Current Generation Games Analysis Technical Discussion [2020-2021] [XBSX|S, PS5, PC]

[iroboto]

Yes, thanks - that was the point I was trying to make - more "how can it render it wrongly - because it supposed to be automated".

For example, in graphics mode we know why something might not look right because it's been manually created, but here we have algorithms that should do it all with extreme precision but it's actually 'accurately' out (if that makes sense)...it's like the PS5 RT 'thinks' the glass is closer than it actually is, but it's just for that one segment of glass - all the others reflect correctly...very odd!?

Its hard to say honestly. It could always be offset wrong entirely, but without the double reflection to prove otherwise, no one would have checked that it was wrong. That's sort of the beauty of these types of mistakes. They fly well below the radar for everyone (developers and QA testers included) until they see a discrepancy.

 

[snc]

When i was i kid and lived in communist Poland there was a news radio program, one day news come that in Russia on red square they are giving away Mercedes cars for free. Few hours later there was correction to that news, that yes in fact in Russia on red square but not giving away but stealing and not mercedes but bikes. What i am trying to say is since we dont know what and how there is no need to call something superior at this stage.

don't know how this comment is related but funny coincidence, I'm from Poland ;d

 

[goonergaz]

The game isnt particularly cpu limited on PC and the difference in photomode fps is big. I'll eat my hat if the xbox isnt running at least like ~5 fps better on average in game mode

Get your salt and pepper ready...NX gamer said maybe a minor advantage and DF said it's a wash - certainly nothing like 'at least 5fps better'...I mean, the footage is there to see, both have minor drops at similar amounts - the photo mode showing the XSX advantage is down to the fact it's just showing the GPU extra grunt.

 

[iroboto]

Yup! Any outright weird behavior is almost always going to be a software issue first, hardware issue second. Your game suddenly hitching and dropping several frames on a modern machine, even a low end one, isn't, usually, because its not powerful enough.

a lot of the posters on this page are console warriors so I want to mention the upside here for sony is huge: why have ps5 games consistently seemed to come out the door in better shape so far? Clearly Sony is doing something (very important!) right.

It's just probably not some super mega powerful hardware secret sauce that's just eating up a several frame hitch that would occur on another high end cpu.

Yea it's unfortunate for MS honestly. People pay for a console experience, and they are getting a pseudo PC experience.

Here is some 'quick fixes on PC' for hitching.
****
I found performance very stable with a GeForce RTX 2060, Intel Core i7-5820K @ 4.2GHz and 16GB DDR4 RAM, although it would appear some folks on PC are experiencing software-side issues. In particular, Control can stutter when running in DirectX 12 mode but this is largely down to a Windows 10 issue, for which there is now a workaround.

If you are experiencing stuttering in Control’s DX12 mode then you can try the following fix:

First of all, make sure you have the latest WIndows (1903) version installed.

Then,

• Open your Windows Search bar
• Search for and run ‘Exploit Protection’
• Click the ‘Program Settings Tab’
• Click on ‘+ Add programs to customise’
• Navigate to your ‘Control_DX12.exe’
• Open its Program Settings
• Find the ‘Control Flow Guard’ section
• Check the box marked ‘Override System Settings’
• Turn ‘Override System Settings from On to Off and then click ‘Apply’.
• Voila, Control’s PC DX12 stutter should now be gone.

And, another thing which some users have found annoying is Motion Blur. There’s no toggle for Motion Blur in the PC version which seems a fairly basic oversight. An update planned for mid to late September will add an On/Off toggle for both Motion Blur and Film Grain.
***

I only just resolved my issue with my keyboard keys getting stuck. I had to throw out my Razer mouse because it was interfering with my keyboard keys. Crazy! It's little things like that, that can ruin your whole expereince. This is one big advnatage that PS5 will have, there's no legacy bullshit in that can cause hitching.

 

[cwjs]

Get your salt and pepper ready...NX gamer said maybe a minor advantage and DF said it's a wash

its capped at 30 dude! They both run at 30 all the time.

Edit: sorry, early on coffee, I dont think i was clear on my first post: I mean hypothetically capable of running ~5fps better if it were an unlocked framerate. The idea that "cpu limitations" would completely wipe away a ~15+% advantage is what i'm calling silly.

 

[function]

How can it run on a higher average of the frame rates where unlocked it it already drops frames the locked 60 to 50s in both machines during gameplay. Photomode can't be used as a benchmark because it doesn't represent gameplay and as u said it takes the cpu out of the picture so what was the point. You can't draw conclusions on photomode when the cpu is out of the picture because you need both a cpu and gpu to play

In performance mode the game is seeing drops when you have lots of alpha. Lowering the resolution to 1080p would reduce resolution by something like 44%. Those drops from 60 to 50 during gameplay would probably go away entirely.

You would then probably be facing a bottleneck from RT, where it's pretty clear the XSX has a bit of an advantage. It's really not much of a leap to make if you think about where the bottleneck is.

Console class CPUs aren't having a problem hitting 60 fps pretty much all the time with RT enabled.

I happen to think both consoles could make a fair stab at a 60 fps RT mode @ 1080p, but it would most likely be a bit easier for the XSX. Remedy probably have their reasons for making the choices they did though. Or maybe two options seemed like enough.

Get your salt and pepper ready...NX gamer said maybe a minor advantage and DF said it's a wash - certainly nothing like 'at least 5fps better'...I mean, the footage is there to see, both have minor drops at similar amounts - the photo mode showing the XSX advantage is down to the fact it's just showing the GPU extra grunt.

I think he was talking about RT mode (30 fps). In performance mode both were indeed said to be very closely matched, and the footage provided by both DF and NXG does seem to back that up.

Personally, I though that maybe the XSX was a touch higher on average (like NXG said), but that the PS5 maybe suffered a little less from alpha is a couple of the more severe dips. Though I've not seen as much footage as these guys, because I've only watched their videos!

 

[Scott_Arm]

The game isnt particularly cpu limited on PC and the difference in photomode fps is big. I'll eat my hat if the xbox isnt running at least like ~5 fps better on average in game mode, the whole "with the very similar cpus out of the picture theres no conclusions we can draw from this test in a mostly gpu bound game" theory is incredibly silly.

That said, it's purely academic: remedy felt neither system performed well enough to warrant a 60fps cap, so they run at 30. For remedy's purposes and this game, the consoles are effectively the same in practice. (But the xbox has those hitches, which I'm sure would exist whether it was averaging 29fps or 45fps)

On pc with ray tracing on it's definitely cpu limited. I tested it out. Some scenes which don't look like they'd be that taxing will cap out at fairly low frame rates for me (70) and even with the resolution set to 640x480 or 640x480 with dlss the frame rate will not increase.

Follow this thread back:
https://forum.beyond3d.com/threads/dxr-performance-cpu-cost.62177/page-2#post-2183785

 

[goonergaz]

its capped at 30 dude! They both run at 30 all the time.

Edit: sorry, early on coffee, I dont think i was clear on my first post: I mean hypothetically capable of running ~5fps better if it were an unlocked framerate. The idea that "cpu limitations" would completely wipe away a ~15+% advantage is what i'm calling silly.

Ok, but on this edition NX Gamer said PS5 had the advantage. Also the corridor of doom was the lowest point at 33 on XSX yet in game we see (very infrequent and minor) dips below 30.

 

[Dictator]

An interesting way to check how the reflection "should" look would be to push Jesse against a glass mirror like that with her gun out and have it partially intersect with the glass geometry. On the platform where it renders "correctly" the reflection of her gun and hand in the glass should start exactly where it intersects, on the one where it does not, it would start adjacent to it.

 

[cwjs]

On pc with ray tracing on it's definitely cpu limited. I tested it out. Some scenes which don't look like they'd be that taxing will cap out at fairly low frame rates for me (70) and even with the resolution set to 640x480 or 640x480 with dlss the frame rate will not increase.

Follow this thread back:
https://forum.beyond3d.com/threads/dxr-performance-cpu-cost.62177/page-2#post-2183785

Thanks for sharing this test, that's informative -- how much does it hold up at higher resolutions though? My immediate speculation if it's only cpu bound with raytracing is that the BVH tree is being constructed on the cpu and that has a flat overhead which is noticeable if you're as far down as 640x480 (remembering that rays are shot per pixel so raytracing performance will scale noticeably with resolution). Generating a bvh tree every frame is definitely a serious calculation but I feel like it might come out in the wash a bit at 1440p or even 1080p.

One other thought though -- that bvh tree generation should happen in photo mode too. Does the pc version have photomode? would be very cool to repeat the test there (with the camera moving around -- might cache if everything is still)

Ok, but on this edition NX Gamer said PS5 had the advantage. Also the corridor of doom was the lowest point at 33 on XSX yet in game we see (very infrequent and minor) dips below 30.

Yeah I'm not saying either console is actually going to run at photomode framerates the whole time if they're uncapped -- they'll both run lower, and i imagine the gap will be smaller than it is in photo mode (if it wasn't, i'd expect something like hitman where the xsx version had higher settings.) That said, i think we can conclude from this test that there would be some gap in the xbox's favor if you took off the cap. Which, again, is a hypothetical that doesn't actually matter in a reality where remedy saw how both consoles performed and was like "yeah, cap them both at 30".

 

[iroboto]

An interesting way to check how the reflection "should" look would be to push Jesse against a glass mirror like that with her gun out and have it partially intersect with the glass geometry. On the platform where it renders "correctly" the reflection of her gun and hand in the glass should start exactly where it intersects, on the one where it does not, it would start adjacent to it.

lol. your job seems fun, it's like science but for games, until you publish results and need to deal with people

 

[function]

Here a screenshot of XSX where I offset the camera so the reflection is not dead-on 90 degree with photomode. It is possible to see the glass reflection (ultra sharp with hard silhouette) and then the opaque reflection underneath it (noisy and diffuse).

Here a more readily parallaxed example of Translucent reflections over opaque reflections on XSX:

Jesses legs as a sharp reflection in the glass, and then offset and behind as diffuse and noisy reflections in the metal casing of that radar dish chassis.

Whelp, there it is. Double reflections on XSX. Perhaps a little more subtle (as in, not exaggerated), but most definitely there.

Great work @higherARC and @Dictator!

Once again Remedy's statement about matched settings has proven to be true.

 

[iroboto]

On pc with ray tracing on it's definitely cpu limited. I tested it out. Some scenes which don't look like they'd be that taxing will cap out at fairly low frame rates for me (70) and even with the resolution set to 640x480 or 640x480 with dlss the frame rate will not increase.

Follow this thread back:
https://forum.beyond3d.com/threads/dxr-performance-cpu-cost.62177/page-2#post-2183785

did you try the downclocking CPU thing and overclocking?

I do agree though, seems like DX12 and Control don't spread the work loads over the Cores very well. This is very much a 2 core - 4 core game.

 

[Scott_Arm]

did you try the downclocking CPU thing and overclocking?

I do agree though, seems like DX12 and Control don't spread the work loads over the Cores very well. This is very much a 2 core - 4 core game.

Yes, if you follow the thread back one post you'll see at both 640x480 and 1080p I was very under utilized on the gpu, but stuck around 70fps. Playing with my clock speed affected frame rate almost linearly. I'm not that interested in spending more time on it. I don't even think I have it installed anymore. There's definitely a heavy cpu cost to ray tracing in that game which can lead to large performance drops in some scenes.

 

[Deleted member 86764]

lol. your job seems fun, it's like science but for games, until you publish results and need to deal with people

Yeah, because you don't publish results and deal with people in science *eye roll*

 

[Scott_Arm]

@cwjs At the time I was playing around with it I only had a 1080p 144Hz display, so I didn't try higher resolutions. I'm sure that if I put the resolution at 1440p I could hit gpu limits in that same scene. I was at ~64% at 1080p, so 1440p would probably do it.

It seemed like if I lowered cpu clock by 10% I'd lose about 10% of my frame rate. Looked linear-ish, though I didn't test extensively. I'm guessing there's a single threaded bottleneck, which could be the bvh build/refit.

I don't have the game installed anymore, but I'm not that interested in going back to test it again.

 

[iroboto]

Yeah, because you don't publish results and deal with people in science *eye roll*

oh lol, never said you don't need to publish results in science
the fun is in the science itself, the unfun is publishing results and dealing with people.

 

[iroboto]

Yes, if you follow the thread back one post you'll see at both 640x480 and 1080p I was very under utilized on the gpu, but stuck around 70fps. Playing with my clock speed affected frame rate almost linearly. I'm not that interested in spending more time on it. I don't even think I have it installed anymore. There's definitely a heavy cpu cost to ray tracing in that game which can lead to large performance drops in some scenes.

its more than enough, I totally get not being that vested in this.
Thanks for the update

 

[Dictator]

An interesting way to check how the reflection "should" look would be to push Jesse against a glass mirror like that with her gun out and have it partially intersect with the glass geometry. On the platform where it renders "correctly" the reflection of her gun and hand in the glass should start exactly where it intersects, on the one where it does not, it would start adjacent to it.

Here we go:

Looks like PS5's transparent reflections have some sort of perspective shift on them which causes them to misalign.

 

[Shortbread]

Here we go:

Looks like PS5's transparent reflections have some sort of perspective shift on them which causes them to misalign.

It maybe for personal or artistic reasons, but this particular set of photos is what I see if you hold an object at an angle in glass (the firearm refraction), you get a reflection that isn't aligned with the image that's being reflected. The PS5 alignment (at least in these set of photos) are quite accurate from a real-world perspective.