• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Dell XPS 7590 with 9750H, possible to increase TDP/OC beyond 45W? Also, is it possible to set TPL based on battery vs AC?

Joined
Mar 3, 2020
Messages
13 (0.01/day)
Using TS UV, repasting, and some thermo mods, I have the 7590 operating nicely. I can sustain 45W (TS reporting around 50) with no throttling under manageable temps. I am now curious whether it is possible to push the 9750H to 60W. If yes, how do I go about doing that with TS? I have the computer plugged in with a 130W adapter, have the Windows performance bar to best, disabled TS FIVR Disable and Lock Turbo Power limits, unclamped TPL, and set Speed Shift to ~60. Still, the computer does not draw more than 45-50W. Is there something else I should do?

On another question, the FIVR profiles are set between battery and AC profiles (which, at this time, are the same). I have the TPL long and short turbo set mostly to conserve battery. Is there a way, when plugged in, to allow the turbo short to run longer and at higher TPL?

TS is still new to me, so I may be barking up the tree in a wrong way. Please feel free to provide any guidance that might help me understand better how TS should be used.

Thank you.
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,059 (1.33/day)
Long term, some Dell laptops have decided to enforce the 45W TDP limit. There is no way around this limit. Other laptops with the same 9750H can go up to 80W or a little beyond.

disabled TS FIVR Disable and Lock Turbo Power limits
This item should be checked. Post some screenshots of how you have ThrottleStop setup. Even with this item checked and the turbo power limits set higher in the TPL window, there is still a third set of hidden turbo power limits that Dell uses to enforce the 45W limit.

At the moment, the turbo power limits cannot be set to different values for each profile. That feature will be added someday soon.

1612380592603.png


Post a screenshot of your settings and post a screenshot while your CPU is loaded running a benchmark like Cinebench R20.


Open up Limit Reasons. Does it show PL1 or PL2 lighting up red under the CORE column? Does ThrottleStop report 45W when this is happening?
 
Joined
Mar 3, 2020
Messages
13 (0.01/day)
Long term, some Dell laptops have decided to enforce the 45W TDP limit.

At the moment, the turbo power limits cannot be set to different values for each profile. That feature will be added someday soon.

Post a screenshot of your settings and post a screenshot while your CPU is loaded running a benchmark like Cinebench R20.

Open up Limit Reasons. Does it show PL1 or PL2 lighting up red under the CORE column? Does ThrottleStop report 45W when this is happening?

Many thanks for your reply. It is great to know that there will be individual TPL settings soon. That will be a great feature to further tune performance based on AC vs battery

I have not yet downloaded Cinebench, but used TS Bench to load the CPU. I am attaching several screenshots and I will caption them as appropriate:

home.png
:

Above is the TS home. I believe the 90.1W Max is erroneous. I have never seen values come near that. It would also not reconcile with the Max temp.


options.png


Above are the options settings. I offset the PROCHOT by 2 degrees for testing.

fivr.png


Above is FIVR, note that Disable and Lock TPL is not selected at this time. This is by default setting for battery life. When I was trying to push the CPU to higher TPL, I did select/check this box.


tpl.png


Above my TPL default setting. Again, designed to conserve battery while maintaining acceptable performance.

limit.png


Above limits appears because I had the Turbo Time Limit set to 28. This is while running TS Bench.


limit 2.png


Above was run with these config changes from default: Speed Shift-EPP to 70, checked Disable and Lock Turbo limits, unchecked TPL long and short power max clamp, set Turbo Time Limit to 1024. It seems to run TS Bench at ~50W while maxing out at ~94 degrees. However, I can never get the chip to draw more power than that.

With the forgoing, I presume the 7590 is capped at 45W? In your post, you wrote "...there is still a third set of hidden turbo power limits that Dell uses to enforce the 45W limit." Is that something that can be changed?

Again, I know (and understand) very little about TS and CPU operations. I would appreciate any guidance. My goal is to conserve while on battery and have two levels of performance while on AC (max TDP power and moderate power, because I use adapters with different power output).

Thanks again.
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,059 (1.33/day)
I presume the 7590 is capped at 45W?
Your last screenshot does not show that. Your CPU is running fully loaded at 52.2W. It is not being capped. Nothing is lighting up in Limit Reasons to indicate any sort of throttling. Your CPU is running at the full 40.00 multiplier. That is as fast as your CPU can run when all 6 cores are active. If you run this test for a longer period of time, does it start to throttle to 45W? Does PL1 light up red?

However, I can never get the chip to draw more power than that.
Why would you want your CPU to draw more power? In the TS Bench test that you are running, it does not need any more power than that. This test is already running at full speed.

If you really want to see your CPU use more power then try running a more demanding test.

Cinebench R20

If you want to see a CPU consume a crazy amount of power then step up to the Prime95 Small FFTs test.

On my desktop CPU, Prime95 draws an obscene amount of power. Surprised that the lights don't dim a little.

1612546479319.png
 
Joined
Mar 3, 2020
Messages
13 (0.01/day)
Your CPU is running fully loaded at 52.2W. It is not being capped. Your CPU is running at the full 40.00 multiplier. That is as fast as your CPU can run when all 6 cores are active.

If I understand it correctly, regarding your comment "That is as fast as your CPU can run when all 6 cores are active", that the 9750H in my computer was running as fast is it can while all 6 cores are running?

I may have a misunderstanding between CPU performance and TDP (regarding my comment about drawing more power). I thought that the more power the chip draws, the faster it is running. Perhaps this is not true?

If you run this test for a longer period of time, does it start to throttle to 45W? Does PL1 light up red?

I will try that later and report back.


Why would you want your CPU to draw more power? In the TS Bench test that you are running, it does not need any more power than that. This test is already running at full speed.

Please see my comment above regarding my possible misunderstanding.

If you really want to see your CPU use more power then try running a more demanding test.

Cinebench R20

If you want to see a CPU consume a crazy amount of power then step up to the Prime95 Small FFTs test.

On my desktop CPU, Prime95 draws an obscene amount of power. Surprised that the lights don't dim a little.

That is an insane amount of power. I don't have a desktop so that number looks unreal. So I see now that more demanding workloads will cause the CPU to pull more power. But in the case of TS Bench where you commented that the CPU was running as fast it can at ~50W, why would more demanding workloads cause the CPU to draw more power (notwithstanding GPU workloads)?

Thank you.
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,059 (1.33/day)
Different tests use different CPU instructions and different parts of the CPU. The TS Bench test is a heavy test but it does not use any of the newer AVX instructions. Software like Prime95 that heavily use the AVX instructions will consume much more power compared to the TS Bench test. Some applications like Cinebench R20 will use some AVX instructions but not as many as Prime95 uses. There is a wide range of apps that all work the various parts of the CPU differently. That is why when stress testing a computer, it is a good idea to use a variety of these testing programs to make sure that all areas of the CPU can run reliably at full speed. You can even combine programs. You could run 4 threads of TS Bench along with 8 threads of Prime95 to fully load a 6 core - 12 thread CPU.

The maximum multiplier possible when a 9750H has all 6 cores active is 40.00. Your CPU is running at full speed. You cannot make it run any faster than that. Your screenshot shows that your CPU needs 52W to run your CPU at full speed during the TS Bench test. No one deliberately tries to make their CPU consume more power for a given test. The opposite is true. The less power your CPU consumes, the cooler it will run. Most people use tricks like undervolting to reduce power consumption.

If I had a magic device that made a 60W light consume 90W but put out the exact same amount of light, would you be interested in that device? I sure hope not. Less power consumption is good. More power consumption is bad.
 
Joined
Mar 3, 2020
Messages
13 (0.01/day)
Different tests use different CPU instructions and different parts of the CPU. The TS Bench test is a heavy test but it does not use any of the newer AVX instructions. Software like Prime95 that heavily use the AVX instructions will consume much more power compared to the TS Bench test. Some applications like Cinebench R20 will use some AVX instructions but not as many as Prime95 uses. There is a wide range of apps that all work the various parts of the CPU differently. That is why when stress testing a computer, it is a good idea to use a variety of these testing programs to make sure that all areas of the CPU can run reliably at full speed. You can even combine programs. You could run 4 threads of TS Bench along with 8 threads of Prime95 to fully load a 6 core - 12 thread CPU.

The maximum multiplier possible when a 9750H has all 6 cores active is 40.00. Your CPU is running at full speed. You cannot make it run any faster than that. Your screenshot shows that your CPU needs 52W to run your CPU at full speed during the TS Bench test. No one deliberately tries to make their CPU consume more power for a given test. The opposite is true. The less power your CPU consumes, the cooler it will run. Most people use tricks like undervolting to reduce power consumption.

If I had a magic device that made a 60W light consume 90W but put out the exact same amount of light, would you be interested in that device? I sure hope not. Less power consumption is good. More power consumption is bad.

Great explanations. I now have a much better understanding. Thank you.

Follow up questions:

1) I read that the 9750H cannot be overclocked. Is this true? That is, could I get the CPU to run faster by overclocking?

2) Based on what you see in my FIVR and other configs, do you have any recommended tweaks or changes?

3) With my current settings, the computer is stable and passes TS Bench with no errors. Because TS Bench does not use AVX instructions, could my UV settings be too aggressive for AVX for when I use an app that uses AVX? That is, should I run Prime95 or Cinebench R20 in addition to TS Bench to verify my UV stability?
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
8,059 (1.33/day)
1) the maximum multiplier when 6 cores are active is locked to 40 in the 9750H. Only the K series have unlocked multipliers that can be increased. With H series, you can decrease the multiplier but you cannot increase it beyond the maximum set by Intel.

2) your settings look good.

3) when reducing the voltage, passing a variety of tests with 100% stability and zero errors is important. Some CPUs will increase the voltage automatically when they encounter AVX instructions. The voltage that you have set that works well in ThrottleStop might also work well in software that uses AVX instructions. You might need a slight adjustment but generally nothing major.

Cinebench R20 is a good full load test. Real world and nothing too excessive. I still like Prime95 but running 12 threads of it on a laptop is overkill in my opinion. Instead of 12 threads, maybe try running 6 threads so it is not such a brutal test. No use melting your CPU into your laptop just to say that you passed Prime95 testing. I think a couple of threads of Prime95 and a couple of threads of TS Bench at the same time might prove to be a good test.
 
Joined
Mar 3, 2020
Messages
13 (0.01/day)
Many thanks for the education and guidance. I appreciate you sticking with this thread and answering my questions. I'd like to help support your projects. Do you have a donation link?
 
Top