Digital Foundry Article Technical Discussion [2023]

Status
Not open for further replies.
This forum is getting exhausting to read -- there are like 15 toxic posts here on one page, all about the same misunderstandings about game tech stack, one from a professional from another part of the industry, all right after an engineer took the time to post a big explanation. What's the point of technical discusison if it's actually "one buzzword I read one time proves you wrong" discussion.
 
Because of Microsoft, they made such a massive deal of this back when they unveiled the Xbox Series consoles and the Velocity Architecture.
I guess so... all I can say to that is that Microsoft maybe doesn't have a great track record in trying to advertise developer features to consumers, so maybe take future things like that with a large grain of salt. Ultimately what consumers need to be concerned with is the games of course.

Edit: alright I went and found a quote...
"The CPU is the brain of our new console, and the GPU is the heart, but the Xbox Velocity Architecture is the soul," stated Andrew Goossen, Technical Fellow on Xbox Series X at Microsoft via Xbox Wire (opens in new tab). "The Xbox Velocity Architecture is about so much more than fast last times. It's one of the most innovative parts of our new console. It's about revolutionizing how games can create vastly bigger, more compelling worlds."
........ yikes. I dunno how a technical guy could say this with a straight face TBH. I know both Microsoft and Sony were going off on their IO marketing for a while but man. To be clear there is legitimately cool stuff, but the coolest thing is still that there's an SSD at all.

From https://news.xbox.com/en-us/2020/07/14/a-closer-look-at-xbox-velocity-architecture/:
Today, developers must load an entire mip level in memory even in cases where they may only sample a very small portion of the overall texture.
That's just... not correct :(
 
Last edited:
The point about UE4 and DX12 "default" struck me as odd. I just loaded up UE4 and packaged the default project and indeed both the editor and game use DX11 by default. Changing the default in UE5 for Nanite/Lumen/VSM was actually a big discussion so while I'm not 100% confident I would be surprised if DX12 was ever the out of the box default on UE4 assuming you don't use any features that require it.

dx11.png

Worth noting that the DX12 RHI (and Vulkan) has seen a ton of updates and improvements as part of the shift for UE5. While it mostly works functionally, there are definitely limitations to what was there in UE4.
 
Last edited:
Calisto Protocoll, Hogwarts and Jedi Survivor use UE4...

/edit: Gotham Knights, too.
Redfall as well. And silly me, I thought Hogwarts was an Avalanche Studios game, so I assumed it used their in house tech. But it's and Avalanche Software game, and definitely uses UE4.
 
https://www.dexerto.com/gaming/acti...-from-pc-gamers-than-console-players-2138822/

I wouldn't say there is a crisis in PC gaming. Crisis for the people like us who "wear the latest fashion"? Maybe. But that's relative. Companies want to get their AAA games out as soon as possible, while giving players a polished console gaming experience. The PC port must get out in the same day, it seems, with a polished gaming experience later. Moral: only buy games like Diablo 4 day one, or most Capcom games.

Yes this was my thought too. In terms of game releases PC gaming is in the best place it's ever been by a pretty comfortable margin. All MS games, most Sony games (albeit with lengthy waits), all cross platform games day and date with the consoles and an absolutely thriving indie scene.

Sure a hand full (and in the grand scheme of things it really is just a hand full) of high profile games release with issues that tech buffs like us find annoying - and these issues are usually reflected on the console versions too, albeit often to a lesser degree - but the majority of these are usually fixed within weeks or months of the release. Could the game have just been delayed on PC until those issues were fixed? Sure, and on balance that would probably be a good thing, but none of us actually gain anything tangible from that vs just waiting to buy the game until its in a decent state.

If there is a crisis with PC gaming right now, I'd say it's isolated to GPU pricing.

Just to be clear though, I love that DF highlight the issues with individual game releases - we absolutely need that. But there's a fine line to draw here between calling out individual game issues (particularly in context of the consoles similar issues if they exist, i.e. Redfall) and setting up a general narrative that PC gaming is broken and people may as well turn to consoles regardless of what hardware they're running - which is only going to have a detrimental impact on the platform. DF obviously aren't doing that, but that narrative is definitely taking hold across the Internet.
 
The point about UE4 and DX12 "default" struck me as odd. I just loaded up UE4 and packaged the default project and indeed both the editor and game use DX11 by default. Changing the default in UE5 for Nanite/Lumen/VSM was actually a big discussion so while I'm not 100% confident I would be surprised if DX12 was ever the out of the box default on UE4 assuming you don't use any features that require it.

View attachment 8889

Worth noting that the DX12 RHI (and Vulkan) has seen a ton of updates and improvements as part of the shift for UE5. While it mostly works functionally, there are definitely limitations to what was there in UE4.
Take a look here at MS GDK - this is what I am referencing. GDK as the Basis for multiplatform releases in PC and Xbox!
 
  • Like
Reactions: snc
If there is a crisis with PC gaming right now, I'd say it's isolated to GPU pricing.

It's a big part of it, but no - not isolated at least. If it was the vast majority of the problem then these issues these AAA games exhibit could be solved by faster GPU's, but they often can't.

Just to be clear though, I love that DF highlight the issues with individual game releases - we absolutely need that. But there's a fine line to draw here between calling out individual game issues (particularly in context of the consoles similar issues if they exist, i.e. Redfall) and setting up a general narrative that PC gaming is broken and people may as well turn to consoles regardless of what hardware they're running - which is only going to have a detrimental impact on the platform. DF obviously aren't doing that, but that narrative is definitely taking hold across the Internet.

It's not DF's job though to worry about how the 'narrative' is formed by less informed sources. There's a reason, while perhaps exaggerated at points, that this narrative has taken hold - it's because so often, PC AAA titles are launching in a very poor state, and that's assuming these problem titles actually get fixed. Not all do.

DF - and frankly anyone - shouldn't be in the business of worrying about platform retention when it comes to delivering evidence-based critique on technical shortcomings of games at launch. The best way to combat the threat of PC gaming losing its appeal is to keep up pressure on publishers to release better ports from the outset. I'm not sure how you combat this 'narrative that is taking hold across the Internet' - like what are you suggesting? If it's not for DF to alter their coverage in any way then I'm not sure what you mean by this being a 'fine line'. How are you suggesting this outlook be combated, if you think it's somewhat disingenuous?

Frankly if more outlets started covering this kind of thing with greater frequency and did it from the outset, we may not be in quite this situation now. Like really, if your rebuttal to concern about the platform is "It's doing fine, you just have to wait 3-4 months after new AAA releases as they're probably be patched", that is...not reassuring!
 
Last edited:
They’re media, putting information out there and putting it in context is literally their whole job. Not taking a side on whether they’re doing it well or not, but come on.

Fair enough, they do have a responsibility to present the information in context like any outlet, but that's why I stipulated it with "...by less informed sources". The clickbait economy of other channels is kind of out of their hands, you're going to have outlets and individuals who shit their pants regardless of how reasoned your argument is.

I think overall though, DF far often falls on the side of measured critique rather than outrage clickbait. Like one of the multiple headlines of the recent DF direct being "PC gaming in crisis" is about as provocative as they get, and Alex clarifies in the video that yes, it can be considered as such - for AAA games only, whereas indie/AA may be better than ever.

I'm saying that if people on forums/other channels than take one video title and then claim the PC is 'dead' as a gaming platform, I don't really put that responsibility on them. If anything, Alex signaled he's going to be more direct in his critique going forward, and I agree. As long as they back up this harsh critique with data, then it's not their problem if some people scream the sky is falling.


All that being said, I would like to see from DF (and perhaps they can get their new PC guy to perhaps do some of these?) to maybe provide quick "Patch updates" for games months down the line that had major issues at launch. It shouldn't change their initial coverage of the game at all of course, as soon as a game is asking $70 from consumers that's when it's going to be reviewed. Question of time/resources as always though.
 
Last edited:
I may be one of the most qualified to talk about these issues on the planet and I'm telling you that you are falling into the trap that I described in the last post
Then we need an explanation on the current sad state of such APIs, Why does performance is often lower than DX11? Why does CPU overhead increase noticeably with such APIs? Why does it prevent GPU configurations from reaching their full potential? Aside from PSO stuttering/VRAM management problem, does the complexity on the CPU side explain all of that?

These are real palpable problems, and they do impact user experience in a hugely negative way, I don't think handwaving them as if they don't exist will help improve the situation.
 
Then we need an explanation on the current sad state of such APIs, Why does performance is often lower than DX11? Why does CPU overhead increase noticeably with such APIs? Why does it prevent GPU configurations from reaching their full potential? Aside from PSO stuttering/VRAM management problem, does the complexity on the CPU side explain all of that?

These are real palpable problems, and they do impact user experience in a hugely negative way, I don't think handwaving them as if they don't exist will help improve the situation.
Bigger games with more teams working on it requires more time to polish. And they don’t have enough budget or time to polish so it’s released it the state it is.

This is often mentioned and often ignored.
 

Andrew, while you're here, could I ask your opinion on GPU VRAM quantities?

I know that 8GB might not be quite enough to match Series X / PS5, but it should be enough to get close with the right choices. But the right choices aren't always made, and so extra helps mitigate issues.

In your experience, is the need for a larger and larger cushion likely to be the trend this gen, for example 8 -> 12 -> 16GB over the next few years, or do you think things have stabilised (or will do)?

It's partly out of genuine interest that I ask this ... but also partly because I need an affordable new GPU that supports RT. And LMAO an 8GB 4060Ti / 7600XT can go slap itself (surely there will be 16GB variants?).

Bigger games with more teams working on it requires more time to polish. And they don’t have enough budget or time to polish so it’s released it the state it is.

This is often mentioned and often ignored.

Amdahl's Law doesn't just apply to silicon based processors it seems.:cry:
 
Yea, I mean I don't really think the APIs are the real issue. I think the current state of things mostly comes down to publishers, and perhaps "the process" not really favoring the PC platform... PC just isn't the primary focus a lot of the time, and even when it is.. I believe there's still a mentality of "it's good enough, ship it".. That mentality has worked for a long time, but due to a bunch of different things.. it's become a bit too much and affecting things too negatively, and so we've got to push back on it. I believe things can be better, and should be. Give them enough time and enough of a reason, and they solve a lot of these issues.

I mean, maybe they didn't properly gauge how well studios would adapt their engines and integrate these newer APIs, and we're paying the price for that right now.. but I still think it was the right choice to push out these lower level APIs and get the ball rolling. It had to be done eventually.
 
Wouldn't program optimization be a huge part of Amdahl's Law or am I misunderstanding it?

I was trying (unsuccessfully) to make the joke that fleshy processor (human) effectiveness doesn't necessarily scale linearly with dev team size. You hit a point where the benefits are much smaller, but the complexity and overheads balloon out to difficult to manage sizes.
 
Take a look here at MS GDK - this is what I am referencing. GDK as the Basis for multiplatform releases in PC and Xbox!
Hmm apologies if I'm being stupid here (I haven't played with the GDK stuff much), but could you elaborate a bit further? Why would a game use GDK target unless it's a windows store game? Or maybe the game pass version of a game I guess?

In your experience, is the need for a larger and larger cushion likely to be the trend this gen, for example 8 -> 12 -> 16GB over the next few years, or do you think things have stabilised (or will do)?
I want to couch this answer especially much with this is just my personal impression, not anything particular to do with job experience. There's also a lot of "it depends on your expectations on resolution, quality settings, etc." of course. That said, as games transition to treating PS5/XSX/XSS as the baseline and no longer supporting the older consoles, 8GB starts to feel a bit too constrained... especially if you're used to running high res desktop with multiple monitors and a zillion chrome tabs and discord and other electron apps all competing for resources. Windows is not great at sharing VRAM in oversubscription situations so you really want to avoid them in the first place. I was sort of shocked when they released the 3080 with only 8GB of RAM. If you're getting a high end GPU I'd really want to have at least 12GB and ideally 16GB if you can.

Obviously PC games will still have to provide settings that work on GPUs with less VRAM, but IMO it sucks to have to sacrifice quality due to VRAM rather than the GPU's core speed. While things like textures, geometry and shadow maps are increasingly virtualized so resident memory is more fixed, they still need a pool of memory to operate in. Add on top all the new stuff coming in that takes quite a bit of room (raytracing structures, GI structures, more complex animation and simulations, etc) and I don't think we're going to go backwards on VRAM needs any time soon.
 
Seem like they were introduced in 4.23 which was 2019? Seems crazy that any current gen title wouldn't be using them, even with the horribly long development times for AAA titles.

Like many things (fast storage on consoles anyone?) it requires explicit support by the developer and to do things differently in their engine. It may not be feasible to quickly change all things in a legacy engine a developer has been using (including UE as they may have legacy customizations they've done for their prior games that they continue to use that rely on certain behavior) for AAA development.

It's easy to be a spectator (gamer) on the sidelines expecting all new tech to be implemented and adopted immediately (say within 1-4 years of introduction, IE the average development time of a AAA title in the previous console generation) when realistically AAA developers might not have the man hours needed (qualified engine architects * time allocated to work on engine before artists and designers need things locked down) to implement/change everything they might want to implement/change.

Regards,
SB
 
Then we need an explanation on the current sad state of such APIs, Why does performance is often lower than DX11? Why does CPU overhead increase noticeably with such APIs? Why does it prevent GPU configurations from reaching their full potential? Aside from PSO stuttering/VRAM management problem, does the complexity on the CPU side explain all of that?
Beyond PSO compilation, I don't think the majority of the issues here are directly due to the graphics APIs to be honest. While it is going to vary game to game I think the DF discussion on this in terms of UE4 stuff is likely a good chunk of the picture - games are just starting to push systems too far that were never really designed to handle the complexity thrown at them. UE4 was never designed to handle open worlds. Unreal in general has always had a tension between being easy to use with stuff like blueprints and very flexible actors and components vs. maintaining good performance. When scaled to large counts these systems simply bog down on serial CPU logic because they were never designed to handle such cases. UE5 has made improvements of course, but as with most engines it is an ongoing process and adherence to supporting legacy content makes it even more complicated.

That said, as we have seen with a bunch of recent launches, a lot of these issues aren't really even technical. If there's low enough hanging fruit to increase performance by double digit percentages within a week of game launch then it's fair to say the schedule was pushed too tight up to the last second. And that's a true issue in AAA games... feature and schedule creep. As a lowly dev I'm not really sure how to address that one fundamentally but it obviously sucks when retrospectively you could have "pretended" that a game launched one week, fixed a pile of issues and then "really" launched it the next week/month in a much better state if you could trick management into believing it. Unfortunately as long as there's hours before launch people will try to keep cramming more stuff in and so it's not that surprising when some last minute change does something like break PSO caching entirely when it had been working fine when QA tested it the day before. There are of course cases where things legitimately only crop up when you ship a game to millions of players, but I don't think that applies to the broad strokes here of recent releases.
 
Status
Not open for further replies.
Back
Top