The 8 Core King – Ryzen 5900HX vs 10980HK vs 4800H vs 5800X
Hidden inside this already impressive machine is AMD’s flagship mobile CPU, the Ryzen 9 5900HX. This is an unlocked, 8 core, 16 thread monster that turbos up to 4.6GHz, and it’s insanely fast for a laptop chip. So much so, I thought I’d compare it against AMD’s Desktop 8 core to see just how close it gets, and to the last gen 4800H, and Intel’s current best, the 10980HK. So, lets get into it. But first, if you haven’t already, consider subscribing for more videos like this one every Monday, Wednesday and Friday.
Before we look at the results, there are a couple of things I should make clear. First, the ‘5800X’ I’m using here is actually a 5900X with 4 cores disabled. I had intended to use my 3700X as that would be incredibly close in performance, but it seems to be on it’s way out and wouldn’t stay booted for long enough to run any tests, so 5900X with cores disabled it is then. I’m also primarily using a 4800H here as while I do have a couple of results I’ll show for the 4900HS, and ‘overboost’ 4800H from the XMG Core 15, I don’t have either of those right now to directly test against.
So, that’s the explaining, lets look at the numbers.
Starting with Cinebench R20 and the single threaded test, the desktop 5000 chip does run away with it, but the 5900HX holds a convincing victory over everything else, including an older result from my 3700X. It’s 17% faster than the 4900HS, and 25% faster than the 4800H in this OMEN 15, and 12% faster than the 3700X with only the simulated 5800X beating it by a further 12%. Not bad eh!
Moving onto the multithreaded test, the results pretty much run out the same, with the simulated 5800X beating the 5900HX by around 10%, but everything else pales in comparison, with a 12% lead over the desktop 3700X chip, and 28% over the 4900HS and a whopping 38% over this 4800H. That’s crazy! Interestingly, in both of these tests, the Intel 10980HK is a pretty close match for the Ryzen 4800H and 4900HS chips, although in my experience it draws significantly more power in doing so. It can peak as high as 135W on full boost, although sits more around 100W, compared to more like 65W for the 4800H.
Moving onto the last set of results I’ve got for all these chips in Blender with the BMW scene. With this result, the 5900HX does fall a little behind the two desktop chips, running 8% slower than the 3700X, and 11% slower than the simulated 5800X. It does still have a significant lead over all the other laptop chips, including 6% faster than the ‘overboost’ 4800H, and a full 17% faster than the 10980HK.
I have a theory for why the 5900HX couldn’t keep up with the 3700X here, and it’s because of thermals. Even in this Asus Zephyrus Duo which has air flow from the top, bottom and sides thanks to the second display lifting to give clearance, it’s still much more thermally constrained than a desktop chip using a 240mm closed loop AIO. As you can see on the graph, the green line is clock speed, which steadily decreases over the run, whereas the blue line is temperature which steadily rises. The portable nature of the 5900HX means unlike their desktop counterparts that can basically remain constant at their max turbo, these laptop parts can’t quite.
At this point I’d like to prove that point with the Blender Gooseberry render test which takes much, much longer, but my 3700X gave up the ghost so lets look at it from a quantitative perspective with the simulated 5900X. In the BMW test, the 5900HX was 11% slower than the desktop chip, but in Gooseberry, it’s more like 13%. That would indicate the issue of heat soak is more prevalent on longer tests like this.
For those that are interested, I’ll also throw up the Cinebench R23 numbers, both single and mutli threaded results. They are pretty much the same as R20 so I won’t spend too much time on this, but if you are interested feel free to pause and take a look.
Finally, my 10 minute Premiere Pro render test. This is a 10 minute sequence made up of 4K 30 clips, rending using the “High Quality 4K” Preset and rendering without GPU acceleration enabled. This always seems to be variable, sometimes it takes 3 – 4 minutes, sometimes 8. What I do know is that it generally tends towards Intel CPUs – even now – which means the 10980HK finally gets a win. Not by too much over the 5900HX, but a win is a win. Strangely, the desktop chip is the slowest here. I re-ran all these numbers multiple times to confirm but like I said Premiere has always given me pretty strange results.
The takeaway from these results, for me at least, is that AMD’s Ryzen 5000 series laptop CPUs are a great step forward in portable performance. They are generally better than both the last gen Ryzen 4000 chips, AND better than Intel’s current top of the line SKU. And the fact it can match, or beat the last gen desktop chip in certain tasks is just icing on the cake.
For those wondering why I’m not including any gaming results here, because these laptop chips come with different GPUs, this HP OMEN has an RTX 2060 whereas this Zephyrus Duo has a mobile RTX 3080, and I can’t control that factor. In theory with enough variants you could do an analysis of all the data to work it out, but I barely have access to one of each let alone multiple models so I’ll leave that sort of analysis to the Hardware Unboxed’s and Gamers Nexus’s of the world.
