If you could drive in busy areas and have some monitoring software running while capturing video that would be great.
Thanks.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Cyberpunk 2077 Ultra performance
- Thread starter g67575
- Start date
Thanks.
but can it lock to 60 fps in bradbury and bran with high crowd density;?
you can find the place on chinatown, it's a fast travel point
you can see reference video, my cpu ***** itself when i sprint around with high crowd density
https://www.youtube.com/watch?v=F3bPlCrihnQ&feature=youtu.be
this place also murders geforce now btw xd (it cant practically go above 25+ fps if you enable rt in that place)
https://www.youtube.com/watch?v=Dk4pbbfAhBU
Last edited:
nice upgrade
Associate
- Joined
- 22 Oct 2012
- Posts
- 1,096
Cyberpunk is very highly multi threaded, GPU usage is 99-100%.
Just looked at CPU scaling chart and it seems to run very well on a 5800X which is only beaten by 5950X.
Anyway I don’t think I’ve seen my 3700X going over 60% usage in this game while gpu usage is dropping to about 70% and with that FPS is going down as well.
A lot of other people are experiencing the same behaviour.
I guess tweaking ram might help but xmp profile on my ram isn’t that bad as it’s 8Pack kit so I don’t expect miracles.
CPU usage on my 2600 @ 4.1GHz was about 90+% non stop on all cores, this is with the 5600x:
Yup wasn't expecting as big of a boost but it has really impressed me, especially since everything is still stock with the 5600x.
nice upgrade
Yup wasn't expecting as big of a boost but it has really impressed me, especially since everything is still stock with the 5600x.
yeah tbh difference between 5800x and 3700x can be up to %35-45 in some certain situations
i guess the actual problem lies on the gimmicky architecture on zen/zen+/zen 2 cpus.
for example, 5800x has no inter-ccx latency, because all cores now actually use the same l3 cache
on the other hand, zen/zen+/zen 2 cpus have inter-ccx latency due to having two seperate ccxs with seperate l3 cache
i read somewhere that this incurs a higher core to core latency between different ccx cores. for example, when core 7 accesses a data core 1 processed, it needs to through infinity fabric, hence higher ccx latency
zen+ is horrible, core0 to core7 latency is usually around 110-125 ns, while zen 2 brings huge improvements and brings it down to 70-80 ns range (these latencies can be influenced by ram timings and ram speed)
for zen 3 though, core0 to core7 has 15-20 ns latency, which is huge. zen 3 actually has the best intercore latencies across the board, even against intel, which usually ranges around 40-45 ns
so in short, in games where 6+ cores are utilized, i'm pretty sure from time to time this latency comes into play, and that's where intel cpus always shined, and that's why zen 3 shine now
i don't know what kind of future high inter-ccx latency cpus will have, considering amd went back to having no inter-ccx latency on their gaming zen 3 cpus (barring two ccd cpus, i talk about 5800x specifically)
consoles seem to have large ccx clusters, which in theory, should make developers try to group thread workload per ccx as much as possible, but since im not a developer, i dont know how possible is that, but again consoles have their special optimization sauce so anything can apply to them, but i cant also know if those special sauces will translate to pc or not (i really wish it would)
i guess the actual problem lies on the gimmicky architecture on zen/zen+/zen 2 cpus.
for example, 5800x has no inter-ccx latency, because all cores now actually use the same l3 cache
on the other hand, zen/zen+/zen 2 cpus have inter-ccx latency due to having two seperate ccxs with seperate l3 cache
i read somewhere that this incurs a higher core to core latency between different ccx cores. for example, when core 7 accesses a data core 1 processed, it needs to through infinity fabric, hence higher ccx latency
zen+ is horrible, core0 to core7 latency is usually around 110-125 ns, while zen 2 brings huge improvements and brings it down to 70-80 ns range (these latencies can be influenced by ram timings and ram speed)
for zen 3 though, core0 to core7 has 15-20 ns latency, which is huge. zen 3 actually has the best intercore latencies across the board, even against intel, which usually ranges around 40-45 ns
so in short, in games where 6+ cores are utilized, i'm pretty sure from time to time this latency comes into play, and that's where intel cpus always shined, and that's why zen 3 shine now
i don't know what kind of future high inter-ccx latency cpus will have, considering amd went back to having no inter-ccx latency on their gaming zen 3 cpus (barring two ccd cpus, i talk about 5800x specifically)
consoles seem to have large ccx clusters, which in theory, should make developers try to group thread workload per ccx as much as possible, but since im not a developer, i dont know how possible is that, but again consoles have their special optimization sauce so anything can apply to them, but i cant also know if those special sauces will translate to pc or not (i really wish it would)
Soldato
I'll will have a go at that section, and see what happens
What CPU/GPU are you running?
im running a 2700x and 3070
i expect yours to perform around 60 but might dip to 50s
(lowest i got was 32, let's assume 1.6x multipler (perf difference), you would have a minimum of 51 frames, or worst case, 1.4x multiplier (since mine is on heavy ram streoids xd), 45 frames (be sure to erratically run like me and turn the camera in quick fashion though, having a slow stroll does not cause such frame drops)
this is a funny topic because i detected this performance issue from a mile away. how did I do it?
they released a "xbox series x" gameplay video a week earlier before the release. i detected something was weird. they deliberately drove the car very slowly, and changed the camera angle very slowly (not a thing you would do while playing a fast paced game with a fast car)
now i know why, because they cheated, they actually turned crowd density setting to high for the video, but they knew that it would drop frames horribly if they speed up with the car or change the camera angle quickly:
https://youtu.be/yhj4qlItES0?t=149
then the game is released, and they set it back to low so that it does not drop frames as hard as it would be on high.
if it can drop to 45-50 fps with top of the line cpu, i dont know what to say to CDPR
again, you may be gpu bound at these frames as well, that's another topic
i expect yours to perform around 60 but might dip to 50s
(lowest i got was 32, let's assume 1.6x multipler (perf difference), you would have a minimum of 51 frames, or worst case, 1.4x multiplier (since mine is on heavy ram streoids xd), 45 frames (be sure to erratically run like me and turn the camera in quick fashion though, having a slow stroll does not cause such frame drops)
this is a funny topic because i detected this performance issue from a mile away. how did I do it?
they released a "xbox series x" gameplay video a week earlier before the release. i detected something was weird. they deliberately drove the car very slowly, and changed the camera angle very slowly (not a thing you would do while playing a fast paced game with a fast car)
now i know why, because they cheated, they actually turned crowd density setting to high for the video, but they knew that it would drop frames horribly if they speed up with the car or change the camera angle quickly:
https://youtu.be/yhj4qlItES0?t=149
then the game is released, and they set it back to low so that it does not drop frames as hard as it would be on high.
if it can drop to 45-50 fps with top of the line cpu, i dont know what to say to CDPR
again, you may be gpu bound at these frames as well, that's another topic
The 3700X would definitely hold you back a lot since Nvidia has higher driver overhead. In Cyberpunk in particular it's even worse because AMD does worse than Intel, and the 3700X itself doesn't do too hot.
See:
https://www.youtube.com/watch?v=JLEIJhunaW8
The 3700X would definitely hold you back a lot since Nvidia has higher driver overhead. In Cyberpunk in particular it's even worse because AMD does worse than Intel, and the 3700X itself doesn't do too hot.
See:
https://www.youtube.com/watch?v=JLEIJhunaW8
overhead in this game is not that much tbh (but it's there)
sorry i misread
Soldato
I've tried to match your settings you listed in one of your videos above, but this isn't how I normally have the game running.
I normally have FOV set at 85, and all the crap blur,bloom,grain rubbish turned off
Mick
PS not sure if YT have finished processing the vid (not watched it yet), but it was recorded at 1440p native res.
wow it handled it like a champ
I've tried to match your settings you listed in one of your videos above, but this isn't how I normally have the game running.
I normally have FOV set at 85, and all the crap blur,bloom,grain rubbish turned off
Mick
PS not sure if YT have finished processing the vid (not watched it yet), but it was recorded at 1440p native res.
5800x is crazy (geforce now video does not belong to me btw, it's a friend of mine's, i also turn those rubbish and play with 100 fov)
https://www.dropbox.com/s/mv9elecxhltd052/QuickSave-1.rar?dl=0
can you use my save and sprint around?
your sprint seems much slower compared to mine, i wonder how much that will affect performanceyou can drop the save file to here, it should be compatible with 1.2 version
C:\Users\---\Saved Games\CD Projekt Red\Cyberpunk 2077
btw you can approach the zen master
Last edited:
it's about your luck
i started the game with 1.02 version and finished the game with 1.06 version and played it around 80 hours in total with gtx 1080 and never crashed once, never encountered a game breaking bug, but surely lots of bugs i have come across
i'm now with 3070 and already crashed 4 times, encountered game breaking bugs in certain places where i was unable to continue with 1.12 version. i didn't play it much with 1.2 yet though
so it seems like it's simply by luck. or that 3000 cards run the game worse in terms of stability
I've tried to match your settings you listed in one of your videos above, but this isn't how I normally have the game running.
I normally have FOV set at 85, and all the crap blur,bloom,grain rubbish turned off
Mick
PS not sure if YT have finished processing the vid (not watched it yet), but it was recorded at 1440p native res.
Yeah that looks good as your gpu was constantly almost maxed out.
If you are doing another video drop resolution scale or change DLSS to performance or even ultra performance so we can see how many frames that cpu can push. Thanks for the video as that shows me the 5800X is a worthy upgrade going from 3700X.
I would rather get 5900X but can’t get those at normal prices.
Associate
- Joined
- 22 Oct 2012
- Posts
- 1,096
I have to agree, I don’t know that memory is the issue. Maybe a driver or GPU setting problem? Had a quick look and this Reddit post seems to have gotten a lot of traction:
https://www.reddit.com/r/cyberpunkgame/comments/kaylnw/comment/gfnjkg5
Soldato
OK, something really strange when using DLSS set to Ultra Performance. Although the FPS goes up slightly, the GPU usage drops to around 75% and the CPU usage hovers around 50%
So why doesnt the GPU usage stay at max to pump more frames? Esp as the CPU isnt saturated?
See the video.
I wonder if GSync has anything to do with this?
So why doesnt the GPU usage stay at max to pump more frames? Esp as the CPU isnt saturated?
See the video.
I wonder if GSync has anything to do with this?