Ionico laptop: Guide how to Undervolt, Power Limit, Fan controlling and Memory tuning

[cedrixman]

@cedrixman I am no expert of GPU undervolting as rarely my GPU risks to throttle, given the ample die size. However when you say 725 mV at 1500 Mhz are you saying that you allows your GPU to any given Mhz but, the only variation you make are, giving the GPU less voltage at each Mhz step? Or are you also hard-limiting the Mhz overall?
I've never done GPU undervolting, so I asks that. Can your GPU still boost to the max frequency Nvidia planned it to reach?

In fact, in Afterburner, you have a graphic where you see the link between frequency and voltage. I setup 2 things :
1) Limiting the frequency (1500MHz is the standard max frequency i can see on game)
2) Undervolt and indicate to the system that at 1500MHz, the voltage will be 725mV instead of 1V

As the CPU and GPU share some of the heat pipe, by reducing the GPU temp, you will reduce the CPU temp and you will also reduce the fan noises.

My experience on Cyberpunk (1440p, ultra), CPU : 75 deg and my GPU : 70 deg

This is for me as important as undervolting CPU (and Nvidia GPU accept a huge undervolting compared to Intel CPU)

 

[barlew]

Here is a question for you. If the max frequency in game is 1500MHz why would you need to limit to 1500MHz?

 

[cedrixman]

Here is a question for you. If the max frequency in game is 1500MHz why would you need to limit to 1500MHz?

You are right. It seams not logical. But, in fact, by limiting the GPU frequency, you avoid the system to go at a too high frequency, which lead to a too high temperature, huge fan noises and probably some throttle. By limiting the frequency, your system will never try to go above it

To be honest, Yes, by limiting the frequency at 1500MHz, you will lower your laptop performance but it is barely noticable. But you will notice that the noise and the temps system are really lower

For exemple, at 1500MHz, CPU at PL1 35W and PL2 35W : Time Spy score : 9985 points

 

[cedrixman]

Please find below some benchmarks (all the benchs with the following CC setup (CC 3.9.18.00 : default overboost setup + unlock configurable TGP : 140 W, dynamic boost desactivated, CPU undervolted at -100mV) :

GPU not undervolted, not overclocked :

GPU undervolted at 950mV, not overclocked (below 950mV, I have an error message) :

GPU undervolted at 950mV, overclocked at +150MHz :

GPU undervolted at 950mV, overclocked at +200MHz :

 

[barlew]

You are right. It seams not logical. But, in fact, by limiting the GPU frequency, you avoid the system to go at a too high frequency, which lead to a too high temperature, huge fan noises and probably some throttle. By limiting the frequency, your system will never try to go above it

To be honest, Yes, by limiting the frequency at 1500MHz, you will lower your laptop performance but it is barely noticable. But you will notice that the noise and the temps system are really lower

For exemple, at 1500MHz, CPU at PL1 35W and PL2 35W : Time Spy score : 9985 points

So you are dropping your TimeSpy score by over 400 points by doing that. That is quite a significant loss in performance.

I go back to my first point you are deliberately reducing the performance of your laptop to keep the temperatures in check which I personally find crazy. If your temperatures are too high and your GPU is thermal throttling I would say there is something wrong with your laptop.

I can understand the attraction of under-volting your GPU to optimise temps in the same way we do a mobile CPU but that needs to happen after the temps have been brought to an acceptable range.

I can never see a scenario however where you would want to cap the frequency of your GPU. The loss in performance your TimeSpy score illustrates speak for itself.

As for the fans. Well you have bought a thin and light performance laptop. They were always going to be loud.

 

[cedrixman]

So you are dropping your TimeSpy score by over 400 points by doing that. That is quite a significant loss in performance.

I go back to my first point you are deliberately reducing the performance of your laptop to keep the temperatures in check which I personally find crazy. If your temperatures are too high and your GPU is thermal throttling I would say there is something wrong with your laptop.

I can understand the attraction of under-volting your GPU to optimise temps in the same way we do a mobile CPU but that needs to happen after the temps have been brought to an acceptable range.

I can never see a scenario however where you would want to cap the frequency of your GPU. The loss in performance your TimeSpy score illustrates speak for itself.

As for the fans. Well you have bought a thin and light performance laptop. They were always going to be loud.

How can I say that... You will never find a laptop so thin without CPU and GPU thermal throttling. Never !!! This is the way those laptops are designed

You can think this is not an ideal scenario and you do not want to understand it but it is basically the case. Try to explain your point to all ingeeners of famous companies like MSI, Asus, Acer, etc... This is unfortunatly the laws of nature

2nd information, the curve GPU Frequency / Temp is not linear. And it is absolutly normal because the curve GPU Voltage / Frequency is not linear. At a point, trying to be at maximum frequency like 2000MHz, you need to be at 1.5/1.6V and the system immediatly throttle. It is the same with CPU. And the system will immediatly go at 1500/1600MHz but with a high GPU temperature

So my point is the following, if you want to use your laptop at maximum performance, you will get 10 249 pts at Time Spy, CPU temp 80° and GPU temp 80° and fans at 100%. If you want to use your laptop at 97.4% of the maximum performance, you will 9 985 pts at Time Spy, CPU temp 70° and GPU temp 70% and fans at 50% (honnestly, in terms of noises, it is the day and the night)

 

[barlew]

How can I say that... You will never find a laptop so thin without CPU and GPU thermal throttling. Never !!! This is the way those laptops are designed

I have two which don't.

So my point is the following, if you want to use your laptop at maximum performance, you will get 10 249 pts at Time Spy, CPU temp 80° and GPU temp 80° and fans at 100%. If you want to use your laptop at 97.4% of the maximum performance, you will 9 985 pts at Time Spy, CPU temp 70° and GPU temp 70% and fans at 50% (honnestly, in terms of noises, it is the day and the night)

Absolutely nothing wrong with those temps. If I owned an Ionico and it was running at full performance at those temps I would be delighted.

As for the rest of your post I have no idea what it has to do with my previous points but that may be the language barrier.

I maintain if you are having to reduce the frequency of your GPU because it is overheating and thermal throttling then something is wrong.

 

[cedrixman]

Ok. It is a dead end with you. And do not sorry about the language barrier, I perfectly undestand you

Anyway, for people who try to understand me, the perf lost vs the heat and the noise, it is absolutly life changing

 

[barlew]

Ok. It is a dead end with you. And do not sorry about the language barrier, I perfectly undestand you

Anyway, for people who try to understand me, the perf lost vs the heat and the noise, it is absolutly life changing

 

[ubuysa]

Ok. It is a dead end with you. And do not sorry about the language barrier, I perfectly undestand you

Anyway, for people who try to understand me, the perf lost vs the heat and the noise, it is absolutly life changing

If I understand what you're saying @cedrixman, it's that a small reduction in the GPU frequency lowers the temperature by a much greater amount than the small loss in performance that results? So you're saying that a small performance drop is worth it to gain a larger reduction in temperature?

That would be a worthwhile trade-off and would make good sense if it were not possible to run the GPU at maximum frequency/performance without excessive temperatures.

I think that what @barlew is saying is that a laptop running at maximum performance with the temperatures you quoted (CPU temp 80° and GPU temp 80°) is actually fine and that no attempt to reduce those temperatures is required. The suggestion is then, that you're overreacting to what are normal laptop temperatures.

Do (both) correct me if I've misunderstood.

 

[barlew]

If I understand what you're saying @cedrixman, it's that a small reduction in the GPU frequency lowers the temperature by a much greater amount than the small loss in performance that results? So you're saying that a small performance drop is worth it to gain a larger reduction in temperature?

That would be a worthwhile trade-off and would make good sense if it were not possible to run the GPU at maximum frequency/performance without excessive temperatures.

I think that what @barlew is saying is that a laptop running at maximum performance with the temperatures you quoted (CPU temp 80° and GPU temp 80°) is actually fine and that no attempt to reduce those temperatures is required. The suggestion is then, that you're overreacting to what are normal laptop temperatures.

Do (both) correct me if I've misunderstood.

Exactly this.

My only other point is that if @cedrixman GPU is consistently thermal throttling there is likely something wrong with the laptop and potentially technical support is required.

This is the first time I have seen reports of the GPU in the Ionico thermal throttling which leads me to the conclusion it is not standard behaviour for that chassis.

 

[cedrixman]

Exactly this.

My only other point is that if @cedrixman GPU is consistently thermal throttling there is likely something wrong with the laptop and potentially technical support is required.

This is the first time I have seen reports of the GPU in the Ionico thermal throttling which leads me to the conclusion it is not standard behaviour for that chassis.

No, it is standard but it is less noticable than CPU throttling as you will not easily see it. The GPU can ran at 2GHz but you will never see it at this frequency. And you will see a stabilisation at around 1,6/1,7GHz. This is the way the laptop works today. A balance between CPU and GPU

 

[barlew]

No, it is standard but it is less noticable than CPU throttling as you will not easily see it. The GPU can ran at 2GHz but you will never see it at this frequency. And you will see a stabilisation at around 1,6/1,7GHz. This is the way the laptop works today. A balance between CPU and GPU

Ok well this is a point we will have to disagree on because I can assure you that GPU throttling is not normal and usually is indicative of a bigger problem.

If it was standard these forums would be full of people discussing it in the same way they discuss CPU throttling.

 

[SpyderTracks]

No, it is standard but it is less noticable than CPU throttling as you will not easily see it. The GPU can ran at 2GHz but you will never see it at this frequency. And you will see a stabilisation at around 1,6/1,7GHz. This is the way the laptop works today. A balance between CPU and GPU

As @barlew correctly says, if the gpu is throttling at all at full frequency, it’s because of a paste application issue or the heatsink isn’t applied correctly, it needs fixing. You’re nerfing performance rather than addressing the underlying issue.

 

[cedrixman]

If I understand what you're saying @cedrixman, it's that a small reduction in the GPU frequency lowers the temperature by a much greater amount than the small loss in performance that results? So you're saying that a small performance drop is worth it to gain a larger reduction in temperature?

That would be a worthwhile trade-off and would make good sense if it were not possible to run the GPU at maximum frequency/performance without excessive temperatures.

I think that what @barlew is saying is that a laptop running at maximum performance with the temperatures you quoted (CPU temp 80° and GPU temp 80°) is actually fine and that no attempt to reduce those temperatures is required. The suggestion is then, that you're overreacting to what are normal laptop temperatures.

Do (both) correct me if I've misunderstood.

You are absolutly right. 80deg both CPU and GPU is perfectly fine. I know that. My only point is that I see, personally, a real interest to undervolt and underclock a bit the GPU as you will see a huge difference in terms of GPU + CPU temp and less fan noise (i mean a lot). And of course, a better longevity of your laptop.

And if you do not want to reduce the frequency, you can still undervolt. Personally at 0,950mV and you will see a difference in terms of temps + fan noise.

And I forgot to say that I want to do that because this laptop is so powerful that I do not need this amount of power. In 2/3 years, it will may not be the same story as I will need maximum power.

That’s all

 

[Macco26]

If I understand what you're saying @cedrixman, it's that a small reduction in the GPU frequency lowers the temperature by a much greater amount than the small loss in performance that results? So you're saying that a small performance drop is worth it to gain a larger reduction in temperature?

That would be a worthwhile trade-off and would make good sense if it were not possible to run the GPU at maximum frequency/performance without excessive temperatures.

I think that what @barlew is saying is that a laptop running at maximum performance with the temperatures you quoted (CPU temp 80° and GPU temp 80°) is actually fine and that no attempt to reduce those temperatures is required. The suggestion is then, that you're overreacting to what are normal laptop temperatures.

Do (both) correct me if I've misunderstood.

Keyword here is: at 80° it required 100% fans, while in the other case it required 50% fans, which is huge (to the ears at least), I guess.

 

[cedrixman]

As @barlew correctly says, if the gpu is throttling at all at full frequency, it’s because of a paste application issue or the heatsink isn’t applied correctly, it needs fixing. You’re nerfing performance rather than addressing the underlying issue.

Ok. Experience is better than worlds. Could you try a Time Spy benchmark for example running Afterburner and look at GPU frequency. You will understand my point. You will not see a frequency at 2GHz during all the test. The system always throttles to maintain correct temps, always as the maximum GPU acceptable temp is around 85deg if I am correct

And it is absolutly right as there is so much power in a small laptop. It is the standard behaviour of today’s laptop

 

[cedrixman]

Keyword here is: at 80° it required 100% fans, while in the other case it required 50% fans, which is huge (to the ears at least), I guess.

Exactly my point + laptop longevity

 

[star_dodger777]

My Elimina Pro RTX 3060 Max-P goes up to 2047MHz at 1080mV , but with temps at 87C (the limit), max fans, max everything, I tested on AC Odyssey. I didn't think it's healthy long term so I tried to undervolt as well and find a 'sustainable' profile. I hard capped the core clock in afterburner to 1830MHz, undervolting to 850mV, the CPU is at 40W on CC, I adjusted the CPU power down until I started seeing big difference in performance, fans at 85%+cooling pad, I get around 73C max GPU temp, the GPU power draw is 100W (maximum is 130W). With this profile I see no big dip in performance playing the game, and the fps drop is really small. Do you think it's sustainable to play for several hours at 87C and max fans?

 

[Eldavo]

I don't think it's healthy in the long term, but I also think you can't avoid 80+ °C temperatures with a gaming laptop.