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

[Globalisateur]

The balanced mode gets as low as that? That's supposed to be the highest resolution mode... doesn't sound right.

The scene I used seems to be very demanding: Big area with fire and weather effects.

It does drop even lower with RT in that scene, but I couldn't find useable aliasing. Just that the resolution is also very similar between PS5 and XSX in that scene with RT. But at least the framerate seems very stable which is the most important.

I also noticed some quite more blurry textures on XSX (balanced mode), more so than the very little resolution difference. The same problem DF noticed in their very short comparison between PS5 / XSX in their video. I also think it's caused by VRS.

 

[iroboto]

At 0:51 the balanced mode has a slightly higher resolution on PS5. I found about 1140p for PS5 and about 1100p for XSX.

it's completely unclear how you are pixel counting an image that isn't the source image. You've gotta get the source files, you can't pixel count 0:51 mark and pass it as accurate with no underlying knowledge of the images that were cropped and put together.

 

[Globalisateur]

it's completely unclear how you are pixel counting an image that isn't the source image. You've gotta get the source files, you can't pixel count 0:51 mark and pass it as accurate with no underlying knowledge of the images that were cropped and put together.

You are right, I should have provided my material. We can find some very visible geometry aliasing on that youtube footage at 0:51. It's early pixel counting but I think there is a very small resolution gap (around 3%) between both images at least in that scene.

Remember it's only one scene. Maybe the XSX will have some edge in another scene.

EDIT: The guy has zoomed the images at 0:51 so all this is worthless, sorry about that.

 

[iroboto]

You are right, I should have provided my material. We can find some very visible geometry aliasing on that youtube footage at 0:51. It's early pixel counting but I think there is a very small resolution gap (around 3%) between both images at least in that scene.

Remember it's only one scene. Maybe the XSX will have some edge in another scene.

What I mean, is that you have no knowledge if the images are downsampled to fit into the crop. You are also working off YouTube compression, which compressed another type of compression. Alone, either issue would throw off any sort of measurement, together, does not make any sort of sense to attempt.

 

[Shortbread]

Judging solely off ElAnalistaDeBits Doom Eternal video, the performance/framerate seems locked across all modes and systems. Since PS5 is running DE quite well without VRS, surely it's warranted [at least in this case] to forgo its usage on the XBSX/S edition?

 

[PSman1700]

Another reason to follow DF's analysis and findings.

 

[Globalisateur]

What I mean, is that you have no knowledge if the images are downsampled to fit into the crop. You are also working off YouTube compression. Alone, either issue would throw off any sort of measurement, together, does not make any sort of sense to attempt

Well, it's always possible indeed that the guy scaled only one of the version when he did the 4 versions comparisons (so the ratio would be wrong indeed). But I have never seen such a bad editing mistake. A very visible aliasing like this one is not modified by Youtube compression.

Anyways it's the same result using one pic posted in the DF thread: Seemingly identical geometrical resolution in both versions with some noticeably blurrier textures on XSX because of VRS.

EDIT: The guy has actually zoomed the images at 0:51.

 

[iroboto]

Judging solely off ElAnalistaDeBits Doom Eternal video, the performance/framerate seems locked across all modes and systems. Since PS5 is running DE quite well without VRS, surely it's warranted [at least in this case] to forgo its usage on the XBSX/S edition?

This is a tougher one to determine. VRS tier2 is very difficult to measure because it’s behaviour is not consistent frame to frame and where it applies itself.

Ie: games with VRS one can find dramatically different values in a single frame, so it makes it very hard to do comparison imo.

one of the reasons I started developing new tools was because I had figured this would become an increasingly difficult problem going forward. You would need newer measurement methods to determine the effect of VRS on the entire frame. Current methods look at spotting a specific area of a frame.

 

[mr magoo]

The biggest problem with new update is it broken quick resume completely. It dosent work at all now for doom eternal.Scary how quickly you get used to good things, playing without QR when you have so little time to play is just pita. Sad face

 

[Jay]

I also noticed some quite more blurry textures on XSX (balanced mode), more so than the very little resolution difference. The same problem DF noticed in their very short comparison between PS5 / XSX in their video.

I don't remember DF noticing any problems with the XSX version?
Remember him being able to find the use of VRS 2, which seemed to work as expected.

You can't just lable the use of VRS 2 as a problem unless it's not implemented well, which that wasn't said in the video.

 

[BRiT]

The biggest problem with new update is it broken quick resume completely. It dosent work at all now for doom eternal.Scary how quickly you get used to good things, playing without QR when you have so little time to play is just pita. Sad face

Pretty certain they put out a quick update afterwards fixing the QR glitch.

 

[Shortbread]

This is a tougher one to determine. VRS tier2 is very difficult to measure because it’s behaviour is not consistent frame to frame and where it applies itself.

Ie: games with VRS one can find dramatically different values in a single frame, so it makes it very hard to do comparison imo.

Yeah, I'm just wondering why the XBSX edition would even need VRS in this case, if the PS5 is capable of powering through towards such locked framerates. Just seems like an odd choice knowing XBSX should power through this as well without VRS.

 

[iroboto]

Yeah, I'm just wondering why the XBSX edition would even need VRS in this case, if the PS5 is capable of powering through towards such locked framerates. Just seems like an odd choice knowing XBSX should power through this as well without VRS.

Feels like marketing. But not sure. We need full frame tools to compare if it’s an improvement or detriment.

The greatest issue I see right now is that if you go with 4K native and run a 2x2 VRS uniformly over the entire image, in theory it should look like a 1080p image for instance.

But we don't apply VRS uniformly, or at the same rates. Which is why our measuring systems for resolution are inadequate for VRS. We have to find oblique angles to count pixels. We can't just measure anything at the moment. With VRS, it applies various shading rates using an algorithm/filter to determine what should receive what. This makes pixel counting incredibly difficult, combine this with dynamic resolution scaling, I feel like this task is nearly impossible

 

[Globalisateur]

I don't remember DF noticing any problems with the XSX version?
Remember him being able to find the use of VRS 2, which seemed to work as expected.

You can't just lable the use of VRS 2 as a problem unless it's not implemented well, which that wasn't said in the video.

Well others will have different opinions about it. But if VRS is implemented, even well, on visible textures, you'll get visibly blurrier textures as a result. It is what VRS does. For me it is a problem because with VRS ON the texture resolution difference is much more than a few percent so for not much performance gain you get noticeably softer textures.

 

[Jay]

Feels like marketing. But not sure. We need full frame tools to compare if it’s an improvement or detriment.

I've not watched any of the other comparisons.
But if the overall resolution maintained is higher which I think the marketing material indicated then the perceptual clarity could be better for when using VRS 2.
So less about locking to 60 and more about resolution (in the parts of the image that matter most).

Will also be interesting how they compare in 120fps modes.

 

[Globalisateur]

What I mean, is that you have no knowledge if the images are downsampled to fit into the crop. You are also working off YouTube compression, which compressed another type of compression. Alone, either issue would throw off any sort of measurement, together, does not make any sort of sense to attempt.

You were actually right here. At 0.51 the guy has actually zoomed the images so my pixel count was actually worthless. Sorry about it.

 

[PSman1700]

You were actually right here. At 0.51 the guy has actually zoomed the images so my pixel count was actually worthless. Sorry about it.

Thumbs up to this, everyones wrong sometimes, the thing is to aknowledge it when (correctly) pointed out.

 

[mr magoo]

Pretty certain they put out a quick update afterwards fixing the QR glitch.

Hmm I played few minutes ago and it’s still not working for me. Have to check if there is an update I am missing. I decided to revisit the game after patch and run dlc1 holy cow this is intense. I feel like I need a break every 30 it requires me to play at my max focus all the time. Damn

 

[iroboto]

I've not watched any of the other comparisons.
But if the overall resolution maintained is higher which I think the marketing material indicated then the perceptual clarity could be better for when using VRS 2.
So less about locking to 60 and more about resolution (in the parts of the image that matter most).

Will also be interesting how they compare in 120fps modes.

It’s not likely that VRS is well coupled with DRS or even used for framerate locking.
VRS can be used in a variety of ways, but the main or most simplest method is to select to a setting level on edge detection (say a sobel filter) and apply it so that the shading rates are varied. We've never see VRS used as a frame rate method, where VRS is being varied to hold a locked 60. Doom is likely to do this through DRS, meaning VRS is actually not providing any real benefit here, and they are likely not well tied together unless the XSX version is running less aggressive DRS.

I think it will be some time for there to be VRS used in a way that makes sense for performance savings, in that VRS in particular does not guarantee a level of load savings like DRS does. As soon as you cut resolution by 30%, you are likely to gain a very predictable amount of performance back. VRS, well vary per frame, you can move the slider on sobel around and may not get the performance you need to lock 60.

So there needs to be a lot of research on VRS, and in particular if performance is a problem, VRS working interlocked with DRS to provide the maximum image quality/performance.

Doom as it looks right now, seems like 2 separate systems working on their own thing. Combined it doesn't seem to offer anything here.

 

[mr magoo]

"PS5 and Xbox Series X in Balanced Mode use a dynamic resolution with the highest resolution found being 3840x2160 and the lowest resolution found being approximately 2720x1530.
Pixel counts below 3840x2160 were found more often on PS5 than Xbox Series X. As an example, in one scene the PS5 dropped to approximately 3456x1944 and the Xbox Series X rendered that scene at 3840x2160.
Xbox Series S in Balanced Mode uses a dynamic resolution with the highest resolution found being 2560x1440 and the lowest resolution found being approximately 1813x1020.

PS5 and Xbox Series X in Raytracing Mode use a dynamic resolution with the highest resolution found being 3200x1800 and the lowest resolution found being approximately 2266x1275. Pixel counts below 3200x1800 were found more often on PS5 than Xbox Series X. As an example, in one scene the PS5 dropped to approximately 2986x1680 and the Xbox Series X rendered that scene at 3200x1800.

PS5 in the 120fps Mode uses a dynamic resolution with the highest resolution found being 2816x1584 and the lowest resolution found being approximately 1992x1120.
Xbox Series X in the 120fps Mode uses a dynamic resolution with the highest resolution found being 3200x1800 and the lowest resolution found being approximately 2266x1275.
Xbox Series S in the 120fps Mode uses a dynamic resolution with the highest resolution found being 1920x1080 and the lowest resolution found being approximately 1360x765.

The Xbox Series consoles use Variable Rate Shading which can result in some double width and double height pixels being visible in the frame, though this coarser shading can often be difficult to perceive. All consoles in all modes seem to often render at or near their maximum resolutions during exploration and undemanding combat encounters.
The resolution appears to drop less often in Balanced Mode than the other two modes.
The Raytracing Mode uses raytraced reflections.
The footage here isn't exactly like-for-like which should be kept in mind when comparing the performance of the consoles here."