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

[thundersteele]

I just installed an update to iStat Menus. The GPU die temperature seems to have dropped by 5C...

Tests run:
Gaming, GPU hit 100C at 2300rpm, where before it would reach 105C
GPU benchmark: 102C at 2300rpm, before would go to 106C or so
GPU + CPU benchmark: up to 103C at 2700rpm, fan drops to 2300rpm when CPU benchmark finishes

All done in OSX. The CPU temps didn't change much from what I remember. Maybe it is just different room temps, so maybe someone else can have a new look?

 

[Astelith]

I just installed an update to iStat Menus. The GPU die temperature seems to have dropped by 5C...

Tests run:
Gaming, GPU hit 100C at 2300rpm, where before it would reach 105C
GPU benchmark: 102C at 2300rpm, before would go to 106C or so
GPU + CPU benchmark: up to 103C at 2700rpm, fan drops to 2300rpm when CPU benchmark finishes

All done in OSX. The CPU temps didn't change much from what I remember. Maybe it is just different room temps, so maybe someone else can have a new look?

This is interesting, in my case I have no difference after the update.
Maybe the difference in the iMacs are just the sensors and we are all running the same real temps

 

[andy9l]

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

At this level, a difference of 3-5C becomes irrelevant.

The thermal throttling exists, regardless of what temperature is reported. The components are still over 100C, the fan is still at max speed, and the core clock speeds of the i7 and M295X are being reduced as they reach these extreme temperatures.

Does it really matter? Probably not, unless it bothers you.

 

[CrankItUp]

I just installed an update to iStat Menus. The GPU die temperature seems to have dropped by 5C...

Tests run:
Gaming, GPU hit 100C at 2300rpm, where before it would reach 105C
GPU benchmark: 102C at 2300rpm, before would go to 106C or so
GPU + CPU benchmark: up to 103C at 2700rpm, fan drops to 2300rpm when CPU benchmark finishes

All done in OSX. The CPU temps didn't change much from what I remember. Maybe it is just different room temps, so maybe someone else can have a new look?

I ll give it a try tomorrow.. Maybe OS X 10.10.3 has an additonal effect? I didnt install the update yet..

 

[WilliamG]

This is interesting, in my case I have no difference after the update.
Maybe the difference in the iMacs are just the sensors and we are all running the same real temps

I see no difference after the upgrade, either.

 

[thundersteele]

I see no difference after the upgrade, either.

Well, as I said it might just be lower ambient temperatures, I'm not claiming that I have a stable testing environment. If nobody else sees any effect it is probably just that.

 

[Astelith]

Well, as I said it might just be lower ambient temperatures, I'm not claiming that I have a stable testing environment. If nobody else sees any effect it is probably just that.

Room temp only affect idle temp, and not so much.

 

[blufrog]

Where can I get iStat? Also, I don't appear to be suffering heat issues when running 3D apps (running 3D apps in Java on my rMBP causes it to become a small heater, but the 5k doesn't seem to care, even after 30+ minutes, and frame rate doesn't drop at all).

 

[aevan]

I may look like the ultimate troll for saying this, but this topic has been around since the iMac 5K came out and a gazillion posts later I don't understand what all of you trying to prove, disprove or achieve. I'm genuinely curious. All I can see is "it runs hotter than 100 degrees, it throttles" and that sentence is repeated, like, 1000 times with some occasional graphs detailing throttling. And after 40+ pages of this, I have to ask, erm, why are you continuing to discuss this?

Here, I'll just wrap this up for you:

1. Yes, the M295X runs hot and it's probably designed that way. Laptop components get hotter then desktop ones. Unless you run, like, Crysis 24/7 for six months straight, it will be fine.
2. Yes, both the GPU and CPU will throttle. They probably throttle less than the base model, but even then, they are faster.
3. You're not judging this iMac for what it is, but for what you want it to be.
4. Apple must've had a reason to go with AMD for this model. Even if it was 'just the price' - that is possibly a good enough reason. If they couldn't make this model for cost X, perhaps we wouldn't even see an iMac 5K this year. Or the next.
5. Comparing this year's model to the 2012 or 2013 iMac is wrong. It is faster (for most things) and it's screen has 4x as many pixels. It is also louder and warmer. If you want to compare the heat and noise, also compare the other specs.


I mean.... and I know a lot of people will hate me for this post but.... aren't you being a little bit overdramatic? Just asking. I mean, you have this marvel of engineering in your hands and you bitc.... complain because it's hot and you can sometimes hear the fan? I don't get it. But then again, I'm aware this is just my opinion. You guys and gals have every right to disagree.

 

[cynics]

I may look like the ultimate troll for saying this, but this topic has been around since the iMac 5K came out and a gazillion posts later I don't understand what all of you trying to prove, disprove or achieve. I'm genuinely curious. All I can see is "it runs hotter than 100 degrees, it throttles" and that sentence is repeated, like, 1000 times with some occasional graphs detailing throttling. And after 40+ pages of this, I have to ask, erm, why are you continuing to discuss this?

Here, I'll just wrap this up for you:

1. Yes, the M295X runs hot and it's probably designed that way. Laptop components get hotter then desktop ones. Unless you run, like, Crysis 24/7 for six months straight, it will be fine.
2. Yes, both the GPU and CPU will throttle. They probably throttle less than the base model, but even then, they are faster.
3. You're not judging this iMac for what it is, but for what you want it to be.
4. Apple must've had a reason to go with AMD for this model. Even if it was 'just the price' - that is possibly a good enough reason. If they couldn't make this model for cost X, perhaps we wouldn't even see an iMac 5K this year. Or the next.
5. Comparing this year's model to the 2012 or 2013 iMac is wrong. It is faster (for most things) and it's screen has 4x as many pixels. It is also louder and warmer. If you want to compare the heat and noise, also compare the other specs.


I mean.... and I know a lot of people will hate me for this post but.... aren't you being a little bit overdramatic? Just asking. I mean, you have this marvel of engineering in your hands and you bitc.... complain because it's hot and you can sometimes hear the fan? I don't get it. But then again, I'm aware this is just my opinion. You guys and gals have every right to disagree.

I don't think people will hate you, objectively and technically you are correct.

However there is a certain level of quality that many people of come to expect of Apple. And when there is a couple unadvertised 'cons' added to the list by spending MORE money that just doesn't sit well with people.

You got a bit subjective with "sometimes hear the fan" because that is a major selling point for many. May seem silly to you and I but its a big deal to others. And to make matters worse many of those people are used to using an iMac so they "remember the time" that their iMac wasn't constantly making a racket.

You'll notice throughout the thread the comparison between 2013 and the 5k is in reference to what it used to do that was identical. "Faster" is relative. For example if I'm playing a game with vert sync on so at 60 FPS on my 2013 iMac and the fan never leaves 1200 rpm while the 5k iMac stays at 2300 rpm how is that better? Its not faster because the FPS is capped, the resolution is the same. The only difference is the whirlwind of noise its producing. Same applies to other things then just gaming that just help show some actual figures.

My question to you would be, if the next generation 5k iMac is dramatically quieter what will you think of this generation? Hypothetical question of course, just curious.

 

[boast]

For example if I'm playing a game with vert sync on so at 60 FPS on my 2013 iMac and the fan never leaves 1200 rpm while the 5k iMac stays at 2300 rpm how is that better? Its not faster because the FPS is capped, the resolution is the same. The only difference is the whirlwind of noise its producing.

If we agree the i7 and m295x are more powerful, and with more power requirements create more heat, I would say the issue is that OS X is ramping the GPU and CPU to maximum when it is not really needed.

The OS would have to know that putting the CPU/GPU at a maximum state will not make anything faster/better, and scale it down to a lower state that produces less power.

On linux I remember the cpu frequency scaling is based on load, and then other parameters like how long the cpu is on a specific load amount before it ramps up to a higher frequency.

I think if we could modify that on OS X and make it very conservative, it would stay quiet longer with the sacrifice in taking much longer to ramp up to full speed and suffering performance for things that are short performance peak based. But for the two extremes, web surfing/video watching it would be fine, and the other end for video encoding it would be long enough that it would eventually reach full CPU peak no problem.

 

[cynics]

If we agree the i7 and m295x are more powerful, and with more power requirements create more heat, I would say the issue is that OS X is ramping the GPU and CPU to maximum when it is not really needed.

The OS would have to know that putting the CPU/GPU at a maximum state will not make anything faster/better, and scale it down to a lower state that produces less power.

On linux I remember the cpu frequency scaling is based on load, and then other parameters like how long the cpu is on a specific load amount before it ramps up to a higher frequency.

I think if we could modify that on OS X and make it very conservative, it would stay quiet longer with the sacrifice in taking much longer to ramp up to full speed and suffering performance for things that are short performance peak based. But for the two extremes, web surfing/video watching it would be fine, and the other end for video encoding it would be long enough that it would eventually reach full CPU peak no problem.

I agree the CPU and GPU are capable of doing more work in a shorter amount of time i.e. faster. However all things aren't equal across the board.

If you have CPU X that is twice as fast as CPU Y at a certain process that doesn't mean it will run at half the temperature doing that process in an otherwise identical manor (to the end user).

I can't speak for every 2013 iMac configuration however with the i5 775m iMac's cooling design is severe overkill. It is near impossible under any circumstance to get the CPU or GPU to pull the fan off idle. Even as I type this I've had Handbrake running for 3+ hours.

So for me its very easy to see why someone would be spoiled by how silently the iMac runs.

However my problem with the 5k is not fan noise its that its an all-in-one that you can make throttle even if its a minute, unnoticeable in the real world amount. That should be virtually impossible and as anyone thats ever built a PC knows is indicative of inadequate cooling.

I use a near maxed out 5k at work and its a very good computer for what I do. It generally runs very quietly and only during rare events does the fan ramp up. I've even debated on multiple occasions replacing my personal 2013 with one. Quite a few times my mouse cursor has hovered above the buy button on a BTO 5k iMac at Apple.com. However just knowing I could make it throttle especially since I tend to game a bit more then your average mac user I just can't pull the trigger on it. Plus I'm not in anyway unhappy with my 2013.

 

[t0bias]

And to make matters worse many of those people are used to using an iMac so they "remember the time" that their iMac wasn't constantly making a racket.

I totally see that and it's true. But it has happened before with apple. I can remember my core2duo 2010 mbp beeing extremely silent. The next generation mbp totally changed that for the worse.

 

[CrankItUp]

I just installed an update to iStat Menus. The GPU die temperature seems to have dropped by 5C...

Tests run:
Gaming, GPU hit 100C at 2300rpm, where before it would reach 105C
GPU benchmark: 102C at 2300rpm, before would go to 106C or so
GPU + CPU benchmark: up to 103C at 2700rpm, fan drops to 2300rpm when CPU benchmark finishes

All done in OSX. The CPU temps didn't change much from what I remember. Maybe it is just different room temps, so maybe someone else can have a new look?

I also tried that - result: nothing changed

Like i said: Lets wait for a smc update

 

[WilliamG]

I also tried that - result: nothing changed

Like i said: Lets wait for a smc update

Oh no, why an SMC update? I'm completely thrilled that watching the Barcelona vs PSG game today in Chrome - full screen - had my 295 GPU at 96C the entire time.

Such a joke...

I love the computer, but the temperatures are just mad.

 

[Melodeath]

I haven't read this thread in awhile, is it confirmed that the 295 and i7 cause more fan noise, but less throttling than the stock configuration? I definitely want the i7 over the i5, but am still undecided about the GPU

 

[andy9l]

More fan noise, more throttling, but still better performance than stock.

 

[blufrog]

I've had my 5k for nearly a week now, and I've been running high-end 3D graphics engines on it without any trouble (yes, it gets quite warm, especially on the stand, and I can sometimes hear the fan) but I haven't seen anything to suggest throttling of either the GPU or CPU, nor any loss of performance after several hours of use.

Ambient temp is around 20 deg. C.

I'm starting to think that there are QA problems with some builds, and you either get a bad unit, or not.

I can only suggest contacting Apple if it is throttling.

Where can I get iStat? I will collect some data from my machine and post it here.

 

[andy9l]

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

Blufrog, your machine will throttle if it's using the i7 or M295X. Carry out the tests yourself if you'd like to see it in action, or review this thread or elsewhere online for demos.

It's just a 'flaw' with this generation. It only really impacts a niche. Doesn't make it right, doesn't make it a huge design flaw. It is what it is.

It clearly doesn't affect you, so no need to worry. If you do find your machine really doesn't throttle, please post the results. There's always that tiny, tiny chance Apple adjust something in the production lines.

 

[blufrog]

I can't find a suitable app that reports system measured temperatures. Reviews are at best inconclusive as to their accuracy.

Looking at it empirically, if my system is throttling, it must be doing it in the first 10 seconds or so, because once I have my software up I don't see any drop in performance over several hours of continuous use.

I *do* sometimes hear the fan running, but only for a couple of minutes, then it goes quiet again (the room is noticeably warm when this happens). There is a good airflow out the back all the time under load, and things are most certainly getting warm (the case is warmer than ambient, no obvious hot spots, whilst the stand is very warm), so I doubt my system is getting that hot (if that was the case, I would guess one symptom I would see of the throttling would be that the fan is running flat-out, which it doesn't appear to be doing).

 

[andy9l]

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

You can use iStat. Temperatures are not really relevant.

It's apparent that your usage does not stress the GPU enough, consistently, for it to negatively impact your usage. Therefore, you can stop worrying about it.

You don't have the i7, so your fan will probably be hitting a maximum of 2,300rpm every now and then to cool the GPU. At such times, the M295X core clock speed will likely be reduced (PowerTune thermal throttle). You don't notice it, the noise doesn't bother you, so why worry?

You can read this thread or search online for more details, tests and results.

 

[blufrog]

Actually yes I do have the i7, and I can't get iStat (I search the App store, and only get something else).

I'd be surprised if my GPU isn't being hammered.

EDIT: OK - I did something that causes the engine to kill the GPU, waited for the fan to ramp up due to heat, then closed what I was doing back to an empty engine rendering nothing, and the FPS was identical at the end with the fan running immediately after high load as it was when I started with a cool machine.

I'm very confident I'm not seeing throttling.

EDIT: There is Temperature Gauge - is that accurate/any good?

 

[andy9l]

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

I thought your signature said i5. My mistake.

Again, you might not 'see' throttling because of what you're doing. Also, without actually monitoring, you're likely going to struggle to notice it.

It exists, though. Watch this from about 2 minutes in...

http://youtu.be/tgTMxB-ffjM

A similar issue applies with the M295X GPU. Simply read this thread for all the proof points you need.

I stress again how unimportant this is. It's not affecting you, so forget about it and enjoy your iMac.

 

[SmallDane]

Actually yes I do have the i7, and I can't get iStat (I search the App store, and only get something else).

Download it here: http://bjango.com/mac/istatmenus/

 

[Tobamory]

Ok lets stop all the anal temp Monitoring the Imac has 4 x more pixels it used the 295 Tonga chipset the first of its kind. Its powerful and yes the fans do go some if playing games, so what ?

but if its actually stopping you buyin it then so be it, but 5k screen is to die for, and using it for photo editing is fantastic.

being a heavy gamer MACs are not good for games end of, the Mac OS just isn't up to it like a PC, Bootcamp and youll see the FPS jump up

Ive tried all things

World of Tanks
Warcraft
StarCraft 2

I also tested a fully loaded Mac Book pro I7 retina and guess what, its overheated and shutdown when gaming, it also throttled a hell of a lot in WOW to cope with the heat..... so its been with us a while this throttling lark,

if you want the crisp screen all in 1 neat Package that can pump out around 60 to 75 FPS in most games at 1920 then this is the 1.

But don't think your gonna game in 5k because most games will run but sub 20FPS......