As far as i know, "Cyber Engine Tweaks" once had an SMT-Patch that did the same thing as this one without having to edit the exe file, but they removed it. So i hereby call for an SMT Fix based on CET so there is no need to edit the exe file every time the game gets updated, and since i can't do that, i ask that someone else who has more knowledge can do it. Or make something like an automatic exe patcher to make it easier to use.DONE.
Feel free to use my findings and data for this purpose, i don't claim them!
Updated to version 2.21 (Feel free to post results)
I would be happy if you could give a short feedback on whether this “Hack” was of any use to you. And if so, what exactly did you find? (Better Frame Times, FPS, Loading Time or Better Core Utilization) Please write down what kind of processor you have so that we can estimate which models are still affected by the SMT problem and we simply have an overview of who it would help. Maybe it will even help CDPR to fix these things in the next update (if there is another one).
Big difference for me. I'm running a 5900x and 4090, path tracing @ 4k DLSS Quality , RR Off, HDR on
Before - Constant stuttering when moving around the game world as well as larger 1-2 second hangs, frame rate around 68/70
After - big pauses and microstutter are fixed, not a big framerate difference. Maybe less occasional dips and holding a more solid 70fps. Just look at that avg utilization, it usually bounces between 60 and 70% for the cyberpunk exe
Well, that's a very clear example^^ For me the whole thing looks similar to my case, my core load is distributed better and I also noticed that all the little stutters have disappeared. I also have to make a picture comparison like this and upload it here when I get around to it. But I'm glad I could help you.
The thing with the dual CCD's is perhaps the solution to the question of why this HEX works. My Ryzen 7 3700X has 2 CCD's. I have CCX 1 clocked at 4.1 Ghz and CCX 2 at 4.0 Ghz. Maybe that's exactly the cause of the unbalanced core load distribution, not my setting of the clock speed but the fact that my CPU has 2 CCDs... That also explains why this HEX works for some and not for others... i think it is CPU dependent.
hey guys my CPU has 8 more processes running and 1500 more handles running but look this mod clearly is using more CPU, not the other things my PC is actually doing . and that uptime screams PC noob
I did some benchmarking and found that this change does definitely increase CPU usage and applies the usage more evenly across all cores but, in my case, actually reduced performance and stability. I'm extremely CPU-bound with RT on which "should" be the best use case for this fix. I'm also still really hit hard by the "fixed" bug that trashes FPS if you change settings which I've heard this should also fix. It did not, in my experience.
Here are my actual results:
I ran the in-game benchmark multiple times and got basically the same results each time. It was also pretty similar in-game.
On my CPU, I have CC0 OC'd to 4.5GHz and CC1 to 4.3GHz. My all-core max clock is 4.3, CC0 can just go higher. I would say this might be why the change made it worse for me... but avg utilization of the "better" cores also went up. So I'm not really sure who would actually benefit. I think it's probably more important for the better cores to be able to take more load on-demand than trying to split it across the entire CPU. At least for most cases and especially if you clock them differently.
More utilization also usually means a lower average clock. 2077 already seems to use cores/threads appropriately and you should get the most out of the highest all-core clock you can. Even pushing one CCD up to 4.5 didn't make much of a difference for me compared to getting them all up to 4.3.
This is less objective but I did notice that, at times, the hex edited benchmark would have up to 10 FPS lower lows and far more instability (including outright crashing) compared to default. On average, it looked like the FPS stayed mostly the same but the lows and maxes were 2-4 FPS higher without the edit. I also noticed some really strange graphical bugs with the hex edit that only really appeared in screenshots specifically with lighting/shadows. With the edit, I also sometimes saw the game pushed to be GPU-bound in GPU-Z... but it still didn't let me get any more FPS with graphics settings changes than before.
Also, my uptime was 6:11:00:00 before I restarted to be a little more scientific while I ran these benchmarks. And that's on the low side for my PC. ;)
on latest patch, i had the same experience. with the "fix", Very smoother when nothing happens, and then it jitters a lot in action and gunfights, 5800x3d... Better without it for me :/
keep smt on AUTO instead of on/off ingame. it fixes everything, and the patcher works.
Just wanted to say there's some misinfo going around (on other websites) about what auto does. All Auto does is turn the setting on if it detects whatever condition requirement it thinks it would benefit (probably simply checks if SMT/Hyperthreading is enabled and the extra cores actually exist).
What it does not do is dynamically adjust anything while the game is running as some people seem to think it does. It's exactly like setting it to On -- if Auto would've turned it on. And Off is also exactly like Auto -- if Auto would've turned it off.
Definitely works with my 7700x. With my RTX4060 I was getting 40-50fps with Pathtracing and everything at High, 1440p HDR with FG and DLSS Balanced, after the fix I now get 55-65fps
will this have benefits if i dont use threads at all. like im have it disabled in bios, but wondering if this mod will benefit just the 8 full cores 9800x3d.
Ok, so. KEEP the ingame SMT on AUTO and use the patcher. SMT off ingame gives the best smoothness with the crowd on LOW, however whenever there is a ton of opponents or crowd increased to medium+, the game will stutter, which is reduced a lot with SMT On, but then it's not smooth all the time! AUTO is much better, well it's automatic for different code....
and the patcher works 100%, much better and steadier fps in conjunction of AUTO.
hi guys, i having a issue here and i hope thi thread could help me out, so i installed the mod, in my utility settingd SMT is on, i have disabled in msi afterburner the GPU power % ( apperently was causing stutter aswell) i have 152 mods in total maybe like max 50-80MB in total, i have FSR 3.13 mod MY spec AMD 7 7700 5.3 rtx 3080 10GB 32 ddr5 6400 AND STILLLLLL STUTTER ! not badly but like i framecapped at 100FPS, when i turn sometimes the camera BOOM 92-97 fps, or the market behind VIK BOOOMM 88 etc....... funny thing is....when only like a handle full of mods are installed like FSR and maybe 20 - 40 other mods no drop !! alll my mods are compatible and updated, my graphical settings are optimesed....could maybe MSI mode uitility solve this ??? thx guys in advance
My game updated itself to 2.21 and when I try to use it I gen an error: There was an error applying the patch: 0x80070057 (the parameter is incorrect). Never used this patched before but my game is running like crap on my 5900X and I'd really like it to work properly!
Didn't reapply after the update and immediately had stutters. Shame on you cdpr. Weird, getting an error trying to apply it. Update 2.21 did something to the exe, delete it and verify files then the patcher will work.
2.21 fix warning (could affect other versions), while overall everything feel smoother, there is a weird effect on lod transition: it's slower and therefore more noticeable, while driving, the "pop-in" effect becomes even more disturbing with this tweak... or is it in my head/placebo? it's in my head.... i guess, it seems like the game dynamically change lod distance depending on the area, especially for npcs.
So I installed the 2.21 patch, which worked but my performance is now WORSE! My CPU is at 50-60& usage, but in the benchmark with the same settings (Ultra, 1440p, RT and DLSS off) I'm getting 53FPS average compared to the 65 I got before. I don't understand why this game won't run properly on my PC.
183 comments
Or make something like an automatic exe patcher to make it easier to use.DONE.Feel free to use my findings and data for this purpose, i don't claim them!
Updated to version 2.21 (Feel free to post results)
Before - Constant stuttering when moving around the game world as well as larger 1-2 second hangs, frame rate around 68/70
After - big pauses and microstutter are fixed, not a big framerate difference. Maybe less occasional dips and holding a more solid 70fps. Just look at that avg utilization, it usually bounces between 60 and 70% for the cyberpunk exe
Here are my actual results:
I ran the in-game benchmark multiple times and got basically the same results each time. It was also pretty similar in-game.
On my CPU, I have CC0 OC'd to 4.5GHz and CC1 to 4.3GHz. My all-core max clock is 4.3, CC0 can just go higher. I would say this might be why the change made it worse for me... but avg utilization of the "better" cores also went up. So I'm not really sure who would actually benefit. I think it's probably more important for the better cores to be able to take more load on-demand than trying to split it across the entire CPU. At least for most cases and especially if you clock them differently.
More utilization also usually means a lower average clock. 2077 already seems to use cores/threads appropriately and you should get the most out of the highest all-core clock you can. Even pushing one CCD up to 4.5 didn't make much of a difference for me compared to getting them all up to 4.3.
This is less objective but I did notice that, at times, the hex edited benchmark would have up to 10 FPS lower lows and far more instability (including outright crashing) compared to default. On average, it looked like the FPS stayed mostly the same but the lows and maxes were 2-4 FPS higher without the edit. I also noticed some really strange graphical bugs with the hex edit that only really appeared in screenshots specifically with lighting/shadows. With the edit, I also sometimes saw the game pushed to be GPU-bound in GPU-Z... but it still didn't let me get any more FPS with graphics settings changes than before.
Also, my uptime was 6:11:00:00 before I restarted to be a little more scientific while I ran these benchmarks. And that's on the low side for my PC. ;)
on latest patch, i had the same experience. with the "fix", Very smoother when nothing happens, and then it jitters a lot in action and gunfights, 5800x3d... Better without it for me :/
keep smt on AUTO instead of on/off ingame. it fixes everything, and the patcher works.
What it does not do is dynamically adjust anything while the game is running as some people seem to think it does. It's exactly like setting it to On -- if Auto would've turned it on. And Off is also exactly like Auto -- if Auto would've turned it off.
on the new patch, it seems to induce stutter when there is a lot of actions, while offering smoothness when nothing is going on. odd.and the patcher works 100%, much better and steadier fps in conjunction of AUTO.
MY spec
AMD 7 7700 5.3
rtx 3080 10GB
32 ddr5 6400
AND STILLLLLL STUTTER !
not badly but like i framecapped at 100FPS, when i turn sometimes the camera BOOM 92-97 fps, or the market behind VIK BOOOMM 88 etc.......
funny thing is....when only like a handle full of mods are installed like FSR and maybe 20 - 40 other mods no drop !!
alll my mods are compatible and updated, my graphical settings are optimesed....could maybe MSI mode uitility solve this ??? thx guys in advance
help
Weird, getting an error trying to apply it.
Update 2.21 did something to the exe, delete it and verify files then the patcher will work.
2.21 fix warning (could affect other versions), while overall everything feel smoother, there is a weird effect on lod transition: it's slower and therefore more noticeable, while driving, the "pop-in" effect becomes even more disturbing with this tweak... or is it in my head/placebo?
it's in my head.... i guess, it seems like the game dynamically change lod distance depending on the area, especially for npcs.