Microsoft doesn't understand FPS
Chapters13
The hosts kick off a very special episode, joking about the world’s most attended standup, unpaid interns, and their CEO‑style affection for the team while hinting at topics to come.
A witty, pointed look at Windows run-dialog latency, the fps vs. ms debate, and why Microsoft’s performance claims spark heated reactions.
Summary
The PrimeTime’s standup dives into Windows 11’s latest run dialog patch and the claim that it’s the fastest it’s ever been, with a median showtime of about 90 milliseconds. Casey and the crew unpack how such measurements are made, what a median really means across machines, and why FPS is both helpful and controversial when talking about UI latency. The conversation veers into how latency is perceived by users, the pitfalls of benchmarking across diverse hardware, and whether a faster run dialog actually translates to a better experience. They debate the practicality of measuring a UI element that’s, at its core, a tiny text box, and how different measurement methods (loads, frames, and monitor refresh) muddy the story. The chat gets personal as they compare Windows’ update rituals, Start menu quirks, and the ongoing struggle to win back long-time Windows users. The panel also riffs on industry norms, from Netflix to game engines, and ends with a candid take on user trust and the real-world impact of software polish. It’s lively, messy, and brutally honest about what “fast” really means for end users.
Key Takeaways
- Microsoft reported a median 90ms latency for the new Windows run dialog, a figure tied to their internal testing methodology.
- Benchmark discussions highlighted the ambiguity of latency measurements: median values, quartiles, and whether timing starts at button press or release.
- A significant portion of the debate centers on using FPS as a mental model for UI latency versus raw milliseconds.
- The group argues that perceived responsiveness hinges on monitor refresh, input processing, and whether a UI action is perceived as instantaneous.
- There’s skepticism about Windows’ reported gains: the exact test setup, hardware mix, and whether ‘fast’ translates into a noticeably better user experience remain unclear.
- Progressive enhancement and pre-loading strategies are suggested as practical ways to improve perceived responsiveness without blocking the UI.
- The conversation closes with strong opinions on user trust: for many creators and long-time users, rebuilding faith in Windows would require substantial, user-first changes beyond a single patch.
Who Is This For?
This episode is essential for developers and UX engineers who design responsive UIs on Windows, as well as Windows power users curious about latency benchmarks and how marketing claims stack up against real-world experience.
Notable Quotes
""The median is 90 approximately 90 milliseconds. It was 94 I think might have been what they said.""
—Casual note on the reported run dialog latency and how numbers can differ in reporting.
""FPS is a very good way for someone to have in their head like, 'Oh, am I close to monitor refresh?'""
—Core argument about why FPS matters for perceived UI speed.
""This is not about a load operation but a UI text box, and 94 milliseconds is still fast by human perception.""
—Defending the relevance of milliseconds over framing the moment as a load.
""You can’t rely on a single median across ten-year-old laptops and expect it to represent modern machines.""
—Critique of test fixtures and hardware diversity in benchmarks.
""I hate to see a response that blames users for reacting to performance claims.""
—Frustration with how companies respond to latency criticism.
Questions This Video Answers
- Why does Windows 11 claim the new run dialog is the fastest ever, and what does that actually measure?
- How should UI latency be measured: milliseconds vs FPS, and what are the pitfalls?
- What is progressive enhancement and how can it improve perceived app responsiveness?
- Do FPS-based explanations help or hinder understanding of UI performance?
- What would it take for Windows to win back long-time users who’ve migrated to alternatives?
The PrimeTimeWindows 11Run dialogLatency benchmarksMilliseconds vs FPSUI responsivenessProgressive enhancementWeb/Netflix/industry anecdotesUser experience vs performanceWindows updates
Full Transcript
Is there any possibility we could intro? We could actually start. Uh, anyways, sorry. Welcome to the standup. Today we have an extremely special episode for everybody that's watching the standup. For those that don't know, this is the world's most attended standup. Oh, congratulations everyone for being a part of it. I hope everybody has their blockers. They're ready. Yep. Thank you. Thank you. Unfortunately, interns still unpaid. Still unpaid. Still unpaid. The interns are always unpaid. But you guys, the things is that I love you with the love of a CEO. Okay? I just want you to know that right now.
We talk about valuable work experience that someday could land you a paying job, right? Someday. Someday you guys will get paid. Just not here. Not today is not one of those days. So go get some coffee. Yeah. Now go give me some coffee. All right. Now, today our very special episode. It has to do Casey. I was absolutely fascinated with some of the tweets you were making because Windows 11 uh actually well to to be fair, we got to back this up a little bit. Uh about a couple months ago, Satia released a note of intent saying that Windows hears you.
We are going to win back the consumer is exactly what he said. And the first major patch since that announcement just came out. And that exact patch contained a new and improved Windows run dialogue. And so this is a part of their major push to make performance more top of mind. At least that is what Windows 11 explicitly said. And as part of their announcement, they said this is the fastest that the run dialogue has ever been. ever. Eve Ver like this is the best it has ever been. Here we go. And they announced it and it takes approximately 90 milliseconds to show on screen on the median.
So the 50 percentile. So we don't know what we don't know what like the 25th percentile. We don't know the quartiles here. So we have no idea what's going on. But we do know that the median is 90 approximately 90 milliseconds. It was 94 I think might have been what they said. Yeah. I don't know how they measure that. That seems really impressive considering most displays don't, you know, they're usually locked to some number that probably doesn't land on 94 every time. I don't exactly know how they're measuring that, but nonetheless, that's what they're saying.
Absolutely impressed with this performance, but then Casey, you did a little bit of snooping. I did. So, I actually didn't. Uh, so I just I I really didn't. I I didn't care because like I said, I don't use Windows really like anymore. week. We kind of we kind of nixed that. Uh pun intended. And uh so like I'm just like it's a it's kind of just a like all right you guys you know you can do your rodeo now of where you decide to try and make Windows performant or whatever. That's fine. Um but what I was going to say is originally I only saw this because Tim Sweeney was commenting on it.
I think he was said something snide like you know it's he said it was 94 milliseconds for this you know performance top of mind rewrite it's like a gaming monitor's like you know 8 milliseconds to refresh a frame like what's go you know what's going on um and so the only thing I really did was just retweet I'm like okay popping up the run dialogue like is is like a drop to 10 fps right uh because that's what you know it's like 11 fps I think would be if you rounded 94 milliseconds to be like it'd be, you know, it' be like 10.6 frames a second or something if that's was your run rate.
And nobody cared about the run dialogue for the most part. Pretty much people just cared about the fact that I used it FPS as a way to like explain what the number was. Now, remind you, I'm not posting anything original here at all. I literally just like here is the screenshot of the thing that Tim Sweeney posted or was replying to rather. Um, and I'm just like and I highlighted it and I was like they were like perf top of mind 94 milliseconds and I was like geez like that'd be really slow. Like 94 millconds is is a long time especially because like when you think about what the run dialogue is it's not really like it's it's a user interface interaction.
Whoa. Hey, is that HTTP? Get that out of here. That's not how we order coffee. We order coffee via SSH terminal.shop. Yeah. You want a real experience? You want real coffee? You want awesome subscriptions so you never have to remember again? Oh, you want exclusive blends with exclusive coffee and exclusive content? Then check out CRON. You don't know what SSH is? Well, maybe the coffee is not for you. in the dream. You hold on right there, Casey. There's processes that they have to create. I don't know what a process is, but apparently it takes 90 milliseconds.
Yeah. So, this was this was the thing um where they were like, uh, a lot of people just had a problem with this. And a lot of people's probably the wrong term. Like most, you know, it was a very popular tweet actually, so I suppose most people didn't have a problem with what I said. Um, but for the people who did have a problem with it, there were two things. One is they were like, well, this is like, you know, loading. It's like how long something takes to load, which to me is ridiculous because it's like, no, it's like a text box and maybe some history.
There's no there shouldn't be any loading. Like this should just be it it's a trivial it's you're talking about like 4K or something of data that should be required to do this thing. like it's so minimal that you there's no question you shouldn't be launching a process for this in my mind, right? Maybe if you have some kind of weird process isolation thing going on because you're scared that someone's hacking your run dialogue, but then at that point we should talk about why that's happening exactly and other things like that. But, you know, there's a whole sort of thing.
So, you could come back with some blog post where you explained like here's why it has to be a process and maybe I'd be like okay. But in general, no. like that should not be if the user doesn't perceive them to be loading something then that's not really acceptable excuse a so I didn't think anyone who replied with that I don't think they have any leg to stand on so I'm sorry like I just completely disagree with you this is not a load operation but that was relatively minor that was I felt like that was just as you as you folks say and as the kids say that was just cope that was just like trying to come up with the number for the 94 that was some cope y um a little bit of cope and also launching node so you can use react that's all just a little bit The bigger the bigger thing was um was they didn't like the use of FPS.
Uh and this to me is like I can if you if you don't want to use FPS as a way to like talk about how things feel and how long they take, right? That's totally valid in my opinion. Meaning I would never go to somebody who quoted milliseconds and say, "Whoa, whoa, whoa, son. You should be saying FPS. Like I wouldn't do that because in this house in this house we measure with FPS. Exactly. Like I would never I WOULD NEVER DO THAT. NO MILLISECONDS. If I did that to somebody and they were like, "Dude, I quote these things in milliseconds.
That's how I do it." I'd be like, "Ah, sorry." Like I was just I Yeah, I was being a dick. I I shouldn't have done that or whatever. Right? Like that. I would feel bad about that if I did that. Right? Not so with these people. These people are like, "You cannot quote FPS numbers unless apparently um because some there's some like internet committee that like must have decided this FPS has to be average FPS." Yes, it's it is. So, Casey, you may not know this, but if in a video game in a video game, if you fire a gun, but you only fire it once and it takes 90 milliseconds, that's just part of that's just Hey, that's just loading.
It only counts if you repetitively fire it. And I thought people were mad though, Casey, because the run thing is kind of like, you know, Anthropic sort of started the trend. Everything's a game engine, Casey. Well, so it's like they were they were so not a fair FPS measure. Okay. Alls I'm saying is So it what I would say is if that's the way you want to quote things, that's totally fine because I like having both. My preference is to see both. So I like to see the milliseconds and then I like to convert that to FPS in my head so I can see like if I was hitting monitor refresh because the way that I usually like to program um anything that's like a user interaction is I like to hit monitor refresh for like what I expect the target monitor refresh to be.
So if someone pushes a button then like from the time that that button comes in I want to try on the frame that it came in I want to try to have the response to it on this frame that's submitted at that time right so you push a button on one frame and the next frame that you see as like should have it now sometimes it's two frames because like there's a compositor in the way sometimes it's more if the TV sucks and it's buffering a lot of like there's reasons why that can start to multiply out.
But the way I like to think of it is like okay at least for the code uh like you know for telemetry which I assume is where this number is coming from. If you timed from like when the the event happens and when someone actually got a frame, I want to see something close to monitor refresh or twice monitor refresh depending on what you're counting there, right? Um, and so this is actually legitimately the way I think about it. Like like that was not I think people were saying or or were trying to uh impugn to me the fact that I was trying to make it sound worse than it is by using FPS.
And I guess what I would say is that is like no and or yes and no. I I do think of it that way. And yes, I suppose the reason I use that number is because I actually believe it's bad. Like when you're not getting monitor refresh that I think you did something wrong probably, right? It's like especially if it's just I'm popping up a text box or some UI thing. I just think you did something wrong. You should you should be at monitor refresh just period, right? And if you don't agree with that, that's fine.
Like you can just be like, "No, I like my things to be more more laggy than that." Um, you can totally make an argument that uh you just don't care. You're like, "Look, 100 milliseconds is what I think is reasonable for response to UI, so that's what I do." Uh, totally fine, too. But like claiming that FPS is a bad way to do it. I totally disagree. I think FPS is a very good way for someone to have in their head like, "Oh, am I close to monitor refresh?" Like, I know I should be hitting 60.
It's just a good like metric to have. And if you want me to convert that so I always say 16.66666667 instead I suppose I could say that for you. Did did that help? If I changed it from FPS to hertz what like I mean well cycles per second presumably they'd say again it has to be average for some reason. I don't know why those are not intrinsic to the measurement. Frames per second just means you took a count of frames and you divided by a time right and it does not imply that there was any more than even one.
actually uh technically you probably need to have two to measure something between but like you know um so anyway I found all of that ridiculous and very disheartening because it just I I was like this is ridiculous all of that all of this for something that as far as I can tell isn't actually relevant anyway like because uh for a couple reasons but here's the first one they said in that same post that the original run dialog experience was 104 milliseconds and the thing they were replacing. Really? Yeah. Was 94 milliseconds. That's what they said in the post, right?
No, no, no. Not not not in that post, right? I thought it was 400 millisecond. That can't be real. No, it was in the in the blog post that I saw. I think it was 104 from on Microsoft's blog. They were like the original one was 104. This one's like 94 or something like that or 98 or I don't like I don't remember the exact numbers, but they were both very close to 100. like one was less than 110, the other was less than 100, right? So between 90 and 100. Um so half half of the measurements are above 100 or above 94, too, cuz remember that's just the median.
Well, sorry. Yeah. Well, I'm just saying. So, um I was curious about that because I was like I I don't know how they're measuring it, right? Because there's different ways you can measure it. If you're if you're talking about like a latency chain, uh, where you press a physical button and then see something, it's pretty easy to get a response time to go up to something like a 100 milliseconds. If you have slow mouse, slow USB processing, the operating system's a little slow, your code is a little slow, the monitor buffers several or the GPU is buffering several frames, the monitor is buffering some more frames.
I could construct a situation where some code that responded potentially like like my little piece of code that responds instantly still is taking a 100 milliseconds because everyone else in the train is screwing me, right? This is a thing that people who have ever measured latency with like a camera know you can easily start to beef that up, right? Um Yeah. Or you've ever accidentally played a game on a TV set to non-gaming mode. Exactly. And all of a sudden you can you can feel it when you're playing Mario. You're like, "Oh, what is happening?
Am I I can't jump anymore. That's why Prime loses in games. That's actually why I've been on TV mode. I didn't know that. I've been on TV mode. I didn't know that. That's why I lost at Fortnite. Yeah, we used to watch TVs. Yeah. When he used to do that doc when he used to be do Dr. Disrespect and do that channel, he had to stop playing competitive games and create this primogen thing specifically because of this, you know, t he had a TV that was buffering too many frames. He couldn't win anymore. There's Call of Duty career went down the toilet and here we are on the standup.
But that's ancient history now. No sense dwelling on the past. The point is um so I don't know where they get this number from. Right. I don't think Casey knows. We'll just keep on going. He Let's just say hit the end of his career in as a Oh, yeah. No, it was not that. No, I I know. Okay. All right. Just Yeah. Yeah. It it it was let's say that it uh anyways it involved trash. It was not Minecraft. But anyway, um the the point is uh when we uh look at these kind of numbers, who knows how they're measured, right?
Like nobody knows, right? Um because they weren't discussed. So we know that they were statistical, meaning they said it was like a median. So meaning we we know that they were talking across a number of machines, right? uh a number of configurations probably meaning it's telemetry meaning it probably comes from end user machines but it's hard to say because again I don't I don't know exactly what their setup is for doing this but I was curious about that because I was like you know that sounds really high actually from my recollection of the run dialogue I was like I don't think it takes 104 milliseconds uh from end of click to showing the run dialogue uh on on like Windows 10 for example By the way, this is a very important part because remember the post said this is the fastest ever, right?
Which it may well which tends to mean forever. It does mean which is right which is uh which good point is what I'm trying to get to. So the part that confused me was I was like where did that number come from? So like on the on Windows 10 I tried testing the existing run dialogue which obviously is not the updated one and I did a 60 hertz capture. Now I know that 60 Hz will make people very uncomfortable. So what I mean to say is a capture where the frames were capped to 16.666667 How many Casey is that in uh decaf frames per hour?
That's my preferred that's the only unit that I prefer discussing. So just you can convert that again. What is 60 in 60th up an hour? Um yeah I don't know. Uh so I so I captured it and captured a bunch of me running the you know doing the run dialogue and just to see what it was and it always responded in two frames which would be you know somewhere in the 30 millisecond range depending on again if you were trying to include mouse clicks or something I don't know but assuming you were doing telemetry from a machine then you can't do that because you don't have a high-speed camera sitting on so I'm assuming they're recording something like that and so I don't so I don't know now So if they're using uh um that number 104 as a median, well I tested this on a machine that's a decade old.
It's 9 years old, right? So it's a very old machine. What are they testing across such that they're able to get a number that's three times higher than that as the median, right? I don't know. And since I don't know now, I have no idea what's going on. So I don't know where the 100 millisecond number goes comes from. But the reason I say this might just be much to do about nothing is because if that number is similarly inflated and the 94 is similarly inflated, the actual number for response time might be much better.
It might be that number divided by three, which would again be two frames, which is fine, right? um much much better than the uh you know drastically increased three times latency one that it sounded like it was going to be which everyone I think would have been correct to kind of freak out about because it's like they are all used to an experience where it's one or two frames to pop up that run dialogue and now it's taking three times longer that would be uh annoying. So I I really don't know what's going on here.
I feel like they needed to um if they were going to make a post about performance and they didn't want people to comment on it, they needed to put more information in there that would be clearer about what's going on. And also blaming people for having a reaction to this, especially when they just quote the like they just put a quote and be like I think what I said is am I reading this right? That's what I said. To get really mad at someone who said am I reading this right? This sounds really slow is ridiculous.
Like that's what makes X so crap. The only response to that should be uh either yes, you're reading it right. This is kind of slow for popping up a run dialogue because that shouldn't really be any work. Or yes, you were reading it wrong. Actually, that 100 milliseconds is this really bad way we measure it where we measure it from the start of the click to when we show it. And since we can't launch till the person releases the mouse button, that's an extra 50 milliseconds that's just padded on there, right? Like that is the answer to this.
Not some snide like you can't use ep. Like it's just it was garbage. Like I really didn't like the responses that I felt like they were ridiculous. And so like I'm pretty I'm pretty thumbs down on what people said. That is my annoyance with it. The happy news though is I think I think uh what's going to happen is the run dialogue will probably be fine. Um I can't say for sure, but I think these numbers are not like they're not measuring either what we think they're measuring or they somewhere in their test suite have 10,000 really old buggy laptops that take an extra 70 milliseconds to do this operation and they're skewing the median or something, right?
like like some explanation like that. So until we get more information, I think this is just a dead issue except for the fact that people out there who think FPS isn't a good way to intuitit response times, I think you're totally wrong. I there Okay, I got to jump in, Casey. So there is something about all of this that you're really missing. And I would just like to I I would love to hear Trash's opinion on this one, which is Oh god, Trash, you've worked at Netflix now since the great fall. Like Trash joined and Netflix uh dropped by 60%.
Uh effectively is what happened. It was more that was might have been 70%. Anyways, they heard that he was coming. Yeah, they heard that he was coming. He joins the company. Netflix went from 690 to 170 or something like that. It was a d it was a disastrous moment for us. Uh do you have like a bunch of like stock from when you worked there? That tra was did trash ruin your retirement? He did everybody a favor. He helped people for a decade. So, thank you trash for that. Everybody trash. Everyone give trash a clap.
This is why he's still at Netflix even though get those options down so they can go back up. Yes, exactly. There you go. Casey stoning right there. Anyway, nailed it. We've been at like you've worked on a product that's sensitive to uh any sort of code, right? Because the TV platform is just it's a very hard platform. It's a much slower platform than everything else. Now, Trash, have you ever once heard anybody at all in any point in life ever talk about response times in frames? No. Yeah. Casey, so this is something that you might not know, but in the web world, uh, nobody has ever once thought of something in terms of frames.
I don't like this is a whole novel concept of you being like, if something happens, I expect next monitor refresh rate something to happen. Like no one's ever said that ever. That would that would explain the disconnect probably because to me so I like obviously I was um raised I don't know if that's the correct term to catch on fire. Um so I was raised in the era of during the the v blink. Yeah. Right. Yeah it was good prime. Nice vocabulary buddy. Okay Casey continue on with saying something smart please. It's not smart it's historical.
uh what you normally would do, right, is you'd try to have it so that you know you update your stuff during the vertical blank of like a CRT, right? It like the scan line comes down, you update your stuff and you come back around to the top and you need to have everything done in that time like for the next Right. So, you're always I mean the the the phrase racing the beam like made it into the title of a book, I think, right? And that's what that was referring to. You're racing the beam the like the the actual like uh CRT like display beam, right?
that the thing that excites the phosphorus on the screen um of the of a cathode ray tube, right? And then went from there to like uh in sort of like the Carmarmac era of gaming like when like the Quake engine kind of ushered in 3D and this sort of stuff in in a big way. um at that point like I believe he had a thing and I don't remember if this was like something that he said in like person that I remember him saying or whether it was something that he posted in like a plan file um but I remember him being like look if you structure your update loop in one way where you like do your rendering it where you issue your rendering you you update the world issue rendering and then process input um you lose a little bit of time because some of the input maybe you could have gotten a little bit be like just before you were about to submit the the render or something like like there was a thing where he was talking about that and I was always like oh okay yeah I I probably should always be very thoughtful of like how can I make sure I get input as close as possible to submission so that I'm introducing the least like the little smallest amount of latency uh especially if you have something like uh vsync on where you're going to like if you like processed input and then waited on vsync then came around and did your render now there's an even bigger gap between those things right and so uh I've always I'm always used to thinking of it is frames so input comes in on a frame you should see it on the next frame as close as I can make it I don't want to be the weak link in that chain I want it to be the TV that's screwing you not me so I've always thought of it that way so I always think of it as FPS and Especially nowadays when like you know when I do ping times on like the network here even on this crappy Comcast connection that I'm on um most uh network on modern broadband internet the actual ping time out from the person uh to like a local to something that's close to them is actually submonitor refresh like you can actually get a network packet to someone and back uh in less time than like a 60 Hz refresh.
That's how good this stuff has gotten. You know, people on fiber are getting sub one millisecond ping times to things like the no l Google node, right? It's incredible. And so nowadays, I just think everything should be thought that way. Like it's just it's just look the fastest we can uh the reason to do it in frames. That's the fastest we can get something out to the user. Like what's the lowest uh uh amount of time this monitor can hit? That's what we should be shooting for. And sometimes it's unrealistic because it's like, okay, they're on like some fi north of 500 uh hertz monitor.
The operation we're doing is too heavyweight. So, we got to set a more conservative frame target. Totally fine. But I I never think of it in terms of anything other than that because that's our limit. That's what I can give you the frames at. Yeah. I mean to be fair we do in the even in the web world we do think about FPS especially for like animated heavy websites when you noticeably see like jank and it's like stuttering that's typically when we'll actually take into account like FPS but honestly most apps like aren't really doing any animations it's all just async crud apps just waiting for something right so it's a more of how long this take to load which we'll use time like milliseconds verse an FPS because that just wouldn't make sense in that case for us will TTF Is that what I'm hearing?
A little bit of that. A little bit of time the first bite. Yeah. A little bit of that. I feel like that is sort of a separate thing too is like um you know to me the user should see the response instantaneously even if the thing takes a little while. So like suppose I'm on some network and I'm like look it's AWS so you know good luck getting this response back in in less than 30 milliseconds because like even though the data center is is like 2 milliseconds of of of hop away or something as soon as it hits AWS it's just grinds to a halt right and like uh that's been my experience and it was like the the late the time it takes light to get to the data center has nothing to do with how long it's going to take that they are just going to chew through those milliseconds doing whatever it is that they do uh with like their switching and you know virtual machine management or whatever it is you know I don't work there I have no idea.
So assuming that you're going to do that you're like look we're not going to hit 60 Hz or 120 hertz because we know this thing is going to take like 40 milliseconds to get back from AWS. I still want to show the thing immediately. So like for the case of a run dialogue let's say for whatever reason I can't predict it. I don't know the is going to do it. I have to fetch their run history from AWS for some reason because like it's oh it's a shared it's like a synced run history across machines.
I didn't know they were on this machine or some I mean I don't know how that doesn't make any sense. I should have sunk this ahead of time but whatever. Um I would fire off that request but display the run dialogue on the next frame right because I know that by the time they're actually able to hit the first key that I'll need to search that run dialogue for I probably will get at least some of that information back. And I so I'm covering a little bit of that latency and giving the user the instant response that I want to give them, right?
Um and then if if you know the response doesn't come back and they do hit a key and I'm sitting there and I'm like as not loading, then I have to put up a little spinner or something, right? Just tell them like look AWS is is [ __ ] me here. So I get it like you know run history not available to WS response the little thing. but you know that's just how I think about it. So like I'm not being disingenuous. That is literally how I think about it. Um, and it pisses me off in games, too.
Like, if I'm sitting there waiting for a loading screen that you I'm like, "Guys, you obviously could have streamed this in." Like, why wasn't this already loaded? Like, what's going on? I get mad at that all the time. I'm just like, the the goal should be zero weight states for the user pretty much ever. So, we we've had a lot of arguments about this over over the years. And who you and uh my previous team uh shall we say? And so one of the one of the big things was what's worse, the initial loading or inapp loading?
Like which one actually makes a user more nuts? Uh which one's going to cause them to be more upset? Would you rather have a bunch of small distributed little loads throughout the app or would you rather have a bigger initial load? So in my mind usually so I would say that if I find myself in that literal position then my answer would be how often do they load this thing but before finding myself in that position I would have asked the question are you sure are you sure that you couldn't find a way to make this work because a lot of times if people are like oh we have no choice choice.
But this long initial load, I'm like, are you sure? Like, are you sure that you actually did everything you could to make that initial package you needed to get started as small as possible? Right. Well, I'm on team bigger load initially. That's if the app is, but obviously within a threshold within a if the app is loaded infrequently, then if you have to make that trade-off, I would say you do that. if the app is loaded all the time because you like typically load this thing, do a little tiny thing, and then close it, then you're kind of in this nasty situation where it's like, okay, well, I don't know.
It's it's sort of just as bad, right? Like because, you know, so I I would say I wouldn't think about those things as different. I would just say there's frequency of operation and cost of operation and you're trying to minimize the total of those, right? The cost times the, you know, the cost times the frequency. And so if the frequency of launching this app is very high and the frequency of things you do in the app is very low, then maybe it would make sense to actually try to minimize your app load time. But in general, I would say like 99% of the time when someone's like, "Oh, I needed to do X, so that's why there's a long load time here." I'm like, "No, you didn't." Like I can think of tons of ways that I could have not done it that way, right?
Like strip this out. Don't do it that way. Have a low quality version of that thing you stream in first. Right? Like think mega texture, right? Or uh level of detail would be a better way to think of it, right? I know that I have something I can display for this thing. It may just not be the best looking version of that thing. So if I'm caught by surprise, I do that right away and then increase its resolution when I can, right? Things like this uh are are you can typically do things to try and make sure that you're very responsive.
And you know, if you then decide like, okay, no, we can't do that because this thing has to be senely. Sure, you can always paint yourself into a quarter where you where you absolutely have to do this thing and then I agree you do, but I would make sure all those avenues are exhausted first is I guess what I would say. Yeah, we do have like some techniques that we call or that we coined call like progressive enhancement which is yeah the shittiest thing but if you have a good machine then we'll slowly enhance your experience right which I think is fair.
I don't go ahead. Go ahead. No, just to try and get that nice responsive feel because, you know, um and that's again that's just because that's the way I like things. I like things to be very responsive. Maybe some people just don't care. They like they push the right button on something, you know, if you open up Netflix, you push the rights button, it takes a while, kind of lagging, it goes, uh I'm like, that sucks. Like, someone should fix this, right? If someone else might be like, I don't know, it was fine, right? And so, and so it really depends on who you're talking to probably what their preference is on those two things, right?
I don't know. Um, does that make sense? Like, you know, some people might be fine with that, some people might not be. Yeah. I mean, it's hard to Everyone has a different threshold of really to tolerate, right? Yeah. So, yeah, it's hard like who do you optimize for? Like, there's nothing I hate worse. I would hate to have a quick experience like load or like a quick loading experience, but then every time I click something, it's like, ah, exactly. Now, now we got to now that thing I think that's like significantly worse in my opinion because the clicking is more frequent, right?
It's that's what I say. Like it's frequency versus cost. If you're going to click a hundred times in this thing but only load it once, you don't want the hundred things to each take the long time and the one thing to be fast, right? That's not helping anybody, right? So, right. That's what all my banking software looks like. I like open like my banking dashboard. It's just like 20 spinners and then I click something else and it's like 40 more spinners. I'm like, who's making these banking apps? Insane. Crash has just got so many zeros in his, you know, it's just such a full bank account.
after spinning forever to figure out what's in his account. Progressive zero is getting added to trash's bank account. It's like, oh my goodness, we have to widen this div. It's crazy. If I had that many zeros, I would not be on this podcast. You know what I'm saying? Happy traveler. That doesn't even make sense. Stretch. It doesn't. You're not working right now. Here's the thing. The wall of Pokemon would be bigger. Pokemon would be much better. That's what I'm saying. Aren't you having fun hanging out with us? Would you still be on the podcast? Yeah, trash.
Trash, you're not paid to be on this podcast. What do you mean? I thought you did it. Whatever. Whatever. Whatever. I said what I said. I don't care. Uh, red button, blue button, I don't I don't even care. Oh, no. That was still the best answer I'm answering. From now on, when people ask that red blue question, I'm giving Trash's answer. It's the best answer I've ever heard and I loved it. It was so good. I don't like it. I was exactly how much thought the question deserved. It was perfect. I like I that was a teachable moment for me and I have learned from Trash and I am a better person now.
That's my opinion. Anyway, I really wanted to for the full episode just like play I thought about just having the edit be it does trash his answer and then we play everything else on 16x and then it just ends with KCU saying I hated this discussion at the end of the episode. It's too fast. I thought people would enjoy the overall conversation, but I did think about like publishing the full episode just with trash answering and then it was and then the answer. It was freaking perfect, dude. I loved I loved it. M Microsoft has a bit of a hill to climb because I I'm just going to throw this out there.
I also get an uh like an update just recently that restarts my computer like six times before the update actually happens. And I just read an article from somewhere where it's just like it's actually totally normal that Windows needs to up, you know, restart your computer 15 times for an update. Like this is actually by design. And so I think part of the problem that I I personally have is that they're like look at the run per dialogue performance and it's just like well that's something I virtually never use because I just don't I mean I don't do development on there so I'm not opening something that needs that.
Typically I'm you know I'm opening a couple things that I've pinned. I use Windows in a very rookie way. And so it's just like none of these improvements really help me. I still run into bugs all the time. Like right now, for whatever reason, we're back into this state where when I create a folder, it's goes to the bottom of the list and not the top of the list. Like all the folders are on the top. I create a new one, it goes to the bottom. And so I have to leave and come back if I want it up at the top again.
And it's just like it's driving me bonkers again that there's so many little things are broken and I just want to like lose my marbles. And then they're like, "Bro, but run dialogue, dog. It's so fast." But they didn't But they didn't say that. What they said is we re we're we're putting in a new run dialogue that was made with performance top of mind. And great news, it's basically the exact same speed as the old one, right? That's what they said. Like nobody said that it actually substantially improved the run the performance. So presumably what they were trying to say was we added some features, right?
I'm guessing. I assume. I mean I don't know because I don't care. This is the other thing that I hate about the fact that this tweet got, I guess, so many likes that people started replying to it with this crap about FPS. But anyway, um I'm assuming that the idea is they added a bunch of features and didn't make it slower, which is good, right? I mean, like I'm happy that they first time in a while didn't do that, but at the same time, it's like I wouldn't mention that, right? I wouldn't have cited some numbers and been like, "Hey, like we're so proud of ourselves that we didn't slow down a run dialogue." Like I'm like, "Yeah, like like what?
That's like that's like par for the course." I MEAN, WHAT ARE YOU TALKING ABOUT? Like of course you shouldn't have slowed down the time it takes to open the run dialogue. That's a blog post now. Like what's going on? Right? So that alone too is just very frustrating cuz you're like now you're getting mad at now you're getting mad at me for replying to somebody who replied to your thing where all you did was say that the run dialogue takes 94 milliseconds which is a hell of a lot of time for something to take. And like how did we get here?
You know it's ridiculous. This shouldn't have even been news. It shouldn't even been posted. Well to be fair the person that got the person that most got mad at you was actually a writer for something. It wasn't that it was like a Odin. They could have got Odin a Wikipedia page, but instead Casey, you're now getting a Wikipedia page being Mr. FPS, you know, falsifier. Um, except it's not false. It is the It is the FPS. But yes, uh, like the question I have um, there would also be like could we, you proposed this trade, uh, I believe to Jimmy Wales on the previous episode.
Let's propose it now. I can go back and delete my original tweet and post a new tweet where I put the millisecond number in place of where the FPS number. Now, mind you, my tweet had both because I quoted the millisecond number. So, it w it wasn't that I didn't include that information, right? It was that they didn't like the even the fact that you could bring FPS into it was just this unholy, you know, uh absolutely uh ridiculous thing to do. But anyway, I'll go back and delete that tweet and only mention milliseconds. I won't mention FPS anywhere in it in exchange for them writing a nice article about Odin um and how it's used at Django Effects.
There we go. There's our deal. That's a good deal. That's a good deal. That's a good deal. Right. We all Everyone wins. Everyone. Um by the way, I don't know if you know this, but a kind of a little update. uh this person named Phil, which you can't really quite see well on my or the stream can't quite see it very well, but Phil over here wrote his own version in which takes approximately 20 milliseconds to show up on screen and so he's able to show a dialogue super ultra quick and his own handwritten one.
Yeah. But the problem with that is that like um I'm just I'm just going to put myself in the mind space of the people who were replying to my tweet. Okay. The problem take a moment, Casey. Take a take a moment. The problem with that is is in that demonstration doesn't mean anything because that person didn't run it on all the same machines and take the median. It doesn't It doesn't matter. It does meaningless. It doesn't matter. It's meaningless cuz he has a different monitor. Also, he's not loading he's not loading the run history from Microsoft One Drive or whatever the heck it is that this thing does, right?
So, he's not doing the the things I always get. They're all ridiculous, but they're that they're like um you know, it's not an apples to apples comparison. Uh you don't have all the features. It's the features. Uh it's um it's the plugins. You just don't you don't have plugins. Plugins. Everyone knows that as soon as there's the concept of a plugin, everything is takes uh several seconds to happen. Um blah blah blah. I think that's what he'll people reply with with that crap to that person even though he did it right. So Casey, you're actually missing kind of like a really fun part of webdev.
Uh which goes a little something like this, which is all right, we're going to rewrite something and we're going to rewrite it. And here's our kind of first implementation of it. And it is 10 times faster, bro. It's so good. the previous version. Like this new version, it's so fast and they release it and everyone's like high-fiving and then about 3 months later after you fully rewritten it, it's effectively baseline or worse. Like typically it actually goes worse. Like this happened with a very major project that I saw actually got significantly worse memory usage. And so it's just like but in the beginning it was so good.
So, when I see like a run dialogue rewrite that came in below budget, I mean, you don't realize what this does to me. Like, I can't I didn't even think it was possible. So, I'm actually very impressed. I mean, like I said, I I don't know. I'm just saying what people will say. I I don't know. I I I thought for Oh, go ahead. No, T, please. I was going to say initially when I saw the tweet I thought for sure I misread it because I thought they were saying the start menu thing. The thing that people actually open.
You know what I'm saying? The one that like sometimes you press the start thing and then it like it just doesn't show up for 5 seconds and then it shows up. And I was like I was like good for Yes. Yeah. It shows me ads for Edge and for Bing and and like I was like wow good for them. a feature that a regular everyday person uses. They rewrote from scratch and took from my experience median of five to 10 seconds to load it feels like. And they brought it down to 90 milliseconds. I'm like, great.
Good for them. It's the opposite of that, right? Every single version of Windows, it gets slower. The start menu is still bad and the run menu that my parents don't know exists is My mom has never seen the run menu. She does not know what that even means. Like if I were to tell, oh, just use the run menu. You'd be like, what is a run menu? I I'm kind of confused about the run menu to be completely honest because um again, this is kind of like for historical reasons. Originally, the start menu did not really have the concept of like type to execute.
So like if you look back at like a Windows 95 or Windows 2000 XP era start menu then what happened if I if I'm remembering correctly is typing would be would be interpreted as like trying to hotkey through that menu. So like you popped up the start menu using like uh the super key or whatever or the Windows key or whatever it would be called. I guess the co-pilot key now Casey co-pilot what? Yeah. Right. Um, you push that, the start button opens and you type a key. I think it just hotkeys through that menu.
It doesn't start uh it didn't originally I at least my recollection is it didn't originally start accepting those as keystrokes to be interpreted as a command of some kind. If you now do it, like if I hit that button, uh, it it does so like if I hit the start menu button and I start typing, it does sort of like a Linux C style thing where you start typing into a kind of ambiguous box that's just like, I don't know, I'm going to do something with this. Yeah, the menu effectively. Effectively, my question is, why is there even a run menu to begin with?
Now, isn't that thing like what why are there two different things? And I'm guessing, and I could be wrong, I'm guessing the answer is because the run menu is actually going to run the thing. Whereas the the box is like we have to keep that as like an idiot box. Like it has to be a thing where you just type if you type in C colon uh back slash uh you know, Windows back slashnotepad.exe exe into that or something. We have to show you ads for our new co-pilot improved uh notepad that you or what right like it has to it has to do all these other businessoriented operations.
So that can't be the run dialogue. We have to have this other run dialogue. Is that the explanation? Because I don't really know like I don't know why there are two. Yeah, Casey. Unfortunately, you're 100% right. It will literally show you internet search results in the start menu in Windows right now. My Windows machine for the internet search results. It's a thing. Uh I find Casey Wait, can I say one more thing about that though, which is that Casey, unfortunately, you'll be really pleased or sad to hear. I don't know. Depending on, you know, are we praying on Microsoft's downfall?
I'm not sure. But okay, often times it shows the internet search result higher than the local program. In my experience on that where you start typing something, I start typing slay the spire 2 and it doesn't pick the game. It's going to pick a web search to bring me to slayer 2.com or something like that instead and opens it up. Oh yeah, it does a bunch of stuff. Edge. Yeah, trash. Hold it up so we can see. Flip flip your, you know, point yourself away from the camera. Hold it. No, you might reveal this one password.
This is This is a work machine. Hold on. It might reveal his one password. Don't do that. Don't do Don't listen to Don't give away your password. Trash. Oddly enough, I do have my password on the screen right now. I saved it. I save it in place. No, you don't. Don't say NO. ARE YOU KIDDING ME RIGHT NOW? Oh no. Was that a joke? I don't think he's kidding. You can't admit that. You cannot admit that out loud. You just You just Bro, it's secure. It's secure. trash. I'm just kidding. All right. So, Daisy, can I make can I make you even more frustrated?
I mean, I I again like I don't care. Windows, do whatever you want. I switched. I don't care. The frustration was the tweets. Okay. So, that's very immature of you. This is my own life. Mhm. So, one thing that's kind of interesting about uh uh what's called the start menu is that you can you can kind of treat it like a fuzzy finder a little bit. So, I can type in like s notes and hopefully I get what I'm wanting. It doesn't quite work that well. It it pulls up results. I can't quite figure it out.
But the run one is definitely not fuzzy findable, So, the run one you have to type in like the correct thing. You have to say the program you're going to run, which I assume is the intent, right? like it's trying it's trying to be more of a command line like experience which is why I this is why I assume that there are two which is kind of weird but like you know yeah so that's that's the that's the interesting one is that you get a more fuzzy find plus internet results one for one and a no fuzzyfind one for the other one and so but the updates to the one that nobody know really generally most people don't use I assume yeah I'm assuming it's more more for people who just like They they back in the day they opened the start menu and selected run which is how you originally got to this or you learned the hotkey of win r or something and now that's just part of your workflow so we tried to keep it which I appreciate them doing like they could have just removed it like you said very few people probably use it in rel relative to their gigantic user base but very few people in this case is still probably in the hundred thousands right because win is so huge.
Yeah, I think I most often see it in um let's see, Kit Boga videos where the guys are trying to scam him and they make him run a command. That's the time I most often see the run menu is they say, "Open up run menu, type," and then they say something insane to type in there. By the way, TJ, just so you know, you could always disable search results if you wanted to. Oh, I can. I don't have any computers left now. Okay, just in case you wanted to if you know it's one of the fantastic features start menu.
You can disable ads. It's like a it's an available option. We talked about this recently. I literally tried to boot up an old Windows computer that I had so I could have it as a Minecraft machine for my wife so that when we play Minecraft together, she just has one computer that we turn on and then it opens to Minecraft. Basically, it got stuck in an update loop where I couldn't download Minecraft because it was waiting for Windows updates, but Windows updates couldn't run cuz it was trying to run Windows Store problems. And so, they fought each other.
And I couldn't get an update or Minecraft installed on the machine. So, I deleted everything and installed Linux and put Minecraft on there faster than I could get the Windows update. Minecraft. A a Microsoft product. A Microsoft product. I had to run the Android emulated version of Bedrock Minecraft on my Linux machine which I installed from scratch faster than I could figure out how to get the Windows thing to work cuz I had to do the because Bedrock one. That's the one that we're running for my kid. So anyways, it's crazy. I think Microsoft should just start shipping Linux.
I think it's over. I think they should just put ruin that too. They should put a Windows logo on Linux and call it Microsoft Linux or something like that. Microsoft Linux Windows 12. That's all you have to do. That's just it's Windows 12. Windows 12 is just Linux and they just and they just like like send all their source code to the wine developers because as far as I'm concerned, it's over. Like it's just it's over, guys. Like it's over. Instead of Windows Subsystem for Linux, it could be window what would be the opposite of subsystem?
something that can other Windows super system. There we go. Linux. Yeah, it's still WSL, but it's also C-pilot. It's called Copilot WSL because people have been trying I mean people have been trying for decades to get things fixed at this point now and it's just it's too late. Like it's just like everyone was forced to go do other stuff now. That's why this is happening. No one wanted to do this. Like like Valve didn't want to make Steam OS. Like I didn't want to change our machines here to Linux. I was fine with Windows. They just they forced us to do it by just never caring about what any of us think.
And so now we're gone. I'm not coming back. It doesn't matter. Oh, great question. Can they win you back? What would Microsoft have to do to win you back? They can't because I was That's what happens. So, when you have a customer who's been a customer for, you know, I don't know, 30 years or something or 25 years, I don't I'd have to go count how long it was. Um, since since like the early 90s probably, I would say, uh, when I was like in high school, I think I started using Windows, but um, doesn't matter like Windows 3.1 and stuff.
Uh, so the point is when you have a customer who's been with you that long and had all of their stuff like built on your system and was used to it and you know that was how they do their work. The amount of work you're asking them to do to leave is pretty substantial. In order to get to that point, right, it's not a like it's not a trial separation. It's not a like we're just gonna take some time to find each other and then we'll come back and see if this marriage still works. It's none of that.
This is a like I find you found you screwing the babysitter and we're getting divorced forever and I will never speak to you again and I'm taking the kids. Like that's that's how the Windows thing went with me and I suspect that's how it went with a lot of people. They were so bad for so long and so egregiously anti-user that the people who left because of that are never coming back. Like I'm never coming back ever. There's nothing you could do to win me back. I will die a user of some alternative operating system.
The question that they would need to ask is how do they stop that from happening to more people? That's the only question left for them. Casey, I I hear all those words you said, the babysitter, all that. Okay. Um but what if just hear me out here. You know, you know how Hollywood stopped having good and new ideas? What if Windows 7 remastered? Windows 7 part two. Windows 7 Performance Edition. Just throwing it out there. What if we went backwards? Well, but the pro here's the thing is that's assuming that uh you know that's to to continue my analogy that's just like saying I've changed I what do YOU MEAN LIKE WHAT HOW DO I know how do I know you're not going to go if I take because now I have to do work to switch back to your platform what what guarantees do I have that you won't just start sucking again immediately right and none you've demonstrated what kind of company you are you've demonstrated how you treat your users.
You've demonstrated institutionally what you think of us, which is you don't think of us. You don't care. When we repeatedly say you need to fix this Windows update situation, when we repeatedly say that like all of these things are broken, the fi like the explorer, if you accidentally drag an icon over a network share, it pauses for several seconds unreoverably. How long has that been in there? Right? Don't fix anything. I don't care anymore. I don't want to rely on you anymore. Right? It's gotten to the point where some random people on the internet who put together a Linux distribution and put up an ISO are more reliable to me than you.
That is where you have gotten right in terms of what I think of your ability to provide software to me that I can rely on. It's over. It's it's it's done. There's no way. There's nothing you could do. I will, you know, I would go start working on Linux to fix it before I would go back to Windows. So, not even XP remaster. Brutal. No, no. I It's the It's the It's a loss of confidence. Complete loss of confidence in the institution, right? Okay. But what about Windows do? Oh my gosh. Shut up. Hey guys, if you like this episode, you can watch the rest of it on the Spotify.
And don't forget to like and subscribe. Woo! See you later. Boot up today errors on my screen. Terminal coffee and living the dream.
More from The PrimeTime
Get daily recaps from
The PrimeTime
AI-powered summaries delivered to your inbox. Save hours every week while staying fully informed.






