1260p performance woes

Framework Laptop 13
#1

I have a Framework with the 1260p motherboard running Kubuntu 22.04. When I try to run a heavier task, the laptop throttles quite quickly.

I first noticed this while running consistency on a model check in FreeCAD .020.1. This is what that looks like:

20221108_160328_freecad
20221108_160328_freecad2162×1504 461 KB

I then did a little stress test with stress --cpu 16

20221108_155259_stress_test_16c
20221108_155259_stress_test_16c2162×1504 300 KB

Both tests were conduced while plugged in and while using the performance power profile.

The length of the graph area is 240 seconds. I think that the laptop starts throttling too soon under load if that is after only 120 to 150 seconds.

Any way I can fix this?

1 Like
#2

I had some success with better thermal compound like other users here:
https://community.frame.work/t/honeywell-ptm7950-phase-change-thermal-pads-sheets-application-tips-and-results/

But I would think we need a bigger cooler. I have yet to try putting an AIO or a bigger heatsink to see if it helps.

1 Like
#3

While your temperatures look a little high, the max clock looks about right. Here are the results from my 1260P running stress -c 16 with results from turbostat. It ends up settling at about 2.5-2.6GHz. I ran this in my normal config, open on an elevated laptop stand on Arch Linux w/ Sway (curent ambient temp: ~25C)

I’ve included the exact turbostat flags I’m using because otherwise its output can be a bit much:

# turbostat -c package -S --show Avg_MHz,Bzy_MHz,TSC_MHz,PkgTmp,PkgWatt,CorWatt

Avg_MHz Bzy_MHz TSC_MHz PkgTmp  PkgWatt CorWatt
5       1460    2496    37      0.91    0.25
11      2079    2496    37      0.99    0.49
5       1668    2496    37      0.95    0.27
5       1550    2496    37      1.00    0.26
523     3269    2496    69      8.61    7.60
3339    3347    2496    82      50.21   48.63
3322    3329    2496    89      51.82   50.00
3240    3248    2496    97      49.72   47.76
3227    3235    2496    100     49.93   47.96
2773    2780    2496    85      35.20   33.22
2556    2562    2496    82      27.74   25.89
2566    2572    2496    81      27.87   26.05
2568    2574    2496    79      27.93   26.09
2573    2579    2496    79      27.97   26.15
2577    2583    2496    78      27.98   26.20
2575    2589    2496    76      27.99   26.25
2579    2585    2496    78      28.00   26.24
2580    2586    2496    76      28.00   26.19
2583    2589    2496    78      28.00   26.28
2582    2588    2496    76      28.00   26.22
2582    2588    2496    76      28.00   26.21
2573    2595    2496    78      28.00   26.26
2584    2590    2496    76      28.00   26.29
2583    2589    2496    78      28.00   26.26
2584    2590    2496    78      28.00   26.29
2583    2589    2496    76      28.00   26.25
2581    2587    2496    76      28.00   26.22
2581    2587    2496    76      28.00   26.21

Things to note: the hard PL1 (sustainted) limit for the framework is 28W. If you are reaching that, then you are not being thermally limited. The PL2 is 64W, but the time window for that is 2440 micro-seconds (~0.0024s) if I’m reading it right. You can see that for my CPU, basically it holds about 50W for about 20s (turbostat samples once every 5s), and then drops down.

AFAIK the most legible way to your PL settings:

# powercap-info -p intel-rapl
enabled: 1
Zone 0
  name: package-0
  enabled: 1
  max_energy_range_uj: 262143328850
  energy_uj: 8034164002
  Constraint 0
    name: long_term
    power_limit_uw: 200000000
    time_window_us: 27983872
    max_power_uw: 28000000
  Constraint 1
    name: short_term
    power_limit_uw: 64000000
    time_window_us: 2440
    max_power_uw: 0
  Constraint 2
    name: peak_power
    power_limit_uw: 90000000
    time_window_us: 0
    max_power_uw: 0
  Zone 0:0
    name: core
    enabled: 0
    max_energy_range_uj: 262143328850
    energy_uj: 5141981221
    Constraint 0
      name: long_term
      power_limit_uw: 0
      time_window_us: 976
  Zone 0:1
    name: uncore
    enabled: 0
    max_energy_range_uj: 262143328850
    energy_uj: 7017071
    Constraint 0
      name: long_term
      power_limit_uw: 0
      time_window_us: 976

Anyway, if turbostat show’s you aren’t able to sustain 28W, then you should try to see if better ventilation helps or a repaste. Otherwise, you’re getting the most performance out of your laptop as is possible.

BTW, you may also want to look at turbostat's full output to see how your idle is (C10%). When my laptop isn’t doing anything, it’s at ~99.8% idle. This is a lot more important for idle power usage but still, if you’re not mostly at idle when you’re not doing anything, you’re probably giving up some performance/thermal headroom.

4 Likes
#4

I’m having the same thoughts than OP and tried to replicate lhl’s turbostat-values with stress -c 16 but mine are looking different:

36	1419	2496	47	2.33	0.70
40	1636	2496	48	2.39	0.83
38	1388	2496	47	2.42	0.75
43	1518	2496	48	2.59	1.03
42	1283	2496	47	2.61	0.78
114	2090	2496	48	4.73	2.69
124	2048	2496	50	5.44	3.04
82	1624	2496	49	4.07	1.71
72	1395	2496	48	3.65	1.37
76	1622	2496	49	3.85	1.63
65	1516	2496	49	3.40	1.30
109	2056	2496	49	4.98	2.84
128	2308	2496	53	6.02	3.79
75	2088	2496	49	3.99	2.12
23	1475	2496	48	1.95	0.65

40	1900	2496	48	2.30	0.89

14	855	2498	48	1.90	0.41
136	885	2640	48	4.71	1.31
139	995	2463	48	5.49	1.78
129	1065	2473	49	3.41	1.15

12	726	2492	50	2.77	0.44
33	1530	2496	48	2.16	0.64
61	2520	2496	48	3.62	2.26
490	3232	2496	86	7.99	6.65
3272	3315	2496	96	42.27	40.24
3229	3236	2496	98	40.81	38.75
3060	3067	2496	100	36.48	34.09
2933	2955	2496	100	34.32	31.56
2884	2905	2496	100	33.32	30.49
2890	2896	2496	100	31.76	29.67
2801	2831	2496	100	31.06	28.56
2838	2848	2496	100	30.67	28.37
2838	2845	2496	100	30.01	28.13
2830	2837	2496	100	29.76	27.94
2796	2802	2496	100	29.33	27.37
2782	2788	2496	100	29.07	27.22
2746	2783	2496	100	28.57	26.69
2765	2773	2496	98	27.81	25.97
2767	2774	2496	100	27.88	26.00
2770	2776	2496	100	27.94	26.11
2747	2759	2496	100	27.96	25.78
2715	2740	2496	100	28.02	25.69
2628	2672	2496	100	27.67	24.94
2650	2657	2496	98	27.34	24.69
2698	2713	2496	100	27.52	25.27
2726	2733	2496	100	27.74	25.81
2732	2738	2496	100	27.71	25.82
2692	2709	2496	100	28.04	25.19
2685	2712	2496	98	26.79	24.81
2665	2692	2496	98	26.83	24.70
2684	2690	2496	100	27.13	24.54
2697	2703	2496	100	26.65	24.76
2674	2690	2496	100	26.85	24.60
2665	2697	2496	100	26.95	24.66
2630	2663	2496	100	27.35	24.20
2649	2658	2496	100	27.20	24.11
2653	2659	2496	100	27.08	23.99
2627	2635	2496	100	26.53	23.71
2632	2643	2496	100	26.53	23.76
2629	2650	2496	100	26.61	23.69
2659	2665	2496	100	26.02	23.90
2672	2678	2496	100	25.89	24.05
2664	2670	2496	100	25.99	24.00
2660	2669	2496	100	26.06	23.96
2649	2655	2496	100	26.20	23.75
2660	2666	2496	98	25.70	23.81
2630	2643	2496	100	25.78	23.56
2648	2672	2496	100	26.03	23.97
2666	2672	2496	100	26.05	23.99
2652	2661	2496	98	25.94	23.73
2649	2660	2496	100	25.66	23.64
2659	2665	2496	100	25.90	23.79
2650	2656	2496	100	25.78	23.60
2661	2667	2496	100	25.74	23.85
2667	2673	2496	100	25.94	24.04
2662	2668	2496	100	25.64	23.90
2631	2638	2496	98	26.34	23.65
2604	2626	2496	100	25.87	23.29
2591	2616	2496	98	25.85	23.17
2632	2638	2496	100	25.93	23.42
2583	2603	2496	100	25.96	23.29
2606	2626	2496	98	25.67	23.35
2464	2472	2496	99	24.26	21.34
2628	2642	2496	100	26.69	23.94
2647	2656	2496	98	26.02	23.71
2625	2638	2496	100	25.92	23.52
2601	2626	2496	100	25.98	23.47
2651	2657	2496	98	25.69	23.64
2649	2655	2496	100	26.11	23.71
2653	2662	2496	98	25.75	23.84
980	2682	2496	99	13.90	11.82
175	2363	2496	73	8.71	6.11
93	1830	2496	68	4.85	2.60
120	2775	2495	67	6.93	4.89
186	2471	2497	82	9.44	6.87
143	1951	2496	66	7.24	4.71
161	2081	2496	70	8.18	5.53
236	2769	2496	100	12.81	10.22
314	3663	2496	99	17.98	15.75
336	3680	2496	100	17.89	15.60
336	3597	2496	98	17.55	15.10
313	3478	2496	72	16.57	14.18
137	1712	2496	66	6.27	3.62
136	1719	2496	62	6.18	3.56
116	1667	2496	60	5.35	3.05
107	2483	2496	59	6.08	4.22
55	1407	2496	58	3.15	1.10
125	2853	2496	62	7.24	5.32
56	2010	2496	56	3.61	2.06
35	1687	2496	55	2.16	0.74
34	1396	2496	55	2.25	0.60
26	1264	2496	53	2.20	0.53
35	1286	2496	53	2.50	0.68

Ambient-Temp around ~20C, Connected to charger, powercap-info is the same, running on NixOS / Kernel 6.4.3. Would you say this indicates thermal throttling? Just got it in Feb 2023 so I would be surprised to already have to service it.