Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Video editing includes playing back several simultaneous local videos, yes. What normal video editing does not include is playing 6 4K flash videos. There's an important distinction there, as playing flash videos uses significantly more CPU power than playing local files does.

doesn't Final Cut pro do that? I may take that back, playing back 6 4k streams isn't anything a video editor wouldn't do.
 
Hey guys. I can confirm that throttling CAN happen, albeit a very minor amount. I have a late 2013 rMBP 2.3/16/512/750M.

Procedure:
1. Ran yes > /dev//null & eight times, one after another in one window.
2. Ran pmset -g thermlog in the same terminal window.
3. Set gfxCardStatus V2.3 for the Intel iGPU
4. Ran the Intel Power Gadget 2.5.2 (Good find btw! Didn't know such a tool existed!)
5. Ran the Heaven Unigine on Preset Extreme. Downloaded from here (http://unigine.com/products/heaven/download/).

Screenshot 2013-12-28 09.09.07 copy.jpg
CPU_Speed_Limit reaches as low as 60. Based off the calculations given by other posters, my Turbo Boost should be 3.5 GHz. 3.5 GHz * 0.60 =2.1 GHz. Results from both Terminal and the intel Power Gadget show a CPU of 2.1 GHz, where as the CPU should be running at 2.3 Ghz. Now, the question is, is this a thermal limit or wattage draw limit. Not a physicist or engineer so I'll let you guys figure that out. Have not changed the thermal past yet but hopefully this should help me determine whether or not it'll improve the longevity and/or performance of my machine. Hope my data helps! Let me know if you have any questions about my methods and procedures.
 
Video editing includes playing back several simultaneous local videos, yes. What normal video editing does not include is playing 6 4K flash videos. There's an important distinction there, as playing flash videos uses significantly more CPU power than playing local files does.

No not flash but with filters, it can really intense; dependent on what kind of filters and how many. While not the same exact sort of thing someone will do, it's similar enough that I would recommend it as a suitable benchmark for our purposes. BUT, I may have a found a possible way to throttle the CPU by utilizing the iGPU in the post just prior to this.
 
No not flash but with filters, it can really intense; dependent on what kind of filters and how many. While not the same exact sort of thing someone will do, it's similar enough that I would recommend it as a suitable benchmark for our purposes. BUT, I may have a found a possible way to throttle the CPU by utilizing the iGPU in the post just prior to this.

OK, but again, the procedure you outlined above is something that video editors, photo editors and any other person who uses this machine on a professional basis are unlikely to do on a daily basis. So it’s not useful in telling us whether or not the MacBook Pro throttles or overheats in normal usage, making it unsuitable for professional work/gaming.

What you have proven is that the MacBook Pro throttles at some point. But so do all computers. Unless there is a way to show that the MacBook Pro throttles more quickly than similarly specced laptops do under the same load, then that information is not very useful.
 
Now, the question is, is this a thermal limit or wattage draw limit. Not a physicist or engineer so I'll let you guys figure that out. Have not changed the thermal past yet but hopefully this should help me determine whether or not it'll improve the longevity and/or performance of my machine. Hope my data helps! Let me know if you have any questions about my methods and procedures.

Both.

I can just peg 60 watts (CPU only) on my MacBook Pro - if that runs for more than a few seconds, the system automatically reduces maximum frequency to 2.8Ghz. This is due to wattage draw, as the thermal subsystem still reports 100% CPU_Limit, only now it is 2.8Ghz max as opposed to 3.3Ghz on my system.

Running 8 yes > dev/null gives me a 97-94% speed limit, so my system holds steady at 3.1 Ghz due to thermal load, until Apple's power system limits my frequency to 2.8Ghz due to wattage concerns.

----------

What you have proven is that the MacBook Pro throttles at some point. But so do all computers. Unless there is a way to show that the MacBook Pro throttles more quickly than similarly specced laptops do under the same load, then that information is not very useful.

Mine never throttles below the stated 2.5Ghz specification. Amazing what proper thermal control can do for a system ;)
 
Mine never throttles below the stated 2.5Ghz specification. Amazing what proper thermal control can do for a system ;)

Your results would only be comparable with calviin's results if your computer were similarly specced to the late 2013 MacBook Pro. Your late 2011 cMBP running at 2.8 GHz is vastly underpowered compared to a late 2013 rMBP with 2GB VRAM running at 2.1 GHz.
 
Your results would only be comparable with calviin's results if your computer were similarly specced to the late 2013 MacBook Pro. Your late 2011 cMBP running at 2.8 GHz is vastly underpowered compared to a late 2013 rMBP with 2GB VRAM running at 2.1 GHz.

Are you stating that you believe a Haswell 2.1 Ghz system to be faster than a 2.8 Ghz Sandy Bridge?

Sandy / Ivy / Haswell Clock for Clock performance

Haswell is, on average in that comparison, ~13% faster than Sandy Bridge clock for clock. Therefore, a 33% faster clocked Sandy Bridge IS faster than Haswell, by approximately 20%.

Now, the GPU (750m vs HD6770M) is a different story.

'Vastly underpowered'? I think not.
 
The overrall system, including PCIe SSDs, faster GPU and faster CPU, is faster than your machine at most tasks -- even at lower clock speeds -- up to a point.

It would be great to have someone who has both systems run an encoding test under the same conditions to see whether or not throttling makes the late 2013 rMBP slower than other machines at certain tasks under normal usage.

Are you stating that you believe a Haswell 2.1 Ghz system to be faster than a 2.8 Ghz Sandy Bridge?



Sandy / Ivy / Haswell Clock for Clock performance



Haswell is, on average in that comparison, ~13% faster than Sandy Bridge clock for clock. Therefore, a 33% faster clocked Sandy Bridge IS faster than Haswell, by approximately 20%.



Now, the GPU (750m vs HD6770M) is a different story.



'Vastly underpowered'? I think not.
 
What temperatures were you seeing when you were running your machine at max. load?

Roughly 102-103 degrees Celcius. You can download the Intel PowerGadget logs in the links I submitted above to get a better idea. You'll see that the temps fluctuate a bit.
 
The overrall system, including PCIe SSDs, faster GPU and faster CPU, is faster than your machine at most tasks -- even at lower clock speeds -- up to a point.

It would be great to have someone who has both systems run an encoding test under the same conditions to see whether or not throttling makes the late 2013 rMBP slower than other machines at certain tasks under normal usage.

I'm running 6Gbps SSD (read/write of ~500MB/s).

Someone set up a standard video for us to test. Now I'm interested to see the results ;)

----------

Roughly 102-103 degrees Celcius. You can download the Intel PowerGadget logs in the links I submitted above to get a better idea. You'll see that the temps fluctuate a bit.

Holy crap, way too hot. I would very highly recommend you either a) have a Genius handle this (warranty) or b) go in and lap/re-apply a proper heatsink compound.

 
Hey guys. I can confirm that throttling CAN happen, albeit a very minor amount. I have a late 2013 rMBP 2.3/16/512/750M.

Procedure:
1. Ran yes > /dev//null & eight times, one after another in one window.
2. Ran pmset -g thermlog in the same terminal window.
3. Set gfxCardStatus V2.3 for the Intel iGPU
4. Ran the Intel Power Gadget 2.5.2 (Good find btw! Didn't know such a tool existed!)
5. Ran the Heaven Unigine on Preset Extreme. Downloaded from here (http://unigine.com/products/heaven/download/).

View attachment 453524
CPU_Speed_Limit reaches as low as 60. Based off the calculations given by other posters, my Turbo Boost should be 3.5 GHz. 3.5 GHz * 0.60 =2.1 GHz. Results from both Terminal and the intel Power Gadget show a CPU of 2.1 GHz, where as the CPU should be running at 2.3 Ghz. Now, the question is, is this a thermal limit or wattage draw limit. Not a physicist or engineer so I'll let you guys figure that out. Have not changed the thermal past yet but hopefully this should help me determine whether or not it'll improve the longevity and/or performance of my machine. Hope my data helps! Let me know if you have any questions about my methods and procedures.

There are limits but we don't know them exactly unless you were to run Throttle Stop under windows to check.
 
Do you have before and after stats?

Prior to performing the 'surgery' I had temps of ~208F (98C), and zero turbo boost.

Now I have temps of ~190F (88C) and 100% Turbo Boost (3.3Ghz on all 4 cores)

Video:
 
Sorry to bring back this old thread (my first post, don't shoot).

Would a kind soul please tell me what tool can I use to log the temperature to a file, besides frequency? I noticed the OP (theSeb) used pmset -g thermlog outputs the frequency (well, converted). Can the Intel Power Gadget monitor and also log the temperature on Mac OS? I've only tested it under windows.

I don't own a Mac but a friend abroad has the same MBP as the theSeb's and is willing to run a few tests for me to find out if it's going to throttle under my workload (maths simulations), so I need to tell him what to run as he's far from being an expert.

I'll change the TIM (with coolab liquid ultra) and possibly lap the heatsink too, regardless of the laptop I get. I'm amazed at how poor this step is implemented by manufacturers in this era ... Apple is not the only one pouring buckets of TIM between the heatsink and CPU core.

Thanks in advance,
Alex.
 
Last edited:
You mention you don't have a Mac, so perhaps you're new to Apple.

If so, be aware that you can return a Mac to the store within 14 days.

No need to get someone else to benchmark; 14 days should be enough for you to convince yourself it's workable or not for your workload/workflow.
 
You mention you don't have a Mac, so perhaps you're new to Apple.

If so, be aware that you can return a Mac to the store within 14 days.

No need to get someone else to benchmark; 14 days should be enough for you to convince yourself it's workable or not for your workload/workflow.

I appreciate the alternative, though it's not what I asked. An answer to my questions will be much appreciated.

Regarding your suggestion (which is good to know), I'd like to avoid the hassle of buying it, returning it and risking them not taking it back if I happen to drop/scratch it, or whatever else. Also, if I decide on the MBP then I'll buy it when I travel to the US later this year as it's so much cheaper than in the UK. My friend can run the tests for me and then send me the freq and temp log files, which is enough to get a good idea.

Alex.
 
That response is a fast route to not getting any help.

Odd. I wasn't trying to be snappy or rude. My apologies if it seemed that way. I actually appreciated the info as I didn't know about it, but am interested in a reply to the question I asked (which was purely technical).
 
Odd. I wasn't trying to be snappy or rude. My apologies if it seemed that way. I actually appreciated the info as I didn't know about it, but am interested in a reply to the question I asked (which was purely technical).

Imagine standing in front of the poster and saying what you said, I can't imagine it not coming over as rude and arrogant. This isn't apple support (or any other support structure), where you can expect a precise answer to your query as you have somehow paid for it via the product.

As it happens I think he is correct, you would get a much better idea of your workload running on a Mac if you were to run it on a Mac, saying that requires some effort on your part doesn't detract from the fact that it is a better quality solution.

What will you do if you buy and it doesn't run your workload as you expect? Berate your friend?
 
Imagine standing in front of the poster and saying what you said, I can't imagine it not coming over as rude and arrogant. This isn't apple support (or any other support structure), where you can expect a precise answer to your query as you have somehow paid for it via the product.

As it happens I think he is correct, you would get a much better idea of your workload running on a Mac if you were to run it on a Mac, saying that requires some effort on your part doesn't detract from the fact that it is a better quality solution.

What will you do if you buy and it doesn't run your workload as you expect? Berate your friend?
I had zero intentions to be cocky or demanding, quite the contrary. I believe I already apologized.

I really do appreciate the alternative/better solutions, but I am honestly more curious to find out what tool was used to log the temperature to a file.

(people don't berate their friends after asking for a favour, well at least I don't, but all that is way beyond the point)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.