So as a current 5950X owner, it appears that any upgrade from here would both consume more power and generate more heat. I get more performance, yes, but I don't like that trade off. And I mostly use the power in games, occasionally for other intense processing purposes. What I'm seeing so far is that at least in games, the latest Zen chips provide only very modest to negligible improvements in terms of framerate (when running at 4K with a RTX 4090, which is my setup).
Am I correct in this assessment?
Depending on the game, you may be correct. I actually went from a 5950X to a 9800X3D. I realize you may instead be looking at a 9950X or wait for 9950X3D, since you mention more power/heat. But I want to lay this out to you:
One of the reasons for my upgrade is better 1% frame rates on 4K (4080S here), and sometimes much higher averages in some games I care about (for instance Baldur's Gate 3, some Factorio.) STALKER 2 ran okay-ish, definitely playable but it's a title where you flick the mouse around a lot and the 5950X had a hard time keeping the frame rate over 70 sometimes. It IS playable, of course, but it was a common theme with games I play where AVG were alright-ish (all in 4K) but lows were not great at all.
The 9800X3D in Stalker 2 mostly bumps against my 144Hz refresh and doesn't go under 110fps that I have seen, and the game feels so much smoother. BG3 is an outlier where the 9800X3D murders everything else downright (again, even in 4K.)
The benefits may be quite a bit higher on the 4090.
I initially bought the 5950X because I work from home, and I compile a lot of code, run large databases, some VMs, just a lot of concurrent tasks. The 16 cores seemed a natural fit. And really that was an awesome workstation that never flailed.
In another thread on ARS someone challenged my perceived need for the 2nd CCX with extra 8 cores and I thought of course I need them, to which they replied "test it." I forget the thread or I'd link it..
I disabled the 2nd CCX in BIOS and did my regular workloads. My specific code compilation load and I will stress this, code compilation is not a single entity, and some builds would definitely use those extra cores, but mine just did not. One CCX was just as quick as two.
One the 9800X3D I can run large databases on local MSSQL, VMs, Visual Studio etc etc perfectly fine (both builds are 64GB) without wishing for more cores. The 8, in my case, are totally fine for now. The improvements in IPC, clock speed and RAM performance completely make up, and surpass losses due to fewer cores, in my use cases.
I also looked at Compile benchmarks on
Phoronix where you can find 9800X3D vs 5950X for Linux Kernel compile. Lo and behold, the 9800X3D is 10-ish% faster than the 5950X. I see similar benefits on my end. I'm already OK with the 5950X level perf there, so no concerns.
The 9800X3D uses WAY less power than the 5950X. My machines are always quiet, I buy the equipment I need to do so, but the 9800X3D is SO very easy to keep very cool. Any half-decent air cooler or AIO are absolutely sufficient.
That chip has only ever seen 80C under extreme stress, and the cooler (AIO in my case) still just laughed at it and kept the fans at ~1000 RPM. The 5950X uses more power; I still had it quiet but it took a lot more effort, and definitely used a good bit more power.
This is a long way to say, carefully evaluate your workloads. You may be surprised by the benefits even if going down half the cores, of course this is WILDLY YMMV.
But specifically for games, people love to bleat about how the GPU is the only limit in 4K. It is if you are only looking at Max FPS for heavily GPU intense workloads, like RT. I personally find the difference very noticable and it's not the sort of rose colored glasses, I always have a FPS indicator up.
