Current Generation Games Analysis Technical Discussion [2020-2021] [XBSX|S, PS5, PC]

Mesh Shaders are great at culling... And so are primitive Shaders. But according to speculation, Sony doesnt have standard primitive Shaders, they have a customized version of it. A version that adds the same capabilities as Mesh Shaders while retaining the native capabilities of the AMD GPUs with primitive shaders. The new GE is refered as culling triangles a lot earlier on the pipeline, making it a lot more efective as a lot of early work done on those triangles that will bem discarded later is not made at all.
This is, of course, speculation. There is nothing oficial about this, but we have Mark Cerny talking about the GE on a presentation where all that was discussed were exclusive customizations for the PS5, and se have Matt talking about earlier culling on a GE on a discussion was consoles. And guess wich console has a new GE, is refered as having early culling, and Matt knows deep?
So, on theory (since nothing of this is official), if this GE can do the same as Mesh, but culls even earlier, it may be able to compete with Mesh Shaders and VRS at the same time.

 

compared to how it was done prior to mesh shaders etc.
At no point did he compare it to what is in RDNA2. How would you even do it before? Even if you some how could, what difference would it make as long as its culled before being worked on?

What do you mean cull earlier than when mesh shaders would?
What does VRS have to do with this?
There performing totally different tasks.

How does this PS5 GE being compared to VRS keep coming up in this kind of context? Pretty sure even the person(was his name matt or something) who originally threw this out there clarified it wasn't meant in that context?

Edit : just realised you quoted the person in post prior.

 

Both replace the front end of the rendering pipeline, so both cull very early. There's not really a difference with respect to the roles they play. The only notable difference between Primitive Shaders and Mesh shaders I believe comes down to scope of functionality and amplification; explicit coding vs non-explicit. But they serve the same role.

 

As stated before, I'm talking about "the word on the street". In no way should we take this as truth, ok.
Now... GE is refered as beeing able to save performance by not generating not visible triangles. Its not a post culling process that removes invisíble triangles, but not generating them at all. This would save processing power since some tasks occur on those triangles before culling.
What does VRS has to do with this? It is explained by Matt. VRS intends to save processing power, same as culling. Question is, according to him, VRS will act on unnecessary triangles, and culling them earlier (as by not generating them at all) would bring more savings than VRS.

 

I think its better than many thought it will be and further in generation it will be less and less important. Its harder for Sony to have same flexibility as Microsoft with xbox using dx.

 

I don't see this happening, the whole point of Sony and AMD building compatibility in at hardware level is presumably because with PS4 their APIs were too near the metal to do much in the way leverage newer hardware. Microsoft have been working on leveraging their solution for decades, first with Windows then Xbox.

If you want to run older games better, Xbox is the platform you want.

 

I really do not get the wish for playing BC games on consoles with more bells and whistles, its a BC game, I would rather have current games with all the bells and whistles that includes.
If it was important to me, I would have gotten a PC to play on.

 

Best user experience for screen output will be PC, right?
They could also just remaster the game completely which they did from PS3 to PS4. This is why I really do not understand it, you want to run the current code with minor tweaks to the representation on screen, to me it feels like a waste of time. That time I would rather see them put into the make new/better stuff.
But for the sake of the argument, how much of the end-user base/market really cares? Everybody on here are super out of sync with the general user-base (my opinion at least), ie we do not spend enough money for it to be worth doing this stuff you care about.
Also BC, to me, is just a minor convenience that I use while waiting for proper PS5/XSX games.
Are there people switching from PS to XSX because they can play last-gen games with some more bells and whistles?

I also want to say, cudos to MS for the work they have done in this area, but to me it really has little to no value, compounded by the fact that the last XB I owned was X360( well i do have an XSX unpacked in the office, but thats another story).

 

Not if yore a console gamer.
Best experience for the platform your on.
Otherwise be easy to say PC gives the worse experience if you just want to play a game without having to work out what are the best settings.

That's for a very limited set of games. What about games that don't get remastered? And this doesn't stop a game from being remastered if that's what a publisher wants to do.

I think part of your misunderstanding comes from your belief its the same team doing everything?
It's the platform team dealing with BC, they don't make games.
By doing BC it's not taking resources or stopping other things from being done. Especially by 3P developers.
It's not stopping them from adding to and improving current or new games.

MS is big enough that they can fund xbox to allow them to have a dedicated platform team that includes a team that deals with BC.

Nothing wrong with you not seeing the value in it as that is strictly your opinion, this isn't a technical discussion it's about option.
But you have to at least appreciate that many people do enjoy what they do with BC and see huge value in it.
I'm pretty sure that has become apparent over the years.

 

Chopping and moving things a bit around with the quotes.

I am onboard with that it's my opinion and that other people have other opinions and are entitled to that, even if I might disagree with those opinions. In this case is merely just a case of not seeing a proper value within what I perceive as valuable. Also it is interesting and a learning experience to discuss these things.

It is in the context of BC, you want to have the best experience, replaying older games without doing remaster. A better PC basically gives you that for free, nobody has to do anything, except for you who have to pay for the PC. As a console gamer only, I understand the PC solution is not viable for everybody, but the argument was "the best end-user experience" without any qualifications about console only (well that is how I read it)

If you want to replay a game or play an old title with more bells and whistles, then remaster is the best way. If the publisher holder wants to give you the best experience then that is the best way. But value for the publisher (money, marketshare, brand recognition, goodwill etc) might not be there in many cases. But the goal is still the best end-user experience.
Sony did not remaster titles just so people could replay or let people play it with nicer graphics. I would guess it was an investment into people buying into the ip/franchise also.
We can never know for sure, but the unchartered and last of us remasters on PS4, most likely contributed to more sales of the PS4 only releases of those titles and again probably more PS4s.

Nope, it's not a misunderstanding on my part. I do know it is the platform holder.
Even MS has a finite pool of resources, so the goal is always what gives the best return for least resources, right? There is also a long and short term perspective to factor in.
Sony has less resources (aka money than MS), which was the original request, ie that Sony assignes resource to make their BC do what MS has done.
In this case you guys might be right, these changes to the platform might give the highest/best return.
Personally I do not belive so, when looking at the people make up the gaming base. I doubt people actually notice it at all, maybe some will say it feels better, but (again) I think its not enough people to make it worth it. But I do not have any facts to back this up with.

By doing BC you are using resources that could have been used somewhere else, like for instance a more mature SDK for XS, would it have made a difference, who knows, but...

Anyway, I am happy people are happy with BC and MS definitely got the best one. And its not like I hate BC, I could not have played Cyberpunk nor Vigor now, without it on my PS5.

 

You are missing the important statement by Matt. He said those features aren't mutual exclusive and if available you'd use both. There is no reason to do a comparison, because using both always wins.

Percentage just for this example and based on nothing.
Using feature A: 30% saving
Using feature B: 20%
Using A and B: ~50%

 

This is true, (and you're right that other posters are completely twisting matts posts), but an important note about matt's tone is that the actual percentages are like, A = 20%, b = 2%, a + b = 22% -- VRS is a fairly minor performance saver -- worth it to implement, but not a showstopper. The point he was making was that the hand wringing over VRS was unnecessary because, while something youd ideally use in combination with other steps, it was a relatively small piece of the pie.

Of course, there are many such As to use, rather than just the one that's unique to one platform!

 

The Coalition has several technical posts about their usages. Including performance improvements from VRS where they didnt have to decrease resolution to keep performance, thus giving increased image quality. For some reason 14%-19% number sticks in my mind.

 

Every engine should be using early culling when possible. It doesn't make sense to compare no-culling to full PS5 GE Culling when discussing performance improvements. It should be AMD GE Culling vs full PS5 GE Culling, AMD GE Primitive Shader Culling vs PS GE Culling, or AMD Mesh Shader Culling vs PS5 GE Culling.

 

You're right, theres this blog post: https://devblogs.microsoft.com/directx/gears-vrs-tier2/ -- up to 4ms, which is quite large, although I wonder if that's actually going to be representative of most uses. Here they save an enormous amount of frame time on screen space reflections and AO, which might be unusually expensive in this test scene (or dialed up to unreasonably high settings because VRS gives them the headroom to not actually render much of it)

Either way, vrs is quite marginal compared to just shading less triangles in the first place -- something many techniques offer.

 

It is also important to note that the 14% figure is what they got from implementing VRS as an addition after game development. They note in their conclusion and future work that they expect more performance savings from VRS when it's taken into consideration from the onset of development.

Point is, all these things are there to help developers exploit more of the hardware and paint a prettier or faster picture. The more of these you have, the easier the work devs have to do. Every little helps.