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

[magbarn]

Guys, surely if this was a serious issue, then ASC and other forums would be swamped with complaints, but I can't see it.

Here's the two issues with this:
1. Currently, if you're using GPU compute like OpenCl, light gaming with older titles, or for general use, no big deal as you'll never stress the GPU enough to saturate the cooling system. Gaming with newer more demanding titles, however, is compromised as the chip will self-throttle in the interest of self preservation.

2. Long term, AMD might have finally figured out how to make a reliable lead free BGA solder that can handle thousands of heating/cooling cycles without eventually falling apart or the stressed GPU's will start to die off several years from now requiring a logic board replacement as Apple no longer puts the GPU on a separate module like pre-2012 days.

There kind of is nothing to fix. The GPU runs hot when gaming, it's a fact. Issue if there will be any repercussions of the cards running at that temperature for extended periods of time.

The desktop AMD R9 290x is running extremely hot (95-100C) and was used in bitcoin mining for the past year, where the cards are pushed to extreme and work 24/7 at 100% and so far I haven't seen anyone complaining that the GPU failed.
On the other hand I had Gigabyte 780ti GHZ edition that physically burned with smoke and everything twice already while the card never made it above 85C, Gigabyte is actually refunding me even though I bought from NewEgg.

I have the Apple Care and I am not worried about it one bit.

The reason why bitcoining actually may be less stressful than regular intermittent gaming/use is the BGA lead contacts aren't dealing with frequent thermal expansion/contraction as the GPU is basically running full throttle constantly with a constant temperature.

 

[imacken]

Here's the two issues with this:
1. Currently, if you're using GPU compute like OpenCl, light gaming with older titles, or for general use, no big deal as you'll never stress the GPU enough to saturate the cooling system. Gaming with newer more demanding titles, however, is compromised as the chip will self-throttle in the interest of self preservation.

2. Long term, AMD might have finally figured out how to make a reliable lead free BGA solder that can handle thousands of heating/cooling cycles without eventually falling apart or the stressed GPU's will start to die off several years from now requiring a logic board replacement as Apple no longer puts the GPU on a separate module like pre-2012 days.

I get what's being said here. My point is, why is the web not swamped with similar threads and stories?

 

[magbarn]

I get what's being said here. My point is, why is the web not swamped with similar threads and stories?

This generation isn't old enough yet for the problems to appear, but Apple's already been sued and lost on the 2008 macbook pro and 2011 iMac GPU heating issues. Also, I would assume the vast majority of rimac users bought the basic configuration which runs substantially cooler than the i7/295x variant considering my local Apple stores only carry the basic configuration.

On the absence of the web being swamped with this throttling... The majority of Apple reviews are basically now marketing fluff pieces considering how Apple has severely dealt with unfavorable reviewers in the past...
I would wait for Anandtech's review, hopefully they get a fully loaded machine and run it till it throttles and documents it.

 

[evilkow]

Throttling Confirmation on OS X?

Getting back to Fenn's original post, has anyone confirmed throttling on the Retina iMac in Yosemite? We've got lots of confirmations that this is happening on Windows, but still nothing that I have seen confirming it in OS X.

 

[magbarn]

Getting back to Fenn's original post, has anyone confirmed throttling on the Retina iMac in Yosemite? We've got lots of confirmations that this is happening on Windows, but still nothing that I have seen confirming it in OS X.

Nope I haven't seen any examples yet, but given that OS X gaming titles are usually on the low end of the graphics scale and much less likely to stress the GPU as much.

 

[Mac32]

Even on the desktop class GPU's, the GTX980 uses much less power and blows away AMD's top of the line chip.

I've been called out for being a Nvidia fanboi by the poster you're responding to, but he's behaving as a cheerleader for AMD likewise. I don't care about either brand but will gravitate towards the superior chip overtime. (Notice he likes to compare the 295 to the prior 880MX which the 980M is much faster still) My wife has a 2012 iMac with 680MX which overclocks quite nicely like yours and doesn't throttle either. We're waiting for the 2nd gen rimac which will either have a proper GPU like the 980 or at least a much cooler running 20nm AMD.

Exactly! I'm really excited about the retina iMac, but I also like to game regularly, and just don't feel comfortable investing money in the present RiMac model. Hopefully 2.gen RiMac will use a Nvidia GPU.

 

[rainydays]

I get what's being said here. My point is, why is the web not swamped with similar threads and stories?

What magbarn said.

If there is a longevity issue due to heat we won't hear about until much later. And as for the throttling there's not a lot of people who would think of even investigating it, let alone know how to.

What I do see a lot of comments about is that the fans run at full speed when people play games. That's as far as most peoples observations will go. Some will accept it, others return it.

Ars technics actually mentioned the potential issue in their recent article about the RiMac and links to this thread.

 

[imacken]

Nope I haven't seen any examples yet, but given that OS X gaming titles are usually on the low end of the graphics scale and much less likely to stress the GPU as much.

Bioshock Infinite, Tomb Raider, Deux Ex, F1 2013, Witcher 2, Max Payne 3, Call of Duty series, Spec Ops, Metro Last Light and many more.

 

[magbarn]

Bioshock Infinite, Tomb Raider, Deux Ex, F1 2013, Witcher 2, Max Payne 3, Call of Duty series, Spec Ops, Metro Last Light and many more.

Shows you how much I'm up to date on OS X gaming as I basically just resigned myself to gaming on bootcamp. Tell me this, are the FPS the same now between the OS X vs the Windows counterpart? 2 years ago the windows version had 25-100% higher frame rates which would stress the GPU more as it's able to work harder instead of waiting to be fed by a slower OS X.

 

[habeebhashim]

You need a heat gun to melt the adhesive around the edges of the screen glass to remove it and a replacement adhesive ring to refit. Then you have to get at GPU to put a sensor probe next to it and having 2012 chassis apart I would find that very difficult to do.

I've yet to see a sensor on the GPU and CPU to seem badly wrong. Put the 5k to deep sleep, and wake it up and if the temperatures are close to ambient you know that it has correct calibration.

I have tested this before and yesterday as well. I have left the 5K to sleep overnight (8 hours) and when I woke it up, iStat showed GPU temp at 50c. Same today when I had switched it off last night and started it this morning.

My room temp is 24c

----------

Shows you how much I'm up to date on OS X gaming as I basically just resigned myself to gaming on bootcamp. Tell me this, are the FPS the same now between the OS X vs the Windows counterpart? 2 years ago the windows version had 25-100% higher frame rates which would stress the GPU more as it's able to work harder instead of waiting to be fed by a slower OS X.

I posted about this earlier...

https://forums.macrumors.com/showthread.php?p=20335245#post20335245

Played Tomb Raider on OS X to compare vs my gaming OS of choice - Win 7. Its not good enough on OSX.

 

[5iMacs]

I have tested this before and yesterday as well. I have left the 5K to sleep overnight (8 hours) and when I woke it up, iStat showed GPU temp at 50c. Same today when I had switched it off last night and started it this morning.

My room temp is 24c

----------

Huh? Why?

All this means is that the GPU temp rose to 50C from the time the iMac came out of sleep to the time that you were able to see the iStat reading.

 

[habeebhashim]

Huh? Why?

All this means is that the GPU temp rose to 50C from the time the iMac came out of sleep to the time that you were able to see the iStat reading.

Don't have a clue.

Just checked with my 2012 rMBP with the 650M... was also on sleep overnight

and iStat reports temp at 38c

 

[magbarn]

Don't have a clue.

Just checked with my 2012 rMBP with the 650M... was also on sleep overnight

and iStat reports temp at 38c

Is powernap on? That might be preventing your iMac from completely sleeping.

 

[ha1o2surfer]

Shows you how much I'm up to date on OS X gaming as I basically just resigned myself to gaming on bootcamp. Tell me this, are the FPS the same now between the OS X vs the Windows counterpart? 2 years ago the windows version had 25-100% higher frame rates which would stress the GPU more as it's able to work harder instead of waiting to be fed by a slower OS X.

I would also imagine DirectX being more demanding than OpenGL as the games in MacOSX don't have the features that DirectX has. You can do a side by side feature comparison of the two and a lot doesn't exist in OpenGL/isn't ported over to OpenGL.

 

[imacken]

Shows you how much I'm up to date on OS X gaming as I basically just resigned myself to gaming on bootcamp. Tell me this, are the FPS the same now between the OS X vs the Windows counterpart? 2 years ago the windows version had 25-100% higher frame rates which would stress the GPU more as it's able to work harder instead of waiting to be fed by a slower OS X.

The frame rate is a bit slower on OS X, but, that is a compromise I am prepared to make as I just can't be bothered booting into Bootcamp. There are some DX features missing like tesselation, but they really aren't that big a deal in my opinion. I get perfectly acceptable fps @2560x1440 so that's ok.

I've even played a fair number of games ok in Parallels, e.g. Dishonored, Skyrim, etc.

 

[Gav Mack]

Don't have a clue.

Just checked with my 2012 rMBP with the 650M... was also on sleep overnight

and iStat reports temp at 38c

The retinas do have a lower baseline temperature, the 2011 mbp are a bit higher. But with the notebooks we are talking about much smaller powered silicon but this m295x is a 125w part and a much bigger die in terms of mm squared for the cooling system to get all that heat out of that chassis.

As I have always found Apple's thermal paste to be consistently inconsistent and you're going to find wildly variable results from user to user as all of the models do have. We will have to wait for the first brave/reckless/stupid owner willing to redo their own thermal paste and polished die plates to see just how much you can minimise the throttling on this chassis to get it completely optimal. I have done more than a few of the retina notebooks with good results for gamers who want the maximum GPU performance and minimised throttling out of their Mac.

But they are easy - this 2012 on 27 inch chassis isn't. Heat gun, adhesive ring required and a lot more effort than a rmbp. I suggest anyone willing to get an Apple experienced tech to help them and not dare contemplate it on their own!

 

[5iMacs]

can anyone post their i5, m290x experience regarding fan noise while gaming (preferably wow)?

since I now gave back my i7, m295x because of unbearable fan noise for every cpu/gpu intensive task but like the imac and screen, I'm considering going with the entry model.

I know that graphics performance is sufficient for wow thanks to barefeats but not sure about heat and fan noise (rpm)?

thanks for any input.

I have the i5/M290X combination.

Can't help you with wow, but subjectively even the maximum GPU heating situation is really not bad. For instance, running the Unigine Valley ExtremeHD preset (easily enough to throttle an M295X) for 45 minutes and then benchmarking I got fan speed ranging from 2004-2124 RPM (checked it 6 times throughout). It was different at different points in the benchmark but you could not hear it changing.

Temperature was 86-88C throughout.

 

[rainydays]

For instance, running the Unigine Valley ExtremeHD preset (easily enough to throttle an M295X) for 45 minutes and then benchmarking I got fan speed ranging from 2004-2124 RPM (checked it 6 times throughout). It was different at different points in the benchmark but you could not hear it changing.

Interesting! Would you mind running it in 1600x900 windowed? Same quality settings. I'm interested in how it compares to the others.

 

[habeebhashim]

The retinas do have a lower baseline temperature, the 2011 mbp are a bit higher. But with the notebooks we are talking about much smaller powered silicon but this m295x is a 125w part and a much bigger die in terms of mm squared for the cooling system to get all that heat out of that chassis.

As I have always found Apple's thermal paste to be consistently inconsistent and you're going to find wildly variable results from user to user as all of the models do have. We will have to wait for the first brave/reckless/stupid owner willing to redo their own thermal paste and polished die plates to see just how much you can minimise the throttling on this chassis to get it completely optimal. I have done more than a few of the retina notebooks with good results for gamers who want the maximum GPU performance and minimised throttling out of their Mac.

But they are easy - this 2012 on 27 inch chassis isn't. Heat gun, adhesive ring required and a lot more effort than a rmbp. I suggest anyone willing to get an Apple experienced tech to help them and not dare contemplate it on their own!

Yes, but does it not also mean that the temp iStat reports is ~10c higher? Why should it report GPU at 38c when ambient room temp is <24c. And this after sleeping for 8 hours.

Does this mean that the peak temp iStat and GPU-Z (under Win 7) reported when gaming is also higher?

----------

Is powernap on? That might be preventing your iMac from completely sleeping.

Powernap is on for the rMBP... but the rMBP wasn't plugged in. So that shouldn't matter.

But with the iMac, this mornings test involved shutting down the computer and cold booting after 8 hours. iStat reported temp on startup was 50c... when room temp was <24c (I had the aircon running last night at 23c)

----------

As I have always found Apple's thermal paste to be consistently inconsistent and you're going to find wildly variable results from user to user as all of the models do have.

This is also true. There was another poster who had trouble with his 5K iMacs fan. It came on a lot quicker and lot louder than what others reported. I think he got a replacement model and that didn't exhibit the same fan speed/noise.

 

[5iMacs]

Interesting! Would you mind running it in 1600x900 windowed? Same quality settings. I'm interested in how it compares to the others.

Here you go. Same fan and temps observed. Although since the average FPS speedup for games from the M290X to the M295X has been shown to be about 20%, and the Unigine Valley benchmark speedup was pretty negligible, so it's a little unsatisfying.

 

[rainydays]

Here you go. Same fan and temps observed. Although since the average FPS speedup for games from the M290X to the M295X has been shown to be about 20%, and the Unigine Valley benchmark speedup was pretty negligible, so it's a little unsatisfying.

Thanks a lot! Yeah there's barely any difference between the different iMac GPUs at those settings. That's really interesting. Especially since both 780M and 295x are supposed to be upgrades. I know, you can't base it all on one benchmark. Still interesting though.

GTX 780M - 34 fps
GTX 775M - 37 fps
R9 M295X - 37 fps
R9 M290X - 35 fps

 

[Gav Mack]

Yes, but does it not also mean that the temp iStat reports is ~10c higher? Why should it report GPU at 38c when ambient room temp is <24c. And this after sleeping for 8 hours.



Does this mean that the peak temp iStat and GPU-Z (under Win 7) reported when gaming is also higher?

----------





Powernap is on for the rMBP... but the rMBP wasn't plugged in. So that shouldn't matter.



But with the iMac, this mornings test involved shutting down the computer and cold booting after 8 hours. iStat reported temp on startup was 50c... when room temp was <24c (I had the aircon running last night at 23c)

----------





This is also true. There was another poster who had trouble with his 5K iMacs fan. It came on a lot quicker and lot louder than what others reported. I think he got a replacement model and that didn't exhibit the same fan speed/noise.

I have seen it setting up numbers of brand new identical Macintosh as well as servicing them. I would get the unit swapped if too hot and would certainly get AppleCare with it as a compulsory purchase. Cos they do not improve down the line more like the opposite.

 

[robotrootbeer]

Fair enough, but the only thing my wife uses her current iMac for is gaming. The arstechnica article already covered that gaming at full 5k resolution was out of the question, and given the reported CPU throttling at scaled down resolution it seemed prudent to get her a new standard iMac with the 780 GPU.

For someone who is ONLY gaming - the riMac is an INCREDIBLY poor choice.
5k gaming is going to be hard on even the most robustly built gaming rigs.

The point of the riMac is the beautiful 5k display - everything else is a cherry on top.

The enhanced GPU w/ extra vram is mainly to drive external monitors.

 

[Serban]

anyone tried battlefield 4 or COD advanced warfare? at high 1440p you get at least 40 fps?

 

[odedia]

Throttling is certainly a possibility, although I find it peculiar neither Barefeats or Arstechnica picked it up and mentioned it in their benchmarking. I wonder if there is a wide difference from machine to machine?

I wish Anandtech would pull finger and get a review out. Does anyone remember how long it took them to publish before Anand left? I saw one of their tweets (can't find it right now) that the iPad Air 2 review is out beginning of next week, and the riMac the week after but I don't know which GPU they got. If they didn't get the M295X I guess we will still be none the wiser.

I asked and they told me they are only reviewing the base model.