Best Non-Native Resolution on 1440p Monitor

[TerribleStomper]

Hey guys,

 

I'm sporting a 1440p (2560x1440) monitor and run into FPS problems in Hunt Showdown. After setting the quality to the lowest setting, I can't get my FPS to 60 so I want to lower my res.

Now, there are a lot of different options possible:

• 2048x1152
• 1920x1440
• 1920x1200
• 1920x1080
• 1768x992
• 1680x1050
• ...
• 1280x720

 

Now I'm wondering, where I lose the least amount of visual quality? Do I run with the highest Res regardless of the aspect ratio or is it better to run with 1920x1440 (because of the correct height).  Or is it even the 1280x720 (Exactly half of the native Res so no rounding errors)?

 

This was bugging me since I got my monitor and didn't really find anything satisfactory after a quick google.

Thanks for answering

[PlayStation 2]

1280x720 will still round pixels because upscaling works by estimating the surrounding pixels to upscale, or something like that.

It's expanding one pixel into four. I'd just use whichever is the highest and still gives you 60fps.

[userzero]

As I understand it, high res monitors do not show well at 1080 or lower resolutions. Push the rest of your spec up slowly to drive the frames you require.

[KarathKasun]

Divide the resolution by 2, this results in perfect scaling on a digitally driven display.  So... 1280x720.

[Glenwing]

21 minutes ago, Dan Castellaneta said:

1280x720 will still round pixels because upscaling works by estimating the surrounding pixels to upscale, or something like that.

It's expanding one pixel into four. I'd just use whichever is the highest and still gives you 60fps.

 

5 minutes ago, KarathKasun said:

Divide the resolution by 2, this results in perfect scaling on a digitally driven display.  So... 1280x720.

 

I suggest you test that on your own display.

[TerribleStomper]

22 minutes ago, Dan Castellaneta said:

1280x720 will still round pixels because upscaling works by estimating the surrounding pixels to upscale, or something like that.

It's expanding one pixel into four. I'd just use whichever is the highest and still gives you 60fps.

 

6 minutes ago, KarathKasun said:

Divide the resolution by 2, this results in perfect scaling on a digitally driven display.  So... 1280x720.

 

Huh... so it is based on which monitor I have? I have the Dell U2715H. With this one, is it better to go 1280x720 or 1920x1080?

Just now, Glenwing said:

 

 

I suggest you test that on your own display.

Is there a good website for that or do I just have to try it in a few games?

[KarathKasun]

If the native resolution is 2560x1440, the next lowest resolution with perfect scaling is 1280x720.

 

Also, depending on your CPU, you may not gain anything significant from dropping resolution.

[Glenwing]

Just now, KarathKasun said:

If the native resolution is 2560x1440, the next lowest resolution with perfect scaling is 1280x720.

 

Also, depending on your CPU, you may not gain anything significant from dropping resolution.

The "perfect scaling at half resolution" thing is a myth. Nearly all monitors do not use that type of scaling. Try it yourself.

[KarathKasun]

Just now, Glenwing said:

The "perfect scaling at half resolution" thing is a myth. Nearly all monitors do not use that type of scaling. Try it yourself.

I have, it works as expected.  960x540 is "pixel perfect" 1:4 scaling on all of my 1080p monitors.

[Glenwing]

Just now, KarathKasun said:

I have, it works as expected.  960x540 is "pixel perfect" 1:4 scaling on all of my 1080p monitors.

Really? I've never seen a monitor that does that before. Which models do you have?

[Mira Yurizaki]

Monitors interpolate regardless of how evenly divisible the resolution fits into the native one. Meaning you're not going to get a clear image if you use say 1280x720 on a 2560x1440 monitor.

[KarathKasun]

Viewsonic VS16261, Several laptops, and a bunch (~8) of random Acer/Asus/NEC monitors.

[KarathKasun]

You can't use fonts to judge either, unless you disable clear type.

[KarathKasun]

You can download this image and display in paint at 100% zoom to see if there is interpolation going on outside of Windows scaling/smoothing.

 

http://i.imgur.com/qBRUlst.png

 

I get perfectly sharp lines when using 960x540 on a 1080p monitor.

[Glenwing]

Here is my Dell UP2516D at 1280×720 (normally 2560×1440)

 

Spoiler

(Microscope is upside down btw, that's why pixels appear to be BGR)

 

Here is my ASUS PA248Q at 960×600 (normally 1920×1200)

 

Spoiler

(Microscope is upside down btw, that's why pixels appear to be BGR)

 

Using display scaling. Not 4:1. Those are just the two I have on my desk right now, but I've never seen it on any display I've tested.

 

No offense, but are you sure you know what you're looking for?

[KarathKasun]

3 minutes ago, Glenwing said:

Here is my Dell UP2516D at 1280×720 (normally 2560×1440)

 

  Hide contents

(Microscope is upside down btw, that's why pixels appear to be BGR)

 

Here is my ASUS PA248Q at 960×600 (normally 1920×1200)

 

  Reveal hidden contents

(Microscope is upside down btw, that's why pixels appear to be BGR)

 

Using display scaling. Not 4:1. Those are just the two I have on my desk right now, but I've never seen it on any display I've tested.

 

No offense, but are you sure you know what you're looking for?

Yeah.  You may want to see if any image enhancements are enabled on your monitor.  Alternatively, you may want to see if the GPU driver is scaling the image, and outputting 1920x1200 to the display.

[Glenwing]

18 minutes ago, KarathKasun said:

You can download this image and display in paint at 100% zoom to see if there is interpolation going on outside of Windows scaling/smoothing.

 

http://i.imgur.com/qBRUlst.png

 

I get perfectly sharp lines when using 960x540 on a 1080p monitor.

So do I. But I don't think that pattern works for this purpose. Try this one:

 

Spoiler

At full resolution it looks like this:

Spoiler

 

But at half/half resolution on my PA248Q (with the same settings that gave me sharp lines in your test image), it looks like this

Spoiler

Which is interpolated, not 4:1.

 

Better results on my Dell, nearly perfect, but some interpolation is visible at the end of the vertical segments. Handling of scaling really depends on the monitor. There are probably better test patterns for this purpose too, this is just what I had on hand.

Spoiler

 

[Glenwing]

17 minutes ago, KarathKasun said:

Yeah.  You may want to see if any image enhancements are enabled on your monitor.

I'll play with the settings later when I get back

[KarathKasun]

I dont have a microscope I can get pictures from, just a shitty semi-macro phone camera... with auto focus.  Its difficult to get a good shot because the AF is hunting all of the time.  On my Viewsonic it appears to have two distinct/uninterrupted rows of green only pixels on the horizontal patterns.

[KarathKasun]

On my laptop the Intel display driver will NOT drive the panel at 960x540, it scales the image in the GPU.  Cant get good verification because the pixels are even smaller.  But the pattern looks different than at full res, so I assume the driver is doing interpolated scaling.

[Mira Yurizaki]

I don't know if test images like that are good at proving or disproving perfect scaling. Depending on the frequency of the transitions, it could line up with the scaling algorithm in a way that produces perfect results. It needs something a little more complicated. Something like the test pattern at https://www.eizo.be/monitor-test/

 

If the monitor does scale a resolution that's a factor of the native one more or less perfectly, then what the monitor displays with that test pattern at the lower resolution should match the same image but scaled up using nearest neighbor at the native resolution.

 

(Also I can't view the imgur one at the moment)

[KarathKasun]

4 minutes ago, M.Yurizaki said:

I don't know if test images like that are good at proving or disproving perfect scaling. Depending on the frequency of the transitions, it could line up with the scaling algorithm in a way that produces perfect results. It needs something a little more complicated. Something like the test pattern at https://www.eizo.be/monitor-test/

 

If the monitor does scale a resolution that's a factor of the native one more or less perfectly, then what the monitor displays with that test pattern at the lower resolution should match the same image but scaled up using nearest neighbor at the native resolution.

 

(Also I can't view the imgur one at the moment)

Its likely going to depend on the monitor.  If Glenwig's results are any indication, the Dell monitor should give nearly perfect 1:4 mapping.

 

On my Viewsonic it gives pretty much identical results, outside of the resolution/line spacing being different.

[Mira Yurizaki]

12 minutes ago, KarathKasun said:

Its likely going to depend on the monitor.  If Glenwig's results are any indication, the Dell monitor should give nearly perfect 1:4 mapping.

 

On my Viewsonic it gives pretty much identical results, outside of the resolution/line spacing being different.

I see it as two factors: what scaling algorithm is being used and what the image being scaled is. If you're feeding a uniform frequency image to the monitor's scaler and it just happens to have a scaling algorithm that plays nicely with that frequency, then you'll get a perfect result. And I'm assuming a perfect result is using nearest neighbor with an integer scaling factor because there's no interpolation.

 

So basically there needs to be a variety of images to judge how good the scaler is for integer based scaling.

[KarathKasun]

Just now, M.Yurizaki said:

I see it as two factors: what scaling algorithm is being used and what the image being scaled is. If you're feeding a uniform frequency image to the monitor's scaler and it just happens to have a scaling algorithm that plays nicely with that frequency, then you'll get a perfect result. And I'm assuming a perfect result is using nearest neighbor with an integer scaling factor because there's no interpolation.

 

So basically there needs to be a variety of images to judge how good the scaler is for integer based scaling.

The grey/green/purple test image is not uniform, and you can see what a bad scaler looks like in the first example.

[Mira Yurizaki]

5 hours ago, KarathKasun said:

The grey/green/purple test image is not uniform, and you can see what a bad scaler looks like in the first example.

The one Glenwing posted is a test pattern with uniform frequency in both directions and the rate of change is a computer friendly value.

 

If the scaler is doing perfect scaling at a resolution that's a factor of the native one, then you should be able to prove it with an image of any kind with sufficient contrast. And again I'm defining perfect scaling to mean that it's effectively the same as if you did a nearest neighbor sampling from the lower resolution to the higher one. Which for example if this were upscaling 1280x720 to 2560x1440, it would do pixel doubling in both directions.