Doom: The Dark Ages Shows Why Microstutter Mega-Matters
video description
Date: 2025-05-10
Related videos
Comments and reviews: 20
CharlesVanNoland
As a hobby game developer for 25 years, I don't have a single clue why a game like Rocket League would have ANY frame spikes. You can literally simulate and render the entire game without sending anything to the GPU other than some model transforms - which entails sending virtually the same number of bytes of data over the system bus every frame. Geometrytextures can all just stay resident during the entirety of a game session, until another map/level/scene/arena loads, so there shouldn't be any kind of asset streaming whatsoever. One arena is just a static set of geometry, textures, and shaders - and you should be able to render virtually the entire frame using ONE SHADER. There definitely shouldn't be tons of little bespoke shaders like the game-making-kit style engines tend to produce, instead of finding where there's similarity/overlap between materials and re-using the same shader for them. DOOM performs so well specifically because it uses so few shaders. I don't know who had the bright idea in the last 15 years to start having games shuttle shaders to the GPU in the middle of gameplay, but they should be hung. Games shouldn't have tons of shaders, they should have dozens that can all be compiled at startup within seconds, and then the application never touches the shaders ever again, except to use them for rendering. Most materials are virtually identical, with only their textures being different, so why do games have THOUSANDS UPON THOUSANDS of shaders! It's ridiculous, it's an insult to gamers and their hardware. It should not be tolerated. It doesn't help anything. It doesn't allow for some kind of special effect that we didn't or couldn't have 10-15 years ago. It's just a huge waste for no good reason whatsoever.
Anyway, Rocket League has zero excuse for having any frametime spikes. For such a simple game with simple scenes and gameplay, there's zero excuse.
reply
As a hobby game developer for 25 years, I don't have a single clue why a game like Rocket League would have ANY frame spikes. You can literally simulate and render the entire game without sending anything to the GPU other than some model transforms - which entails sending virtually the same number of bytes of data over the system bus every frame. Geometrytextures can all just stay resident during the entirety of a game session, until another map/level/scene/arena loads, so there shouldn't be any kind of asset streaming whatsoever. One arena is just a static set of geometry, textures, and shaders - and you should be able to render virtually the entire frame using ONE SHADER. There definitely shouldn't be tons of little bespoke shaders like the game-making-kit style engines tend to produce, instead of finding where there's similarity/overlap between materials and re-using the same shader for them. DOOM performs so well specifically because it uses so few shaders. I don't know who had the bright idea in the last 15 years to start having games shuttle shaders to the GPU in the middle of gameplay, but they should be hung. Games shouldn't have tons of shaders, they should have dozens that can all be compiled at startup within seconds, and then the application never touches the shaders ever again, except to use them for rendering. Most materials are virtually identical, with only their textures being different, so why do games have THOUSANDS UPON THOUSANDS of shaders! It's ridiculous, it's an insult to gamers and their hardware. It should not be tolerated. It doesn't help anything. It doesn't allow for some kind of special effect that we didn't or couldn't have 10-15 years ago. It's just a huge waste for no good reason whatsoever.
Anyway, Rocket League has zero excuse for having any frametime spikes. For such a simple game with simple scenes and gameplay, there's zero excuse.
reply
renereiche
Will, there is a problem with this video: What makes a game feel fast and good is not so much frame consistency, it's responsiveness.
Let's not go into the controverse frame-gen debate, let's just take two games, both running at 120fps, one has micro-stutter the other has a higher Pre-Rendered Frames setting (basically a frame-buffer that updates the shown frame later to smooth out frame-pacing inconsistencies). The micro-stutter game will feel faster and likely better, while the other will feel like you attached a big weight to your mouse... mass that increases the characters rotational inertia and prevents you from turning fast in a game. Why is it the case Latency from slower updated screen.
Best way to test this subjectively for me: Wiggle your mouse left-to-right-to-left real fast again and again, you will feel it in your brain.
Now going into the frame-generation debate: Take the same game running at 65fps without frame-gen and the game running at 120fps with frame-gen and an added latency of 2ms and do the wiggle test and the frame-gen game will feel a bit better. But compare it to the wiggle-test on the game at 120fps native without frame-gen and it will feel MUCH better. (in other words: I really like frame-gen, but I wouldn't call it free performance if I were nvidia)
To make a game feel amazing, you need consistent responsiveness: (1) high-refresh-rate display with VRR, (2) game engine that has low latency even at higher frame-rates [latency doesn't change linearly with higher fps, double fps doesn't mean half latency, this depends on the engine and all engine have their sweet-spots for the linearity], (3) smooth frame delivery timings / good frame-pacing, and I want to include (4) higher FOV that still feels right for how far you sit away from your screen.
reply
Will, there is a problem with this video: What makes a game feel fast and good is not so much frame consistency, it's responsiveness.
Let's not go into the controverse frame-gen debate, let's just take two games, both running at 120fps, one has micro-stutter the other has a higher Pre-Rendered Frames setting (basically a frame-buffer that updates the shown frame later to smooth out frame-pacing inconsistencies). The micro-stutter game will feel faster and likely better, while the other will feel like you attached a big weight to your mouse... mass that increases the characters rotational inertia and prevents you from turning fast in a game. Why is it the case Latency from slower updated screen.
Best way to test this subjectively for me: Wiggle your mouse left-to-right-to-left real fast again and again, you will feel it in your brain.
Now going into the frame-generation debate: Take the same game running at 65fps without frame-gen and the game running at 120fps with frame-gen and an added latency of 2ms and do the wiggle test and the frame-gen game will feel a bit better. But compare it to the wiggle-test on the game at 120fps native without frame-gen and it will feel MUCH better. (in other words: I really like frame-gen, but I wouldn't call it free performance if I were nvidia)
To make a game feel amazing, you need consistent responsiveness: (1) high-refresh-rate display with VRR, (2) game engine that has low latency even at higher frame-rates [latency doesn't change linearly with higher fps, double fps doesn't mean half latency, this depends on the engine and all engine have their sweet-spots for the linearity], (3) smooth frame delivery timings / good frame-pacing, and I want to include (4) higher FOV that still feels right for how far you sit away from your screen.
reply
YourMom-ll6ku
Yes! This is exactly right. For example i was impressed how smooth one game was playing at 50 fps But another felt horrible at 60 fps , granted i could totally tell it was 50fps but the way you test the smoothness is to move the view(ie move mouse around) really slow. If everything was perfect, then This smoothness factor is proportional to the framerate and your mouse movement. For instance moving the mouse fast at 50fps makes it feel slow but moving the mouse slo w makes it smooth. Now of course that’s if frame times are perfect. So this(frame pacing /microstutter) is the metric of why one game at 50fps feels much smoother than another at 50fps. And this is more important than fps. Cpu makes a difference here for example sometime one cpu vendor’s cpu may have 300 fps and the other vendor have 250 in reviews, making one look better when in fact the lower fps cpu is putting out more consistent frametimes and so it feels’ better. Reviewers need to start measuring this , i guess it would be like the weather’s feels like’ temperature:)
reply
Yes! This is exactly right. For example i was impressed how smooth one game was playing at 50 fps But another felt horrible at 60 fps , granted i could totally tell it was 50fps but the way you test the smoothness is to move the view(ie move mouse around) really slow. If everything was perfect, then This smoothness factor is proportional to the framerate and your mouse movement. For instance moving the mouse fast at 50fps makes it feel slow but moving the mouse slo w makes it smooth. Now of course that’s if frame times are perfect. So this(frame pacing /microstutter) is the metric of why one game at 50fps feels much smoother than another at 50fps. And this is more important than fps. Cpu makes a difference here for example sometime one cpu vendor’s cpu may have 300 fps and the other vendor have 250 in reviews, making one look better when in fact the lower fps cpu is putting out more consistent frametimes and so it feels’ better. Reviewers need to start measuring this , i guess it would be like the weather’s feels like’ temperature:)
reply
sergeleon1163
The biggest difference is ID_TECH is an optimized game Engine vs for example Unreal Engine which is badly optimized and gives many devs problems game performance wise, specially in version 5.x and in regard to multiplayer also things like net code come into play where Blizzards Overwatch is one positive standout vs many others. Many devs not only in the gaming industry get lazy with generic tools, compilers, engines with will result in games/software/websites being easy(easier) to make but badly optimized code, the amount of devs making optimized code, able to code to the bare metal etc. are getting more and more rare.
reply
The biggest difference is ID_TECH is an optimized game Engine vs for example Unreal Engine which is badly optimized and gives many devs problems game performance wise, specially in version 5.x and in regard to multiplayer also things like net code come into play where Blizzards Overwatch is one positive standout vs many others. Many devs not only in the gaming industry get lazy with generic tools, compilers, engines with will result in games/software/websites being easy(easier) to make but badly optimized code, the amount of devs making optimized code, able to code to the bare metal etc. are getting more and more rare.
reply
wiard
I just put the audio on when hearing criticism on doom, since I want to minimise visual spoilers but these commentaries help me decide if the game is going to work moderately to high on a 14400FB580.... according to their own requirements its enough but it's something which would TOTALLY BE FINE IF the game was optimised in a traditionally id-software type of design. I'm nervous that this might be a money cow production, since when id moved to microsoft it's genius could quite easily be smashed by their fat
reply
I just put the audio on when hearing criticism on doom, since I want to minimise visual spoilers but these commentaries help me decide if the game is going to work moderately to high on a 14400FB580.... according to their own requirements its enough but it's something which would TOTALLY BE FINE IF the game was optimised in a traditionally id-software type of design. I'm nervous that this might be a money cow production, since when id moved to microsoft it's genius could quite easily be smashed by their fat
reply
rebootcomputa
This is the main reason why I game on Linux, even with Proton the frametime on most games its better than windows, specially win11 it was soo bad, it infuriated me, my system is very high end and I couldnt stand how bad it run with windows, sure I would get higher FPS in most games but the frametime was horrid, specially on ported games specially from Playstation, on Linux is rock solid. Cant wait to play the new Doom
reply
This is the main reason why I game on Linux, even with Proton the frametime on most games its better than windows, specially win11 it was soo bad, it infuriated me, my system is very high end and I couldnt stand how bad it run with windows, sure I would get higher FPS in most games but the frametime was horrid, specially on ported games specially from Playstation, on Linux is rock solid. Cant wait to play the new Doom
reply
decibal7532
I noticed that not having a locked multiplier and voltage on the CPU causes frame times to vary. When your CPU increases frequency and decreases frequency latency is more sporadic. Same with voltages. Need to control frequency, voltage, and temperatures to better control latency and micro stutters. Same with GPU. Haven't had issues with stutters and latency since properly undervolting and tuning my system years ago.
reply
I noticed that not having a locked multiplier and voltage on the CPU causes frame times to vary. When your CPU increases frequency and decreases frequency latency is more sporadic. Same with voltages. Need to control frequency, voltage, and temperatures to better control latency and micro stutters. Same with GPU. Haven't had issues with stutters and latency since properly undervolting and tuning my system years ago.
reply
OrlandoCaba
That sucks, I can't stand Microstutters that is why I removed all the bloat off of my Windows 10 os. I mean everything... From windows defender, game presence bar, cortana, store, etc way too much to list. I have always removed bloat off of my os installs but I came from dos when I was a teen so I am just sensitive to latency. Miss the days when my computer was MY COMPUTER, Now I have to fight for it.
reply
That sucks, I can't stand Microstutters that is why I removed all the bloat off of my Windows 10 os. I mean everything... From windows defender, game presence bar, cortana, store, etc way too much to list. I have always removed bloat off of my os installs but I came from dos when I was a teen so I am just sensitive to latency. Miss the days when my computer was MY COMPUTER, Now I have to fight for it.
reply
unknownuser3000
Video game design is one of the hardest things I've tried. Even in unity with one character model I've had dx12 crash chainswap errors that cause my entire pc to shut down. I can't imagine how much optimization, occlusion, culling takes place to get your character even rendered in 99 percent of games. People have no idea how hard it is until they try.
reply
Video game design is one of the hardest things I've tried. Even in unity with one character model I've had dx12 crash chainswap errors that cause my entire pc to shut down. I can't imagine how much optimization, occlusion, culling takes place to get your character even rendered in 99 percent of games. People have no idea how hard it is until they try.
reply
ZeroB4NG
I would find it interesting to see how these graphs with the bad spikes change if you cap the framerate at a level that the GPU and CPU can easily handle. (so that the GPU load isn't at 100% but maybe 80%)
Will the spikes appear at the same moments with the same severity or will there be enough free overhead for the engine to absorb them
reply
I would find it interesting to see how these graphs with the bad spikes change if you cap the framerate at a level that the GPU and CPU can easily handle. (so that the GPU load isn't at 100% but maybe 80%)
Will the spikes appear at the same moments with the same severity or will there be enough free overhead for the engine to absorb them
reply
SamOnKBD
Interesting, especially the histograms at the end. If something becomes difficult to discern in the time domain it can sometimes helpful to analyze it in the frequency domain. I have taken some frame time traces, normalized the data so the mean becomes 0, and then run it through a Fourier Transform and plotted it, it looks pretty cool
reply
Interesting, especially the histograms at the end. If something becomes difficult to discern in the time domain it can sometimes helpful to analyze it in the frequency domain. I have taken some frame time traces, normalized the data so the mean becomes 0, and then run it through a Fourier Transform and plotted it, it looks pretty cool
reply
Ben-Rogue
I'd also suggest taking a look at racing games for why it matters so much also. Dirt Rally 2.0 vs EA Sports WRC, both rally games made by Codemasters, with the first made in their in-house Ego engine, and the second made with Unreal, the difference in smoothness is massive!
reply
I'd also suggest taking a look at racing games for why it matters so much also. Dirt Rally 2.0 vs EA Sports WRC, both rally games made by Codemasters, with the first made in their in-house Ego engine, and the second made with Unreal, the difference in smoothness is massive!
reply
DarkHenryGoneDark
Great vid. I would only add that, I've been told, setting a cap on the framerate to contain the studders makes the game feel smooth again. It would have been nice to see frame rate cap each game would require to have an equivalent frame time consistence to Doom.
reply
Great vid. I would only add that, I've been told, setting a cap on the framerate to contain the studders makes the game feel smooth again. It would have been nice to see frame rate cap each game would require to have an equivalent frame time consistence to Doom.
reply
germ0nik
I would love to see this chart and comparison include Monster Hunter Wild, I have a 4080 but was having godawful frametime issues with that game, and it was pretty widely reported to have consistent stuttering issues across a wide variety of hardware profiles.
reply
I would love to see this chart and comparison include Monster Hunter Wild, I have a 4080 but was having godawful frametime issues with that game, and it was pretty widely reported to have consistent stuttering issues across a wide variety of hardware profiles.
reply
MikeBob2023
'Microstutter' is the reason why frametime is infinitely more important to me than raw FPS performance, because high FPS numbers are meaningless if the video isn't _consistently_ being rendered smoothly.
reply
'Microstutter' is the reason why frametime is infinitely more important to me than raw FPS performance, because high FPS numbers are meaningless if the video isn't _consistently_ being rendered smoothly.
reply
screatyswote3035
Just a bit of feedback. Whilst I understand what you mean, it makes it hard to compare the frametimes you're talking about when the scale changes between every game.
reply
Just a bit of feedback. Whilst I understand what you mean, it makes it hard to compare the frametimes you're talking about when the scale changes between every game.
reply
impuls60
I've known about this issue for over 20 years and still its an issue in most games. Now test 9800x3d oc to 6400Mhz ram cl28 vs Intels Core ultra and decent ram!
reply
I've known about this issue for over 20 years and still its an issue in most games. Now test 9800x3d oc to 6400Mhz ram cl28 vs Intels Core ultra and decent ram!
reply
dotxyn
I'm getting the impression that you've actually put quite a bit of time into playing Fortnite. Could you do a comparison between DX12 and Performance Mode
reply
I'm getting the impression that you've actually put quite a bit of time into playing Fortnite. Could you do a comparison between DX12 and Performance Mode
reply
WSS_the_OG
Am I on glue, or aren't you the guys who defended the 8 GB GPUs being released in 2025 And now you're pointing out micro-stutter. I'm confused.
reply
Am I on glue, or aren't you the guys who defended the 8 GB GPUs being released in 2025 And now you're pointing out micro-stutter. I'm confused.
reply
InstantCassette
Oblivion was showcased in the best case, a small dungeon. Try that again in the overworld moving quickly and panning the camera around.
reply
Oblivion was showcased in the best case, a small dungeon. Try that again in the overworld moving quickly and panning the camera around.
reply
Add a review, comment
Other channel videos
