English Community

ThinkPad NotebooksThinkPad: P and W Series Mobile Workstations
All Forum Topics
Options

129 Posts

11-12-2019

IL

225 Signins

3495 Page Views

  • Posts: 129
  • Registered: ‎11-12-2019
  • Location: IL
  • Views: 3495
  • Message 1 of 2

P53 CPU power management and cooling bugs

2019-12-10, 4:35 AM

Hi,

 

I have just received 2 identical P53 computers for my clients. Both computers have the following specs:

i7-9850H CPU, RTX3000 GPU, 2x 16GB RAM, Samsung 512GB NVME SSD, Windows 10 Pro 1903

 

I updated all the drivers on both machines and changed a few minor settings in BIOS but nothing related to CPU power. I also updated the BIOS to the currently latest version - 1.18. All the relevant power settings in Windows are configured correctly. The active power scheme is "Lenovo Intelligent Power", and the Power management slider is set to "Best Performance".

 

 

While testing the computers, I noticed several identical problems with both of them.

 

1) Occasionally the CPU power gets stuck at 50W long-term limit. Normally the allowed long-term power limit for CPU appears to be 60W and usually both machines can maintain this long-term in CPU stress tests. However sometimes, for some reason, they decide to change that limit to 50W and once that goes into effect, there is no way to make them go back to the 60W power limit. Even if I leave them alone for 5-10 minutes and everything cools down to 40C, the 50W power limit stays in effect and shows up in the next test. The CPU will in such case boost to the proper short-term allowed maximum (sometimes they temporarily boost to 87W) but after several seconds the CPU drops to 50W and stays there regardless of the temperature. The problem does not show up every time and despite my attempts, I could not figure out what triggers this behavior. Sometimes this happens at first run of the stress test, sometimes doesn't happen at all despite a very long stress test, sometimes it shows up if starting a test while the machine is relatively hot. The only trigger that I've noticed is a start of a test. The 50W limit never shows up if a test has been running for a minute or two and the CPU has settled at 60W. It may show up later if a running test is stopped and a new one started. Also in some cases it showed up when starting a CPU test after a prolonged GPU test.

 

2) Fan curves are not right. There is a very long delay between a temperature event and the fan speedup or slowdown. Effectively, the fans follow a temperature that was actual a long time ago, so in some cases if I stop a stress test, the fans will continue to run very loudly for several minutes while the CPU is completely cool, and if I run a stress test again, that is when the fans suddenly decide to slow down. Sorry Lenovo but this makes no sense at all. You can do much better than that. It's fine if the fan ramp up logic is not very abrupt. However at some point the fans need to catch up to what's going on with the temperatures. If the temperature just went up to 100C, it makes no sense for the fans to slow down.

 

3) The Lenovo Intelligent Thermal Solution (ITS) service is not functioning properly. This has been discussed at length in other threads and I can confirm the problem with these 2 new machines. Once the computer has been picked up and moved a bit, either the CPU power gets stuck at 20W long-term and does not recover from that, or the CPU frequency gets stuck at 800MHz and does not recover. The only way to make it recover is to restart the Lenovo Intelligent Thermal Solution Service. Sorry but this is just silly. 800MHz in a workstation machine? 20W CPU power limit that does not recover in a 45W CPU? Did anybody do any testing at all? Is there any QA for the software that Lenovo produces?

 

4) There is A LOT of potential to improve performance in these machines - it's only a matter of proper fan curves, basic testing of all the too intelligent software and of firmware tuning to allow the CPU to max out the thermal headroom. Currently all of that is not implemented properly and hence a P53 with a big and heavy cooling system, huge fans and proper big vents can not compete in CPU performance against a crimped machine such as a Macbook. You can do much better than that and users of your machines expect you to do better. We are not your QA department! Please hire coders who know what they're doing and QA who can actually do testing!

 

P.s. both machines came with power supplies that suffer from loud and obnoxious coil whine. This is not acceptable as well and those power supplies will have to be replaced. But at least that has nothing to do with the performance.

Reply
Options

1 Posts

12-10-2019

US

1 Signins

15 Page Views

  • Posts: 1
  • Registered: ‎12-10-2019
  • Location: US
  • Views: 15
  • Message 2 of 2

Re: P53 CPU power management and cooling bugs

2019-12-10, 11:57 AM

Yes, the fan control on this machine has a very long delay. It takes a few minutes for the fans to start, and the fans keep spinning for a few minutes even after the temperature is back to normal.

 

A fix for this would be nice.

 

Reply
Forum Home

Community Guidelines

Please review our Guidelines before posting.

Learn More

Check out current deals!

Go Shop
X

Save