AMD Radeon R9 M295X Core Clock Throttling, Heat, and Performance

[steve62388]

You need to read the article again, slowly.

Here is a great summation:

"Ultimately there is always going to be a longevity cost to increasing temperatures..."

He then goes on to say that warranty and longevity issues are AMD's worry.

You need to read my post again, slowly. I clearly wrote 'I do not know whether running at these temps are healthy in the long term.'. But I also said 'But you can't answer that either, only AMD will know those figures. But they're obviously comfortable enough with failure rates to push forward on this path'.

I actually remember you from the other thread and I will point out now what I said then for the benefit of other readers, you have a dog in this fight. In the previous thread on discussion of the 5k you were recommending people to buy an old Mac Pro and upgrade the video card instead of the 5k. Guess what? You sell those upgrades, so you will have to forgive me for being a little cynical when these comments are coming from someone that stands to gain from such advice. It makes you a salesman not a tech expert. I could be wrong but if I had to place a bet than I would err on the side of the technical expertise of Apple and AMD than a salesman from a tiny company attempting to convince you to buy his product.

But hey, maybe I have made a mistake. Perhaps you have been designing GPUs for years and have a Masters degree in Electrical Engineering (or whatever you need to make these things).

With this information I will allow others to make up their own minds.

Oh, and that stuff about Newton's Laws of Cooling, let me sum it up for you. You're heading up the Tejon Pass, at 5,000 ft you pass an Escalade boiling over, it's radiator water at 260F. Meanwhile in your Subaru, the radiator is only at 200F. Newton's Law of Cooling that is being referenced is saying that the Escalade can give off more heat on a 100F day since it is hotter.

Or, more simply put. A frying pan on the stove heated to red hot can give off more heat than one that is less hot. Even when there is no heat coming from the burner, the hotter pan can give off more heat and thus cool faster. So, do you really want your $3,000 iMac to be like the Escalade/red hot pan and then brag about how easy it is to give off heat? That's not really a good thing.

Hmmm. Car analogies are all too common on the net and they should be banned, most of the time they are stupid.

 

[Astelith]

I don't agree with this. This is a hugely misplaced faith statement. Apple absolutely CAN afford to have the first retina imac burning after 1 or 2 years. Theyve had many many similar huge issues that have required recalls and extensions of warranty etc. You know what else they have? Some of the best PR and media agents available. They also have a massive base of die hard fans, who are happy to have a five minute grumble and then put Apple back onto the pedestal of 'can do no wrong'. A story could probably some out tomorrow where every time Apple sold a retina iMac a puppy was murdered to help make the CPU, and people would still buy anything and everything Apple sells.

I mean an economical "can't afford" not a financial, we all know Apple is one of the richest company in the world but to stay rich they need to avoid or calculate carefully the risks, if the M295X wasn't a reliable GPU why put it in a brand new model ? Also Dell/Alienware is going to use it in the new series of laptop so the logic is telling me all this mess is purely a paranoid of some users, the machine works well, nobody got heat issues or real throttling where the performance is clearly reduced due to the thermal protection

Also, is anyone able to help with my previous question of M290 vs M295x - will the less good model still be fine for gaming with an i7 CPU? As mentioned, my top spec 2012 iMac can handle my games max spec, so I assume the newer M290X will still be able to hand that, even if its not as good as the M295X?

The fact is, 5K resolution is hard to handle in gaming without compromises, the M295X is able to do the job in a fair way but if you play often, going for the M290X it means taking a lot of compromises, read the benchmarks of the games you play and evaluate the numbers

 

[andy9l]

I mean an economical "can't afford" not a financial, we all know Apple is one of the richest company in the world but to stay rich they need to avoid or calculate carefully the risks, if the M295X wasn't a reliable GPU why put it in a brand new model ? Also Dell/Alienware is going to use it in the new series of laptop so the logic is telling me all this mess is purely a paranoid of some users, the machine works well, nobody got heat issues or real throttling where the performance is clearly reduced due to the thermal protection

Apple will not be paying for any GPU replacement programmes. AMD's hardware, AMD's problem. Just like with the 2011 AMD GPUs that are failing and being replaced now. Apple will only potentially lose brand loyalty of the affected customers.

I signed up to an Alienware forum a few weeks ago to ask users of the Alienware 15 to check their heat. The hottest we've had yet is 79°C. Now, I'm not jumping to conclusions, which is why I haven't posted findings yet. We're looking to gather more data.

 

[tmorterlaing]

Is there any chance it's such a new component that the measurements are wrong? Perhaps there's some configuration issue and actually its not reaching 105 degrees?

 

[cynics]

Is there any chance it's such a new component that the measurements are wrong? Perhaps there's some configuration issue and actually its not reaching 105 degrees?

Unlikely, users have recorded throttling around the temps that it should throttle due to excessive heat. Plus the baseline temps when coming on and idle seems accurate.

----------

Apple will not be paying for any GPU replacement programmes. AMD's hardware, AMD's problem. Just like with the 2011 AMD GPUs that are failing and being replaced now. Apple will only potentially lose brand loyalty of the affected customers.

I signed up to an Alienware forum a few weeks ago to ask users of the Alienware 15 to check their heat. The hottest we've had yet is 79°C. Now, I'm not jumping to conclusions, which is why I haven't posted findings yet. We're looking to gather more data.

Is that accurate? Most manufacturers have very strict guidelines on warranty especially involving heat. If I built a PC with AMD components, then didnt do an adequate job keeping them cool that is no longer AMD's problem.

Point in case would be the Alienware system you mentioned. From what you know so far it does a better job cooling the GPU. Btw keep us posted on that, interesting stuff.

 

[tmorterlaing]

You know what? I'm out. For one reason- I don't care what the GPU shouldn't and should not be doing in terms of heat - it runs loud. The old one didn't run loud, playing the same game on the same specs, which means there's NO EXCUSE for my entire game playing experience to be taken over by the fact that the fan is now pretty obviously audible. Whether it's 'loud' is subjective, but compared to the 2012 on here it's MUCH louder doing the exact same thing. This doesn't make sense to me - unless someone can justify that? This super hot GPU thing has been worrying me anyway, so that plus the fan noise, not to mentioned the backlight bleeding would suggest Apple just don't care any more. I'll wait (and hope) until the next one methinks, unless by some massive stroke of luck Apple release a patch in the next few days that sorts this all out. Given their track record, that's highly unlikely.

 

[WilliamG]

Agreed- the rest of the machine does stay relatively cool, even the other GPU sensors stay cool. What's going on with the diode? Is anyone else able to check? Are we CERTAIN the 100+degrees is accurate? Is there some way of physically checking? It also seems odd for it to cool down so quickly....air doesnt conduct heat quite like that usually.

I don't agree with this. This is a hugely misplaced faith statement. Apple absolutely CAN afford to have the first retina imac burning after 1 or 2 years. Theyve had many many similar huge issues that have required recalls and extensions of warranty etc. You know what else they have? Some of the best PR and media agents available. They also have a massive base of die hard fans, who are happy to have a five minute grumble and then put Apple back onto the pedestal of 'can do no wrong'. A story could probably some out tomorrow where every time Apple sold a retina iMac a puppy was murdered to help make the CPU, and people would still buy anything and everything Apple sells.

Also, is anyone able to help with my previous question of M290 vs M295x - will the less good model still be fine for gaming with an i7 CPU? As mentioned, my top spec 2012 iMac can handle my games max spec, so I assume the newer M290X will still be able to hand that, even if its not as good as the M295X?

If you don't need the 5K screen, the M290X won't be a gaming upgrade. It's *probably* a downgrade. The late 2012 680 GPU is very quiet, as you know, and can be overlocked really easily with no overheating issues at any point.

At this point in time, for gaming especially, unless you really love the 5K screen (I can't go back personally...), stick with your 2012/2013.

 

[ioxakoyba]

This discussion got so interesting, so partly because I just want to investigate this heating issue and possible fixes to it myself, I yesterday finally decided to order riMac loaded with 4Ghz i7 CPU, 3 TB Fusion Drive, m295X GPU and 3 years of Apple Care. Plus from the local store I got a free insurance for 6 months covering all kinds of accidents for the product up to 1k euros (for example if an infant throws something on the screen and brokes the glass etc.) ***Edit: I also chosed Fusion drive because I am planning to defuse it to 120 GB SSD and 3 TB regulad HDD (installing OS X on 60 GB chunk of SSD, Windows 10 on another 60 GB chunk of SSD and formatting the remaining 3 TB HDD to exFAT for both system to utilize).

Anyways, in a 3-4 weeks of time I will get the Mac and do some tests and post the results. For me it's not a problem to use 3rd party fan control like Macs Fan control (on the contrary, I have always liked to tinker with my gaming PC's with such gadgets to optimize their cooling, overclocks and so on).

If the claim of Astelith is true and the cooling can keep GPU under 100C (or at least between 99-101 C) with fan on maximum (2700-2800 rpm) I don't see a much problem of fixing this issue with such way by setting more aggressive fan profile to keep the temperature on acceptable levels. In case Astelith is wrong and the temp still rises to 106-108 C, well then, I've got a nice 3,5k euros aluminium table ornament...

But currently I am more conserned about the display's possible uneven light bleed, than the temps of the GPU. Hopefully that will be acceptable on minimum.

In the meantime I got some replies to some quotes:

The fan sensor of GPU-Z is not working with the 5K and in Windows there is an issue preventing the fan spin over 1200rpm without any fan control software, as far as we know you may had this problem with no fan spinning.

Astelith: If this claim is true (and by your screenshots it seems to be), I am happy to admit that I've been wrong saying that "This is a harware problem, not software". On the contrary, it seems that the problem may after all being in the software side, if the fan does not spin up to maximum of 2700-2800 rpm, but instead stays on 2300 rpm when full load.

This also would explain the GPU temps of 106-108C where people also believe that fan is running on maximum (when it is not, as it's just running 2300 rpm where as it should be 2700-2800rpm).

OF course this is also not acceptable. People should not need to do anything about their fan controls. But I am just saying that this throttling problem may be eased or even fixed with 3rd party fan control software.

And my findings? Nonsense. Just fired up some Diablo 3, and the fans eventually found their way to 2300rpm, which APPEARS to be Apple's lock for max fan speed. Perhaps the fan could hit max speed (2700rpm), but it looks like that would probably be a really worst-case scenario if that happened, and it's never happened for me under OS X or Windows 8.1. I'd recognize that sort of fan noise as it's quite loud (when manually set to 2700rpm).

WilliamG: To me it seems that even if the fan spins up to 2300rpm, Apple has not used the full potential of the fan. THE FAN MUST GO TO MAXIMUM OF 2700 RPM before the throttling (and 105C temp is exceeded).

Apple seems to have made this so that they prefer lower noise over performance and lower temps. I do not, so I'll try to fix this with 3rd party fan control.

On my late 2013 iMac, when I'm playing 3D games on Windows 8.1, fans rises to 2300 or 2700 rpm, I can hear them.

Alesc: There's a quite a bit difference does the fan spin 2300 or 2700rpm, so making such assumption simply "by your ears" is not much of a proof of anything. So why don't you install Macs fan control and just test it?

Also, is anyone able to help with my previous question of M290 vs M295x - will the less good model still be fine for gaming with an i7 CPU? As mentioned, my top spec 2012 iMac can handle my games max spec, so I assume the newer M290X will still be able to hand that, even if its not as good as the M295X?

I would say that with such configurations riMac offers, the GPU is always the bottleneck in gaming. So it doesn't much matter for framerates do you choose i5 or i7 for CPU but it matters is there m290x or m295x for GPU.

Even i5 3.5Ghz can easily handle the CPU load for modern games, but even m295X is just "so and so" to do this job.

So it would not even be much of a bad configuration to select riMac with i5 + m295X if the gaming is the main purpose.

 

[ioxakoyba]

Btw, one more question to Astelith:

You claim that riMac owners who has problems with GPU temps going over to 105 C are having possibly some defect in their temp cesor (or possibly some other flaw in their fan controls), as the fan RPM won't go higher than 2300 (or in case of Windows 10 they stay on the default 1200-1300 rpm). But you did had this also, before installing any fan control software, did you?

Were you able to fix this issue simply by installing Macs fan control and leave its default fan settings to "auto".

I am asking this because in my opinion this setting should not change anything. It just says to Macs fan control to do nothing and leave fan control to system at is already is. (It is even stated in the Macs fan control settings, at least in Windows version: "Automatic (Controlled By Computer)").

Or did you experienced (that for some unexplained reason) this fixed the OS X / Windows 8.1 maximum rpm from 2300 to 2700-2800, thus bettering the cooling automatically. So you did not need to do anything but just to install the software?

Or did it just fixed the issue with Windows 10 (that is still officially unsupported by bootcamp) and fixed the fan control from keeping on 1200rpm on all temps, to rise up to 2300 rpm as Apple has set it?

I am still wondering why fan doesn't spin faster than 2300rpm on OS X / Windows 8.1 even the GPU temp rises higher than 105 C and GPU throttling starts.

If this is a fact, then this issue could be fixed by Apple by some kind of softare update that would set fan to spin up to maximum of 2700rpm when required (and this issue could be fixed user by himself right now just by using 3rd party fan controller like Macs fan control).

Also I am wondering the first post of this thread where Fenn firstly claimed that the throttling point starts somewhere 75 C (meaning immediately)? Was this false and proved wrong at some point of this discussion? Because what I have understood the GPU should work just fine on maximum 850Mhz until the 105 C is reached, which after the GPU starts to throttle.

So if fan spinning on 2700 rpm can keep GPU temps between 99-101C (or even somewhere between 90-93 C as Astelith claims), this m295X GPU should not throttle at any point at maximum load, correct?

So using if using Macs fan control this can be archieved, there should be no question is this more powerfull GPU also in practise compared to m290X, 780M or 680MX, correct? (And I don't take opinion on the noise, that's irrelevant for me).

 

[Astelith]

Btw, one more question to Astelith:

You claim that riMac owners who has problems with GPU temps going over to 105 C are having possibly some defect in their temp cesor (or possibly some other flaw in their fan controls), as the fan RPM won't go higher than 2300 (or in case of Windows 10 they stay on the default 1200-1300 rpm). But you did had this also, before installing any fan control software, did you?

Were you able to fix this issue simply by installing Macs fan control and leave its default fan settings to "auto".

I am asking this because in my opinion this setting should not change anything. It just says to Macs fan control to do nothing and leave fan control to system at is already is. (It is even stated in the Macs fan control settings, at least in Windows version: "Automatic (Controlled By Computer)").

Or did you experienced (that for some unexplained reason) this fixed the OS X / Windows 8.1 maximum rpm from 2300 to 2700-2800, thus bettering the cooling automatically. So you did not need to do anything but just to install the software?

Or did it just fixed the issue with Windows 10 (that is still officially unsupported by bootcamp) and fixed the fan control from keeping on 1200rpm on all temps, to rise up to 2300 rpm as Apple has set it?

I am still wondering why fan doesn't spin faster than 2300rpm on OS X / Windows 8.1 even the GPU temp rises higher than 105 C and GPU throttling starts.

If this is a fact, then this issue could be fixed by Apple by some kind of softare update that would set fan to spin up to maximum of 2700rpm when required (and this issue could be fixed user by himself right now just by using 3rd party fan controller like Macs fan control).

Also I am wondering the first post of this thread where Fenn firstly claimed that the throttling point starts somewhere 75 C (meaning immediately)? Was this false and proved wrong at some point of this discussion? Because what I have understood the GPU should work just fine on maximum 850Mhz until the 105 C is reached, which after the GPU starts to throttle.

So if fan spinning on 2700 rpm can keep GPU temps between 99-101C (or even somewhere between 90-93 C as Astelith claims), this m295X GPU should not throttle at any point at maximum load, correct?

So using if using Macs fan control this can be archieved, there should be no question is this more powerfull GPU also in practise compared to m290X, 780M or 680MX, correct? (And I don't take opinion on the noise, that's irrelevant for me).

At the beginning I had some fan issue and probably because when I installed the OS after the unboxing I've restored from an old Time Machine backup of an iMac 2011 with Macs Fan control already on it and maybe it messed up something.
I reinstalled it and everything after was normal.

A quick recap, over all the test I run and posted here I can say that if there are some units going over 100° is because of a software issue (displaying a wrong reading or controlling poorly the fan) not for the real capacity of the hardware, the system can stay cool (below 100°) without any software controller and rarely over 2200rpm, in most cases of games the fan does not exceed even 2000.
If I force the fan at maximum (2700) the temps are below 90°, but doing so I proved that if the system want it can keep everything cool.

I read that one guy almost fixed the 104° C temp issue by resetting the SMC, maybe there's something delicate at sensor level and a bad software installation can make it going crazy

This GPU have an architecture with a variable clock that can increase UP TO 850MHz based on application and TDP (watt consumed) so is perfectly normal to have the clock ranging between 7xx to 850, the throttling (thermal protection) start to enter over 105° and it cut almost in half the frequency.

 

[WilliamG]

At the beginning I had some fan issue and probably because when I installed the OS after the unboxing I've restored from an old Time Machine backup of an iMac 2011 with Macs Fan control already on it and maybe it messed up something.
I reinstalled it and everything after was normal.

A quick recap, over all the test I run and posted here I can say that if there are some units going over 100° is because of a software issue (displaying a wrong reading or controlling poorly the fan) not for the real capacity of the hardware, the system can stay cool (below 100°) without any software controller and rarely over 2200rpm, in most cases of games the fan does not exceed even 2000.
If I force the fan at maximum (2700) the temps are below 90°, but doing so I proved that if the system want it can keep everything cool.

I read that one guy almost fixed the 104° C temp issue by resetting the SMC, maybe there's something delicate at sensor level and a bad software installation can make it going crazy

This GPU have an architecture with a variable clock that can increase UP TO 850MHz based on application and TDP (watt consumed) so is perfectly normal to have the clock ranging between 7xx to 850, the throttling (thermal protection) start to enter over 105° and it cut almost in half the frequency.

I'm sorry but I don't buy it. I've reset SMC, done fresh installs etc. GPU hits105-106 pretty quickly and easily with fan at 2300rpm. I'm just about willing to bet all M295Xs do. My system is not faulty. Because if it is, then everyone's is. I don't believe you can run e.g. Diablo 3 full screen for more than 30 seconds at 5120x2880 (for test purposes) in OS X without the fans hitting 2300rpm and the GPU temp hitting ~105C.

 

[Astelith]

I'm sorry but I don't buy it. I've reset SMC, done fresh installs etc. GPU hits105-106 pretty quickly and easily with fan at 2300rpm. I'm just about willing to bet all M295Xs do. My system is not faulty. Because if it is, then everyone's is. I don't believe you can run e.g. Diablo 3 full screen for more than 30 seconds at 5120x2880 (for test purposes) in OS X without the fans hitting 2300rpm and the GPU temp hitting ~105C.

105° and 2300rpm doesn't make sense, one of the two are wrong, a functional machine will never and never let the GPU go above 100°C (or 101° if you consider some peak) and at 2300 there's other 400 rpm to use to keep it cool, something is not wright in your machine

If you want I can make a video of gaming fullscreen and logging the temps, no problem at all

 

[Astelith]

Diablo 3, 5K, High settings, Fullscreen, Os X

https://www.youtube.com/watch?v=MhO24PCbD84&feature=youtu.be

But considering that in Os X is not that easy to track the temp I made a better test under Win 10 with better tools.

I played the chapter "Separation" in Metro Last Light, 4K, High Details, first with the fan at max (2700) and then in "Auto", so controlled by the system (Apple default), pay attention to the fan noise and look at the frame per second, exactly the same, 93-94° vs 99-100° = no difference of performance (look at 1:17 and 7:50), and the temp is not passing 100°C, even with the GPU constantly at 100%

https://www.youtube.com/watch?v=EWLuCQpt5Kc&feature=youtu.be

This is for sure the most demanding game I tried on this machine but the fan doesn't go over 2300 as you can hear, yes, a bit loud, but hey, we are talking about 4K on High detail and 25+ Fps !
If I lower the resolution and/or the detail I have an average clock higher (800-850) and temps around 95-96° due to the GPU usage mostly below 75% and so the fan at lower rpms.

I can keep testing and making videos but I hardly think nobody here will change their minds.

You (or who with high temps) should do a similar test to compare and understand better this issue...

 

[andy9l]

Thanks for posting this Astelith. This is much more thorough.

So, let me get this right, are you suggesting that any iMac that lets the GPU get to 105C+ and throttle under automatic fan control is faulty?

 

[steve62388]

Thanks for posting this Astelith. This is much more thorough.

So, let me get this right, are you suggesting that any iMac that lets the GPU get to 105C+ and throttle under automatic fan control is faulty?

Erm.... Might be the other way around. The following is all supposition so should be taken with a large pinch of salt but it's something to consider.

Higher fan speed keeps the GPU cooler, that's no surprise and is probably all as it should be. Apple chooses not to use full 2700rpm on the fan because it's noisier, this is an aesthetic decision (by all accounts it maxes out at 2400rpm).

Astelith's GPU is throttling, you can see this in the GPU speed stepping up and down. Throttling in itself is not bad, the GPU is supposed to work this way. The thing is his/her GPU is throttling before it hits 105C. It's throttling early because the GPU is a lower quality binned part.

 

[Astelith]

Thanks for posting this Astelith. This is much more thorough.

So, let me get this right, are you suggesting that any iMac that lets the GPU get to 105C+ and throttle under automatic fan control is faulty?

I don't know, it's not an easy thing to sort out, I show how a healthy machine should work but to go further we need data from the "bad" ones.

----------

Erm.... Might be the other way around. The following is all supposition so should be taken with a large pinch of salt but it's something to consider.

Higher fan speed keeps the GPU cooler, that's no surprise and is probably all as it should be. Apple chooses not to use full 2700rpm on the fan because it's noisier, this is an aesthetic decision (by all accounts it maxes out at 2400rpm).

Astelith's GPU is throttling, you can see this in the GPU speed stepping up and down. Throttling in itself is not bad, the GPU is supposed to work this way. The thing is his/her GPU is throttling before it hits 105C. It's throttling early because the GPU is a lower quality binned part.

If you go some pages backward you can find in one of my post an AMD white paper who explain how the boost works, and I simulate the throttling, it's another story, I can have 100 degrees and 850mhz straight if the application is built to use only some part of the gpu calculation unit (look to a benchmark I did, 100degrees/850sharp for minutes)

 

[WilliamG]

Diablo 3, 5K, High settings, Fullscreen, Os X

https://www.youtube.com/watch?v=MhO24PCbD84&feature=youtu.be

But considering that in Os X is not that easy to track the temp I made a better test under Win 10 with better tools.

I played the chapter "Separation" in Metro Last Light, 4K, High Details, first with the fan at max (2700) and then in "Auto", so controlled by the system (Apple default), pay attention to the fan noise and look at the frame per second, exactly the same, 93-94° vs 99-100° = no difference of performance (look at 1:17 and 7:50), and the temp is not passing 100°C, even with the GPU constantly at 100%

https://www.youtube.com/watch?v=EWLuCQpt5Kc&feature=youtu.be

This is for sure the most demanding game I tried on this machine but the fan doesn't go over 2300 as you can hear, yes, a bit loud, but hey, we are talking about 4K on High detail and 25+ Fps !
If I lower the resolution and/or the detail I have an average clock higher (800-850) and temps around 95-96° due to the GPU usage mostly below 75% and so the fan at lower rpms.

I can keep testing and making videos but I hardly think nobody here will change their minds.

You (or who with high temps) should do a similar test to compare and understand better this issue...

I don't know, it's not an easy thing to sort out, I show how a healthy machine should work but to go further we need data from the "bad" ones.

----------



If you go some pages backward you can find in one of my post an AMD white paper who explain how the boost works, and I simulate the throttling, it's another story, I can have 100 degrees and 850mhz straight if the application is built to use only some part of the gpu calculation unit (look to a benchmark I did, 100degrees/850sharp for minutes)

FYI, your OS X video proves nothing at all.

First, something is wrong with your iMac. Your screen should not go black for that long when command-tabbing to the desktop, so you should probably get your iMac exchanged (only half serious, but in line with your comment about how ALL our iMacs except yours are faulty). But seriously, I can command+tab to the desktop in literally milliseconds from Diablo 3 at 5120x2880, so what's wrong with your iMac?

Secondly, you do realize that during that really long black screen you get when you are command+tabbing - your GPU temp is cooling off considerably. Look at the time stamp 1:03, and you'll see your GPU is at 99C. Now, before the screen goes black for that inordinately long time, your GPU temp is HIGHER than 99C, quite likely 5-7C higher if your fan is running at only 1900rpm, so I think YOU are the one with a defective iMac. Notice how between 1:03 and 1:04 in your video your GPU temp drops from 99C to 92C. That's a 7C drop in under half a second! Do you seriously think your GPU temp isn't above 99C before your desktop eventually shows up?

I'm sorry, but something is wrong with your iMac, and of that I'm pretty darn sure.

Feel free to show us the iStat history of your GPU temp by hovering over the GPU Die in iStat Menus, after Diablo 3 has been running for 10 minutes at 5120x2880.

 

[Astelith]

FYI, your OS X video proves nothing at all.

First, something is wrong with your iMac. Your screen should not go black for that long when command-tabbing to the desktop, so you should probably get your iMac exchanged (only half serious, but in line with your comment about how ALL our iMacs except yours are faulty). But seriously, I can command+tab to the desktop in literally milliseconds from Diablo 3 at 5120x2880, so what's wrong with your iMac?

Secondly, you do realize that during that really long black screen you get when you are command+tabbing - your GPU temp is cooling off considerably. Look at the time stamp 1:03, and you'll see your GPU is at 99C. Now, before the screen goes black for that inordinately long time, your GPU temp is HIGHER than 99C, quite likely 5-7C higher if your fan is running at only 1900rpm, so I think YOU are the one with a defective iMac. Notice how between 1:03 and 1:04 in your video your GPU temp drops from 99C to 92C. That's a 7C drop in under half a second! Do you seriously think your GPU temp isn't above 99C before your desktop eventually shows up?

I'm sorry, but something is wrong with your iMac, and of that I'm pretty darn sure.

Feel free to show us the iStat history of your GPU temp by hovering over the GPU Die in iStat Menus, after Diablo 3 has been running for 10 minutes at 5120x2880.

Seriously ?

Maybe my iMac is in need of a format but after the windows video it seems pretty clear to me how the things are.

I cannot find the history graph for the GPU die, if you tell me how can I use it I'll show you

 

[andy9l]

AMD Radeon R9 M295X Core Clock Throttling, Heat, and Performance

Yeah it's difficult to measure temperatures on OS X. The Windows video is what has my interest.

Your GPU does seem to be cooler than the majority. My old M295X didn't even need to be running a full screen app to smash 105C+ (as shown in my video).

----------

*snip*

Hmm, an interesting thought.

Edit: the hottest I'm yet to see in a Alienware 15 is 88C...

 

[WilliamG]

Seriously ?

Maybe my iMac is in need of a format but after the windows video it seems pretty clear to me how the things are.

I cannot find the history graph for the GPU die, if you tell me how can I use it I'll show you

You just click the GPU die in the dropdown of the temperatures list in iStat Menus. It will pop up a graph showing you the highest temp your GPU has reached.

And yes, seriously. Why would I not be serious in my post?

 

[Astelith]

You just click the GPU die in the dropdown of the temperatures list in iStat Menus. It will pop up a graph showing you the highest temp your GPU has reached.

And yes, seriously. Why would I not be serious in my post?

I don't have it there, is the only one missing inside the dropdown, could you run a test like this and post it here ? So I can make the same and we compare ?

Tonight I'm going to record another video showing how the boost and the real throttling works so everything will be easier to understand.
I will run various stress test that can use different area of the GPU to have max GPU %, max Temp and different Clocks boosts plus the fan settings, I will use the automatic settings, full speed and I will force 1200 rpm to simulate a throttle, with stress test and games.

I only ask everybody that thinks I'm wrong to contribute to the discussion with tests, we can really understand what's going on but we must cooperate.

 

[Astelith]

Thanks for posting this Astelith. This is much more thorough.

So, let me get this right, are you suggesting that any iMac that lets the GPU get to 105C+ and throttle under automatic fan control is faulty?

I read your thread in the Alienware forum and It seems too early to have a good feedback from them yet.
But I watched again your video several times and from the behavior (core clock, temps, fps and the speed it reaches that temp) you are not throttling, 76x to 850 with costant 90fps is perfectly normal, and if you run the game in full screen you will notice a lower clock due to the bigger workload (the GPU lowers the clock to avoid surpassing the TDP watt limit).

So far I think your 106 (with some 107) is the exact same of my 100 (with some 101), at this point I think there's a bug on certain sensors that show more than it really is.
All the sensors are cool except this, maybe even my GPU run cooler than 100°.

A good test for who have this higher temps will be forcing the fan to 1200 and watch if they have the real throttling (loss of 80% of the fps, clock reduced to 6xx) around 110°/112° it means that is the sensor the problem.

Do you agree ?

 

[steve62388]

I only ask everybody that thinks I'm wrong to contribute to the discussion with tests, we can really understand what's going on but we must cooperate.

I'm happy to help but I think a couple of things need to be set down. Any testing needs to be consistent and repeatable.

The O/S and drivers need to be identical. For example I don't have Win10, only Win 8.1. Win10 has a completely different graphics subsystem, I was just reading about it over at Anandtech.

Temperature needs to be measured in app because the switching time from full screen can be enough time to skew results as the GPU cools. A possibility is windowed Furmark or something, what is that Windows app you're using with the graphs on? Furmark normally stresses out a GPU pretty hard.

Any benchmarking needs to be done in an app on a self running loop, actual gameplay will result in inconsistencies because nobody is going to do the exact same stuff.

All tools need to be freely available, virtually nobody is going to buy a game or app just for this. I know I'm not.

 

[Astelith]

I'm happy to help but I think a couple of things need to be set down. Any testing needs to be consistent and repeatable.

The O/S and drivers need to be identical. For example I don't have Win10, only Win 8.1. Win10 has a completely different graphics subsystem, I was just reading about it over at Anandtech.

Temperature needs to be measured in app because the switching time from full screen can be enough time to skew results as the GPU cools. A possibility is windowed Furmark or something, what is that Windows app you're using with the graphs on? Furmark normally stresses out a GPU pretty hard.

Any benchmarking needs to be done in an app on a self running loop, actual gameplay will result in inconsistencies because nobody is going to do the exact same stuff.

All tools need to be freely available, virtually nobody is going to buy a game or app just for this. I know I'm not.

Perfect, at this point we don't have much options, I have a Win 8.1 full license so I will use it to have the same base.
Let's write down the drivers we are going to use, I suggest the last, normal or beta ? But not a driver that need system tweaking, too time consuming.

For the software i suggest Geek3D GPU Test GUI 0.7.0 as it contains all the stress test we need (TessMark, PixMark, FurMark etc.) and they behave very different to each other on the GPU boost clock, TessMark for example can push max temp and max frequency (850) but it uses only a medium part of the GPU so the TDP is below range, FurMark is heavier and the GPU average is 834.

Look here: https://forums.macrumors.com/posts/20584267/
And let's pick one or more of the tests.

To measure with extreme accuracy let's use MSI Afterburner, it's free and it can provide us the graph and the OSD.

There's nothing actual capable of logging the fan speed, we need to make a video with Macs Fan control or iStat menu to see the fan behavior

 

[iMacC2D]

Removed.