiator is best kept outside of the PC case.
Unfortunately PC case manufacturers do not make it easy to allow the radiator & hoses outside of the case.
So unless you want to take a metal cutter to a PC case (I don't) then removing the side panel and placing the radiators outside is the only option.
RX Vega 64 Liquid using 360 Watts with the radiator inside the PC Case really does push up the ambient temps inside the case.
I have machines running 6 R9 Fury X for Blender. When I run one of those I can turn off the central heating.