CPU Benchmarks and TDP

Framework Laptop 16
1

I just did a couple of Cinebench 24 runs, and got 898 points.
While running the CPU was reported at 100°C and 40W TDP. The laptop pulls 80-90W from the wall.
I expected the TDP to be above 45W.
Could someone run the same benchmark and report?

I used AMD adrenaline for monitoring.

7840HS and no dGPU.

1 Like
2

Power plan?

3

Balanced and Max Power produce the same result. Energy saving uses less power, produces less heat and less points.

4

Could be throttling back at 100C? AVX512 can be a hard load.

Don’t have my FW16 yet, although it’s tantalisingly close…

5

I expected it to thermal throttle, I’m wondering about the reported 40W TDP, because that is below spec

6

TDP is a thermal envelope measure for sizing cooling requirements, rather than a performance measure. You may have got a better binned chip that needs less voltage. Less voltage = less power = better battery life.

How are the clock speeds looking when under load?

7

I know, and I got 4.1Ghz.
That’s why I’m asking for people to run the benchmark so I can see how it should perform.

1 Like
8

In case anyone ever cares: I opened the laptop up again and tightend the 4 heatsink screws. Now I get 45W TDP, 4.2-4.4 GHz and 904 points in Cinebench.
Lot less performance increase than I expected, but I guess it’s fine.

1 Like
9

Running cinebench r23 in wine under Ubuntu 22.04 with all drivers installed, I got 14928 points multi-core and 1711 points single-core. Temperatures were rather low (75°C range, power around 45W), the fans were barely audible. 7840HS, 1x 32GB RAM, no GPU
Single core power consumption was between 15W and 22W.

1 Like
10

Maxon revised the scoring system for Cinebench 2024 for it is completely different compared to Cinebench r23 scores. If you have the time, could you send the Cinebench 2024 scores?

11

Sadly, r24 doesn’t run with wine (yet) and I don’t have windows on the FW16.

ekkehard@FW16:~/.wine/drive_c/Program Files/cinebench24$ wine Cinebench.exe 
0084:fixme:hid:handle_IRP_MN_QUERY_ID Unhandled type 00000005
0084:fixme:hid:handle_IRP_MN_QUERY_ID Unhandled type 00000005
0084:fixme:hid:handle_IRP_MN_QUERY_ID Unhandled type 00000005
0084:fixme:hid:handle_IRP_MN_QUERY_ID Unhandled type 00000005
0110:fixme:ntdll:NtQuerySystemInformation info_class SYSTEM_PERFORMANCE_INFORMATION
0110:fixme:process:SetProcessPriorityBoost (FFFFFFFFFFFFFFFF,1): stub
0110:fixme:combase:RoGetActivationFactory (L"Windows.System.Profile.SystemIdentification", {5581f42a-d3df-4d93-a37d-c41a616c6d01}, 00007FFFFE1FD980): semi-stub
0110:err:combase:RoGetActivationFactory Failed to find library for L"Windows.System.Profile.SystemIdentification"
0110:fixme:nls:get_dummy_preferred_ui_language (0x8 0x407 00007FFFFE1FDA30 00007FFFFE1FD1E0 00007FFFFE1FDA34) returning a dummy value (current locale)
0110:fixme:process:NtSetInformationProcess (0xffffffffffffffff,0x0000004d,0x7ffffe1fe5b0,0x0000000c) stub
0118:fixme:win:DisableProcessWindowsGhosting : stub

(wine:88568): GStreamer-Video-CRITICAL **: 06:25:22.307: gst_video_info_from_caps: assertion 'gst_caps_is_fixed (caps)' failed
12

Win 11 Pro 64 GB 5600 Crucial 7840HS

  • I ran and was on my Dell 130W 3 runs avg was 764 which i thought was low so
  • I ran and was on the FW 180W 3 Runs avg was 839

it WAS NOT thermal throttling on the 130 but was on the 180.
will check my HS screws tomorrow.
HS screws where as tight as i was willing to go.

13

Pretty close to what I had in Fedora, using proton in steam.

1668 single core and I think it was ~14205 multicore (didnt want to run again).

22w in single core
42w in multi core

Tctl temp showed 100… but Laptop never got hot… I dont think tctl is very accurate.

I will say, if I play helldivers 2, I get 54w out of the CPU and it holds there as long as I play it.

14

Win 11 Pro 64GB 7940HS no dGPU

Cinebench 2024 - 825 multi and 104 single.
Cinebench R23 - 13529 multi and 1766 single.
Cinebench R20 - 5235 multi and 692 single.

Comparing these numbers to what I can see over at notebookcheck tells me I am thermal throttling like crazy. I am able to hit the expected single core benchmarks but I fall very short of the multi core numbers. I was watching in HWInfo and HWMonitor to see what was going on and core #4 would hit 100C within a few seconds. I also noticed cores #2 and #6 would also get very hot at 97-98C. The remaining cores were anywhere from 78-92C. The package power was never able to climb over 31W when using the 180W charger in any port on the High Performance power profile. Considering this is marketed as a 45W TDP package, I am hoping this is just a case of a bad thermal interface somewhere.

I was originally using Ubuntu 22.04 and watching temps in Tctl and thought the report of 100C was incorrect since the laptop didn’t seem that hot. I wanted to investigate further since my clocks really couldn’t get past 3.6-3.7Ghz, so I installed Windows 11. This allowed me to see more granular data and I was very surprised to learn Tctl was in fact reporting an accurate package temp.

Now that I am running Windows, I was able to run the tests on a few different power profiles. I found that with my current thermal throttling issues, I would get near the same result using the Balanced profile as I did with the High Performance profile. I even observed core #4 hit 100C in Balanced mode while disconnected from the AC power. During the benchmarks, I observed about 91-92% CPU usage due to the thermal throttling issues.

15

May be worth it to test what power load your CPU can sustain with a non-AVX load.
IIRC AVX layouts have pretty high density on the die, which means that there can be hotspots which cause the throttling.

Are your fans also ramping up correctly?

16

I went ahead and tried Prime95 with AVX-512 disabled and I can sustain about 39W that way. Still seeing 2 cores throttle at 100C due to hot spots. I also tried with AVX2 disabled and saw the same result. Completely disabling AVX saw my power drop to about 37W. I did all tests using smallest FFTs.

My fans do ramp up after a few seconds.

I understand AVX is hard on the processor, but tools I use each day like x265 utilize it. Considering there are benchmarks showing what the 7940HS can do under an AVX-512 load, I am hoping there is some headroom to be found in the current cooling solution. I would be curious to know if AVX loads were considered during the cooling validation process.

For what it is worth, I am communicating with support about this and they are taking it seriously.

17

In addition I would look for the status on the thermal paste, whether the laptop itself get ample cooling (maybe elevate the intakes slightly), whether the removal of the palmrest helps (it should, consider these are turbine fans)

I recall in LTT’s video the FW16 thermal solution helped by Cooler Master does pretty well. Are both fans spinning? Are both fans spinning fast?

18

removing the palmrest is not gonna do anything, you would need to remove the midplate.

19

Not removing the heatsink they are using Liquid Metal and don’t want to screw it up. Not that worried about it currently.

It games fine right around where I expect it to be.

fans seem to work fine as well. They never seem to spinning loudly to me at least.

20

Phoronix made some new benchmarks with the FW16 - Overall, under Ubuntu 24.04 16% faster then Windows 11 …