The Finals is a CPU usage monster

Associate
Joined
26 Jan 2024
Posts
569
Location
North Yorkshire
After crawling through CPU and GPU usage logs I've confirmed that my 7700X is definitely holding my RTX5080 back in The Finals, even at 4K (albeit with some DLSS). I've always known the performance wasn't great, but always wrote it off as the price of playing at that resolution.
This is in Bazzite BTW. I wonder if Windows would be much different.

Is anyone else running this game at high framerates on a seemingly GPU-limited setup? I'm curious to know where the balance is.
 
After crawling through CPU and GPU usage logs I've confirmed that my 7700X is definitely holding my RTX5080 back in The Finals, even at 4K (albeit with some DLSS). I've always known the performance wasn't great, but always wrote it off as the price of playing at that resolution.
This is in Bazzite BTW. I wonder if Windows would be much different.

Is anyone else running this game at high framerates on a seemingly GPU-limited setup? I'm curious to know where the balance is.

Does CPU-Z show 100% CPU usage x 8?
 
Does CPU-Z show 100% CPU usage x 8?
I'm told The Finals will use a maximum of 12 threads, which is 75% of my 16 thread CPU, and my usage bumps up against 75% usage frequently, especially when there's a lot going on in the game. At the same time my GPU usage drops to around 60~65%.

I'm using MangoHud to log all this while in game.
 
Last edited:
I would think if you do any sort of regular gaming, especially at 4k and have a high end GPU, you'd be wanting to look at an X3D CPU to make the best of it. Google suggests a 7800X3D would offer a 10-20% FPS increase.
 
Bazzite will be the problem, the game is not that intensive, 5080 should make mincemeat of it even with a 7700x at 4k, this is not a bad CPU.
 
  • Like
Reactions: HRL
Now that I have detailed frametime/framerate/cpu+gpu usage/memory usage data I'm keen on installing Windows and comparing numbers. Might spend some time logging data for a few othr games first for a more comprehensive comparison.
 
  • Like
Reactions: HRL
The finals is crammed with large scale destructible physics right? Play it and graph CPU usage and blow a bunch of stuff up. See what happens to CPU usage. My gut feeling is that in order to not let the physics have a performance impact on play when things get heavy, objects are simulated fully in the scene even when still. This just means that when explosions and forces are made in game to move everything around, it doesn't dip frame rate. That would account for an abnormally high CPU usage most of the time.
 
I just had a few rounds on my handheld attached to a 4K screen on Epic with quality FSR (its a 4060 level igpu in 395 Max) plays fine no issue, I do have 32 threads but it was only using ~10-20% on W11 ~45fps no stuttering, I had CPU boost off to feed power to GPU, TDP limited handheld tradeoff so only running 2.2Ghz on that CPU. Obviously I'd drop settings for a high frame rate.

What is your framerate target, 144 or more?

I'll try my 5080 in a bit, but it's got a 9950X3D as backup, I wouldn't expect an issue if my toy can play that well just loads more frame.

EDIT - All maxed 5080 is doing 120fps@4k epic with no scaling, CPU usage ~20-25%

I could try my 7840u similar Gen CPU and config 8/16 but with a 9070XT see if it does high framerates

EDIT - Tried this one, CPU 50-60%, not as fast as proper 5080system, but it's a 30w mobile CPU running 9070xt over USB, maxxed out it needed FSR Quality to hit half the frame rate of 5080 @ ~60fps.
 
Last edited:
The finals is crammed with large scale destructible physics right? Play it and graph CPU usage and blow a bunch of stuff up. See what happens to CPU usage. My gut feeling is that in order to not let the physics have a performance impact on play when things get heavy, objects are simulated fully in the scene even when still. This just means that when explosions and forces are made in game to move everything around, it doesn't dip frame rate. That would account for an abnormally high CPU usage most of the time.
That makes a lot of sense.

I just had a few rounds on my handheld attached to a 4K screen on Epic with quality FSR (its a 4060 level igpu in 395 Max) plays fine no issue, I do have 32 threads but it was only using ~10-20% on W11 ~45fps no stuttering, I had CPU boost off to feed power to GPU, TDP limited handheld tradeoff so only running 2.2Ghz on that CPU. Obviously I'd drop settings for a high frame rate.
45fps isn't crazy but that's still really impressive for a handheld at 4K.

What is your framerate target, 144 or more?
MOAR. I'm using a 240Hz oled so I'll take every frame I can get.

I'll try my 5080 in a bit, but it's got a 9950X3D as backup, I wouldn't expect an issue if my toy can play that well just loads more frame.
EDIT - All maxed 5080 is doing 120fps@4k epic with no scaling, CPU usage ~20-25%
20~25% of a 9950X3D...yeesh. I must try epic settings with no scaling and see what I get.
I'm currently running a mix of medium and maximum settings with Balanced DLSS and getting 110fps average, 71fps 1% low and 58fps 0.1% low :(

Thanks for the info. Really useful comparison!
 
Last edited:
45fps isn't crazy but that's still really impressive for a handheld at 4K

It's peak handheld :D iGPU is plenty good enough for its native res of 1920x1200, that it can make a half decent desktop does impress me having done years of APU/iGPU gaming Strix Halo is next level.

MOAR. I'm using a 240Hz oled so I'll take every frame I can get.

Yeah, its going to need a lower settings or framegen to get up there :D

I'd look for an optimisation guide see what settings can be dropped with minimal impact to visuals to push higher frame rates at 4K, obviously the faster the frame rate the faster the CPU is likely needed to feed it but yours is decent. I didn't try dropping settings, i should have just to see where the CPU goes to get there, could the handhelds do it. 7840 is old handheld, 395 new.

25% of 9950 is only 8 threads.
 
Last edited:
Yeah, its going to need a lower settings or framegen to get up there :D
I'm not really expecting to hold 240fps in the Finals at 4K on a 5080 without dropping settings, but I'd like at least 140 with 1% lows above 100. The fact that you can crank your settings, turn off DLSS and get roughly the same FPS as me with the same GPU means I'm doing it wrong.

I'd look for an optimisation guide see what settings can be dropped with minimal impact to visuals to push higher frame rates at 4K, obviously the faster the frame rate the faster the CPU is likely needed to feed it but yours is decent. I didn't try dropping settings, i should have just to see where the CPU goes to get there, could the handhelds do it. 7840 is old handheld, 395 new.
I've put extensive time into this, both with online guides and a TON of A/B testing one setting at a time myself. I think the latter has resulted in a setup that reduces CPU load more than anything but there are a few GPU-only sacrifices made too.
If I can get 120fps with everything cranked like you, and then tweak settings to get 140+ I'd be happy. DLSS alone might get me there.
 
It's the bazzite effect, Linux is slow, or something else in the background hammering his CPU as the game doesn't really ask for a lot as I am running same GPU, if I cut my CPU down to 8 threads and run balanced DLSS as he does I'm doing 200+ fps
 
It's the bazzite effect, Linux is slow, or something else in the background hammering his CPU as the game doesn't really ask for a lot as I am running same GPU, if I cut my CPU down to 8 threads and run balanced DLSS as he does I'm doing 200+ fps
You still have the x3d cache though, in the old 3d mark physics test my 5800x3d does way better than the standard part.

Could totally be bazzite too, just saying your cache WILL be helping
 
I do more or less the same fps as him using similar scaling settings via FSR on windows with both Strix Halo as CPU and a 7840U handheld chip, neither have 3d vcache and are weak TDP limited things in handhelds that are driving a USB 9070XT eGPU, this perhaps half the power of a 5080 when crippled like this, but yes the 9950X3d does have vcache and is strong, I could turn the cores with vcache off but I don't know which CCD that is.

I say my 9950 is strong, looking in the X3D thread as some of the numbers those guys are putting in, I've a stack of tuning to do as mine is well down of on performance in comparison :eek: :D
 
Last edited:
Back
Top Bottom