Automated tool to reveal throttling and overheating - GitHub

[mamarrash]

That being said, the amount of throttling in the GpuTest test with the Iris Pro is still a bit much (compare to my 2.3 GHz for instance). Try running the GpuTest test with the Nvidia GPU, and have it run for longer than 10 minutes. See if it throttles then.

@SVTmaniac I would completely agree. Thankfully I didn't pay a penny for my upgrade otherwise I would be a bit frustrated!

I have attached my 10 minute nVidia results. No throttling

You should try and do the GPUtest for more than 10 minutes. Both GPU test (Nvidia and integrated)

 

[hexxed]

I clearly can't read, ill do them for longer than 10 minutes /facepalm.

 

[mamarrash]

What I've gotten from all this so far is that the 2.6 may not really be worth the price increase. When it matters it runs at about the same speed as a 2.3 does.

You´ve got a point. You know you´re paying a premium basically for an overclock chip.

The 2.6 Ghz does throttle in Gputest with Iris Pro (at least with the results here) The 2.3 not so much.

2.6 Ghz GPUtest Iris Pro results: 2.8 CPU frequency (steady. No throttling)
2.3 Ghz GPUtest Iris Pro results: 3.3 CPU frequency (steady. No throttling)

here the 2.3 performance is better. By 0.5 Ghz .

Now:

2.6 Ghz GPUtest NVidia results: 3.8 CPU frequency (steady. No throttling)
2.3 Ghz GPUtest Nvidia results: 3.3 CPU frequency (steady. No throttling)

Now the 2.6 performs better. By 0.5 Ghz. But if you see in the test, and i checked myself when running it, the laptop runs a little bit cooler.

Here the throttling occurs due to GPU stress, not CPU stress. It´s just overheating prevention that not allows the CPU to turbo its speed.

In the CPU tests, basically (and to my disappointment) they perform about the same.

 

[hexxed]

Ok here are my longer tests. 20 minutes each on the nVidia 750 and Iris Pro graphics cards.

Room temperature of 20 degrees centigrade.

No throttling on the nVidia test however the CPU does seem limited when using the Iris Pro.

 

[graemenail]

@graemenail: I'm glad to see that you're on your way to fixing the problem. Let me know if Apple sorts it.

So similar to mine then. Well at least they've diagnosed and repairs are proposed, so that's great.
Mine failed the re-paste twice which was when they decided to replace the logic board and the heatsink. Maybe yours will be the same. Whatever, as long as the repairs make the unit pass the test that's fine.
Another result for qnxor's little gizmo

Apologies for the delay, one replaced logic board later and here are my new results. Substantial improvements, even on a hot day! Thanks again!

 

[Quackers]

Much better graemenail, though the overall temp gets a little high for my taste.
Glad they stepped up for you.

 

[qnxor]

Apologies for the delay, one replaced logic board later and here are my new results. Substantial improvements, even on a hot day! Thanks again!

That's much better. Well done! I can definitely now say I haven't wasted my time coding this little tool. Helped revigorate two laptops so far . 105C max is usual for i7-3xxx, so don't worry.

Was thinking at one point to have it upload the results to some public website to build a public database (similar to other benchmarking tools). It would paint a better picture of Apple machines and the Intel chips used. If there is interest and I find time then I'll give that a go.

On the 2.3 vs 2.6 GHz performance. So far I too noticed that wrt CPU performance, when pushed, they either perform the same or the 2.6 GHz gets slightly worse. So far I'm glad I went for the 2.3 GHz as I actually push mine quite a lot and often in Matlab.

 

[romesk90]

Below are my results:

2012 Macbook Pro
2.9ghz Intel i7
8gb ram
750gb HDD
OSX 10.9.2

Ambiant Air was about 40 Celsius.

Pictures of graphs are attached.




===== issue running program was fixed, left here for anyone else =====
When I run the program (either direct command line from examples, or from the old school menu) I get this:

Fectching ImageMagick into /Users/****/macoh/bin ...
#######
tar: Unrecognized archive format
tar: Error exit delayed from previous errors.

edit: scratch that, i had to change the url_im in the conf file. i changed the darwin13.1.0 to 13.2.0 and it worked

=== end of issue (now fixed) ===

 

[qnxor]

@romesk90: ambient 40 Celsius? Were you at the beach in direct sun? Results look good.

About the imagemagik issue, that's odd ... v13.1 should work just fine on 10.9.x, like it did for all of us. It sounds like there was a temporary glitch in the download and the archive was corrupted.

 

[romesk90]

@romesk90: ambient 40 Celsius? Were you at the beach in direct sun? Results look good.

That's what bresink and istats is giving me. It seems high to me...I'm in my room and it is not warm lol

 

[qnxor]

That's what bresink and istats is giving me. It seems high to me...I'm in my room and it is not warm lol

Ambient temperature means your room tempetature, not the idle CPU temperature (which is what those tools measure)

 

[romesk90]

Ambient temperature means your room tempetature, not the idle CPU temperature (which is what those tools measure)

Well..that explains that lol thanks!

 

[Maximus434]

Just to add my own results on the new 2014 MBPr with upgraded 2.8GHz Core i7:

Heavy throttling as expected but I am surprised to see it throttle below stock speed to between 2.5 and 2.6 GHz....

On a side note: Very good program Bogdan

 

[yjchua95]

Just to add my own results on the new 2014 MBPr with upgraded 2.8GHz Core i7:

Heavy throttling as expected but I am surprised to see it throttle below stock speed to between 2.5 and 2.6 GHz....

Image

On a side note: Very good program Bogdan

Were you using the Iris Pro or NVIDIA during the tests?

On my 2.6/16/1TB/750M machine, I find that it throttles in Iris Pro, but not in NVIDIA.

 

[Maximus434]

Were you using the Iris Pro or NVIDIA during the tests?

On my 2.6/16/1TB/750M machine, I find that it throttles in Iris Pro, but not in NVIDIA.

I tried it with both, getting exactly the same result. Interestingly, I also run the BOINC distributed computing app (for Einstein@Home) and tracked that with the intel power gadget. That doesn't seem to throttle the CPU and it stays at between 3.0 and 3.1 GHz with all 8 threads active....

Anyone else with the 2.8GHz Core i7 who can try this tool?

 

[ha1o2surfer]

Were you using the Iris Pro or NVIDIA during the tests?

On my 2.6/16/1TB/750M machine, I find that it throttles in Iris Pro, but not in NVIDIA.

Intel uses a power balance to give power to the GPU and CPU. Under full CPU and iGPU load, the CPU can NOT reach full turbo and this appears to be throttling when in fact it's not.

This is totally normal. Even with my unlocked TDP set to 85 watts, my 3840QM will not reach above 3ghz when my intel HD 4000 is under full load.

 

[Diversion]

Intel uses a power balance to give power to the GPU and CPU. Under full CPU and iGPU load, the CPU can NOT reach full turbo and this appears to be throttling when in fact it's not.

This is totally normal. Even with my unlocked TDP set to 85 watts, my 3840QM will not reach above 3ghz when my intel HD 4000 is under full load.

Yep, you can see this physically happening under Windows with Intel XTU.. It will show Power Limit Throttling = 100% typically..

Only a few laptops in existence that can run without starving for power. Mostly big 10+ lb gaming monster machines with 300+ watt supplies

Prime95 will throttle pretty much any laptop.

 

[ha1o2surfer]

Yep, you can see this physically happening under Windows with Intel XTU.. It will show Power Limit Throttling = 100% typically..

Only a few laptops in existence that can run without starving for power. Mostly big 10+ lb gaming monster machines with 300+ watt supplies

Prime95 will throttle pretty much any laptop.

In the Mac world you are right, I do not throttle under a Prime95 load and a FurMark load. It's all about what the laptop is made for.. I even have my CPU overclocked and it draws 69 watts under full load... off topic but just wanted to put that out there. It's a 14 inch G46VW haha

 

[qnxor]

@Maximus434 has tested Prime95 which is a CPU only test, so all the subsequent discussion about nvidia vs intel graphics does not apply in his case.

It throttles quite a lot but I wouldn't be worried about Prime95 (read my first posts to understand why). The more relevant CPU test is the x264 test. Try running that instead of Prime95 in order to spot any relevant throttling or overheating.

Bogdan.

 

[yjchua95]

@Maximus434 has tested Prime95 which is a CPU only test, so all the subsequent discussion about nvidia vs intel graphics does not apply in his case.

It throttles quite a lot but I wouldn't be worried about Prime95 (read my first posts to understand why). The more relevant CPU test is the x264 test. Try running that instead of Prime95 in order to spot any relevant throttling or overheating.

Bogdan.

Are you the same Dr Bogdan Ramon at Cambridge University's Department of Applied Maths and Theoretical Physics?

 

[qnxor]

Are you the same Dr Bogdan Ramon at Cambridge University's Department of Applied Maths and Theoretical Physics?

If you meant "Roman" (don't worry, I've been called worse), then that's what my signature points to in the first post of this thread. Why do you ask? Have I directly affected you at cambridge? If yes, particularly if it was bad, then by all means do let me know (privately if possible) ... it's hard to get my students to do that

 

[yjchua95]

If you meant "Roman" (don't worry, I've been called worse), then that's what my signature points to in the first post of this thread. Why do you ask? Have I directly affected you at cambridge? If yes, particularly if it was bad, then by all means do let me know (privately if possible) ... it's hard to get my students to do that

Oh no, I'm a computer science student in the same university, except that I'm currently on an exchange program and will be at Melbourne University for quite a while.

That said, once I return to Cambridge, it'll be great to meet you (if you have the time!)

 

[Maximus434]

@Maximus434 has tested Prime95 which is a CPU only test, so all the subsequent discussion about nvidia vs intel graphics does not apply in his case.

It throttles quite a lot but I wouldn't be worried about Prime95 (read my first posts to understand why). The more relevant CPU test is the x264 test. Try running that instead of Prime95 in order to spot any relevant throttling or overheating.

Bogdan.

Well I just did the x264 encode:

Fairly happy overall - 3.0-3.1Ghz. Scratch that - I see hexxed's 2.6Ghz 2013 machine (previous page) is actually beating this in GHz and FPS. Clearly it's throttling - is it something to be concerned about?


Just curious as to why you think the Prime95 test is not indicative of an realistic scenario? Does it stress CPU cores a different way to other tasks such as encoding (BSc in Computing student here!)?

 

[qnxor]

@maximus, please see some of my first posts where I gave some explanations on Prime95. In short, the stress levels of Prime95 is almost never reached in a normal real-world scenario. Unless you run specialized computing apps on your Mac that can reach that kind of stress, you shouldn't worry about it. Your cpu throttles quite a bit in x264. To an extent, it's expected (that cpu is practically an overclocked version to begin with), but I would be somewhat unhappy. My 4850HQ, which is suposedly quite a bit slower than yours, keeps a steady 3.3 GHz in x264 ... it has been said in this thread that these high clocked CPUs are simply not worth the money as they end up performing worse than their cheaper counterparts when it matters. Apple does cash in on the "many GHz" illusion though

@yjchua, I also have an affiliation at the Computer Lab as you can see on my webpage (been with the CL since 2006). Let me know once you get here and we can meet, sure.

 

[ha1o2surfer]

@maximus, please see some of my first posts where I gave some explanations on Prime95. In short, the stress levels of Prime95 is almost never reached in a normal real-world scenario. Unless you run specialized computing apps on your Mac that can reach that kind of stress, you shouldn't worry about it. Your cpu throttles quite a bit in x264. To an extent, it's expected (that cpu is practically an overclocked version to begin with), but I would be somewhat unhappy. My 4850HQ, which is suposedly quite a bit slower than yours, keeps a steady 3.3 GHz in x264 ... it has been said in this thread that these high clocked CPUs are simply not worth the money as they end up performing worse than their cheaper counterparts when it matters. Apple does cash in on the "many GHz" illusion though

@yjchua, I also have an affiliation at the Computer Lab as you can see on my webpage (been with the CL since 2006). Let me know once you get here and we can meet, sure.

I do agree with you on the GHz illusion when talking about Apple. In Luxmark (again off top sorry) no macbook pro to date can beat my G46VW with a 3840QM/HD4000/GTX660m due to power and temp throttling. I own a 2013 Macbook pro that scores 536 while using all OpenCL CPUs and GPUs where as my G46VW gets above 1000 using the sala scence. 1041 to be exact.

As far as again with the Intel Power balance again. I can't find any papers on the TDP budget but the creator of ThrottleStop states "The Intel Power Balance feature lets you decide whether the Intel CPU or Intel GPU should get the biggest share of the TDP budget. You can't get full CPU turbo and full GPU clocks while both are being heavily utilized no matter how good your cooling system is/what the cTDP you have your CPU set to.

I like to make sure even a non real world scenario can be sustained as I run BOINC for 24/7 on this laptop. I read your post so I shouldn't have brought it up. Just wanted to share some facts with that poster.