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

[vagos1103gr]

i did the gputest i choose 1440 with 8xx anti temps max 100, and i kept it for 20 min

 

[h9826790]

Here for you some test

Hi, do you mind to run OCCT GPU 3D test with your native screen resolution in full screen mode? Some of my friend is considering this RiMac, and would like to know more about it before pull the trigger. Thanks!

 

[ioxakoyba]

Here for you some test..

I would'n want to be spoiling your results, but running GPU test on 800 x 500 pixel resolution wouldn't necessarily push the GPU to it's limits, even the tester says that it's 100% usage of the GPU. Could you please repeat the test and run it with like 1080p or 1440p resolution at least, if not full 5k?

However, I must admit that your results are comparable proof, if they remain same on the bigger resolution, that this problem may not appear as bad as reported on all iMac 5k's and they may be something wrong on the some of the products. Btw, when did you buy this riMac? Is it very new? (I am asking because I am thinking could it be possible that Apple has already changed something in its production line, possibly more affective cooler on the newly purchased producst compared to products from the last Autumn).

However, I still would wait for the bigger resolution GPU tests. (If you cannot select higher resolution for the test, please download and run Furmark from http://www.ozone3d.net/benchmarks/fur/ - and if possible GPU-Z from http://www.techpowerup.com/gpuz/ and post "Sensors" page results after 30-60 minutes of running furmark. (When clicking on GPU Temperature you can select "max" to show the top GPU temperature it has reached during the test).

Then please post the sensors page from the GPU Z for us the see. I would really like to see if the throttling won't appear with your riMac if pushed Furmark at least for 30 minutes with at least 1080p resolution. (If this will happen, I am really considering changing my mind complitely and possibly purchasing this product after all.)

 

[andy9l]

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

Thanks Astelith!

So did you achieve this without manually controlling the fans? Also, same as above, any chance you could do 1440p resolution tests?

I appreciate your time.

 

[Astelith]

Thanks Astelith!

So did you achieve this without manually controlling the fans?

The fan is managed by the system, Os X and Windows

 

[RageDaddy]

The fan is managed by the system, Os X and Windows

********! PC Bios, Mac SMC. It should be clear to anyone who has at least some knowledge that M295X heat problems are related to voltage. Same goes for GPU cards (vBios). vBios has hardcoded vcore, mem and so and so on...vBios is going to need to be updated!

I bet my sweet 44 year old ass, that there is a SMC update on the way!

http://support.apple.com/en-us/HT202040

Someone with windows could use GPU-Z and get that M295X vBios out for closer looking?

6970M use v1.100 and that's a lot. I think i saw some iStat shots that GPU vCore was v1.17 or similar...and M295X should draw much less than 6970M?

http://www.techpowerup.com/vgabios/109915/apple-hd6970m-2048-110128.html

 

[andy9l]

********! PC Bios, Mac SMC. It should be clear to anyone who has at least some knowledge that M295X heat problems are related to voltage.


I bet my sweet 44 year old ass, that there is a SMC update on the way!

What could a SMC update achieve? Surely the only way to fix the extreme heat is to blast the fan more, or further throttle the M295X?

I'm genuinely asking, not trying to be smart.

 

[AlifTheUnseen]

Thanks Astelith!

So did you achieve this without manually controlling the fans? Also, same as above, any chance you could do 1440p resolution tests?

I appreciate your time.

Indeed, thanks Astelith, it's good to have these results.

Sadly, my confusion whether to buy it or not is more profound now, but the good thing is, it's based on more facts… ;-)

Any chance to redo the test in 1440p?

 

[RageDaddy]

What could a SMC update achieve? Surely the only way to fix the extreme heat is to blast the fan more, or further throttle the M295X?

I'm genuinely asking, not trying to be smart.

New vBios for M295X? Too bad that GPU-Z in windows won't show VDDC (voltage)..

 

[andy9l]

New vBios for M295X?

Sorry buddy - what could that mean to the end user? Same performance, less heat?

 

[Astelith]

2nd Round of tests + Forced Throttling Test:

You will see that the result at higher resolution are the same as the previous I did, what matters is GPU % + GPU Core Clock + CPU % (and Fan speed!)
The most important thing is in every test the Temp. was stabilized by the system to not exceed 100/101° C with the fan at 1800 rpm more or less, so not loud (at least to me), forcing the fan at 1200rpm at 105°C we start experiencing a light GPU Core fast Throttling, with also the CPU at 100% the system start the heavy GPU clock reduction going below 680MHz

OCTT: 4K Resolution - not able to push the GPU to the limit



FurMark: 4K Resolution - GPU Almost 100% (core 834 average), max Temp 100° C - Fan <1700 rpm



TessMark: 2880x1800 Resolution - GPU Almost 100% (core 847 average), max Temp 101° C - Fan <1900 rpm



PixMark: 2880x1800 Resolution - GPU 100% (core 850 straight), max Temp 99° C - Fan <1600 rpm



GilMark: 2880x1800 Resolution - GPU Almost 100% (core 834 average), max Temp 100° C - Fan <1740 rpm



Valley Bench: 2560x1600 Resolution - GPU not 100% (core 786 average), max Temp 100° C - Fan <1820 rpm



Throttling : Fan @ Manual 1200rpm - TessMark 2880x1800 - OCTT CPU - Max Temp 108°C - GPU Core dropping below 680, and probably more but I'm not going to fry my iMac for you

 

[RageDaddy]

And similar to M290X with windows laptop (7970M). v1.100 max, average v.1.007 and 78c. Can someone see where i am going here?

M295X, not showing volts at all..yet!

 

[Vapor matt]

Very good tests indeed, and like any GFX card mac or PC based as in this statement


********! PC Bios, Mac SMC. It should be clear to anyone who has at least some knowledge that M295X heat problems are related to voltage. Same goes for GPU cards (vBios). vBios has hardcoded vcore, mem and so and so on...vBios is going to need to be updated!

I bet my sweet 44 year old ass, that there is a SMC update on the way!

http://support.apple.com/en-us/HT202040

Someone with windows could use GPU-Z and get that M295X vBios out for closer looking?

6970M use v1.100 and that's a lot. I think i saw some iStat shots that GPU vCore was v1.17 or similar...and M295X should draw much less than 6970M?

http://www.techpowerup.com/vgabios/1...48-110128.html

GPU voltage and memory speeds are controlled by the 295m on board bios and as we used to do on pc gfx cards was to use an overlock bios to increase volts to increase core speed and heat, you can also reduce voltage and slack off the timings on the ram to reduce heat, the baseline is what voltage these cores need to be able to run 850mhz is the question.

also drivers are what push these cards, so driver improvements can also improve speeds and temps as we have seen on PC desktop cards.

those tests are pushing that card, driver and bios updates could improve it further yet, after all it was a brand new chip and hardly optimised yet.

 

[RageDaddy]

Compare to 780M:

8970M (pretty much the same as M290X)

 

[ioxakoyba]

Astelith, thank you very much for these tests and screenshots of them. They have been helpful.

So what I can see here is that at least with Astelith's iMac 5k the problem of the overheating does not emerge, at least not so clearly, than with the other cases we have seen on this forum. (Of course CPU going to 100-101C with full load of GPU is still troublesome and inconvenient, but it's much more better result than others reporting 105-110 C temperatures).

Also it seems that with Astelith the fan does not spin up to maximum when holding such temperatures ca. 100 C, but somewhat convenient 1800 rpm. The others, as far as I have understood, have reported fan spinning the maximum 2800 rpm and temperatures still rising above 105 C with full load.

I do not understand what could explain this difference. You are not running your iMac in a freezer are you? (However the ambient room temperature would be also interesting to hear).

Also, it is still unanswered, when did you bought this iMac? (This is important, because it may be possible that Apple have already changed something in the production line to address this issue or at least easen it.)

Astelith, is there possibility to request a one more test from you: Please run simultaniously Furmark with full resolution + stress CPU with Prime 95 torture test running all 8 threads. (You can download Prime95 from here: http://files.extremeoverclocking.com/file.php?f=103)

This is the case that puts the GPU on 100% load + CPU on 100% load. If the machine can hold temp above 105 C and not throttling much more, I believe Apple must have fixed something with the newer machines (in case Astelith's is one of them).

(Also I must point out that no game will stress machine so that CPU is constantly on 100% load, so this final test is more like the biggest torture a computer can ever get).

 

[Astelith]

Astelith, thank you very much for these tests and screenshots of them. They have been helpful.

So what I can see here is that at least with Astelith's iMac 5k the problem of the overheating does not emerge, at least not so clearly, than with the other cases we have seen on this forum. (Of course CPU going to 100-101C with full load of GPU is still troublesome and inconvenient, but it's much more better result than others reporting 105-110 C temperatures).

Also it seems that with Astelith the fan does not spin up to maximum when holding such temperatures ca. 100 C, but somewhat convenient 1800 rpm. The others, as far as I have understood, have reported fan spinning the maximum 2800 rpm and temperatures still rising above 105 C with full load.

I do not understand what could explain this difference. You are not running your iMac in a freezer are you? (However the ambient room temperature would be also interesting to hear).

Also, it is still unanswered, when did you bought this iMac? (This is important, because it may be possible that Apple have already changed something in the production line to address this issue or at least easen it.)

Astelith, is there possibility to request a one more test from you: Please run simultaniously Furmark with full resolution + stress CPU with Prime 95 torture test running all 8 threads. (You can download Prime95 from here: http://files.extremeoverclocking.com/file.php?f=103)

This is the case that puts the GPU on 100% load + CPU on 100% load. If the machine can hold temp above 105 C and not throttling much more, I believe Apple must have fixed something with the newer machines (in case Astelith's is one of them).

(Also I must point out that no game will stress machine so that CPU is constantly on 100% load, so this final test is more like the biggest torture a computer can ever get).

My iMac was built in November so one of the first release, so far I still waiting screenshot of a riMac going at 105 at full fan speed, and I truly believe that is impossible.
If you see all my test you can find the one at full load for CPU and GPU, in this case the system put the fan at maximum (2700) because the high CPU temp at almost 98, so the GPU, even in full load, goes at 85, showing a very good dissipation power.
I'm living in the south of Spain, 200 miles away from Africa, outside we have a wonderful sunny day with almost 20 degrees and the room temp not much lower than that

I've done a lot of tests showing what I'm saying, now it's time to see the same test, with also fan speed, of all those "faulty" iMac, if you can't provide any, at least please stop writing BS

And regarding driver updates, smc update and so on, I don't think we will have any, or with some game changing improvements, the machine is perfectly working, the GPU was made to operate at this temps, the fan start spin up at 95 degrees for a reason.

Facts guys, show me a proof, or this thread is going to be ridiculously useless.

 

[RageDaddy]

So what your saying is absolut truth and these guys are all wrong, cause your always right, right?

https://forums.macrumors.com/threads/1835628/

Like here:

Where is the fullscreen and 8XAA shots? Same goes for the rest of your stuff?

Guys, remember to turn on the fans


Fans in Auto ( Macs Fan Control) Max 101°C

[url=http://s13.postimg.org/s9nxkj6bb/Valley_Fan_Auto.jpg]Image[/url]

Fans at Max, 91/92° C

[url=http://s24.postimg.org/qa5j51w5h/Valley_Fan_Max.jpg]Image[/url]

Fans at min, 106°C + GPU Throttling @ 6xx MHz

[url=http://s18.postimg.org/96urnu0mx/Valley_Fan_Min.jpg]Image[/url]

 

[Astelith]

So what your saying is absolut truth and these guys are all wrong, cause your always right, right?

https://forums.macrumors.com/threads/1835628/

Like here:

Where is the fullscreen and 8XAA shots? Same goes for the rest of your stuff?

I made tests with screenshot, so yes, I'm right, until somebody will post another test showing a retina iMac going above 105° C and throttling with fan speed at maximum.

Where is "all these guys" ? I just need only one screenshot, is not that hard, isn't it ?

Fullscreen or not doesn't change anything.

 

[RageDaddy]

But your screenshots are in windowed mode and without 8XAA! Fullscreen and 8XAA like the rest of us.

Get my point at last?

AA is very demanding and you dont use it all? Then you whine when others have 103c and you dont? Furmark in windowed mode?

Now this is where you come close, add fullscreen and voila....102c with 2300rpm like the rest. Care to test and show a fullscreen shot?

Here you go (Mac):

http://support.apple.com/en-us/HT201361

And this? Taken after the run, it cools down in seconds and still 97c and 1616rpm? Get a shot on 17/18 part of the test, fullscreen.

In Os X was even better, 99° average with 1700 rpm

[url=http://s1.postimg.org/7drz505lr/valley_bench.png]Image[/url]

Your stuff basically sucks hairy balls...

Valley Bench: 2560x1600 Resolution - GPU not 100% (core 786 average), max Temp 100° C - Fan <1820 rpm

Throttling at 100c to 784c and afterburner shows 334mhz clocks many times, so that's that. 850mhz is max and it never ever even got there? You wanna now why? It doesn't have to..windowed mode...so you are running tests with lower core clock/vcore (784/334 vs 850 with fullscreen/full vcore like the rest of us).

Now, who's talking BS

 

[Astelith]

But your screenshots are in windowed mode and without 8XAA! Fullscreen and 8XAA like the rest of us.

Get my point at last?

AA is very demanding and you dont use it all? Then you whine when others have 103c and you dont? Furmark in windowed mode?

Now this is where you come close, add fullscreen and voila....102c with 2300rpm like the rest. Care to test and show a fullscreen shot?

Here you go (Mac):

http://support.apple.com/en-us/HT201361

Image

And this? Taken after the run, it cools down in seconds and still 97c and 1616rpm? Get a shot on 17/18 part of the test, fullscreen.



Your stuff basically sucks hairy balls...



Throttling at 100c to 784c and afterburner shows 334mhz clocks many times, so that's that. 850mhz is max and it never ever even got there? You wanna now why? It doesn't have to..windowed mode...so you are running tests with lower core clock/vcore (784/334 vs 850 with fullscreen/full vcore like the rest of us).

Now, who's talking BS

Enjoy:

 

[RageDaddy]

And back to the subject in hand...

Remember that old 6970M took v.1.100 max and it's made 40 nm process. Now take a look at this, just a few pages back:

v.1.13 and M295X is made 28nm process.

Nvidia explains it like this:

 

[joema2]

...yes, I'm right, until somebody will post another test showing a retina iMac going above 105° C and throttling with fan speed at maximum...Where is "all these guys" ? I just need only one screenshot, is not that hard, isn't it ?

I think many of those were reporting the indicated temp from *within* Valley itself -- not a consistent, independent temp tool. On my 2013 iMac the Valley benchmark reports temps about 8-10C higher than iStat Menus GPU die temp. I don't know if this behavior is common across all iMacs or not.

It would be more consistent if all testers used exactly the same method of reporting temps, else you could be chasing a difference in instrumentation.

 

[carlob]

If you run Valley fullscreen there is no way to check temperature precisely with istat. In the time needed to switch to istat and check temperature, even if you are very quick, gpu has cooled down a couple of degrees. Maybe running it windowed at full screen.

 

[RageDaddy]

I am still keeping firmly on my theory. M295X is operating with too high vCore at the moment.

Hope i can shine here someday if they fix this and it actually is hot for that reason.



----------

If you run Valley fullscreen there is no way to check temperature precisely with istat. In the time needed to switch to istat and check temperature, even if you are very quick, gpu has cooled down a couple of degrees. Maybe running it windowed at full screen.

I did it 4 times to get it right.

Set a mark, lets say 2700 frame. Get a shot at fullscreen in 2700 framish, near the end. Do another test and when in 2700 frame, get the rpm for macs fan control or istat or what ever.

Now, you got two shots, for two different runs...but in fullscreen and rpm's are about there where they really are. Temps, frames etc are from the same test (from a fullscreen shot taken), but the final result is from another run. So, both runs are done in fullscreen from start to finnish without alt+tab desktop stuff or taking shots..

----------

Enjoy:

[url=http://s3.postimg.org/v2n4yjek3/Valley_Fulscreen_2560x1440.jpg]Image[/url]

[url=http://s22.postimg.org/tpdfpzr6p/Valley_Fulscreen_4_K.jpg]Image[/url]

And where might be this fullscreen shot that we are all talking about? Why windows? I dont use windows. Macs fan control shows 1205rpm with 56c GPU? That's not taken in the middle of the run? So 4k resolution keeps your GPU cooler than 1440p windowed run earlier? Pretty much BS.

Screen shots after screen shots, but the one that counts is missing? I wonder why...

 

[carlob]

Please don't post screenshots with the 290 that are going to further confuse ideas (that are already confused), we are debating the 295 gpu. The 290 has no "issues" as far as I know.