Question i5-5675c and GT3e reducing power of iGPU

[Fury13]

Hello forum!

I hope this is the right forum as my issue relates mainly to the iGPU as opposed to the CPU.
I am surprised this use case hasn't come up before, but i loooooove the i5-5675c and the performance from the GT3e graphics given this is now a 5 year old processor and still has one of the best iGPU's available.
So here's my drama...

I am using this 65w processor as a HTPC/light gaming rig in a passive case (Streacom FC10a) with an AsRock Z97E-ITX/AC mobo.
I have de-lidded the CPU and have Cool Laboratory Liquid Metal between the die and heat spreader (prior to this temps hit 90c+ within minutes).
Under normal gaming, eventually thermals rise to 90c+ and the system hangs (screen goes black, and i have to hold power to reset).
Thermal transfer to the chassis is good (the chassis gets hot) and i am using thermal grizzly kryonaut between GPU and all heatpipes / chassis.

I have undervolted the CPU by ~120mv and the CPU draws at around 40w running a XTU CPU Stress Test.
Temps stay around 60-65c after several hours of stress testing, so this result i would deem as OK.

Unfortunately, when the iGPU is added, i start to run into the before mentioned issues.
I have also undervolted the iGPU by ~120mv.
The iGPU will still try and draw as much as it can (up to 60w) while the CPU will draw up to 15w when the iGPU is loaded, giving a total package of ~75w.
This causes temps to steadily rise over 90c and subsequently the system crashes.

Does anyone have any suggestions as to why this might be happening, and how to resolve it?
I am happy to reduce the iGPU's performance slightly as i am only gaming at 1080p 60hz.
My mobo (AsRock Z97E-ITX/AC) does not appear to let me reduce the CPU's TDP, or reduce the iGPU's clock.
I am concerned about reducing the GPU volts any lower (unless anyone has any comments on suitable voltages for the iGPU), however this has definitely increased performance.

Any help much appreciated!

 

[Stuka87]

Well, in short, it means your cooling is inadequate. How warm does the case get where the heat pipes contact it?

I am not a huge fan of the way that case has the "some assembly required" heatsink and heat pipe setup. Typically the heatpipes make direct contact with the CPU. But in this case, they are placed inside a cast aluminum block. So you have a thermal resistance area between the CPU and the aluminum, and then again between the aluminum and the copper, and then between the copper and the aluminum case. I am sure this works fine for low load use cases, or with a mobile class CPU.

So, I think you need to verify the heat is getting from the CPU, to the case. And it sounds like it is not.

 

[Fury13]

The chassis is getting warm (too hot to touch).
Heat is moving from CPU/iGPU to chassis as it now takes ~1-2hrs at full tilt to go from ambient to 90c. Ambient temp is around 18-20c.
During this time the iGPU is drawing ~60w and the CPU is drawing 15w for a total package power draw of ~75w.
I understand that i am pushing the limits of the thermal handling of the case (even though the manufacturer swears 65w processors should be no issues and it can handle 95w under the right conditions).

My understanding was that the CPU should thermal throttle as opposed to just shut down?
It does not appear to do that, or atleast the GPU does not.
However, if i can limit the TDP of the process to 60w or less (from the 75w it wants to peak at now) then I believe the chassis will be able to cope with dissipating the heat (as it can cope with 50w from the CPU only).
Undervolting the CPU by 120mv has reduced CPU max draw from 65w to 50w.
I've undervolted the iGPU by 120mv as well and believe it's reduced the max draw from GPU as well (can't recall the original draw, but now it's peaking at 60w...
But still the total package draw is too much, hence wanting to limit the iGPU in other ways if possible....

 

[ZGR]

I have a 5000 series processor, but I have that giant Noctua D15 which solves all heat issues. You basically made your own NUC with that tiny case. Pretty cool.

On my Asus Z97, I can overclock or underclock the iGPU in BIOS. I have never even bothered using that feature, but that would probably help you a lot... The iGPU does suck a lot of juice for me as well.

In this case, I would see how low of a CPU clockspeed you can get away with in the games you want to play. In a lot of games 2.0 GHz will suffice. But as you say, the iGPU is what is causing things to warm up so much.

But this case says it can handle 65w. It shouldn't be throttling so much. The review I saw has a 6700k in it.

 

[Fury13]

The iGPU won't clock below 1100mhz in the bios... Can go all the way up to like 2500mhz, but can't go any lower
I don't think the CPU clock is the issue at all here sadly.

As a note, if i set any short term or long term power limitations (eg: limit it to 75w) i get crashing within 5mins... sometimes even BSOD.

I think a CPU handles thermal throttling far better than an iGPU...

 

[Stuka87]

The chassis is getting warm (too hot to touch).
Heat is moving from CPU/iGPU to chassis as it now takes ~1-2hrs at full tilt to go from ambient to 90c. Ambient temp is around 18-20c.
During this time the iGPU is drawing ~60w and the CPU is drawing 15w for a total package power draw of ~75w.
I understand that i am pushing the limits of the thermal handling of the case (even though the manufacturer swears 65w processors should be no issues and it can handle 95w under the right conditions).

My understanding was that the CPU should thermal throttle as opposed to just shut down?
It does not appear to do that, or atleast the GPU does not.
However, if i can limit the TDP of the process to 60w or less (from the 75w it wants to peak at now) then I believe the chassis will be able to cope with dissipating the heat (as it can cope with 50w from the CPU only).
Undervolting the CPU by 120mv has reduced CPU max draw from 65w to 50w.
I've undervolted the iGPU by 120mv as well and believe it's reduced the max draw from GPU as well (can't recall the original draw, but now it's peaking at 60w...
But still the total package draw is too much, hence wanting to limit the iGPU in other ways if possible....

Ok, it sounds like the case is heat saturating then, which means not enough airflow around it. Is this inside an entertainment center or the like where there is no air flow?

You may need to add low speed fan to the outside of the case where the heat pipes contact. A fully passive cooler still needs airflow, just a lot less.

Another way to limit the GPU's power consumption is to frame cap any games you are playing. Do you have VSYNC enabled for whatever it is you are playing, so that it wont go over 60fps? If so, you may need to frame cap at 30fps.

How are you setting the power limitation? It may be enabling that, and also undervolting is causing the crashing. If you are setting the power limitation by undervolting, it just means you undervolted too far. Intel isn't known for having their voltages higher than they should be, so not a lot of leeway there.

 

[Fury13]

The unit is sitting on a desk. There is nothing around it obstructing it.
I do feel the chassis is reaching heat saturation but Streacom assure me it should take a 65w load with out any issues.
In my testing it does apprar to take a 60w load for several hours and keep temps around 65c. As soon as the iGPU is asked to work as well, the package is drawing 75w and this is where temps start rising slowly to eventually hit 90c+.

I used short term/long term power in the bios set to 65/75w respectively where i experience crashing.

Today i removed the undervolting and ran XTU GPU stress test (note: i also open a web browser and load a web based game in it to load the CPU as well) and i hit at around 80-82w constant current draw for this 65w processor, which just makes me hit 90c much quicker.

Unfortunately the main game I play (TeamFortress2) has horrible latency when vsync is enabled which makes it unplayable. I can limit this ingame to 60fps however this isn’t the only game I play. I’d rather find a global way to stay within 65w current draw than have to adjust this on a game by game basis....

 

[ZGR]

I would blow a fan directly onto the heatsink at this point. The 5000 series has such spotty BIOS support. A shame that ASRock didn't add iGPU underclocking in BIOS.