AMD Ryzen 9 7900X3D Linux Gaming Performance

[phoronix]

Phoronix: AMD Ryzen 9 7900X3D Linux Gaming Performance

After earlier this week providing the initial Linux benchmarks of the AMD Ryzen 9 7950X3D across many Linux gaming tests as well as nearly 400 other tests, in today's article I am looking at the AMD Ryzen 9 7900X3D as the 12-core / 24-thread processor with the hefty 128MB L3 cache on this Zen 4 desktop processor. Due to having less time with the 7900X3D thus far, today's article is just getting things started in looking at the Linux gaming performance -- both native Linux games as well as many Windows games running on Linux thanks to Valve's wonderful Steam Play (Proton + DXVK / VKD3D-Proton) software.

https://www.phoronix.com/review/ryzen9-7900x3d-gaming

[rhavenn]

All I really saw was get a 7700X and save yourself some money unless you REALLY need all those cores. A 8C/16T processor (the 7700X) is going to be plenty for 99% of games. Most games have no clue on how to truly multi-thread outside of a few processes and the higher clock rate per core is better than more cores once you get past 4 cores or so for gaming.

Put your money towards a GPU.

[schmidtbag]

I'm surprised how efficient this CPU is. Makes me wonder how much better the 7800X3D will be.

[Barley9432]

Out of curiosity did AMD provide you with the 7900x3D?

[NeoMorpheus]

I'm surprised how efficient this CPU is. Makes me wonder how much better the 7800X3D will be.

And I am surprised how often that fact is conveniently ignored by all the other reviewers that only seem to want Intel cpus hence not providing such info.

[ms178]

The frametimes of the 13900K look not that great in Hitman even though the frame rate itself is fine, why so? Or has Raptor Lake still issues in some games on Linux?

[muncrief]

I'd love to buy a Zen 4 CPU but since I just bought a new X570 motherboard, 32 GB of DRAM, and 5700X CPU at the end of 2019 I can't justify the expense of buying a whole new system again. Especially with prices as they are.

But wow, the Zen 4 CPUs are just awesome! And in another 2 or 3 years, if prices come down, it might be worth it. On the other hand Zen 3 works so well, and Zen 4 requires a completely new system, so it's certainly put AMD users in a bind as far as upgrades go.

[lamka02sk]

These things are incredibly efficient. I would like to see their idle power consumption too.

[qlum]

I am personally kind of curious about other gaming workloads on linux, such as how fast time passes in paradox games, factorio benchmarking, civ turn time, things of that nature.

[stormcrow]

I am personally kind of curious about other gaming workloads on linux, such as how fast time passes in paradox games, factorio benchmarking, civ turn time, things of that nature.

Game world time should not materially change between supported CPUs as that's a function of an algorithmic governor in the program. The governors are there to make games playable for humans because CPUs can make decisions far faster than either the screen can display and the human player can respond. For one known example try The Ur-Quan Masters and turn Super Melee's time governor off. Even on an ancient 486 machine turning off the time governor resulted in a game no human could play, but it was fun watching the computer fight itself.

I understand what you're getting at. Some games like the X series and Civilization, etc have a computer AI in which decisions are made either in the background (as in X4, RTSes) or at specific turns (Civilization, Gal Civ, MOO). I just don't see how you can automatically measure that kind of thing. You'd have to play the game or connect it to a test harness. Decision trees are going to vary based on the situation presented to the game's AI at any given time or turn. How do you know if a given problem is taking too much time because of a bug, or the decision tree is validly complex? Most people aren't going to have the skills or patience to map out the logic flow in those systems to know if a benchmark is a valid result or not even if the code is available. They'll just make assumptions that may or may not be valid. You may not even be benchmarking the CPU itself but some logic decision that depends on other code paths before it finishes.