About 8KX performance. Performance comparison of two different native resolutions

Pimax Hardware P2: 8kX
,
1

It seems to me that some still do not believe in acceptable 8KX performance on current video cards.

Well, I did a practical study on the difference in load between two different native panel resolutions, but with the same SteamVR SS resolution.

I have Pimax 4K, which supports two native resolutions: 1920x1080 and 2560x1440. (2 million pixels vs 3.6 million pixels, i.e. the difference is almost 2 times)
Actually, I just compared the load on the GPU in both cases with the same resolution in SteamVR Home in the same place.

I think everything will be clear on the screenshots.

1080p%20vs%201440p%20native
1080p%20vs%201440p%20native3112×1294 612 KB

In fact, I compared 2 different native resolutions 1080p vs 1440p (2mln vs 3.6mln pixels) for GPU loading with the same SS. As a result, it turned out that there is no difference in load between 1080p and 1440p.
The conditions were clean, the helmet did not even move, the picture is the same, same resolution.

update:

Thanks to @Risa2000, it turned out that SteamVR Home gives the wrong idea of ​​resolution, and, as a result, of loading the GPU.

There is a performance difference between different native resolutions in games.
In the case of 1080p and 1440p, the difference was 15%! (Hellblade)

That is, the loading of the GPU has increased in 1440p mode by 38% compared to 1080p mode.

1080p%20vs%201440p%20native%20
1080p%20vs%201440p%20native%203113×1273 566 KB

In 1440p mode, native resolution provides 1.8 times more pixels than 1080p.
In the 2160p mode, the number of pixels increases by 2.3 times in comparison with the 1440p mode.

From this we can conclude that 8KX in native mode will require more than 40% (approximate ) more performance than 1440p (8K +) mode.

10 Likes
2

Thanks for the independent confirmation!

1 Like
3

SteamVR Home is not a good application to compare the GPU load, because it adjusts the GPU load dynamically and ignores the recommended render target resolution from SteamVR.
https://community.openmr.ai/t/recommended-render-target-resolution-is-just-that-recommended/16832

The chances are, the settings in your case were both ignored and the app run in both cases at the same GPU load with the corresponding resolution.

Unless you can confirm the resolution it uses by other means, do not rely on what SteamVR setting says.

3 Likes
4

The native resolution of the headset was different in both cases.
Well, tomorrow I will do the same in some game, if you consider SteamVR Home an inappropriate place for this.

Of course, I am an amateur in this regard, but I think it should work something like this. The only thing is that the 8K + / 8KX scaler may have slightly different algorithms, and there will be a slightly different difference. But in general, the principle is probably the same, I think so.

Therefore, you should not consider this as some kind of expertise, but it can be taken into account as some indirect indicator of the difference between two different native resolutions for the same LCD RGB panel, which is some analogy to the 1440p and 2160p modes for 8KX.

5

I am sorry, I misread your post (images), I thought you were comparing two different rendering resolutions. Now I see that you run SteamVR Home at the same render target res.

You are right, in that case, this should have exactly the same performance impact (GPU load). It does not matter what is the native resolution of the display (as the final pre-lens warp transformation should have minimal impact).

I would however still recommend not using SteamVR Home (and possibly any other Valve app as The Lab), because it will give you false suggestion about the rendered resolution (for the reason mentioned above).

4 Likes
6

Great, thanks for the confirmation :beers:
Yes, I’m sure to repeat the same thing in some game.

1 Like
7

You were very right, the game has a completely different situation with GPU performance. The difference between 1080p and 1440p was as much as 15%!

1080p%20vs%201440p%20native%20
1080p%20vs%201440p%20native%203113×1273 566 KB

I was tricked by SteamVR Home :eyes:
I will redo the first post now.

1 Like
8

Well, actually, it seems I was wrong :slight_smile: as I did not expect there to be any difference. Unless the game also does some “resolution guessing”, the only other possible explanation is that the pre-lens warp transformation for higher res panel takes much more time than for the other one.

Which, technically is probably true (that it takes more time), but I would not expect a difference of this magnitude. It says frame time is 11.4 vs 8.8 ms, which is like 25% more.

1 Like
9

I can also confirm existing GPUs are up to the task, running DCS World at 1.46x Pimax 8kX native resolution.

3 Likes
10

We’ll have a completely new generation of video cards out by the time the 8KX releases anyway.

2 Likes
11

Maybe, not.

To be meaningful, a new GPU would need to be 70% faster than an NVIDIA Founders Edition ‘overclocked’ RTX 2080 Ti.

Because EVGA RTX 2080 Ti Cards are already reliably overclockable beyond a steady 2085MHz, which is probably already ~30% faster than any other available GPU.

1 Like
12

My Asus strix 2080ti is running stable at 2100hz on water . I hope the next Nvidia Gpu’s will also be easy to overclock . From what I have read their next gpu being 70% faster is highly unlikely though .

3 Likes
13

70% faster in total, including any ‘overclocking’ from pre-binned cards

That would give us something 30% faster than what we already have, which is just enough to unlock slightly higher resolution and framerates.

I can also get 2100MHz. I just don’t trust it enough yet at that speed.

2 Likes
14

FPS? 20702080280TI3080chars

15

Not sure what you are asking. In any case, at the moment, I am currently adding maximum load statistics to my spreadsheet for a few apps. Already, I have exceeded 1.5x super-resolution for the Pimax 8kX under a variety of DCS World loads.

My GPU is this.

By now I guess I have posted these links more than a few times around here… hope that is ok.

2 Likes
16

^^but what FPS are you getting? I have that same GPU (although x2 but doesn’t work in VR sadface)

17

Above the 32Hz or 37.5Hz smart smoothing cutoff is what I generally try to achieve. Though DCS World in some places can run without smart smoothing, and in others, is within a few percent of missing rendering requirements to avoid severe double framing. So every bit of performance counts.

2 Likes