i7-10875H > 4800H? Really? That and other surprises comparing Aorus 15G, Apex 15, Recoil IV and TUF A15

Andy O. · August 5, 2020

First of all, I should mention that this is actually probably most suitable for enthusiasts/engineers who use laptops for professional reasons. In my article that I will link to at the end (it's quite long and wouldn't be pretty after copy/pasting), I even go so far as to restrict to "front end engineers". But I've only made that restriction because of the real world scenarios I tested and the tools I work with. Although based on what I've discovered, I see no reason why this shouldn't apply to most use cases.

I'm also not a professional reviewer. When I watched Linus' video about the other Linus' new machine recently, it reminded me of myself - I rarely pay attention to hardware news but every few years, I dig deep. But I arrived at a different conclusion compared to the other Linus. Because of my different use case.
And something that surprised me a lot related to that was the apparently distorted reporting on the current situation around at least laptop CPUs. Maybe CPUs in general? As I'm not doing this professionally (and thus lack both the audience and hardware), I won't do more research on it. But I thought I should at least take the time to write an article about my journey, so some others who happen to stumble across it won't have to go through the same. And maybe even some reviewer will notice it and reconsider their methods and rhetoric.

Here are some benchmarks I ran (based on the holes in them you can see that I didn't intend to publish something about my experiences ) that demonstrated to me not only that Ryzen 4800H is not only quite a bit inferior to the i7-10875H for my particular use case but it also seems that the reports on the alleged efficiency of AMD CPUs are inflated. Possibly because people rely on hwinfo (I measured at the wall) and don't put scores in relationship with power consumption. The 4800H certainly is more efficient but apparently by not nearly as much as reviews tend to make it seem.

There were also PCMark scores that were consistent with the CB20 single core results (i.e. 10875H beating 4800H in all categories. Aside from "video rendering" which at least on the 10875H said that "Intel UHD" was used for that test when I looked at the details.) but I unfortunately started running that way too late in the process and didn't keep track of it thoroughly. Which is why that's not in the chart.
https://docs.google.com/spreadsheets/d/1U83MGUTr2DFsbDqYuRpblUtETdlc4oi_Sq8rn-NqqRY/edit#gid=0

A wrong assumption that I made going into this was that multi-core performance matters more than single-core. Why did I assume that? Constant talk about how AMD 4000 mobile CPUs "destroy" Intel 10th gen. But as I had to realize - no, most applications still don't perfectly parallelize their workloads. The only people for whom this holds true are 3D artists, video editors, people using C compilers I guess (Linux kernel), etc. - which I'm guessing isn't more than 20% of all laptop users. Come to think of it - how many of that target audience might do at least most of their work on a desktop anyway?

Anyway, here is the full article if you'd like to read about my journey in more detail. It also contains a links to fan control tools that allow to fully customize the fan speeds for the Gigabyte Aorus 15G as well as comments on the various machines (and their manufacturers/retailers) I had to try out before realizing that the 15G is the best compromise, really mostly thanks to those custom fan control options.