Defiance Temps

[Brask]

I have a PCSpecialist Defiance V RTX (Clevo P970EF), with undervolt to -125mV on CPU core and cache and -50mV on iGPU I have lowered my temperatures, adding a cooling base also has helped, installing Clevo Software Center pushes the fans to the maximum but I still feel the temperatures are too high.

The CPU has no problem in being applied Liquid Metal Thermal paste, I am thinking in repasting it with Thermal Grizzly Conductonaut and the GPU as it does have a problem with liquid metal I am thinking in repasting with Thermal Grizzly Kyronaut.

What do you guys think?

Repasting voids warranty?

 

[Brask]

Hi all.
Well you have seen the response from PCS.
'' this kind of temperatures is not fixable through us. It is a design detail with the slimmer laptops. ''
I have decided to give this expensive piece of junk laptop to my daughter to play Roblox. Core TEMP should copy with that game.
And I'll buy proper pc/laptop elsewhere.
IF you read this and doing research about RECOIL RTX my advice to you is DON'T BUY IT!
CPU reaching 99 degrees with cooling pad and so called high grade thermal paste is not acceptable!

On CPU try Liquid Metal Thermal paste, put a drop in a plate, pick it up with the ear stick and apply it to the CPU only, on the GPU use high quality normal thermal paste.

 

[ubuysa]

Repasting voids warranty?

AFAIK it doesn't, as long as you do no damage.

 

[phitol]

I have a PCSpecialist Defiance V RTX (Clevo P970EF), with undervolt to -125mV on CPU core and cache and -50mV on iGPU I have lowered my temperatures, adding a cooling base also has helped, installing Clevo Software Center pushes the fans to the maximum but I still feel the temperatures are too high.

The CPU has no problem in being applied Liquid Metal Thermal paste, I am thinking in repasting it with Thermal Grizzly Conductonaut and the GPU as it does have a problem with liquid metal I am thinking in repasting with Thermal Grizzly Kyronaut.

What do you guys think?

Repasting voids warranty?

I think the only way you will void the warranty is if you either
- Physically damaged things during the re-paste
- Applied Liquid Metal but got it on the surrounding components and shorted things out, causing electrical damage

Grizzly Kryonaut is OK, (12.5W/mk) but Conductonaut (Liquid Metal) is 73W/mk), which is why people love the LM, although I'd always be worried about application shorting things, or with the heatsink corroding over time (Not too sure which metals its not recommended to be used with).

With your defiance, maybe post some AIDA64 stress tests (just use stock settings in AIDA, it does stress memory as well by default but useful for comparisons), also see if you can determine your ambient room temperature somehow, as obviously for comparisons sake that is important.

I can certainly post mine in the same tests (Just use performance mode from Clevo Control Centre) and run the test for a minimum of 5 minutes to ensure its nicely heat-soaked.

As I've pointed out in another thread, Intels power limiting logic and ability make it hard to know if you have thermal issues as these things are designed to effectively use every last ounce of cooling ability and there is no way in a thin/light gaming laptop that you can have all 6 cores 100% max'd out in a stress test @4.0GHz indefinitely as that requires 80W or so of TDP , which would require a massive cooling solution.

 

[Brask]

I think the only way you will void the warranty is if you either
- Physically damaged things during the re-paste
- Applied Liquid Metal but got it on the surrounding components and shorted things out, causing electrical damage

Grizzly Kryonaut is OK, (12.5W/mk) but Conductonaut (Liquid Metal) is 73W/mk), which is why people love the LM, although I'd always be worried about application shorting things, or with the heatsink corroding over time (Not too sure which metals its not recommended to be used with).

With your defiance, maybe post some AIDA64 stress tests (just use stock settings in AIDA, it does stress memory as well by default but useful for comparisons), also see if you can determine your ambient room temperature somehow, as obviously for comparisons sake that is important.

I can certainly post mine in the same tests (Just use performance mode from Clevo Control Centre) and run the test for a minimum of 5 minutes to ensure its nicely heat-soaked.

As I've pointed out in another thread, Intels power limiting logic and ability make it hard to know if you have thermal issues as these things are designed to effectively use every last ounce of cooling ability and there is no way in a thin/light gaming laptop that you can have all 6 cores 100% max'd out in a stress test @4.0GHz indefinitely as that requires 80W or so of TDP , which would require a massive cooling solution.

The problem I mainly see is appart from the temperatures, when I am gaming the CPU throttles (power limit) as well as the graphics, I am now looking for help from a Defiance V RTX owner to get updated BIOS and EC firmware as PCSpecialist rejected giving me the BIOS and EC firmware.

 

[Oussebon]

I've moved these to a new post to avoid hijacking of the Recoil topic.

You can get updates directly off the Clevo site, which I believe has been pointed out multiple times.

 

[phitol]

The problem I mainly see is appart from the temperatures, when I am gaming the CPU throttles (power limit) as well as the graphics, I am now looking for help from a Defiance V RTX owner to get updated BIOS and EC firmware as PCSpecialist rejected giving me the BIOS and EC firmware.

As mentioned in the other thread, whilst gaming your cpu is probably not used as much. If your clock speeds are all 4ghz, but the package is only using 20w, then the game is not using all the available hardware in the cpu, (eg AVX)
I have just been playing VR games on mine and the cpu package averages at 22w, but all cores average at 4ghz. Remember your cpu multiplier for all core turbo is 4ghz, so with light loads it will easily max out all core speeds to the 4ghz limit without needing to use much power. You could increase the all core multiplier in throttlestop, but whilst that might work for gaming, it’ll struggle thermally in stress testing where the cpu will be loaded sych that thermal abs power limits trip.

[edit]
Just to update, doing more monitoring whilst playing various VR games it does highlight (to me at least) that power/thermal throttling is going to occur in many situations normally.

HW INFO confirms the Turbo PL1/PL2 values are 50W/78W for the CPU in performance mode.
Lightweight games - Average TDP 22W, Average Clocks 4.0Ghz, Max TDP 56W, Max Clocks 4.6GHz, Average Temps 67
Demanding games - Average TDP 42W, Average Clocks 3.6Ghz, Max TDP 72W, Max Clocks 4.6Ghz, Average Temps 82
AIDA64 - AVX ON - Average TDP 53W, Average Clocks 3.4Ghz, Max TDP 77W, Max Clocks 4 Ghz, Average Temps 84 (No GPU test running)

I think this highlights what I mentioned above,
Lightweight games don't tax the CPU heavily, and so pretty much can sustain max all core speeds on relatively low power, the predominant 'limit' being reached in this case is the All-Core turbo 4.0Ghz design limit. You can also see times when the CPU load is very low, and the odd core will turbo up to the single core 4.6Ghz limit, and sometimes hit thermal throttling, but that is very briefly and also infrequently.

Demanding Games present a very mixed workload, and actually I found that often the cores would max out, the package TDP reaches the 50W PL2 limit fairly often and thus clock speeds drop accordingly (Power limiting), then it goes through a lighter section and clocks creep up again as the TDP reduces. In this case, heavy GPU usage transfers a bit of heat in to the CPU (being in a confined space, this will occur) and the temps average above 80. In addition, you also get periods where lighter/mixed loads cause single cores to turbo to 4.6Ghz and occasionally thermally throttle but again, very briefly.

AIDA64 - Being a stress test this is probably the one to use for consistency, initally the PL2 (78W) is used and clocks briefly start at 4Ghz, but thermal/power throttling can occur and these decline until PL1 is reach (50W), at that point 4.0Ghz can't be achieved on 50W on all cores, so clocks reduce as the CPU is very frequently power limiting to keep within 50W, although note that only in the PL2 initial few seconds do I get thermal throttling, whilst in sustained PL1, no thermal throttling occurs.

To summarise, you really have to be careful when assessing your system, as the CPU is designed to reach one of three limits
TDP - 50 W, exceeding this will flag power limit throttling and clocks speeds drop
Thermal - 95 Degrees, exceeding will flag thermal throttling and clock speeds drop
Max Clocks - depending on number of cores, worse case all cores is 4Ghz, hittting that will hold the TDP/Thermals at their current limits.

Those limits all interact to govern overall performance at any given time/situation.
In general, for sustain loads you are realistically looking at what the processor can do with 50W, if every hardware unit of the CPU is utilised (ALU/FPU/AVX/CACHE, etc) 50W is not enough to achieve 4Ghz on all cores even when temps are not throttling.
Then you have instances where mixed loads lead to the CPU briefly allowing single core turbo's to kick in and 4.6Ghz can sporadically be seen on cores, but that will be brief and note that if the rest of the CPU is fairly well loaded, that might lead to small instances of thermal throttling.

I've seen some games kind of work against the limits and it bounces between 4.6Ghz on a single core, thermally limit, then use all cores, power limit, then rinse and repeat. It looks quite bad, but in reality its just working within the rules.

Conclusion:
To see if your laptop is performing OK, use AIDA64 as a stress test, this will get you a constant/sustained PL1 (50W) scenario where you want the processor to be power limiting at that value. With no Undervolt, you should not expect to see thermal throttling, although do expect reasonably high temps (80-90) and you may also see sustained lower clock speeds (such as 3Ghz).
Undervolting in this scenario will allow a higher clock speed for the same 50W limit, although don't expect temps to drastically reduce in this situation.

Where undervolting will reduce temps is in normal gaming scenarios where the power limit is not being reach (i.e. average TPD is 20-40W) and it is holding 4Ghz on all cores, in this situation, the undervolt will reduce the average TDP and that will lead to less temperature.