Gaming Laptop; core count vs clock speed / DDR5 5600 vs 6400

[LmnSour]

Performance investigation on the intel 13900HX using Warhammer II benchmarks.

 

Theory: Due to limited power available, and with the GPU able to dynamically boost up to 175W, reducing CPU cores should enable more GPU power and higher boost clocks, not only directly, but indirectly by reducing the heat load and fan speeds.  

 

Laptop: Eluktronics Mech 17 GP2, 13900HX, RTX 4090, 32GB DDR5 5600 CL40 and 6400 CL38

 

Tests done: 6-P cores and 4-E cores with 5600 and 6400 Ram vs all cores (8p and 16e) with 5600 and 6500 ram.  Ran the GPU undervolted at 2335 and +900 core.  Only did one run, no averages; I'm not getting paid for this 

 

Ran the Warhammer 2 benchmarks Battle, Campaign, and Skaven, at Ultra settings, DX11, and 2k resolution with AA disabled an Gsync/vsync disabled.  Max fans with the LPP watercooling and used Throttlestop to keep P-core clocks at 5200-5300 and E core clocks at 3900.  P-core undervolted by .145v and e cores by .130v.

 

First, ram benchmarks:

 

5600CL40 1.1v

 

6400CL38 1.35V

 

---------------------------------------------------------------------------------------

 

For the results we have the individual results plus the averages of the 6e, 4p and the averages of the 5600 and 6400 ram speeds:

 

Battle Benchmark:

 

 

*Discussion: I was not expecting the 6400 memory to perform so poorly here.  I think the issue with this benchmark is the power draw and the DDR5 6400 memory at 1.35V might be reducing the available power budget to the CPU/GPU?  Cant be sure but I could try in the future to set the fans to auto and see how that affects performance in this benchmark. But we can clearly see that disabling the two p-cores and 12 e-cores netted a performance improvement.

 

Campaign Benchmark:

 

 

*Discussion:  This is about what I was expecting; disabling 2 p-cores and 12 e-cores definitely saw an improvement in performance and the higher ram speed also saw a considerable improvement.

 

Skaven Benchmark:

 

 

*Discussion: Here we see the opposite of the Battle Benchmark where the ram speed saw an improvement but the extra cores helped more than the extra power budget.  I think this benchmark is more CPU intensive than the other two (Skaven battles always throttle my CPU) and the extra cores and extra memory bandwidth certainly saw a large improvement.

 

I might experiment with 8 p-cores and 4 e-cores.  I know inevitably someone will ask to disable all e-cores but TBH, I like the idea of some e-cores for background applications.  I'm trying to figure out how to overclock this laptop because there is a LOT of headroom in the boost clocks.  For some reason, clocks seem to be capped at 5.3.