Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Intenditore

macrumors member
Original poster
Aug 25, 2018
86
8
Hi!
I've got A1398 Late 2013, I7 4960HQ and discrete GPU. It's overheating and can't boost higher than 2.8ghz despite it should go up to about 3.4. It overheats, despite the fresh thermal paste. I even changed the heatsink and it caused NO CHANGES.
Here are the temps (with old heatsink).
1. Before CB
fuOLl6WUlDkYrIBw.huge

2. While second run launched instantly after the first
VhJYXsFWNSyAcSKC.huge


Now after I installed the new one It's even worse for some reason, idle cpu is about 65, CPU PCIE Proximity is 70 constantly! Airflow 38 in the same time. What the heck is that??
Thermal paste is Cooler master V1 IC value. I apply it a proper amount, I screw it properly, no way I could screw it up.
I found some similar things in one place only, but there's no solution fixing-a-boiling-hot-macbook-retina-2015
Can anybody suggest me what could it be please?
 

Intenditore

macrumors member
Original poster
Aug 25, 2018
86
8
I applied liquid metal to CPU - 0 results. It throttles on 3.0 ghz now even with fans maxed out
What is this?
 

Webster's Mac

macrumors 6502
Dec 18, 2016
341
284
Boost clock is exactly as it sounds...a boost. Do turbo charged cars constantly engage the turbo charger? No. Same applies here. And it's even worse on MacBook Pros because they have pretty skimpy coolers compared to some laptops. You generally have to choose between thin and light and sustained performance. I'd be more concerned if it went under the base clock under load. Under full load with fans maxed out, it should be maintaining at least base clock, and ideally still be boosting some. My Mid-2015 doesn't stay on its max boost clock for very long, and under sustained loads, tends to stay 100-300mhz above base clock in a cool environment with the dedicated GPU engaged as well. This is after the repast. These heatsinks are razor thin. They just don't have the thermal capacity to handle the CPU and GPU at full load *and* at full boost clock for very long.
 

Intenditore

macrumors member
Original poster
Aug 25, 2018
86
8
I get this point, but I compare my experience to what other users have. And even when I start Cinebench cold with fans blasting out at max, I get a 20% lower score then everybody else. And this is clearly a sign of some kind of malfunction I can't understand
 

Webster's Mac

macrumors 6502
Dec 18, 2016
341
284
I get this point, but I compare my experience to what other users have. And even when I start Cinebench cold with fans blasting out at max, I get a 20% lower score then everybody else. And this is clearly a sign of some kind of malfunction I can't understand
Are you manually controlling the fans? I've got a friend of mine who was doing that as well and was having slowness issues. Told him to uninstall any fan control software, reset the SMC, and then use it. No complaints since. Apple integrates the cooling system tightly with macOS, as well as with the CPU/GPU. If you mess with one of those, the others are thrown out of whack.

Try completely uninstalling any fan control software, and reset the SMC (power off, hold down Shift + Control + Option + Power at the same time for 10 seconds, then release all keys and power on the machine) Reset PRAM for the hell of it as well, just to rule that out (Command + Option + P + R). Run the benchmark again and let me know if that helps.
 

Intenditore

macrumors member
Original poster
Aug 25, 2018
86
8
Hm.. :cool:
I tried both automatic control (not through the software) and maxing it out with FanControl.
Ok, I will try it though
 

Webster's Mac

macrumors 6502
Dec 18, 2016
341
284
Hm.. :cool:
I tried both automatic control (not through the software) and maxing it out with FanControl.
Ok, I will try it though
MacsFanControl is still controlling the fans even if it's on automatic...atleast that's my understanding. Give it a try without MacsFanControl and do the resets I mentioned. Hopefully that helps!
 
  • Like
Reactions: Intenditore

Intenditore

macrumors member
Original poster
Aug 25, 2018
86
8
What I found out:
In Windows 10 it's FASTER THAN IT SHOULD BE! 665 CB instead of 640 intended. Probably, because of LM on CPU
But it still throttles way lower in ANY MacOS, Mojave nor High Sierra, with or without custom fan control, on automatic nor full fan blast.
I've ran ASD testing and everything is fine
What da heck?
 

Webster's Mac

macrumors 6502
Dec 18, 2016
341
284
What I found out:
In Windows 10 it's FASTER THAN IT SHOULD BE! 665 CB instead of 640 intended. Probably, because of LM on CPU
But it still throttles way lower in ANY MacOS, Mojave nor High Sierra, with or without custom fan control, on automatic nor full fan blast.
I've ran ASD testing and everything is fine
What da heck?
When you say “with or without fan control” did you fully uninstall the fan control software? And did you reset the SMC and PRAM and see how it does
 

Intenditore

macrumors member
Original poster
Aug 25, 2018
86
8
When you say “with or without fan control” did you fully uninstall the fan control software? And did you reset the SMC and PRAM and see how it does
yes I did. I have installed fresh HS and tested it there. Than installed iStats and tested again. Than turned the fans to full throttle and tested. No difference.
SMC was reset, cmd+option+P+R, cmd+option+shift+power and so on were performed. No results. Scores are ***
 

Webster's Mac

macrumors 6502
Dec 18, 2016
341
284
There is a first time for everything.
But it's working under Windows on full power!
usually macs perform worse under windows. Like my rMBP mid 2015 throttles much more under windows than it does on macOS because macOS undervolts the CPU and controls the fans better.

Try re-applying the thermal paste again. Maybe pickup some Arctic MX-4. That’s a solid paste and is what I used in my mid-2015. Non-electrically conductive does it won’t short out stuff around it if it runs a little.
[doublepost=1565882901][/doublepost]I can’t really think of anything except for it’s performing by design or the paste isn’t good enough. Most people say paste doesn’t matter, but they are applying it on systems that have much higher temperature/throttling tolerances. Apple runs these up to the temperature limit of the CPU and then starts throttling. A degree or two improvement with a better paste is enough to significantly increase benchmark scores.
 

Webster's Mac

macrumors 6502
Dec 18, 2016
341
284
The problem is it's liquid metal already *_*
Did you clean the dust out of the heat sinks and fans when you repasted it? Also, liquid metal has to be perfectly applied to work properly. I’d try getting MX-4, remove the liquid metal, apply MX-4, and see if that helps. Liquid metal is more efficient *if* it is applied properly. And that can be difficult to do. I never use liquid metal in anything. Not worth the risk of damage and corrosion that happens to the heatsink surface
 

cynics

macrumors G4
Jan 8, 2012
11,959
2,156
Did you measure power consumption in watts under MacOS and Windows? Did you measure power consumption, temp and frequency using single core mode in Cinebench?

I'm not familiar with your particular model or the performance results people get but hitting the max cTDP on all cores for the entire duration of a benchmark seem unreasonable to me, at least for a laptop. For example this is my 2018 MBP 13" with I5-8259U running Cinebench all cores.


Screen Shot 2019-08-15 at 7.53.34 PM.png Screen Shot 2019-08-15 at 6.52.01 PM.png

Temp is high however I'm at the power limit (PL) for the package (35w). 3.10ghz to 3.20ghz is maintained after the initial turboboost. If power limit wasn't reached then temperature would be.

Single core mode in Cinebench maintains ~3.7ghz for the length of the run.

There are way too many variables to consider when comparing benchmarks. Ambient temp is a big one with a laptop. I don't know what sets the Cpu power limiting states between MacOS and Windows however a change can cause a performance difference. Background task. OS and app version. Etc.

I have 150 points difference between my highest and lowest Cinebench results on the same system.

I've seen different results but I haven't see overall improvements with LM in my iMac. Maybe briefly however the high sink gets saturated causing heat to not transfer as fast as it could with a heatsink that could dissipate heat better. I'm curious about the difference between your core temps though. I have 5 degrees celsius between the hottest cores max temp on both my Macs. A large discrepancy could indicate uneven heat transfer. However that could be the nature of that CPU.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.