Digital Foundry Article Technical Discussion [2024]

The other advantage crt's have over lcd's is black levels, I remember when I got my first lcd I spend the first few days reading calibration guides because black looked like gray to me. It drove me nuts but over time you get used to it
and if someone said to me today black looks dark grey I wouldn't agree because I've been using a lcd for years.
 
While I agree with Alex that it would be nice to get rid of FPS as a metric for a game's performance or smoothness.. I still think it's the best way to show the typical viewer an easily quantifiable difference in performance between two or more platforms at any given second. MS and frametimes tells a better story for smoothness and consistency over time though obviously.

I'd say get rid of the FPS graph portion and just post the FPS of each platform in the top corner like Rich does in his reviews, and then have the frametime graph the entire ength of the screen on the bottom. (y)
3 frametime stats (framerate mean, 5th and 1st percentile frametime) would be enough IMO. And they could even show those in real-time instead (or on top of) current framerate / frametime. NXGamer was often showing real-time framerate average in his analysis.
 
Last edited:
While I agree with Alex that it would be nice to get rid of FPS as a metric for a game's performance or smoothness.. I still think it's the best way to show the typical viewer an easily quantifiable difference in performance between two or more platforms at any given second. MS and frametimes tells a better story for smoothness and consistency over time though obviously.

I'd say get rid of the FPS graph portion and just post the FPS of each platform in the top corner like Rich does in his reviews, and then have the frametime graph the entire length of the screen on the bottom. (y)

I personally think these graphs are the easiest way for me to visualize performance (both from capframex)

1723736718420.png
1723736749758.png

The box and whisker plots in digital foundry's written content are good too.
 
Last edited:
When did frame time become the talking point?

I know on PC it was back in 2012'ish (Or was it 2013??) when it came out that poor frame times were the reason that AMD Crossfire didn't feel as smooth as SLI, and there was a big deal made of it with every PC site running articles about it.

Was it this that also sparked the conversation on console too? Or is it more recent?
 
When did frame time become the talking point?

I know on PC it was back in 2012'ish (Or was it 2013??) when it came out that poor frame times were the reason that AMD Crossfire didn't feel as smooth as SLI, and there was a big deal made of it with every PC site running articles about it.

Was it this that also sparked the conversation on console too? Or is it more recent?
IIRC it was around then, perhaps slightly earlier.. I found some emails between myself and Scott Wasson from the Tech Report where we were discussing various ways to capture and present the data from early 2013 - indeed some were linked to a B3D post I made at the time with my thoughts. Unfortunately that seems to fallen *just* off the end of where the B3D history seems to cover, but sounds like the right time-frame. I know other folks were experimenting with similar methods around the same time as well but I'm fairly sure that TR was one of the first to start putting things like frame time graphs and 99%/1% percentile type measurements routinely into their reviews.

I don't think it was specific to crossfire/SLI, but that was certainly one place where it was very obvious that the numbers said things were good but the visual smoothness did not align. There were games that exhibited poor frame pacing at the time as well that were other motivations to try and find a better way to quantify the issues.

[Edit] Found that this is still (sort of) online, which IIRC was one of the most often quoted manifestos around this topic, although there were several follow-ups. Dated 2011:
 
Sublime. My game for the next months along with Age of Mythology.


Great update and a great video, the way they multithreaded the rendering into 4 quads was extremely interesting.

Unfortunately, I'm wondering if it has something to do with the microjudder. I can confirm Alex/John's observation about the improper frame interpolation, at least with the PC version (and switch apparently). There's noticeable microjudder when rotating the camera, even at just 60fps.

(Goddamn if was ever in a position of authority wrt a games release date "Rotate the camera with a gamepad to ensure correct frame pacing" would be a QA blocker)
 
IIRC it was around then, perhaps slightly earlier.. I found some emails between myself and Scott Wasson from the Tech Report where we were discussing various ways to capture and present the data from early 2013 - indeed some were linked to a B3D post I made at the time with my thoughts. Unfortunately that seems to fallen *just* off the end of where the B3D history seems to cover, but sounds like the right time-frame. I know other folks were experimenting with similar methods around the same time as well but I'm fairly sure that TR was one of the first to start putting things like frame time graphs and 99%/1% percentile type measurements routinely into their reviews.

I don't think it was specific to crossfire/SLI, but that was certainly one place where it was very obvious that the numbers said things were good but the visual smoothness did not align. There were games that exhibited poor frame pacing at the time as well that were other motivations to try and find a better way to quantify the issues.

[Edit] Found that this is still (sort of) online, which IIRC was one of the most often quoted manifestos around this topic, although there were several follow-ups. Dated 2011:
Ryan Shrout at PCPerspective was also one of the earliest to recognize the important of frame pacing. Though even he credits Scott with kicking things off.

Both Scott and Ryan got picked up by major IHVs. They did great work. RIP TechReport 🕯️
 
Last edited:
Great update and a great video, the way they multithreaded the rendering into 4 quads was extremely interesting.

Unfortunately, I'm wondering if it has something to do with the microjudder. I can confirm Alex/John's observation about the improper frame interpolation, at least with the PC version (and switch apparently). There's noticeable microjudder when rotating the camera, even at just 60fps.

(Goddamn if was ever in a position of authority wrt a games release date "Rotate the camera with a gamepad to ensure correct frame pacing" would be a QA blocker)
the only issue I found, playing at 165fps is that the gamepad movement has too much inertia for me. Maybe it's my cheap gamepad, maybe it's the game. Gotta tweak it. Currently playing the secret level of the first campaign of Doom 1. I still remember most of the secrets of the first campaign, not many from the rest.

Another thing I preferred to do is enabling HDR, and for that I had to switch the graphics API from Vulkan to Direct3D 11. With Auto HDR enabled the levels look a lot more like I remember, without Auto HDR they look too dark compared to what I remember. HDR wasn't used then, maybe tweaking the contrast/brightness would do the trick, but still.

Haven't tried the Nuts level, but maybe if you aren't using the software renderer performance isn't an issue?
 
Ryan Shrout at PCPerspective was also one of the earliest to recognize the important of frame pacing. Though even he credits Scott with kicking things off.
Both Scott and Ryan got picked up by major IHVs. They did great work. RIP TechReport 🕯️
Indeed! I had the pleasure of meeting them both at various GDCs over the years and had some great chats about all this stuff. Both great and passionate people, but I certainly can't blame them for wanting to apply their skills in a different context. Hope they are both still doing well!
 
0:01:48 News 01: Black Myth: Wukong benchmark drops
0:22:35 News 02: Zen 5 launch unravels
0:44:24 News 03: Switch 2 possibly “delayed” further into 2025
0:54:49 News 04: Unannounced game Deadlock hits 18,000 concurrent players
1:06:33 News 05: FF7 Remake trilogy third title to switch to UE5?
1:12:32 News 06: Man connects 444 consoles to one TV
1:20:39 News 07: Tango Gameworks acquired by Krafton
1:24:26 News 08: Forza Horizon 5 potentially headed to PS5
1:34:40 News 09: Valve plans to offer SteamOS to other handhelds
1:44:58 Supporter Q1: As Microsoft increasingly goes multiplatform, what’s the point of Xbox?
1:49:45 Supporter Q2: Should Microsoft be putting older exclusives on Switch?
1:52:10 Supporter Q3: What does the DF audience have to look forward to in 2025?
1:56:12 Supporter Q4: Which graphically intense ports should come to Switch 2?
2:00:10 Supporter Q5: What are some technical annoyances you have with modern gaming?
2:04:34 Supporter Q6: What is the best way to achieve good frame-rates for PS3 titles?

 
Last edited:

PC video by Alex is up.

Good news on the relative lack of shader stutter, but of course our old friend traversal stutter is not surprisingly here. Never held up much hope for any game with traversal stutter to actually see some significant improvement with patches and that pessimism has unfortunately turned out more accurate than not, but the team is at least communicative.

That volumetric smoke battle, gorgeous.
 

PC video by Alex is up.
Very very impressive, so it's essentially Nanite (for everything except foliage) + ray traced lighting (ReSTiR) + ray traced shadows + ray traced reflections + ray traced particled + ray traced caustics, the visual result is a true treat.
 
Very nice. Shadow resolution seems pretty high. I wonder how close the BVH geometry is to the highest Nanite LOD. They might be using screen space shadows to fill in any cracks.
 
Very nice. Shadow resolution seems pretty high. I wonder how close the BVH geometry is to the highest Nanite LOD. They might be using screen space shadows to fill in any cracks.
No thanks, I tried playing Witcher 3 next gen with RT shadows, the extremely obvious lack of shadow animation anywhere further than like 2 meters away and constant shadow LOD pop (now shadows get to very obviously lod pop too!) made me turn RT shadows off entirely after a while.

Until seamless LOD geometry is possible with RT I'm keeping RT shadows off.
 
Low framerates are better on CRTs because the display flashes and this makes it more smooth compared to oled where (without bfi) the image is always on until the next frame

So while lcd would be the best for lower framerates, crt comes next and at the very end comes oled which is always choppy unless you insert some soap opera of black frame is insertion
 
Back
Top