G751jy Temperature almost the same after changing thermal paste

[andreacos92]

Ok, now it's pretty clear.

In the second video, before cleaning and repasting, your CPU was throttling because of CPU temps. Intel CPU starts throttling at 93-94 °C, and this kind of throttling reduces clock speeds just a little bit, like we can see in the second video. CPU clocks stay in the 3.1 - 3.3 GHz range, instead of 3.4 GHz (this is the turbo clock with all 4 cores active).

In the first video, after repasting, CPU temps stay in the low 90s, that is quite high but not enough to start CPU thermal throttling.
Now your throttling is much more aggressive, falling down from 3.4 GHz turbo clock to 2.6 GHz that is the base clock for the 4720HQ: this is the thermal VR throttling. You can see it in HWiNFO stats, at IA: VR Thermal Alert, and in your screenshot it's marked as YES.
You should see also an entry under CPU [#0]: Intel Core i7-4720HQ: Enhanced called VR VCC Temperature (SVID). If you can't see it, just put under load your CPU and then restore original order in HWiNFO, that temp sensor doesn't work if it's cold and show 11 °C or less (grayed out).



If you will monitor that sensor when you play games, you will see it goes up to 96 °C and then your CPU will start to throttle at base clock (2.6 GHz).

The paradox is that before repasting your CPU was thermal throttling because of CPU temp, and that was a "soft" throttling, cutting down just 200-300 MHz to keep CPU in the low 90s. Now, after repasting, your CPU is a bit cooler, it doesn't reach throttling temperature, but this allows the voltage regulator to run hotter and trigger another throttling system, that is more aggressive than the previous one.

Bad news n° 1: repasting has worked, but maybe not enough. Low 90s are still quite high temps for stock clock and voltages, with a fresh applied thermal paste and a clean laptop.
What's the temperature inside your room? I'd try to use a custom fan profile, to let them spin a bit faster. High 80s, with a demanding CPU game like Battlefield is already a quite good result in the middle of the summer.

Bad news n° 2: CPU VRs are not cooled, they are abandoned to their heat. I put some little heatsink over them just to have a sort of passive cooling, they help but don't work miracles.
Also in this case, if the fans spin a little faster, they should create more airflow in the chassis, that "could" cool VRs a little better. But without heatsink and active cooling, this is a very minor effect.
Anyway it's strange that you hit VR thermal at 40-45 W of CPU power, IIRC even with stock speeds I experienced this at higher power consumption (at least 60 W); but if your room temp is around 28-30 °C this can be possible (I simply don't play games/use intensive apps in summer lol).

You can also try to disable HyperThreading, if you don't use your laptop for rendering/encoding, heavy multitasking. From my experience, many games (even recent ones) will run better with HT off, and CPU stay cooler, with less power draw.
If you can't disable it through Windows or CPU affinity, as last resort I would try to set a limit on CPU multiplier, something like 3.1 - 3.2 GHz through XTU. 200-300 MHz should not affect your gaming performance, but can be enough to not suffer from aggressive throttling, that can cause FPS drops or inconsistency.

[Cannon.19]

[andreacos92]

[Cannon.19]

"Well". 29° C is not low anyway. I remember some tests where it has been shown how ambient temperature has a direct influence on CPU/GPU temps. So, if room was 22 °C, your system temps should be about 7 degrees lower.

Before disabling HT, I would try undervolting: it's the simplest, safest solution, with no performace reduction. Undervolting (and disabling HT, but that is a bit of weird solution) should reduce CPU power draw and accordingly CPU VRs temperature. You should see some improvements.

Increasing fan speed will reduce mainly CPU and GPU core temps, as said.

From what I understand from the other post, K5 Pro Thermal paste can be used as thermal pads replacement, because it's more viscous than normal thermal paste, so you can use it for GPU VRAM chips and GPU VRMs. But CPU VRMs are not cooled in our laptop, so it would be useless in your case.

Try starting with a -50 mV undervolting, and do some tests. It's not the best time to do so (because of ambient temperature), but you can try playing games straight, to combine business with pleasure.


This are mine (sorry for bad focus-old photo), with the little passive heatsinks. Just to say, in some test with keyboard popped out and a little third fan pointed on the VRMs, I could cool them of about -20 degrees, and I can mantain about 85-90 W of long-term CPU power draw without throttling. Of course that was just for test, then i closed my laptop
I tried also to "link" the VRM chokes (R36 modules) to the copper heatpipe (the gap is few millimeters), but then I had higher and bad core-spreaded CPU temps, because the pressure on heatpipe tended to bend it a little bit, preventing the right plane pressure between CPU die and copper heatsink (this is one of the reasons why I pointed on your first repasting with pads, it takes very little to preventing right contact).

Plus, since I have no battery inside and sothere is plenty of space, I want to try to put a little third fan inside the laptop, between the vents, to blow air to the CPU VRM heatsinks. But this is a bit too far.

I tried running RealBench stress test on -50 mV and laptop tuned off after about 10-15 minutes (even though it survived 1 hour on Prime95 before that), so I increased it to -45 mV and it survived 1 hour on RealBench so I guess that's the limit. I also applied the cache undervolt afterwards, and surprisingly it survived at -100mV for 1 hour on RealBench (together with -45 mV on core). I noticed about 2 degree difference in Battlefield (90 degrees was the highest), just about enough to eliminate VR throttle. Then I set both fans to 100%, got about 3250 rpm for cpu fan and 3450 rpm for gpu fan. It decreased cpu and gpu temps additionally by about 4 degrees, so in the end CPU was running at ~86 and GPU at ~73. I used this NBFC profile, which I found HERE.
Btw, undervolting was done through Intel XTU, and after all that, I wanted to see how it would perform at max turbo and cache multiplier, at 50-60 Watts and 128 second power boost window, and there was still no throttling, and it never reached 90 degrees (with fans at 100% of course), but it wasn't stable and turned off after ~30 mintues, which probably means -40 mV should be stable at those speeds.

And those heatsinks look nice in your picture, good job . How much difference did you notice after installing them?

[andreacos92]

I tried running RealBench stress test on -50 mV and laptop tuned off after about 10-15 minutes (even though it survived 1 hour on Prime95 before that), so I increased it to -45 mV and it survived 1 hour on RealBench so I guess that's the limit. I also applied the cache undervolt afterwards, and surprisingly it survived at -100mV for 1 hour on RealBench (together with -45 mV on core). I noticed about 2 degree difference in Battlefield (90 degrees was the highest), just about enough to eliminate VR throttle. Then I set both fans to 100%, got about 3250 rpm for cpu fan and 3450 rpm for gpu fan. It decreased cpu and gpu temps additionally by about 4 degrees, so in the end CPU was running at ~86 and GPU at ~73. I used this NBFC profile, which I found HERE.
Btw, undervolting was done through Intel XTU, and after all that, I wanted to see how it would perform at max turbo and cache multiplier, at 50-60 Watts and 128 second power boost window, and there was still no throttling, and it never reached 90 degrees (with fans at 100% of course), but it wasn't stable and turned off after ~30 mintues, which probably means -40 mV should be stable at those speeds.

Ok for undervolting. Seems that you didn't get a very lucky chip, because usually Haswell can go a bit lower with undervolting, but -40 mV is better than nothing.
For CPU stability, it's a bit of test-fail-change-repeat process. Maybe -40 mV will be full stable, maybe it will crash next week while watching a video on YouTube and you'll have to bump up another +5 mV.
For cache, try using AIDA64 with CPU + Cache test (no FPU), or prime95 with some specific data size for cache (make sure to use latest prime95 version and disable AVX options in GUI). Anyway, CPU vCore affects temperature and power draw much more than Cache voltage, so don't go too low with cache, because then you won't be able to recognize if any crash or instability is due to vCore or vCache.

Anyway, VR temp is affected mainly by CPU Package Power (that is mainly affected by CPU vCore), if undervolting you see that Power goes down, it's still worth
Did you see Package Power is less than before while playing?


About fan profile, that profile is "wrong", because it uses 151 and 152 registers, that are Fan States, and not Fan Speed.
If you want to better control your speed, you have to use 165 and 166 registers in the NBFC profile.
The difference is that 151/152 registers can't max out fans speed, while 165/166 can (for "max out" I mean speed you have never seen until now )
And in register values 147-148 and 149-150 you can see and calculate actual fans speed (should be the same thing HWiNFO does with EC).

I know it seems a bit confusing, if you want you can read these (it's not that difficult, and can be interesting):
https://web.archive.org/web/20170725...en/threads/69/
http://forum.notebookreview.com/thre...ux32vd.705656/
https://web.archive.org/web/20170929...ven/threads/92

Anyway, if you max out fans you will see another nice temperatures reduction, but it's better to be alone in the room, and max out only if temps goes high (more than 80/85).

And those heatsinks look nice in your picture, good job . How much difference did you notice after installing them?

As I said, not so much difference, maybe 3 degrees. They, as passive coolers, don't affect much the final steady temperature, but more the time VRs take to reach final steady temperature, slowing down the heating rate.
With a fan pointed to them, they can do a much better job

[Cannon.19]

Anyway, VR temp is affected mainly by CPU Package Power (that is mainly affected by CPU vCore), if undervolting you see that Power goes down, it's still worth
Did you see Package Power is less than before while playing?

Yeah, I think it went down by 1-2 watts on average, and VR temps maxed at 92 degrees, so 4 degrees less than before.

About fan profile, that profile is "wrong", because it uses 151 and 152 registers, that are Fan States, and not Fan Speed.
If you want to better control your speed, you have to use 165 and 166 registers in the NBFC profile.
The difference is that 151/152 registers can't max out fans speed, while 165/166 can (for "max out" I mean speed you have never seen until now )
And in register values 147-148 and 149-150 you can see and calculate actual fans speed (should be the same thing HWiNFO does with EC).

I know it seems a bit confusing, if you want you can read these (it's not that difficult, and can be interesting):
https://web.archive.org/web/20170725...en/threads/69/
http://forum.notebookreview.com/thre...ux32vd.705656/
https://web.archive.org/web/20170929...ven/threads/92

Anyway, if you max out fans you will see another nice temperatures reduction, but it's better to be alone in the room, and max out only if temps goes high (more than 80/85).

Really? Fans can go even faster than this? That's great . If I understood you correctly, I just need to change the register values in that profile to 165 or 166, and everything else can be left as it was? I'd like to try it out, even though this profile was enough to stop the throttling, but I just want to test the limits I guess .

[andreacos92]

Yeah, I think it went down by 1-2 watts on average, and VR temps maxed at 92 degrees, so 4 degrees less than before.

Well! Actually more than 4 degrees less than before, because before it maxed out at 96 °C due to throttling, but it could go higher if it wasn't limited

Really? Fans can go even faster than this? That's great . If I understood you correctly, I just need to change the register values in that profile to 165 or 166, and everything else can be left as it was? I'd like to try it out, even though this profile was enough to stop the throttling, but I just want to test the limits I guess .

Yes, you should be fine just changing register values: 166 for CPU and 165 for GPU.
I set EC poll rate to 300 ms too.
Then I suggest you to change thresholds according to your need. You can play a little with values and threshold, should be safe.
Quoting my post of 4.5 years ago (crying inside), CPU fan speed ranges should be:

The percentage values have been set to obtain a change in fan state at every step, so I have:
0% = 0 rpm
50% = 2200-2300 rpm
65% = 2400-2500 rpm
75% = 2600-2700 rpm
85% = 2800-2900 rpm
91% = 3000-3100 rpm
95% = 3200-3300 rpm (at first it will go up to 3400 rpm then it reduces)
98% = 4000-4100 rpm

So, with 0-49% value CPU fan doesn't spin, with 50-64% it should spin at 2200-2300 rpm, and so on.

GPU fan speed values are a bit different IIRC, but you should find them pretty easily.

Last, set 0 as min fan speed and 100 as max for each fan as they should already be, but then in thresholds use 98% as max value instead of 100. It should be enough to trigger max speed for both fans, but I remember I had some strange behaviour with 100%.

Let me know

[Cannon.19]

Well! Actually more than 4 degrees less than before, because before it maxed out at 96 °C due to throttling, but it could go higher if it wasn't limited




Yes, you should be fine just changing register values: 166 for CPU and 165 for GPU.
I set EC poll rate to 300 ms too.
Then I suggest you to change thresholds according to your need. You can play a little with values and threshold, should be safe.
Quoting my post of 4.5 years ago (crying inside), CPU fan speed ranges should be:



So, with 0-49% value CPU fan doesn't spin, with 50-64% it should spin at 2200-2300 rpm, and so on.

GPU fan speed values are a bit different IIRC, but you should find them pretty easily.

Last, set 0 as min fan speed and 100 as max for each fan as they should already be, but then in thresholds use 98% as max value instead of 100. It should be enough to trigger max speed for both fans, but I remember I had some strange behaviour with 100%.

Let me know

Well, I tried doing that, changing cpu from 151 to 166 and gpu from 152 to 165, and fans turned off. I panicked and held the power button for 10 seconds. After turning the laptop back on, it ran disk error check. I guess I shouldn't mess with it anymore

[andreacos92]

Well, I tried doing that, changing cpu from 151 to 166 and gpu from 152 to 165, and fans turned off. I panicked and held the power button for 10 seconds. After turning the laptop back on, it ran disk error check. I guess I shouldn't mess with it anymore

Wait, did you change Max Speed Value from 6 to 100 too? Otherwise of course they turn off, because the first speed threshold is at about 50, as I said in the previous post.

Try this, I use it since 4 years now: https://mega.nz/file/N0oFEbKY#7Y22iK...Lx1PZaT7XgoGjc

[Cannon.19]

Wait, did you change Max Speed Value from 6 to 100 too? Otherwise of course they turn off, because the first speed threshold is at about 50, as I said in the previous post.

Try this, I use it since 4 years now: https://mega.nz/file/N0oFEbKY#7Y22iK...Lx1PZaT7XgoGjc

Thank you, that one works great! Why is there no manual control slider though? With the other profile I could set the fans to 100% myself regardless of temperature. Would it be possible with this one too?
EDIT: Actually I can't do it with old profile now either, I messed something up didn't I