Ultra vs Low Textures in Deus Ex Mankind Divided Frametime testing

[Bacon1]

As to not pull another thread off topic, I'm posting the results here.

There was no noticed in game stutter while I ran around for 3 minutes in Prague while it recorded. I used PresentMon with 16.9.2 drivers and DX12 version of the game.

As you can see there is no additional spikes going from low to very high and the 4GB Fury handles it very well.

This is at 3440x1440.

Anyone else care to do their own testing? I just ran around prague for 3 minutes while recording with PresentMon and Afterburner.

PresentMon flags used: PresentMon64.exe -process_name dxmd.exe -timed 180 -scroll_toggle

Full settings:

DX12
3440x1440
Ultra / Low Textures
16xAF
Shadow Quality: High
AO: On
CHS: Off
POM: High
DOF: On
LOD: High
Volumetric Lighting: On
Screenscape Reflections: On

TAA - On
Motion Blur - Off
Sharpen - Off
Bloom - On
Lens Flares - On
Cloth Physics - On
Subsurface Scattering - On
Chromatic Aberration - Off
Tessellation - On

Edit: Same spikes are shown with medium + low textures (3.3GB VRAM)

 

[Carfax83]

There was no noticed in game stutter while I ran around for 3 minutes in Prague while it recorded.

Looking at the graph, that's rather hard to imagine

As you can see there is no additional spikes going from low to very high and the 4GB Fury handles it very well.

Are we looking at the same graph? The blue represents ultra textures, and it's clear that it has many more frame time spikes into the 30 and 40ms range than the orange. Low textures seems to stick around the 15-22ms range, but ultra is much higher in the mid to upper 20s for the most part, but with a lot more spikes into the 30s and 40s.. The MSI afterburner readout tells the same thing..

I don't get it. Looking at the data which you have presented, you seem to have proven MY case more than anything, but yet you don't see a problem?

 

[Bacon1]

Do your own testing. Presentmon shows data like that and its not representative of the final output but much more spiky than what you actually see in game.

The point is, you can clearly see spikes in the Low textures version same as the Ultra. There is no VRAM limitation in the Low Texture version, yet it still has spikes. That shows that the spikes are not related to VRAM issues.

Stop being so rude

 

[Carfax83]

The point is, you can clearly see spikes in the Low textures version same as the Ultra. There is no VRAM limitation in the Low Texture version, yet it still has spikes. That shows that the spikes are not related to VRAM issues.

Random spikes like the ones that go above 90ms aren't what I'm looking at as those could be caused by pausing the game, or perhaps some random stutter that isn't VRAM related but might be associated with shader compilation. I'm looking at the averages as shown by your graph, and the amount of spikes into the 30 to 40ms range. Going by your own graph, it's a fact that ultra quality textures has increased the frame time latency significantly over using low textures...

And your framerate has also suffered as well, despite the fact that framerate typically isn't affected by texture quality, PROVIDED the GPU has a large enough frame buffer.. If it doesn't have a large enough frame buffer, then framerates can definitely be affected due to increased frame latency..

Stop being so rude

I'm not rude. I merely find it exasperating that you keep continuing to gloss over the OBVIOUS 4GB limitation that the Fury cards have.. I mean, even after proving what I've been saying to yourself, you're still trying to find some excuse as to why it is what it is..

 

[Red Hawk]

I've disagreed with Carfax before but I think the hostility is uncalled for here...disagreeing with you over the data of your tests isn't rude. I mean, what is so hard to believe about DXMD having issues trying to run ultra textures on a 4 GB card? The game itself warns that the ultra setting is meant for cards with more than 4 GB. You think you know the game better than the people who made it? If anything's rude, that is.

 

[Bacon1]

And your framerate has also suffered as well, despite the fact that framerate typically isn't affected by texture quality, PROVIDED the GPU has a large enough frame buffer.. If it doesn't have a large enough frame buffer, then framerates can definitely be affected due to increased frame latency..

Then prove me wrong and do your own testing and provide it here.

There are lots of spikes in the graph but I experienced no stuttering in the game at either, while there are spikes in both low and ultra. Look at the image I provided from pcperspective.

I'm not rude.

Using stupid similies all over your post is rude.

Please provide your own testing.

 

[Red Hawk]

Using stupid similies all over your post is rude.

Please provide your own testing.

That's a strange definition of rude.

 

[Carfax83]

Then prove me wrong and do your own testing and provide it here.

I don't have to prove you wrong. You proved your own self wrong! You're just beating a dead horse with a stick at this point. Why should I waste my time proving something that is already known to me, and something that you yourself have already shown with your own data?

There are lots of spikes in the graph but I experienced no stuttering in the game at either, while there are spikes in both low and ultra. Look at the image I provided from pcperspective.

Not everyone is sensitive to micro-stutters. Some people see them easily, others don't. I guess you're one of the latter. But even if you don't see the micro-stutters, you can surely see the framerate which is lower when ultra quality textures are used right?

Or are you going to explain that away as being caused by some anomaly with MSI Afterburner?

Using stupid similies all over your post is rude.

I love these smileys and will continue to use them, and you can't stop me

Please provide your own testing.

I told you, I can't be bothered. It's a waste of my time (and yours as well), because I already know what the result is going to be, AND because that guy on YouTube already did it.. The only person that is surprised by the results of this data is you apparently.

 

[Bacon1]

I don't have to prove you wrong. You proved your own self wrong!

No I haven't. The spikes are present even at Medium settings + Low Textures, when VRAM isn't NEAR being maxed out.

Also I ran it with medium + Ultra and memory didn't cap like it did with my higher settings + ultra.

So please, explain to me why there are the same spikes with medium + low textures with 3.3GB of VRAM used if its a VRAM problem...

That test on youtube is outdated. Please provide your own testing.

The spikes seem to happen when I do other things like setting off car alarms while running / jumping around.

 

[Bacon1]

That's a strange definition of rude.

Would you think it was rude if that was my response to you?

 

[Red Hawk]

Would you think it was rude if that was my response to you?

The eyeroll smilie could be rude, yes, because that expresses condescension. The others express confusion and surprise, which I do not consider rude. Carfax's response to you did not solely consist of smilies.

 

[antihelten]

Instead of arguing about what the graph may or may not show, it might be more useful if you posted the raw data Bacon1. That way we could easily ascertain exactly what the data shows.

 

[Bacon1]

Instead of arguing about what the graph may or may not show, it might be more useful if you posted the raw data Bacon1. That way we could easily ascertain exactly what the data shows.

http://pastebin.com/Yx7zvjzp

 

[antihelten]

http://pastebin.com/Yx7zvjzp

Thanks. Here's a quick couple of comparisons of the data, that should hopefully make it a bit clearer exactly how the differents settings compare to each other performance wise:

Note that frametime variance is calculated similarly to PCPER's methodology, with a 20 frame running average.

 

[Carfax83]

Congrats @ Bacon1, you've irrefutably proven my case with the raw data dump!

Now can we finally put this silly issue to rest?

 

[Headfoot]

Congrats @ Bacon1, you've irrefutably proven my case with the raw data dump!

Now can we finally put this silly issue to rest?

There's still one more comparison that needs to happen - it's clear that moving to Ultra textures on the Fury gets you a net loss of frametimes. What we need to see though is if that same behavior happens on a card with a bigger VRAM buffer (e.g. an 8gb card such as 390x since it is closest in speed and same architecture). If both an 8gb card and the 4gb Fury both show the same drop in frametimes from going to Ultra, then we would know that Ultra is just a taxing setting altogether. If it only drops on Fury, then it is likely because of VRAM limitations.

We need the control data - does anyone with a 390 or 390x want to contribute? I only have a 4gb 290 myself...

 

[Bacon1]

Congrats @ Bacon1, you've irrefutably proven my case with the raw data dump!

Now can we finally put this silly issue to rest?

No it shows that it takes more FPS, not that it causes stuttering and issues. Moving up settings should take more FPS. If you or someone else wants to reproduce the test I've done with a 6/8GB card please, please feel free to prove me wrong and show that changing those settings has 0 Fps hit for you.

 

[antihelten]

No it shows that it takes more FPS, not that it causes stuttering and issues. Moving up settings should take more FPS. If you or someone else wants to reproduce the test I've done with a 6/8GB card please, please feel free to prove me wrong and show that changing those settings has 0 Fps hit for you.

Actually the frametime variance graph does indicate that stuttering increases with ultra textures, however it's by a small enough amount that we can't conclusively say whether it's due to VRAM or whether it's just the expected drop in performance from an increase in settings. So as both you and headfoot mention, we need data from 6/8 GB cards.

 

[Phynaz]

Do your own testing. Presentmon shows data like that and its not representative of the final output but much more spiky than what you actually see in game.

The point is, you can clearly see spikes in the Low textures version same as the Ultra. There is no VRAM limitation in the Low Texture version, yet it still has spikes. That shows that the spikes are not related to VRAM issues.

Stop being so rude

So you say the data is not representative (accurate) and then use it to support a hypothesis. Ummm

 

[Headfoot]

I sincerely believe only 390 or 390x will be a valid comparison. An Nvidia card introduces a lot of variables around drivers, different card architecture, etc. It's a valid comparison for buyers but its not very helpful if you just want to zero in on VRAM issues.

In fact the absolute best comparison would be 290 or 290x 4GB @ 1000/1250 Mhz versus 390 or 390x @ 1000/1250 mhz. We can compare only VRAM in this manner. Similarly 1060 3GB vs 1060 6GB would be instructive and further clarify how it works on nVidia.

290x vs 390x vs Fury would be the closest comparison you could make to see how effective (or not effective) AMD's specific Fury optimizations stack up versus a non-optimized 4GB card (290x) and a non-optimized 8GB card on the closest possible architecture.

 

[Bacon1]

I sincerely believe only 390 or 390x will be a valid comparison. An Nvidia card introduces a lot of variables around drivers, different card architecture, etc. It's a valid comparison for buyers but its not very helpful if you just want to zero in on VRAM issues.

In fact the absolute best comparison would be 290 or 290x 4GB @ 1000/1250 Mhz versus 390 or 390x @ 1000/1250 mhz. We can compare only VRAM in this manner.

Yep that would be best. However if Nvidia cards show the same small fps drop, then it proves that it is the setting itself, not VRAM. It would be if Nvidia cards showed something different that we'd then need to double check if it was AMD vs Nvidia or 4 vs 6/8.

 

[Bacon1]

So you say the data is not representative (accurate) and then use it to support a hypothesis. Ummm

I'm saying that the spikes aren't actually visible when playing and that they look worse than how the gameplay actually is, and that is both with Nvidia and AMD (hence why I linked a 980 Ti). I don't have two systems to run FCAT with so I'll have to do with PresentMon.

Do your own testing and post your results, thats why this thread exists.

 

[Phynaz]

Be glad to, you're going to gift me the game, right?

 

[AtenRa]

Bacon 1, try medium settings with Ultra Textures so we can compare vs medium low textures.

 

[Azix]

from op it doesnt look like vram. i think you would expect a quality difference, i.e wider spread in frametimes and more spikes. it just looks like the low