It's a well known fact that Apple likes to run their stuff hotter for the sake of quietness/thinness. AFAIK, there's no other laptops or AIO using the 650M/6750M/6770M/6970M that are experiencing failure rates necessitating a class action lawsuit.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
AMD Radeon R9 M295X Core Clock Throttling, Heat, and Performance
- Thread starter Fenn
- Start date
- Sort by reaction score
It's a well known fact that Apple likes to run their stuff hotter for the sake of quietness/thinness. AFAIK, there's no other laptops or AIO using the 650M/6750M/6770M/6970M that are experiencing failure rates necessitating a class action lawsuit.
The AMD cards have a dynamic clock speed based on either 'prediction' of power draw OR thermal throttling. In the case of a benchmark or a consistently demanding game, the desired power draw will be near enough 100% - ALWAYS. Therefore performance and clock speed should be at 100% "in theory". Theory doesn't exist, heat does. The M295X 5K iMac fails to dissipate this heat well enough to avoid thermal throttling - the other piece of the clock speed puzzle.
In the videos we see here and elsewhere online, the FPS of a game or benchmark are DIRECTLY linked to the core clock speed. This, by definition, is the AMD card thermal throttling NOT reducing clock speed based on power draw. If it were reducing based on predicted power, the FPS should not be affected.
The M295X thermal throttling appears to be somewhere between 6-11% in a typical 5K iMac. Again, based on data from this thread.
Read this (from 'AMD GPU power use over time') to better understand: http://semiaccurate.com/2013/12/16/amds-powertune-2-0/
I've lurked on this board for quite a while, ever since the idea of having a Mac came to me about 5-6 months ago. Having used all kinds of OS's over the years (and I've also built a number of machines) it's almost strange actually having a Mac - especially since I would have been one of the people laughing at Apple fanboy's 12 months ago! 
This thread was one of many that I read intently before making my purchasing decision. Having made the error of seeing the Retina screen first when I went to the Apple store for the first time I found it too hard to resist, despite taking another month before making my purchase - and yes, I do have the M295X.
So what I have learnt? I'd do screen-grabs but I'm still learning how to do stuff! But, running my iMac with something that's fairly graphically intensive (but not a stress test) I have this data to add to the discussion.
- GPU temp @ 95-96 degrees: fan sits @ 1200 rpm or so (standard speed). Clearly (despite the concerns that some people have) Apple and AMD feel this is a "normal' temperature, else the fan would kick in.
- GPU temps @ 98-99: fan moves up to around 1500 or so. Personally I'd like the fan to be a little more aggressive, but it's holding the temp stable. Again, working as intended?
- GPU nudges to 100: fan picks up again to 1800 rpm. This does pull the temp down below 100 again (98 seems typical), but the fan does slow down fairly quickly. Again, smarter drivers would perhaps keep the fan speed up a little higher.
I did run the benchmark test that was on another thread - the results were consistent with what other people reported. The GPU temp got to 103. Keeping it running saw it eventually get to 105, but this was after running it continuously for 20 minutes. From my observations the system fan had more headroom to increase cooling should it need to.
Ultimately it's user experience that counts the most. I'm thoroughly impressed with this machine. It does what I want it to do, I'm getting to grips with many of it's nuances and I'm expecting it to continue like that.....and yes, I do have AppleCare
This thread was one of many that I read intently before making my purchasing decision. Having made the error of seeing the Retina screen first when I went to the Apple store for the first time I found it too hard to resist, despite taking another month before making my purchase - and yes, I do have the M295X.
So what I have learnt? I'd do screen-grabs but I'm still learning how to do stuff!
But, running my iMac with something that's fairly graphically intensive (but not a stress test) I have this data to add to the discussion.- GPU temp @ 95-96 degrees: fan sits @ 1200 rpm or so (standard speed). Clearly (despite the concerns that some people have) Apple and AMD feel this is a "normal' temperature, else the fan would kick in.
- GPU temps @ 98-99: fan moves up to around 1500 or so. Personally I'd like the fan to be a little more aggressive, but it's holding the temp stable. Again, working as intended?
- GPU nudges to 100: fan picks up again to 1800 rpm. This does pull the temp down below 100 again (98 seems typical), but the fan does slow down fairly quickly. Again, smarter drivers would perhaps keep the fan speed up a little higher.
I did run the benchmark test that was on another thread - the results were consistent with what other people reported. The GPU temp got to 103. Keeping it running saw it eventually get to 105, but this was after running it continuously for 20 minutes. From my observations the system fan had more headroom to increase cooling should it need to.
Ultimately it's user experience that counts the most. I'm thoroughly impressed with this machine. It does what I want it to do, I'm getting to grips with many of it's nuances and I'm expecting it to continue like that.....and yes, I do have AppleCare
Well, finally my Retina iMac arrived (or something like week ago, but I have been very busy with defusing the 3TB fusion drive, partitioning it with seperate SSD + HDD partitions, installing Windows 10 with bootcamp, and then repartitioning the remaining OS X partition and remaking a chomp of 40 GB size of SSD + 460 GB size of HDD, and refusing it to my own OSX Fusion partition to get the optimal performance (500 Gt fusion drive for OS X, 80 Gt SSD for Windows 10 and remaining HDD for shared files with Ex-fat partition). I have to say "diskutil coreStorage" command is something magic.
Anyways, long story short, finally I have had some first tests.
First of all - unfortunatelly I must say - that what Astelith claimed proved wrong. Even with fan set to maximum of 2700 rpm the temp cannot be kept under 100 C on any modern games.
I downloaded Macs Fan Control (no desperate tries to make any bad taste humour of it this time), set the fan to constant maximum of 2700 rpm, it took only 5 minutes of play with Tomb Raider (2013) to achieve GPU temp 104 C (and possibly it's not even top of it).
Also note that this temp was reached with fan running maximum (2700 rpm) which iMac nearly never puts (as the automated controlled maximum is usually 2300 rpm). But even with that maximum what fan is capable of this cooling system of Retina iMac seems complitely insufficient for any modern PC gaming. Also the air output in the back of this machine that is a size of an small matchbox is just a joke.
Sorry to say, I have just purchased a nice 3500 euros paper weight to my desktop
What comes to performance itself, I was impressed how m295X was able to run Tomb Raider with 4k resolution pretty smoothly (aproximatelly 30-35fps) with medium settings. With Ultra settings it was just constant 60fps on full hd resolution and even 2.5k resolution something like 50-55fps.
This is compared to my current "Mac-inspired", watercooled Mini-ITX gaming rig running Radeon HD 7850 (built by myself, see here - sorry about poor Google translations, it doesn't seem to work very nicely with Finnish language). (The thing is that in that thread I was very concerned that my build's GPU temps exceeded 70 C at some cases, which now seems ridiculous low when comparing to iMac retina's 105-108 C).
The resolution scaling seems to be decent as well. The 4k resolution on Windows desktop doesn't look much different compared to 5k on OS X. (Of course if you take a magnification glass you can see a very small blurring affect on there, but in generally the desktop is extremely sharp even scaled to 4k). Also in games the 2560 x 1440 resolution seems to scale very sharp (and luckyly it's playable on most games) as full hd 1920 x 1080 seemed to be a little more blurry, but okay as well. I would definetely play with 2560 x 1440 when ever possible.
Also the noise level of 2700rpm doesn't seem to bother me so much (as I tend to wear headset, but even without them I wouldn't call this iMac very noisy compared to proper gaming rig. Of course it outputs some noise and possibly this is too much for some user's with earlier iMac gen's, but I still would not call this thing noisy at all, even with 2700rpm. Also the noise quality is mostly windy humming - there is no hearable fan bearing noise or so.
Anyways, I will continue some further testing and we'll see will I manage to use this aluminium chunk to anything else that keeping paper's on hold on my desk The 5k display is beatiful, I can say that. Shame that currently it runs only 4k on Windows, but hopefully some day the AMD Windows drivers will also support 5k.
But I am still quite afraid that the gaming performance of this machine takes a hit on any gaming session longer than 5-10 minutes: GPU starts to throttle more and more as the temp keeps pushing over 105-108C and after constant 1-2 hours of gaming sessions the framerates will drop dramatically. We'll see as I am going to test this all trough in forcoming days and weeks to come. Hopefully I will prove myself wrong, but looks bad to me...
Anyways, long story short, finally I have had some first tests.
First of all - unfortunatelly I must say - that what Astelith claimed proved wrong. Even with fan set to maximum of 2700 rpm the temp cannot be kept under 100 C on any modern games.
I downloaded Macs Fan Control (no desperate tries to make any bad taste humour of it this time), set the fan to constant maximum of 2700 rpm, it took only 5 minutes of play with Tomb Raider (2013) to achieve GPU temp 104 C (and possibly it's not even top of it).
Also note that this temp was reached with fan running maximum (2700 rpm) which iMac nearly never puts (as the automated controlled maximum is usually 2300 rpm). But even with that maximum what fan is capable of this cooling system of Retina iMac seems complitely insufficient for any modern PC gaming. Also the air output in the back of this machine that is a size of an small matchbox is just a joke.
Sorry to say, I have just purchased a nice 3500 euros paper weight to my desktop
What comes to performance itself, I was impressed how m295X was able to run Tomb Raider with 4k resolution pretty smoothly (aproximatelly 30-35fps) with medium settings. With Ultra settings it was just constant 60fps on full hd resolution and even 2.5k resolution something like 50-55fps.
This is compared to my current "Mac-inspired", watercooled Mini-ITX gaming rig running Radeon HD 7850 (built by myself, see here - sorry about poor Google translations, it doesn't seem to work very nicely with Finnish language). (The thing is that in that thread I was very concerned that my build's GPU temps exceeded 70 C at some cases, which now seems ridiculous low when comparing to iMac retina's 105-108 C).
The resolution scaling seems to be decent as well. The 4k resolution on Windows desktop doesn't look much different compared to 5k on OS X. (Of course if you take a magnification glass you can see a very small blurring affect on there, but in generally the desktop is extremely sharp even scaled to 4k). Also in games the 2560 x 1440 resolution seems to scale very sharp (and luckyly it's playable on most games) as full hd 1920 x 1080 seemed to be a little more blurry, but okay as well. I would definetely play with 2560 x 1440 when ever possible.
Also the noise level of 2700rpm doesn't seem to bother me so much (as I tend to wear headset, but even without them I wouldn't call this iMac very noisy compared to proper gaming rig. Of course it outputs some noise and possibly this is too much for some user's with earlier iMac gen's, but I still would not call this thing noisy at all, even with 2700rpm. Also the noise quality is mostly windy humming - there is no hearable fan bearing noise or so.
Anyways, I will continue some further testing and we'll see will I manage to use this aluminium chunk to anything else that keeping paper's on hold on my desk
The 5k display is beatiful, I can say that. Shame that currently it runs only 4k on Windows, but hopefully some day the AMD Windows drivers will also support 5k.But I am still quite afraid that the gaming performance of this machine takes a hit on any gaming session longer than 5-10 minutes: GPU starts to throttle more and more as the temp keeps pushing over 105-108C and after constant 1-2 hours of gaming sessions the framerates will drop dramatically. We'll see as I am going to test this all trough in forcoming days and weeks to come. Hopefully I will prove myself wrong, but looks bad to me...
Last edited:
More or less, with a high demanding game or GPU application the power draw is at 100%, but the clock is going to be calculated by Power Tune 2.0 to keep the power draw in spec and to keep the temperature below the limit also controlling the fan speed, pushing it to the limit, and in case, reducing further the clock (and the voltage accordingly).
In some cases, the power demand is too high that the clock can't be at the maximum rate, to keep power draw in the limits, Power Tune 2.0 set the clock to a lower value, let's make an example:
Scenario 1 - iMac 5K, M295X, I launch valley benchmark, the GPU receive instructions, based on that, the clock is set to 834MHz (to draw 125W max), the temp rise quickly reaching 96C, the fan goes at 1800rpm, the temp rise again to 100, the fan goes at 2300 (max limit for the GPU controller, only CPU can drive it to 2700), the clock goes to 815, the temp hit 101, the clock goes to 784, the temp goes to 99, the clock goes at 815, the temp goes at 100, the clock goes to 805, the temp goes at 99, etc.
Scenario 2 - iMac 5K, M295X, I launch WoW in low settings, the GPU receive instructions, based on that, the clock is set to 850MHz (the estimated draw is slightly less than 125W), the temp rise reaching 96C, the fan goes at 1600rpm, the temp rise again to 100, the fan goes at 1800, the temp hit 101, the fan goes at 2000, the temp goes to 99, the fan goes at 1800, the temp goes at 100, the fan goes to 1900, the temp goes at 99, etc.
*you can change 101 to 105 or even 108 for the "hot iMacs", the mechanism is the same
A 6-11% of clock difference (constantly going up and down) it's not really perceivable at naked eye, in your video showing high temps, the fps was in the 90's remaining there with 2-3 fps dancing up and down, I would't call it a performance issue due to high temperature
The article Is saying more or less what I do (in a very long way
) and explained with the examples above.More or less, with a high demanding game or GPU application the power draw is at 100%, but the clock is going to be calculated by Power Tune 2.0 to keep the power draw in spec and to keep the temperature below the limit also controlling the fan speed, pushing it to the limit, and in case, reducing further the clock (and the voltage accordingly).
In some cases, the power demand is too high that the clock can't be at the maximum rate, to keep power draw in the limits, Power Tune 2.0 set the clock to a lower value, let's make an example:
Scenario 1 - iMac 5K, M295X, I launch valley benchmark, the GPU receive instructions, based on that, the clock is set to 834MHz (to draw 125W max), the temp rise quickly reaching 96C, the fan goes at 1800rpm, the temp rise again to 100, the fan goes at 2300 (max limit for the GPU controller, only CPU can drive it to 2700), the clock goes to 815, the temp hit 101, the clock goes to 784, the temp goes to 99, the clock goes at 815, the temp goes at 100, the clock goes to 805, the temp goes at 99, etc.
Scenario 2 - iMac 5K, M295X, I launch WoW in low settings, the GPU receive instructions, based on that, the clock is set to 850MHz (the estimated draw is slightly less than 125W), the temp rise reaching 96C, the fan goes at 1600rpm, the temp rise again to 100, the fan goes at 1800, the temp hit 101, the fan goes at 2000, the temp goes to 99, the fan goes at 1800, the temp goes at 100, the fan goes to 1900, the temp goes at 99, etc.
*you can change 101 to 105 or even 108 for the "hot iMacs", the mechanism is the same
Great explanation - absolutely agree. Assuming Scenario 1 is a consistently demanding process, it is an example of 'thermal throttling'. The core clock speed is being dynamically limited due to heat, and heat alone. The card will lower the clock rate with no regard to the impact on performance - which may well be noticeable to a human eye - to keep the card within safe thermal limits.
The game I was playing limits the FPS to 90. Had I been able to remove this limitation the game could have been running at over 100 FPS therefore dropping ~10-15 FPS during the 'cool-off periods'. It's a shame my video is limited in that way, so I admit that's all theoretical - equally it may have run at 91 FPS without the game limit.
If a 10% performance hit is enough to bring FPS below ~35, then it suddenly becomes 'noticeable' whilst in-game. I agree with you partially, though, it's subjective and relatively minor. It also mostly affects gamers - not the target market for iMacs.
At 105°C the fan should be at 2300rpm and, no, it doesn't have headroom to increase – the GPU can only push the fan to 2300rpm. You need to get your CPU really hot too in order to push it any higher (max 2700rpm).
So when your fan is stuck at 2300rpm and your GPU temperature flickers between 105°, 104°, 103°, 106°, etc. on the benchmark, that's more than likely down to thermal throttling - your GPU clock speed is being dynamically reduced in order to avoid overheating.
Regardless, if you're happy - who cares? If it doesn't affect you, it doesn't matter! Enjoy your iMac
And a point to be taken in consideration is the equation behind the clock/fan control is always strictly related to the power draw, what does it means ? It means that even with low temps (let's say 70 degrees) is the power draw that pull down the clock, so the thermal limit is in most of the cases only anticipating what the power draw was about to do, or all the way around.
Can be thermal or power draw but this is the way this card works, and the performance seems constant over the cool and the hot 5Ks confirming what states above.
Can be thermal or power draw but this is the way this card works, and the performance seems constant over the cool and the hot 5Ks confirming what states above.
At 105°C the fan should be at 2300rpm and, no, it doesn't have headroom to increase the GPU can only push the fan to 2300rpm. You need to get your CPU really hot too in order to push it any higher (max 2700rpm).
So when your fan is stuck at 2300rpm and your GPU temperature flickers between 105°, 104°, 103°, 106°, etc. on the benchmark, that's more than likely down to thermal throttling - your GPU clock speed is being dynamically reduced in order to avoid overheating.
Regardless, if you're happy - who cares? If it doesn't affect you, it doesn't matter! Enjoy your iMac
What you're describing, yes - that the system deciding that it can't keep the temp down (assuming we're talking about a consistent load; a more 'real life' scenario could see some elements of that, given that the load on the GPU is varying somewhat) and adjusting the MHz accordingly.
But, what I saw on the benchmark was something different. I stayed at 102 for a good few minutes, then began to see 103 appear (iterating between the two values). Eventually it stayed at 103......then, after another few minutes 104 started to appear in the same pattern. It took a good 15-20 minutes before I started to get to 105 - if you like you could imagine it as the fan needing to go just that little bit quicker, then it would have stabilised the temp.
I almost get the impression that if the drivers were better, or if the system fan was a little more aggressive in how quickly it ramped up the speed then things may be a lot better. That doesn't exclude the points you've raised (they featured a lot in my decision making) - ultimately for me I know I won't be pushing a lot of games through this, especially since I've not bought anything new for about 4 years! All told - it's a great machine!
Two regrets. One: I wish I bought a Mac a few years ago. Second, why won't Safari remember the password for this forum account! It remembers everything else and even suggests passwords for this site. It just doesn't want to remember said login!!
For those looking to run more benchmarks, a new version of GPU-Z appeared recently that now officially supports the M295X: http://www.techpowerup.com/downloads/2463/techpowerup-gpu-z-v0-8-2/
@Astelith
Seems I'm finding interesting stuff today. Here is a link to a review of a R9 290X (custom configuration). The relevant part:
Different card, different vendor, different version of GCN and PowerTune. Yet it is interesting to note that customers (like Apple) could decide to use thermal throttling instead of PowerTune.
The piece of information that seems most clear at the moment is that under load, the M295X in the iMac reaches a stable temperature (around 105C) without pushing the fans to their max RPM. This doesn't look like thermal throttling to me, but more like a behavior that is controlled by software and desired to be that way by Apple or AMD (or both).
@Astelith
Seems I'm finding interesting stuff today. Here is a link to a review of a R9 290X (custom configuration). The relevant part:
Different card, different vendor, different version of GCN and PowerTune. Yet it is interesting to note that customers (like Apple) could decide to use thermal throttling instead of PowerTune.
The piece of information that seems most clear at the moment is that under load, the M295X in the iMac reaches a stable temperature (around 105C) without pushing the fans to their max RPM. This doesn't look like thermal throttling to me, but more like a behavior that is controlled by software and desired to be that way by Apple or AMD (or both).
Last edited:
Whoa mate, it gets that high because you were trying 4K! I know it's sort of "playable", I've done it myself. Better save that for older games and non-demanding games though, the experience will be much better. 1440p is probably the most optimal and I found most games would be playable at 1440p "High". I don't really experience throttling though, past weekend I had a 5-hour long Witcher 2 sessions and did not experience slowdowns.
The maximum rpm is being reached at 105C - the maximum the GPU can push the fan to is 2300rpm in the iMac. Only the CPU AND GPU can push it higher (ie. they both need to be extremely hot).
Even when manually forced to 2700rpm, the M295X still reaches 105C on most of the iMacs in this thread, and settles there. Therefore it still thermal throttles to achieve this.
You are absolutely correct though - Apple have opted to rely on thermal throttling rather than to redesign their cooling system. This is preferred because it makes for a quieter Mac. Simple as that.
The i7 CPU thermal throttles about 15% after 40 seconds of load in the 5K iMac. The M295X about 10%. It will be a niche of gamers and perhaps video editors who experience this, though.
I agree, but it's not a big compromise, as written above, that GPU would have cut the clock anyway due to TDP constrain (125W), so a better cooling system won't help.
With my "lucky" 5K, by forcing the fan at 2700 i can achieve the GPU at 90C in full load but the clock goes down even to 784 if the application is too demanding, so 90, 100 or 105 does not really make the difference
Astelith:
I still wonder how did you manage to keep temp at 90 C when GPU is in full load and fan is 2700rpm, because when I did the very same test just by playing Tomb Raider for 5 minutes, it simply wasn't enought to do the same but 104 C was the top (by the latest GPU-Z).
I still doubt you have not actually stressed your GPU properly. Maybe the WOW (or whatever the game was) idling in your videos actually do not stress GPU 100%. Or even furmark whatsoever, even though it tends to say so.
With my experience even 5 minutes of running Fireflies HTML5 test on OSX Safari makes GPU temp ramp to 102 C with automatic fan settings. Even if I set fan to the maximum possible 2700 rpm, it was just here-and-so to keep the GPU temp between 98-99C.
So what I have experienced here it simply seems that the fan system of the iMac 5k Retina (at least with i7 + m295X) is just insufficient even if the fan is set manually to keep 2700rpm constantly.
It think the resolution is irrelevant, if we just try to see if the cooling can keep up after 100% GPU load (+ some additional CPU load modern games can cause). It should not matter if the GPU renders it to 4k, 2k or fullhd resolutions, if the other settings are maxed out and GPU has to work 100%.
I will do some more testing, but to me it seems that sub-100C temps are impossible with gaming even with 2700rpm fan. At least if you don't game inside a freezer (my ambient room temp here is something like 23-24 C).
Cooling system of this device just cannot keep up dissipating the heat this CPU + GPU combination outputs, and this leads to thermal throttling at some point. For me it's not yet clear is this a major problem and does it affect my gaming experience noticeably, but I will see that in the near future.
But I know that Apple should have redesigned the cooling system for iMacs when selecting such high TDP CPU and GPU models for their top iMac model. This would have not necessarily meant more noise, but even less. (As you know a bigger and more effective cooler may move more air even with lower rpms than the current rather small one. Maybe also the air outputs should have been a little larger too.)
I still wonder how did you manage to keep temp at 90 C when GPU is in full load and fan is 2700rpm, because when I did the very same test just by playing Tomb Raider for 5 minutes, it simply wasn't enought to do the same but 104 C was the top (by the latest GPU-Z).
I still doubt you have not actually stressed your GPU properly. Maybe the WOW (or whatever the game was) idling in your videos actually do not stress GPU 100%. Or even furmark whatsoever, even though it tends to say so.
With my experience even 5 minutes of running Fireflies HTML5 test on OSX Safari makes GPU temp ramp to 102 C with automatic fan settings. Even if I set fan to the maximum possible 2700 rpm, it was just here-and-so to keep the GPU temp between 98-99C.
So what I have experienced here it simply seems that the fan system of the iMac 5k Retina (at least with i7 + m295X) is just insufficient even if the fan is set manually to keep 2700rpm constantly.
It think the resolution is irrelevant, if we just try to see if the cooling can keep up after 100% GPU load (+ some additional CPU load modern games can cause). It should not matter if the GPU renders it to 4k, 2k or fullhd resolutions, if the other settings are maxed out and GPU has to work 100%.
I will do some more testing, but to me it seems that sub-100C temps are impossible with gaming even with 2700rpm fan. At least if you don't game inside a freezer (my ambient room temp here is something like 23-24 C).
Cooling system of this device just cannot keep up dissipating the heat this CPU + GPU combination outputs, and this leads to thermal throttling at some point. For me it's not yet clear is this a major problem and does it affect my gaming experience noticeably, but I will see that in the near future.
But I know that Apple should have redesigned the cooling system for iMacs when selecting such high TDP CPU and GPU models for their top iMac model. This would have not necessarily meant more noise, but even less. (As you know a bigger and more effective cooler may move more air even with lower rpms than the current rather small one. Maybe also the air outputs should have been a little larger too.)
Unfortunately, the 5K comes with random quality heat dissipation and I'm a very lucky owner, also in an Italian forum we are trying to make some statistics and it seems the average is around 104°C as max temp.
This is the real topic we should discuss about, as I told you, we have the same performance, the temperature is not lowering alone the clock, there's also power draw that will activate this mechanism of clock dynamical adjustment.
The shame for Apple is that units are relatively quiet (like mine and some others) and others are louder (like yours and many others), the cause might be thermal paste, bad assembly or a not well calibrated sensor.
The only difference between 5K iMacs are the fan noise under the same load, I can run at 1850rpm something that you will do at 2300
Unfortunately, the 5K comes with random quality heat dissipation and I'm a very lucky owner, also in an Italian forum we are trying to make some statistics and it seems the average is around 104°C as max temp.
This is the real topic we should discuss about, as I told you, we have the same performance, the temperature is not lowering alone the clock, there's also power draw that will activate this mechanism of clock dynamical adjustment.
The shame for Apple is that units are relatively quiet (like mine and some others) and others are louder (like yours and many others), the cause might be thermal paste, bad assembly or a not well calibrated sensor.
The only difference between 5K iMacs are the fan noise under the same load, I can run at 1850rpm something that you will do at 2300
Unfortunatelly - as already discussed in this thread - I cannot agree with your statement, unless you post me a screenshot of GPU-Z showing maxed GPU temp after playing like 30 minutes of Tomb Raider (or any other modern PC game title).
To me it seems impossible that there could be such big quality variations between the same models. It's just not possible that very minimal differences in thermal paste applying in Apple's manufacturing plant alone could lead to such large differences in heat dissapation.
I think it's more like the one's reporting sub 100C temps have not such gamed with this machine for several hours and kept GPU-Z recording the max GPU temp there. Its obvious that it gets past 100C even with fan set to 2700rpm manually with Macs Fan Control.
I don't know how demanding Divinity: Original Sin is but I can play it with everything on Ultra at roughly 3k resolution. I don't monitor FPS (don't see the point if it's just for fun) but it seems to play smoothly.
It crashes completely at 4/5k but I'm pretty certain that's a glitch in the game and nothing to do with the card not handling it.
It crashes completely at 4/5k but I'm pretty certain that's a glitch in the game and nothing to do with the card not handling it.
I made a video playing Metro Last Light in 4K with High details:
https://www.youtube.com/watch?v=EWLuCQpt5Kc
First part with maxed out fan (2700), I got temp around 93/94° and then with auto settings I have temp at 99/100°.
Sad but true, Apple have a very big problem with quality for this unit.
Very well, the same game is also in my Steam collection, so I will try to reduplicate the same and report the temps I will get here. I cannot comment yet anything, before seeing your video troughly and testing the same with my own machine. I'll get back to this issue later.
if you're going to do that, you should both make sure you record room temperature as well
I was using Win 10 with the last beta driver from AMD, with afterburner you can have that type of logs, let me know if you need other informations.
----------
I made a quick test also with that, I kept the heater on for a night reaching 30° in the room but with no difference under load (only a slightly higer ilde temps), consider that I live in a hot place where even in winter is difficult to have less than 10°C outside and I keep usually an inside temp around 23°C
Are you refering the new AMD Omega drivers, that does not yet officially support m295X but can be installed by instructions here?
Or possibly the AMD's official Bootcamp drivers published in Dec 2014, listed here?
Currently I am using the original bootcamp drivers Apple provides, but it's not a problem to upgrade them. However I hardly believe drivers would affect the temperature at all.
So to me it seems that ambient room temperature has no much affect on this, but we'll see. Mine's always 23-24C.
Unfortunatelly I didn't yet had time to test run Metro Last Light, but I run something like 15-20 minutes of Bioshock Infinite and clocked a top temp of 106 C with my iMac
I noticed almost instantly the micro stutters mentioned also in the first page of this thread. Very annoying! It's like 5 seconds of smooth playing and then half second stuttering (like dropping the frame rate from 50fps to 15 fps for half a second, and then coming back to 50fps). Fan set the 2700rpm at all times.
When this occured the temp was oscillating between 105-106C and the GPU Mhz's was between 783Mhz and 832Mhz, but there were some short drops to low as 530Mhz as well.
The weirdest thing is that also the GPU load didn't constantly run on 100% but dropped to half or less when gaming Bioshock Infinity. But when I get to menu, the GPU load is constantly 100%, temps 105 C and GPU MHz between 780-830Mhz.
When I did the same test with my old computer (having watercoolder HD7850) the overal performance was little less, but it's constantly same: There is no micro stuttering at all. So it's subjectively much more enjoyable to play with my old gaming rig that on this new iMac.
So to me it seems that there is something really wrong with these iMac Retinas. These are just not suitable for gaming at all. At least my one. The micro stuttering is the most annoying side affect of GPU throttling that is very noticeable, and troughoutly kills any enjoyment of the gaming.
I wish I could return this peace of ****. I am afraid in Finland we just don't have such luxury, at least what the retail stores and Apple support says: "Custom macs cannot be returned".
Well, at least I have a nice 3500 euros paper holder on my desk that you can surf the web (but even that sometimes leads over 100C of GPU temps).
Seriously: I will try to do some more tests asap (in forcoming days). What is the software Astelith is using to see the fps, gpu temp and load directly on gaming screen? To me those temps 93-94 C are just amazing compared to my 105-106C (both running fan at 2700 rpm). You don't experience any micro stuttering on games?
PS. It's sarcastically funny that this iMac idles the GPU on the same temps I have my maximum temps on the old machine, 65-70C My old rig's idle GPU temp is between 37-38 C. There've you got something to learn of cooling in Cupertino.
I noticed almost instantly the micro stutters mentioned also in the first page of this thread. Very annoying! It's like 5 seconds of smooth playing and then half second stuttering (like dropping the frame rate from 50fps to 15 fps for half a second, and then coming back to 50fps). Fan set the 2700rpm at all times.
When this occured the temp was oscillating between 105-106C and the GPU Mhz's was between 783Mhz and 832Mhz, but there were some short drops to low as 530Mhz as well.
The weirdest thing is that also the GPU load didn't constantly run on 100% but dropped to half or less when gaming Bioshock Infinity. But when I get to menu, the GPU load is constantly 100%, temps 105 C and GPU MHz between 780-830Mhz.
When I did the same test with my old computer (having watercoolder HD7850) the overal performance was little less, but it's constantly same: There is no micro stuttering at all. So it's subjectively much more enjoyable to play with my old gaming rig that on this new iMac.
So to me it seems that there is something really wrong with these iMac Retinas. These are just not suitable for gaming at all. At least my one. The micro stuttering is the most annoying side affect of GPU throttling that is very noticeable, and troughoutly kills any enjoyment of the gaming.
I wish I could return this peace of ****. I am afraid in Finland we just don't have such luxury, at least what the retail stores and Apple support says: "Custom macs cannot be returned".
Well, at least I have a nice 3500 euros paper holder on my desk that you can surf the web (but even that sometimes leads over 100C of GPU temps).
Seriously: I will try to do some more tests asap (in forcoming days). What is the software Astelith is using to see the fps, gpu temp and load directly on gaming screen? To me those temps 93-94 C are just amazing compared to my 105-106C (both running fan at 2700 rpm). You don't experience any micro stuttering on games?
PS. It's sarcastically funny that this iMac idles the GPU on the same temps I have my maximum temps on the old machine, 65-70C
My old rig's idle GPU temp is between 37-38 C. There've you got something to learn of cooling in Cupertino.Unfortunatelly I didn't yet had time to test run Metro Last Light, but I run something like 15-20 minutes of Bioshock Infinite and clocked a top temp of 106 C with my iMac
I noticed almost instantly the micro stutters mentioned also in the first page of this thread. Very annoying! It's like 5 seconds of smooth playing and then half second stuttering (like dropping the frame rate from 50fps to 15 fps for half a second, and then coming back to 50fps). Fan set the 2700rpm at all times.
When this occured the temp was oscillating between 105-106C and the GPU Mhz's was between 783Mhz and 832Mhz, but there were some short drops to low as 530Mhz as well.
The weirdest thing is that also the GPU load didn't constantly run on 100% but dropped to half or less when gaming Bioshock Infinity. But when I get to menu, the GPU load is constantly 100%, temps 105 C and GPU MHz between 780-830Mhz.
When I did the same test with my old computer (having watercoolder HD7850) the overal performance was little less, but it's constantly same: There is no micro stuttering at all. So it's subjectively much more enjoyable to play with my old gaming rig that on this new iMac.
So to me it seems that there is something really wrong with these iMac Retinas. These are just not suitable for gaming at all. At least my one. The micro stuttering is the most annoying side affect of GPU throttling that is very noticeable, and troughoutly kills any enjoyment of the gaming.
I wish I could return this peace of ****. I am afraid in Finland we just don't have such luxury, at least what the retail stores and Apple support says: "Custom macs cannot be returned".
Well, at least I have a nice 3500 euros paper holder on my desk that you can surf the web (but even that sometimes leads over 100C of GPU temps).
Seriously: I will try to do some more tests asap (in forcoming days). What is the software Astelith is using to see the fps, gpu temp and load directly on gaming screen? To me those temps 93-94 C are just amazing compared to my 105-106C (both running fan at 2700 rpm). You don't experience any micro stuttering on games?
PS. It's sarcastically funny that this iMac idles the GPU on the same temps I have my maximum temps on the old machine, 65-70C
Use the osd on screen installed with afterburner (https://gaming.msi.com/features/afterburner)
Honestly I don't know if mine stutter, I'm not a gamer, I play just a couple of hours a week or so, I enjoy the performance but maybe it's me less demanding.
For the record, my idle temp are very low:
If I were you I will try to go in an apple store and ask to repaste the GPU/heatsink, we should have the same components, what can make such a difference ?