110°C Hotspot Temps "Expected and Within Spec", AMD on RX 5700-Series Thermals

[killster1]

Common sense and physics. Use your brain.

A device has a max rated limit. This is the max it can take before IMMEDIATE damage occurs. Long term damage does not play by the same rule. Whenever you are dealing with a physical product, you NEVER push it to 100% limit constantly and expect it to last. This applies to air conditioners, jacks, trucks, computers, tables, fans, anything you use on a daily basis. Like I said, my car can do 155 MPH. But if I were to push it that fast constantly, every day, the car wouldnt last very long before experiencing mechanical issues, because it isnt designed to SUSTAIN that speed.

Every time the GPU heats up and cools down, the solder connectors experience expansion and contraction. Over time, this can result in the solder connections cracking internally, resulting in a card that does not work properly. The greater the temperature variance, the faster this occurs. This is why many GPUs now shut the fans off under 50C, because cooling it all the way down to 30C increases the variance the GPU experiences.

What AMD is doing here is allowing the GPU to run at max tjunct temp for extended periods of time and calling this acceptable. Given the GPU also THROTTLES at this temp, AMD is admitting it designed a GPU that cant run at full speed during typical gaming workloads. Given AMD also releases GPUs that can be tweaked to both run faster and consume less voltage rather reliably, it would seem a LOT of us know better then RTG engineers.

Would you care to explain how AMD's silicon is magically no longer affected by physical expansion and contraction from temperatures? I'd love to hear about this new technology.

really what damage to your car would happen at 155mph daily? Do you perhaps have 3 gears? small motor struggling to get to 155? id say letting a car sit idle would be more damage then most cars at 155


getting from 0 to 100mph is where you’re going to be doing the most ‘damage’ - if you do it in a quarter mile, you’re really stressing the car, but if you take 20 miles to get to that speed, your wear and tear is much less, due to less torque. Once you get to that speed, it doesn’t much matter if you’re driving a muscle car or a Prius, as long as the overdrive gear is set up to sip fuel (or pull juice from the battery) just enough to overcome 100mph drag.

 

[efikkan]

I don't care what excuses they come up with, any sustained temperatures in the range of 100-110°C can't be good for long term reliability of the product. And this goes for any brand, not just AMD.

We have to remember that most reviews are conducted on open test benches or in open cases, while all customers will run these in closed cases, and even the best of us will not keep it completely dust free. That's why it's important that any product have some thermal headroom when reviewed under ideal circumstances, since real world conditions will always be slightly worse.

 

[Vayra86]

Actually, talk was about thermal design and horrors that nvidia GPU owners feel, for some reason, for 5700 XT ref GPU owners.

Now that we've covered that, NV's 2070s (I didn't check others) AIBs aren't great OCers either, diff between Ref and AIB performance is also similar between brands.

You're right, Turing clocks right to the moon out of the box as well, but still gets an extra 3-6% across the whole line - its minor, but its there, and it says something about how the card is balanced at stock. The actual 'OC' on Nvidia cards is very liberal, because boost also always punches above the specced number. And I have to say, the AIB Navi's so far look pretty good too, on a similar level even - sans the OC headroom - AMD squeezed out every last drop at the expense of a bit of efficiency, and it shows. Was that worth it? I don't know.

The problem here is AMD once again managed to release ref designs that visibly suck, and its not good for their brand image, it does not show dedication to their GPUs much like Nvidia's releases are managed. The absence of AIB cards at launch makes that problem a bit more painful. And its not a first - it goes on, and on. In the meantime, we are looking at a 400 dollar card. Its not strange to expect a bit more.

Oh and by the way, I said similar stuff about the Nvidia Founders when Pascal launched, but the difference there was that Pascal and GPU Boost operated at much lower temps. And even thén the FE's still limited performance a bit.

 

[jmcosta]

You're right, Turing clocks right to the moon out of the box as well, but still gets an extra 3-6% across the whole line - its minor, but its there, and it says something about how the card is balanced at stock. The actual 'OC' on Nvidia cards is very liberal, because boost also always punches above the specced number. And I have to say, the AIB Navi's so far look pretty good too, on a similar level even - sans the OC headroom - AMD squeezed out every last drop at the expense of a bit of efficiency, and it shows. Was that worth it? I don't know.

The Turing chips have a small OC headroom but the performance gain is almost 10%, which is the opposite of what you see with Navi

 

[Sithaer]

Honestly I don't care about this 'issue' and I don't belive it for a second that Nvidia or Intel doesn't have the same stuff going on anyway.

In the past ~10+ years I only had 2 cards die on me and both were Nvidia cards so theres that.

Don't care about ref/blower cards either,whoever buys those should know what they are buying instead of waiting some time to get 'proper' models.

I'm planning to buy a 5700 but I'm not in a hurry,I can easily wait till all of the decent models are out and then buy one of them 'Nitro/pulse/giga G1 probably'.

 

[Aerpoweron]

Every Vega has the T-juction temp sensors. GPU-Z showed them, which confused a lot of people. So they choose for one GPU-Z version not to show it by default. But you could activate that.

And don't forget there is still the usual GPU-Temperature. Which is showing the temperatures we are used to.

We could only compare Nvidia and AMD cards, when we have a sensor with with a lot of temp zones, which we could put between the GPU-Die and the cooler. We could see the hot spots no matter who built the card.

 

[Th3pwn3r]

Ultimately what I and others should care about is power consumption to performance ratio BUT with AMD being lower nm than Nvidia I hoped and expected much better results. Oh well.

 

[Zubasa]

First of all: is this your intuition or are there some publications to support this hypothesis?

Second: you seem a bit confused. The passive cooling does not increase the number of times the fan starts. The fan is not switching on and off during gaming.
If the game applies a lot of load, the fan will be on during the whole session. Otherwise the fan is off.
So the number of starts and stops is roughly the same. It's just that your fan starts during boot and mine during game launch. So I don't have to listen to it when I'm not gaming (90% of the time).

In fact it actually decreases the number of starts for those of us who don't play games every day.

First of all, motors draw the maximum amount of current when they start, this heats up the wire winding in the motor.
Extra start and stops means extra thermal cycles for the wires. This is similar to the concern that other members have raised about the solder joins of the GPU.
The there is the wear on the bearings depending on type. Rifled Bearing and Fluid Dynamic Bearings require a certain speed to get the lubricant flowing.
This means at start up there are parts of the bearing with very little lubrication which cause extra wear on the bearing than otherwise.

Now because the the fan blades are rather light loads, the motor gets up to speed quickly and the effects are minimal.
Therefore I said it is only slightly detrimental to fan life span.
Shutting the fans off at idle is for noise reasons and nothing else, that is exactly what I said in my post thanks for repeating my point.

No not a fact, it certainly doesn't decrease the number of starts, the GPU fans will spin up at least once on boot.
Also depending on the design of the card, some GPUs will start the fans on video play back due to the the gpu heating up under load for hardware acceleration.
So the best case scenario is it is the same number of start cycles.

 

[R0H1T]

You really need to clarify whether you actually have a point or just want to keep this slowchat going with utter bullshit. The numbers speak for themselves, what are you really arguing against? That AMD is a sad puppy not getting enough love?

So that's your argument huh? What's your data point for 110°c "hotspot" temperature being (always) bad given we have absolute no reference, especially at 7nm nor do we know if 110°c is sustained for n length of time? Do you for instance have data about hotspots on 9900k @5GHz or above & how about this ~ Temperature Spikes Reported on Intel's Core i7-7700, i7-7700K Processors

So you come with absolutely no data, lots of assumptions & then ignoring historical trends & call whatever I'm saying as utter BS, great

Could AMD have done a better job with the cooling ~ sure, do we know that the current solution will fail medium - long term? You have absolutely 0 basis to claim that, unless you know more than us about this "issue" or any other on similar products from the competitors.

 

[Bytales]

It is hard to compare to the competition, because nVidia GPUs do not have a TJunction sensor at all.
Without knowing where the Temp sensor on nVidia GPUs is located, there really is no valid comparison.
The edge temp on AMD gpus aka "GPU" read out is much closer to what you typically expects.

Edit: It is not a single TJunction sensor, the TJunction / Hotspot read out is just the highest reading out of many different sensors spread across the die.
In the case of Radeon VII there are 64 of them. It is not necessary the the same area of the GPU die that is getting hot all the time.

Yeah, People never bother to read the whole text, they only see, AMD 110 Degrees. Then start complaining.
You moronic faqs, why bother to waste energy commenting at all ! You are in no measure to understand squat, so please go do something else with your life, instead of spamming us on the Forums here !

 

[Anymal]

Almost no OC headroom since 7970 Ghz ed. OC dream Watercooled Fury X? No OC AIB cards. 7nm brand new Navi and here we are, underwhealming already OCed straight from production lines. RDNA 2.0 to the rescue in 2020.

 

[medi01]

You're right, Turing clocks right to the moon out of the box as well, but still gets an extra 3-6% across the whole line - its minor, but its there, and it says something about how the card is balanced at stock.

When I checked it was 4% for ASUS and 3% for MSI.

The actual 'OC' on Nvidia cards is very liberal, because boost also always punches above the specced number.

Boost doesn't count as OC in my books. It's part of the standard package, and wizard keeps explicitly stating the clock range in recent reviews.
The performance we get at the end of the day, includes that boost.
You can't count it as something being added on top.

The problem here is AMD once again managed to release ref designs that visibly suck, and its not good for their brand image, it does not show dedication to their GPUs much like Nvidia's releases are managed.

More of a PR, nothing practical. We don't even know what "spot" temps of NV are.

The absence of AIB cards at launch makes that problem a bit more painful. And its not a first - it goes on, and on. In the meantime, we are looking at a 400 dollar card. Its not strange to expect a bit more.

That's simply caused by playing catch-up game.
And, frankly, I'd rather learn what's coming 1-2 month in advance, rather than wait for Ref and AIB cards to hit together. (I don't even get what ref cards are for, other than that)

Oh and by the way, I said similar stuff about the Nvidia Founders when Pascal launched, but the difference there was that Pascal and GPU Boost operated at much lower temps. And even thén the FE's still limited performance a bit.

Ok, let me re-state this again:
1) AMD used a blower type (stating that is the only way they can guarantee the thermals)
2) Very small perf diff between AIB and Ref proves that even ref 5700 XT is not doing excessive throttling, despite being 20+ degrees hotter.
3) "Spot temperature" is just a number, that makes sense only in pair with ref cards (who buys them) and even there, is not causing practical problems, although I admit that @efikkan has a point and it might have bad impact on card's longevity, still, "ref card, who cares"

In short: possibly bad impact on card longevity, but we are not sure. Definitely not having serious performance impact. We don't even know what values are for NV, as there is no exposed sensor.

 

[BorgOvermind]

That's exactly what was said about the 1070 GPU, which indeed could exceed 100C, but that in turn for notebooks overheated the CPU too much, so maintenance was due anyway.

Radeons have always ran hot, but this is ludicrous.

And nVs didn't ?
lol

 

[Frick]

I don't care what excuses they come up with, any sustained temperatures in the range of 100-110°C can't be good for long term reliability of the product. And this goes for any brand, not just AMD.

We have to remember that most reviews are conducted on open test benches or in open cases, while all customers will run these in closed cases, and even the best of us will not keep it completely dust free. That's why it's important that any product have some thermal headroom when reviewed under ideal circumstances, since real world conditions will always be slightly worse.

The thing is we don't know that. If these cards start to drop dead in a year or so an it is confirmed to be temperature related (as opposed to sloppy manufacturing), then we know. Until then it's more or less qualified guesswork. And modern chips rarely sustain any load, the voltage regulation and boost thingies are way too sophisticated for that.

 

[B-Real]

The GPU is not the only thing using power...

Galax GeForce RTX 2060 Super EX Review

The Galax GeForce RTX 2060 Super EX retails for the same price as the NVIDIA Founders Edition, yet comes overclocked out of the box, and features a much better cooler. The card is actually the coolest RTX 2060 Super we tested so far, and it includes idle fan stop, too.

www.techpowerup.com

5700 XT uses more power than 2070 Super in gaming on average, while performing worse. 5700 XT is slower, hotter and louder.

The RTX 2060 uses more power than an RX 5700 in gaming on average while performing worse. So what did you want to say?

 

[notb]

The thing is we don't know that. If these cards start to drop dead in a year or so an it is confirmed to be temperature related (as opposed to sloppy manufacturing), then we know. Until then it's more or less qualified guesswork. And modern chips rarely sustain any load, the voltage regulation and boost thingies are way too sophisticated for that.

Well. Many people on this forum are convinced that high temperatures are killing Intel CPUs. Do you want to tell them that AMD GPUs are magically resistant to 100*C? :-D

 

[Xuper]

Read this :

Why 110-Degree Temps Are Normal for AMD's Radeon 5700, 5700 XT

AMD's new 5700 XT and 5700 can hit higher temps than some people are comfortable with, but the company changed how and where it measures temperatures with Navi.

www.extremetech.com

This hotspot 110'c was there long before Navi/Radeon VII but people couldn't understand thing.here one of people said very clear :

Under the old way of measuring things, AMD had one value to work with. It established a significant guard band around its measurements and left headroom in the card design to avoid running too close to the proverbial ragged edge.

Using this new method, AMD is able to calibrate its cards differently. They don't need to leave as much margin on the table, because they have a much more accurate method of monitoring temperature. The GPU automatically adjusts its own voltage and frequencies depending on the specific characteristics of each individual GPU rather than preprogrammed settings chosen by AMD at the factory.

It is also possible that pre-Navi AMD GPUs hit temperatures above 95C in other places but that this is not reported to the end-user because there's only one sensor on the die. AMD did not say if this was the case or not. All it said is that they measured in one place and based their temperature and frequency adjustments on this single measurement as opposed to using a group of measurements.

I hope one day Intel/Nvidia/AMD follows path and allow us to see temp of all array of sensors.

Views of Pluto Through the Years - NASA

This animation combines various observations of Pluto over the course of several decades.

www.nasa.gov

this is best example for those who don't understand.
From 1995 to 2015.It took them 20 years to get a SHARP image of Pluto Planet.

 

[Aerpoweron]

Boost doesn't count as OC in my books. It's part of the standard package, and wizard keeps explicitly stating the clock range in recent reviews.
The performance we get at the end of the day, includes that boost.
You can't count it as something being added on top.

I think the differences between Boost and Overclock have started blending together for some time now. And AMD with it's Junction temp and the Precision Boost can get close to a good Overclock with it's Boost features.
I find it amazing that they have 32 sensors on the Vega and 64 on the Radeon VII. I wonder how many the Navi GPUs have. And if we get a tool to see the temps in a colored 2d-texture we could see where and when what part of the GPU is utilized.
And think about the huge GPU dies of the Nvidia RTX cards. They likely have a lot of headroom with a junction temp optimization.

People complaining about the "not so good" 7nm GPUs. It is a new process, and it will take some time to get the best out of that manufacturing node. And we will see how good Nvidia's architecture is scaling on 7nm when it will be released

 

[Frick]

Well. Many people on this forum are convinced that high temperatures are killing Intel CPUs. Do you want to tell them that AMD GPUs are magically resistant to 100*C? :-D

I've never seen that claim. And yes, if that is what the specs says.

 

[ssdpro]

110 seems pretty hot. Not so hot it dies within warranty though.

 

[ZoneDymo]

rubbing eyes

so how many ppl are still running 7970s/r9 2xx cards around here,which are 6-8 years old.

My sisters are still running my old HD6950's sooo yeah, oh and a friend uses an HD7950 still today purely because the prices are ridiculous so upgrading does not make sense.

 

[Vayra86]

So that's your argument huh? What's your data point for 110°c "hotspot" temperature being (always) bad given we have absolute no reference, especially at 7nm nor do we know if 110°c is sustained for n length of time? Do you for instance have data about hotspots on 9900k @5GHz or above & how about this ~ Temperature Spikes Reported on Intel's Core i7-7700, i7-7700K Processors

So you come with absolutely no data, lots of assumptions & then ignoring historical trends & call whatever I'm saying as utter BS, great

Could AMD have done a better job with the cooling ~ sure, do we know that the current solution will fail medium - long term? You have absolutely 0 basis to claim that, unless you know more than us about this "issue" or any other on similar products from the competitors.

You oughta scroll back a bit, I covered this at length - Memory ICs reach 100C for example, which is definitely not where you want them. That heat affects other components and none of this helps chip longevity. The writing is on the wall. To each his own what he thinks of that, but its not looking comfy to me.

By the way, your 7700K link kinda underlines that we know about the 'hot spots' on Intel processors, otherwise you wouldn't have that reading. But these Navi temps are not 'spikes'. They are sustained.

We can keep going in circles about this but the idea that Nvidia ref also hits these temps is the same guesswork; but we do have much better temp readings from all other sensors on a ref FE board - including memory ICs. And note: FE's throttle too but I've seen GPU Boost in action and it does the job a whole lot better; as in, it will rigorously manage voltages and temps instead of 'pushing for the limit' like we see on these Navi boards. This is further underlined by the OC headroom the cards still have. There are more than enough 'data points' available...

Besides, nothing is really new here - AMD's ref cards have always been complete junk.

"ref card, who cares"

In short: possibly bad impact on card longevity, but we are not sure. Definitely not having serious performance impact. We don't even know what values are for NV, as there is no exposed sensor.

We are never sure until after the fact. I'll word it differently. The current state of affairs does not instill confidence. And no, I don't 'care' about ref cards either, but I pointed that out earlier; AMD should, especially when AIB cards are late to the party. These kinds of events kill their momentum for any GPU launch, and it keeps repeating itself.

 

[John Naylor]

and its 100 - 150 dollars cheaper.... so why are you comparing the two?
If anything you should compare it to the RTX2060 Super (like in your link...was the 2070 a typo?) and then the 5700XT is overall the better option.

To my eyes the 5700 XT should be compared with the 2070 .... the 5700 w/ the 2060... no supers.
With both cards overclocked, the AIB MSI 2070 Gaming Z (Not super) is still about 5% faster than the MSI Evoke 5700 XT ... so if price difference is deemed big enough (-$50) I can see the attraction ... but the 5700XT being 2.5 times as loud is a deal breaker at any price. The Sappire is slower still but it's significantly quieter

MSI 2070 = 30 dbA
MSI 5700 XT = 43 dba .. 13 dba = 2.46 times as loud

https://tpucdn.com/review/msi-radeon-rx-5700-xt-evoke/images/relative-performance_2560-1440.png

MSI 5700 XT = 100%
Reference 2070 = 96%

https://tpucdn.com/review/msi-radeon-rx-5700-xt-evoke/images/overclocked-performance.png

MSI Evoke Gain from OC = 100% x (119.6 / 115.1) = 103.9

https://tpucdn.com/review/msi-geforce-rtx-2070-gaming-z/images/overclocked-performance.png

MSI 2070 Gain from overclocking = 96% x (144.5 / 128.3) = 108.1

108.1 / 103.9 = + 4.85%

The Gaming Z is a $460, the Evoke suggested at $430.... will likely be higher for the 1st few months.

If we ask, "Is a 5% increase in performance worth a 7% increase in price ?" It would be to me. But with a $1200 build versus a $1230 build, that's a 5% increase in speed for a 2.5% increase in price, and that's a more appropriate comparison as the whole system is faster and the card don't deliver any fps sitting on your desk. However, the 800 pound gorilla in the room is the 43 dbA 2.5 times as loud thing.

I think the issue here is, from what we have seen so far most of the 5700XT cards are not true AIB cards but more like the EVGA Black series ... pretty much a reference PCB with a AIB cooler. Asus went out and beefed up the VRMS with an 11 / 2 + 1 design versus the 7 / 2 reference . They didn't cool near as well as the MSI 5700 XT or 2070, The did a lot better on the"outta the box" performance but OC headroom was dismal. As the card was so aggressively OC'd in the box, manual OC'ing added just 0.7% performance.


Asus 5700 XT Strix = 100% x (118.3 / 117.4) = 100.77
MSI 2070 Gaming Z = 95% x (144.5 / 128.3) = 107.00

107.00 / 100.77 = + 6.18 %

Interesting tho that Asus went all out, spending money on the PCB redesign when MSI (and Sapphire) looks like they used a cheaper memory controller than the reference card and yet MSI hit 119.6 in the OC test where as Asus only hit 118.3. Still, tho it will surely cost closer to what the premium AIB 2070s costs due to the PCB redesign and tho it's 7C hotter and 6% slower than the MSI 2070 ... it's only 6 dbA louder (performance BIOS). To get lower (+2 dbA), the performance drops and temps go up to 82C.

Tho the Asus is 6% slower and the MSI is 5% slower than the MSI 2070.... if I couldn't get a 2070, and was looking to choose a 5700 XT, it would have to be the Asus. ... but not at $440.

As for the hot spots, I'm kinda betwixt and between ... Yes, I'm inclined to figure that I have neither the background nor experience to validate or invalidate what they are saying .... but in this era ... lying to everybody seems to be common practice. In recent memory we have AMD doing the "it was designed that way" routine when the 6 pin 480's were creating fireworks .... and then they issued a soft fix , followed by a move to 8-pin cards. EVGA said "we designed it that way" when 1/3 of the heat sink missed the GPU on the 970 .... and again, shortly thereafter they issued a redesign. Yet again, when the EVGA 1060s thru 1080s started smoking, the "we designed it that way" mantra was the 1st response and then there was the reacall / do it yaself kit / redesign with thermal pads.

All I can say is "I don't know ... I'm in no position to judge. Ask me again in 6 monts after we get user feedback. But I also old enough to remember AMDhaving fun at nvidias expense frying an egg on the GTX 480 card.

 

[Prima.Vera]

Just admit it AMD. You overclock and overvolt those GPUs like crazy just to gain 5-6% more performance in order to barely compete with nVidia.
How much is the power consumption when those GPUs pass 100°C ??

 

[Zubasa]

Just admit it AMD. You overclock and overvolt those GPUs like crazy just to gain 5-6% more performance in order to barely compete with nVidia.
How much is the power consumption when those GPUs pass 100°C ??

They don't the GPUs average reach high 80s to 90-ish degrees, and the power consumption are in the reviews.
Every modern GPU has a target set in the bios, without any overclock it just reach the power target and stays there.
It is not like the old days where GPUs run themselves into the ground when you run Furmark.

Spoiler