Asus Zephyrus GX531 Cooling Options Assistance Needed

[Kostiz]

Hey All

Laptop:
Asus Zephyrus GX531
RTX 2070Max-q / i7-8750H/ 16GB RAM / 1TB SSD
Warning some photos, apology in advance if I uploaded them incorrectly..

Ok so I purchased this 2nd hand and quickly noticed this was extremely hot doing absolutely nothing more than browsing and file transferring. CPU temps peaking in the high 90DegC, I figured it needed a good clean and new thermal paste.

While I was going to open it up I figured I'd give Liquid Metal cooling a go and add in a larger nvme since this only takes 1 and boy am I glad I did this!

I took some photos on the progress and couldn't believe the state of the thermal pads they're practically melted away.


Before

After

Prepping & Application of LM

But why stop there right..So went on the hunt for better thermals all round which led me to Throttle Stop, thank you for an awesome tool!

A few nights of youtube, reading this forums, got me to a decent thermal drop, but I wanted more and I am not sure I am doing this right or fully understand some of the settings..

Because I didn't know where to start I didn't get the best logs or photo's as I fumble my way, and still not sure what to log TBH but here is a few photos of before and after..I did update BIOS which locked me out of these features, but I managed to roll back to give me control again, Phew..

Before Changing settings

After some tweaking

So a few quick questions if the CPU TDP Limit is 45 watts how safe is to hold this much high or is that the TPL only function is running this for a set time in this case 28sec period then it would throttle?

If limiting the voltage to the CPU is the lowest the better option before for a BSOD, so Cache is more important then core voltage, they don't have to match, so I can go as low as possible before CB23 fails?

Ideally my goal is for no throttling of any kind (no red or yellow 'lights'), without to much sacrificing the CPU/GPU Clock Speeds and Jet engine fans..

Am I expecting to much and the FANS, HEAT, THROTTLING, is all part of the design?

Once I get my head around the CPU tweaks I have to tackle the GPU, while MSI Afterburner is the best option here it knda has a mind of its own when trying to set limits and power curve. Undervolt is going to be a challenge here for sure as I would love to achieve the same sweet post for FPS Mhz vs Temp vs Fan Noise..

I look forward to feedback and options to try and how to best find my balance

Cheers
Kosti

 

[AOne]

These are no pads. It's thermal paste. Asus does this for a few years. Mine is also with paste instead of pads.

 

[Kostiz]

These are no pads. It's thermal paste. Asus does this for a few years. Mine is also with paste instead of pads.

Oh snap!! it didn't feel like thermal paste at all, was like a putty when I started to remove it and I added .5mm pads

So should I change and add thermal paste? Maybe LM?

 

[AOne]

LM is impossible for replacing pads, as these chips are too uneven. It would be too hard to isolate them too. I've tried using LM for CPU and GPU on my previous ASUS, but it was not working well. Removed it and placed Noctua NT-H2 and its been an year and three months so far, I have no yellow or red throttle warnings neither in games or CAD/CAM nor in tests.

Last score in CB20 since yesterday is 3231.

 

[Kostiz]

For me LM made a massive improvement but I agree that a new paste would've yielded a similar improvement..

I am more keen now on how to tweak to achieve zero to minimum throttling

 

[unclewebb]

@Kostiz - When running Cinebench, are you seeing any power limit throttling, PL1 or PL2 in Limit Reasons. What does ThrottleStop show for power consumption when this happens?

The 45W TDP rating allows manufacturers to design their cooling solutions. If your cooling is good, you can safely go beyond 45W. Intel uses the exact same technology in their desktop CPUs and people run them at way over 100W without thinking twice about it. Running your CPU at 60W or 70W is perfectly safe for your CPU as long as it is not thermal throttling.

I would check the FIVR - Disable and Lock Turbo Power Limits box. Some computers use a secondary set of power limits so it is best to turn these off if they are being used.

For the core and cache voltages, many people get better results by leaving the cache at an offset of -125 mV and then increasing only the core offset up to -200 mV. Slowly bump the core in steps of -25 mV. I prefer using Cinebench R20 when testing different voltages. I find it is quicker. R23 should be OK too if you want to compare to your previous scores.

MAXON Cinebench R20.0 Download

CINEBENCH is a real-world cross platform test suite that evaluates your computer's performance capabilities. CINEBENCH is based on MAXON's award-winn

www.techpowerup.com

 

[Kostiz]

Hey Uncle Webb
Thanks for chiming in!

OK TDP makes more sense now thanks, I always thought that was a safe zone not to be pushed past this value but with ample cooling if you can dissipate the heat away from the package then running more power to it makes sense, thanks for clearing that up

Yes I see PL1 and PL2 for the CORE and EDP other for the RING

I probably am not doing this test correctly and I have just been playing with the undervoltage settings and managed to get it stable using these values -160mv core and -155mv cache and also took -50mv off the iGPU, but PL1 & 2 are RED

I will try and Disable and Lock Turbo Power Limits box ( the one found under memory)? I will run tests again using R20, I've also unchecked the CLAMP and increased it from the TPL Long Boot, honestly I am fumbling my way around, so thanks for helping out

Will be back ASAP with n update
Cheers
Kosti

 

[unclewebb]

-155mv cache

Most 8750H CPUs start losing stability somewhere around -125 mV or -130 mV. If you get any crashes, reduce your cache offset first. If you have an uber CPU and it never crashes mid game, ignore this advice.

CORE and EDP other for the RING

The ones lighting up red under the CORE column are the important ones.

PL1 & 2 are RED

Keep track of what the power consumption is when PL1 or PL2 are lighting up in red. Better yet, turn on the Log File option. That way you will have a record of CPU performance, power consumption and throttling reasons from while you are testing. The log file data makes it a lot easier to get a CPU sorted out.

 

[Kostiz]

Rightio, so this is getting exciting..I turned on log file too so hopefully I can attach what I done, I did run this in CB23 only as I wanted to see if it beat the score, it did, but it ran hot too.

These temps scare me and it took a while for my fans to go full jet engine too, what am I doing wrong as power was up in the 60W range and CPU Freq was all over the place, I assume this is normal behaviour for the CPU to boost

EDIT - I noticed MSI afterburning has my GPU ramped up too, for me this seems to have a mind of its own but that's another journey..Lets cool down my 8750H and I will move onto the next adventure.

EDIT 2 - I looked over the logs,
PL1 came on most likely cause the FANS took a while to spin up right?
As for VRTEMP it doesn't make sense on those values it triggered

Am I reading these correctly?

 

[unclewebb]

PL1 came on

PL1 came on because your CPU was throttling down to the 60W PL1 power limit that you set in ThrottleStop. Raise that power limit and it will not power limit throttle anymore.

VRTEMP

You do know that denial is not a good reaction to a problem.

Your voltage regulators are overheating. When this happens, the CPU briefly turns off Intel Turbo Boost. This drops the CPU down to its base frequency which is 2200 MHz. This cools down the voltage regulators so the CPU can go back up to full speed. This overheats the voltage regulators so they trigger throttling again and again and agan. Your computer stays in this endless cycle until you stop fully loading it.

Definitely not a good sign. You need to find a way to cool your voltage regulators. If you cannot do that, you cannot run your CPU at 60W. You will need to go back to the default 45W turbo power limit (PL1) and see if you are still having these voltage regulator throttling issues.

 

[Kostiz]

Thanks for the information and guidance UncleWebb, I am wondering if these VR's are transferring to the shared heasink after I applied LM to the CPU & GPU and new thermal pads on the rest of the components as to me it didn't look like paste but melted pads..

So I change the TurboBoot Long to 45 instead of the default 60 right..OK, let me try this for a few runs

BTW should CLAMP be checked or unchecked, thanks

Kosti

 

[AOne]

Is it possible that you placed too thin pads, or the thickness varies between chips? Maybe this is why it was intended to be with paste/grease instead of pads I would fix this mistake as soon as possible and would not use tha laptop before that. (You can not melt pads, it's not a 300 C oven).

 

[Kostiz]

@AOne, thanks for the feedback
OK, sure its possible the pads may not be making proper contact? however what I removed felt like blue-tack not thermal paste it was very sticky, so really confused if this was paste or pad now?
I'm pretty sure the replacement pads I used are not thinker than what's needed as I went with .5mm pads on all VR and Memory around the GPU which was the same thickness as the one seen in this 1st photo near the GPU on the left which remained unaffected when i removed the HSF for the first time.

This goo stuff took ages to removed between the VR and other components around the VR's, so which Thermal Paste would be ideal for these components?

Hate to strip this down again...Damn it!

 

[AOne]

This is 100% paste on your first two pictures. Actually my Asus looks pretty much the same.
If you've placed pads not touching the heatsink it would not dissipate heat at all. If it's too thick, the CPU and GPU would have air gaps and then you would have extreme thermal throttling. Just use paste and all would be good.
Actually, today I decided to replace my paste after a year and 3 months as one of the cores was starting to become hotter with more than 10 degrees difference than the others (although none exceeded 90C in any case). The paste over the gray chips was more dry then the rest, which means they're producing quite a lot of heat and the paste I used there is aging too fast. Back than I used Noctua NT-H2 for CPU and GPU, and MX4 2019th for the rest. Well, just as expected, MX4 degraded over time much faster then the Noctua. Now this error is fixed and there's Noctua everywhere. It took me 20 minutes to remove the old paste and 2 more for new paste and putting all back together. Use alcohol and ear buds for easier cleaning.

http://imgur.com/jXluLBA

3-4 tests (TSBENCH+CB20) in a row and you see the results - Black Limit reason's box and temp different of 4 degrees max.

 

[Kostiz]

Thanks for sharing AOne, that's pretty awesome temps! OK I will order some new thermal paste and have a go at reapplying, probably take a while since I am going away next week for 3 weeks.. Damn was hoping to use the laptop while away

 

[unclewebb]

should CLAMP be checked or unchecked

When a CPU reaches one of the turbo power limits, if the Clamp option is checked, the CPU will slow down as much as necessary so the CPU does not exceed the power limit. If Clamp is not checked, when it power limit throttles, the CPU will reduce the amount of turbo boost but it will not go below the base frequency which is 2200 MHz for the 8750H.

I do not like throttling so I never check Clamp. The less throttling the better.

In a laptop, using Clamp can be useful if you want to be absolutely sure that power consumption does not exceed the PL1 power limit that you have set.