Is there something wrong with my PC, or is it just how the game is?

[CallMeLeonidas]

Hello everyone, I have a new-ish PC with Windows 11 24H2, Ryzen 5 5600, RX 6600 and 16GB DDR4 3200MHz RAM, and I'm experiencing some CPU-related performance issues when I use FSR 2.1 on Quality and set graphics to mostly High. Setting things to Ultra results in less severe drops, but it still happens occasionally.

https://steamcommunity.com/sharedfiles/filedetails/?id=3395878913

Is there something wrong with my setup, or is it just how the game is?

[CallMeLeonidas]

Apparently, changing the resolution to 1440p via AMD VSR and then using FSR 2.1/XeSS 1.3 on Balanced results in a smoother frametime and better GPU utilization even when using High settings and even in Dogtown, interesting.

Nevermind, it still drops in PL.

[Swull]

maybe try turning down population density, i have a 4060, same cpu, 3600 ddr4 and i can get 100 ish frames on high settings using dlss quality @ 1440p. Cpu utilization sits around 60% most of the time.

[CallMeLeonidas]

Originally posted by Swull:

maybe try turning down population density, i have a 4060, same cpu, 3600 ddr4 and i can get 100 ish frames on high settings using dlss quality @ 1440p. Cpu utilization sits around 60% most of the time.

I think it was set to medium at the time of making that screenshot, right now I'm using low alongside the SMT fix mod from Nexus. Running the game at 1440p VSR with XeSS on Balanced seems to yield around 70-80 FPS with 70-90% CPU utilization. It's more stable in the main game for sure, but there are still dips below 60 FPS in Dogtown.

[Pistol Packin' Mama]

Since the LOD setting doesn't make any real impact, crowd density is the only CPU bound thing you can adjust.

As far as the standard raster settings go, the only ones with major impact are volumetric fog and screen space reflections, with volumetric clouds being a distant second.

For reference:

SSR should never be set to anything past high—ultra and psycho end up performing WORSE than ray traced reflections for no visual gain whatsoever, and even then, the differences between low, medium, and high are so visually indistinct that you can, for all intents and purposes, leave it on low and likely never notice a difference. I know I never have.

Volumetric fog and clouds scale with the render resolution. At native res, fog at medium and clouds at high is the best balance in my experience, but if you're downscaling with DLSS/FSR/XESS, you can usually bump those up a notch for a much more reasonable performance trade off.

If those alone don't get you above 60, then dropping local shadow quality, cascade shadow range, and ambient occlusion to medium will save you a few more frames without much fidelity loss. In comparisons, only local shadow quality is noticeable to me, and to be honest, I think the softer, more diffused appearance of medium looks better to begin with.

All the other settings? Leave them maxed, they have no significant (or any) performance impact whatsoever.

2077 is sadly very poorly optimized and tends to get worse with every patch, so don't feel bad if your system struggles with it.

[CallMeLeonidas]

Originally posted by I'm just a girl 🎀:

Since the LOD setting doesn't make any real impact, crowd density is the only CPU bound thing you can adjust.

As far as the standard raster settings go, the only ones with major impact are volumetric fog and screen space reflections, with volumetric clouds being a distant second.

For reference:

SSR should never be set to anything past high—ultra and psycho end up performing WORSE than ray traced reflections for no visual gain whatsoever, and even then, the differences between low, medium, and high are so visually indistinct that you can, for all intents and purposes, leave it on low and likely never notice a difference. I know I never have.

Volumetric fog and clouds scale with the render resolution. At native res, fog at medium and clouds at high is the best balance in my experience, but if you're downscaling with DLSS/FSR/XESS, you can usually bump those up a notch for a much more reasonable performance trade off.

If those alone don't get you above 60, then dropping local shadow quality, cascade shadow range, and ambient occlusion to medium will save you a few more frames without much fidelity loss. In comparisons, only local shadow quality is noticeable to me, and to be honest, I think the softer, more diffused look of medium looks better to begin with.

All the other settings? Leave them maxed, they have no significant (or any) performance impact whatsoever.

2077 is sadly very poorly optimized and tends to get worse with every patch, so don't feel bad if your system struggles with it.

That's a lot of words, but these settings do seem to be consistent with BenchmarKing's video on optimizing the game who also said in his first CP2077 video that Cascade Shadow Range is pretty CPU-intensive, so I set to medium. SSR on Psycho is somehow more intensive than RT Reflections, which is really strange.

I'm mostly CPU-limited in this game, so Crowd Density and Cascade Shadow Range are the only settings that somewhat mitigate low GPU usage. Lowering GPU settings actually makes the game less smooth and more erratic in terms of frametime and FPS.

[Pistol Packin' Mama]

BenchmarkKing is who I compared to! My notebook performs better than the desktop he uses for his videos, and my findings were still entirely consistent.

It should also be noted that some people (including me) are having performance issues with the latest patch. 2.2 reintroduced the microstutter issue for me, as well as drops in scenarios I never had problems with before.

[CallMeLeonidas]

Originally posted by I'm just a girl 🎀:

BenchmarkKing is who I compared to! My notebook performs better than the desktop he uses for his videos, and my findings were still entirely consistent.

It should also be noted that some people (including me) are having performance issues with the latest patch. 2.2 reintroduced the microstutter issue for me, as well as drops in scenarios I never had problems with before.

That could be it. Pretty common issue with games that receive updates often.

I remember having a laptop with GTX 1660 Ti and i7-10750H, and I think it ran the game pretty well until I got to Dogtown where GPU usage dropped to 50% and I got 30 FPS.

[Angry Gandalf]

Originally posted by CallMeLeonidas:

Originally posted by I'm just a girl 🎀:

BenchmarkKing is who I compared to! My notebook performs better than the desktop he uses for his videos, and my findings were still entirely consistent.

It should also be noted that some people (including me) are having performance issues with the latest patch. 2.2 reintroduced the microstutter issue for me, as well as drops in scenarios I never had problems with before.

That could be it. Pretty common issue with games that receive updates often.

I remember having a laptop with GTX 1660 Ti and i7-10750H, and I think it ran the game pretty well until I got to Dogtown where GPU usage dropped to 50% and I got 30 FPS.

Well you have a low end system tbh.
The 6600 is so bad, that even your old 1660TI could keep up with it.

If I may ask, why did you buy this system? I guess it was a prebuild and sold as "gaming PC", am I right?

[CallMeLeonidas]

Originally posted by Angry Gandalf:

Originally posted by CallMeLeonidas:

That could be it. Pretty common issue with games that receive updates often.

I remember having a laptop with GTX 1660 Ti and i7-10750H, and I think it ran the game pretty well until I got to Dogtown where GPU usage dropped to 50% and I got 30 FPS.

Well you have a low end system tbh.
The 6600 is so bad, that even your old 1660TI could keep up with it.

If I may ask, why did you buy this system? I guess it was prebuild, wasn't it?

No, I built it myself. 5600 and 6600 should be a match made in heaven, right?

The 5600 is way stronger than i7-10750H, it should be way better for Cyberpunk.

[Angry Gandalf]

Originally posted by CallMeLeonidas:

Originally posted by Angry Gandalf:

Well you have a low end system tbh.
The 6600 is so bad, that even your old 1660TI could keep up with it.

If I may ask, why did you buy this system? I guess it was prebuild, wasn't it?

No, I built it myself. 5600 and 6600 should be a match made in heaven, right?

The 5600 is way stronger than i7-10750H, it should be way better for Cyberpunk.

Well, yes and no. It's a good match because you could use SAM since both components are from AMD, but still pretty weak. I'd even say that your old notebook is almost on par with your gaming rig.

[CallMeLeonidas]

Originally posted by Angry Gandalf:

Originally posted by CallMeLeonidas:

No, I built it myself. 5600 and 6600 should be a match made in heaven, right?

The 5600 is way stronger than i7-10750H, it should be way better for Cyberpunk.

Well, yes and no. It's a good match because you could use SAM since both components are from AMD, but still pretty weak. I'd even say that your old notebook is almost on par with your gaming rig.

I wanted to switch from my laptop to a desktop since I wanted something that I could later upgrade down the line.

I play Payday 2 and L4D2 a lot, my FPS there with DXVK improved dramatically when I built this PC. I can now also play Metro Exodus Enhanced Edition, which wasn't possible with my GTX 1660 Ti laptop. I'd say it's a good upgrade.

I also want to point out that my FPS in Dying Light 2 increased by a huge margin with my RX 6600 even when compared to the previous A580 before I swapped it.

[shoopy]

6600 is better than 1660 TI, not by a ton but enough to matter. Though you probably could have gone for a 6600 XT at least.

[CallMeLeonidas]

Originally posted by shoopy:

6600 is better than 1660 TI, not by a ton but enough to matter. Though you probably could have gone for a 6600 XT at least.

I bought a used 6600 for $190 to replace my Arc A580 due to driver issues and power consumption concerns, the market for computers is messed up here.

Also, wouldn't the 6600 XT make the CPU bottleneck worse? It is slightly more powerful than my 6600 for sure, but that leaves less breathing room for the CPU.

[shoopy]

Originally posted by CallMeLeonidas:

Originally posted by shoopy:

6600 is better than 1660 TI, not by a ton but enough to matter. Though you probably could have gone for a 6600 XT at least.

I bought a used 6600 for $190 to replace my Arc A580 due to driver issues and power consumption concerns, the market for computers is messed up here.

Also, wouldn't the 6600 XT make the CPU bottleneck worse? It is slightly more powerful than my 6600 for sure, but that leaves less breathing room for the CPU.

I don't think it matters that much. I have a Ryzen 5 3600 compared to your 5600 with a 6650 XT and this is what I can get with your settings
https://steamcommunity.com/sharedfiles/filedetails/?id=3396115014

[CallMeLeonidas]

Originally posted by shoopy:

Originally posted by CallMeLeonidas:

I bought a used 6600 for $190 to replace my Arc A580 due to driver issues and power consumption concerns, the market for computers is messed up here.

Also, wouldn't the 6600 XT make the CPU bottleneck worse? It is slightly more powerful than my 6600 for sure, but that leaves less breathing room for the CPU.

I don't think it matters that much. I have a Ryzen 5 3600 compared to your 5600 with a 6650 XT and this is what I can get with your settings
https://steamcommunity.com/sharedfiles/filedetails/?id=3396115014

How is that possible? Are you running Windows 11 24H2? With the A580, which is less powerful than your 6650 XT, and 3600 combo the game was dropping frames really hard a lot. Upgrading to a 5600 was a huge upgrade.

Something is clearly wrong with my setup if my 5600 and 6600 combo has worse GPU utilization than yours.