News AMD Ryzen 4000 Benchmark Results Show 30 Percent Performance Gap Between 4800U and 4900HS

Gillerer

Distinguished
Sep 23, 2013
312
33
18,890
30
I wish AMD would release a 15W TDP processor for socket AM4. Fanless computers are so much nicer.
You can get pretty close with an Athlon 3000G (35W, Zen 1) - in TDP-down mode it's 25W.

If it's used for light tasks like browsing or video playback, just slap on a large sparse heatsink in a well-ventilated case and let it simmer at 70 - 95°C. If the temperatures exceed that, the APU will throttle itself before it gets damaged.
 
Last edited:

watzupken

Notable
Mar 16, 2020
333
116
870
1
I feel if the 15W chip comes with a dedicated GPU, it is possible the CPU gap will be smaller. Problem with the 15W APU is that the CPU and GPU don't have a lot of power legroom to stretch. As a result when under load, both suffers. With a dedicated GPU, the CPU can have all the resources and power to give it a higher boost.

Anyway, it is good to see that the gap between the 2 is actually quite small. Shows how much benefit in terms of power savings is realized as a result of a move to 7nm. After all, Zen 2 chips is supposed to be a die shrink from the Zen 1 architecture with some tweaks to improve IPC.
 

Gillerer

Distinguished
Sep 23, 2013
312
33
18,890
30
Anyway, it is good to see that the gap between the 2 is actually quite small. Shows how much benefit in terms of power savings is realized as a result of a move to 7nm. After all, Zen 2 chips is supposed to be a die shrink from the Zen 1 architecture with some tweaks to improve IPC.
I think Zen 2 is far more than that. It was Zen+ (12nm, found in the 3000 series 2nd gen Ryzen APUs) that was a shrunk and optimized version of Zen 1 (14nm, found in the 2000 series 1st gen APUs).

*

Cooling and heat mass of laptops will heavily affect results from short intensive tests.

If a laptop CPU has a large effective heatsink (from being connected to a metal chassis, for instance), it can retain high clocks for a couple of minutes. Fan speeds or curves will also affect results. If the test has a limited duration and the system hasn't been pre-heatsoaked, that test can't be a realistic measure of performance under longer loads.

Since we don't know what kind of systems the CPUs were running (no doubt different systems and different size systems), or whether fan speeds or fan noise were normalized (very likely not), these results don't have any other meaning than "depending on the systems and setup, these CPUs can have this kind of relative performance".
 

ASK THE COMMUNITY