New Nvidia Cards

[gilesgoat]

( parenthesis "stupid question for Dio" ) ...

Not sure if you know anything of this but .. WHY apparently my Nvidia GTX 1070 "supports multiview extension" ( i.e. I can perform multiple 'passes' with a pipeline, very useful in VR for example ) but apparently "depends on where it outputs" ..

By absurdity when I output on the Oculus Rift ( where multiview would be VERY useful ) it tells me "that extension is not available" but if I render on a standard PC screen it IS available.

Does it make any sense ? Surely "an output device" should not affect this ? Isn't Oculus "Just another HDMI display" ??

[Dio]

I still think ( from my point of view ) that "not having a real separation" between shaders stages is ... "odd" I mean I don't know at this point "if it makes any sense" to talk about VS/FS/xxxS and/or in the end "gets all into a single compiled mess" and/or if we should call them alL "compute something shader" and that's it.

The whole point of the PSO is that everything can potentially be converted to code and compiled up front. How many and what kind of blobs it gets divided into is arbitrary. While it's unlikely that anyone will go to software scan conversion or depth buffering (since it costs very little hardware) it is perfectly possible to do it.

Not sure if you know anything of this but .. WHY apparently my Nvidia GTX 1070 "supports multiview extension" ( i.e. I can perform multiple 'passes' with a pipeline, very useful in VR for example ) but apparently "depends on where it outputs" ..

By absurdity when I output on the Oculus Rift ( where multiview would be VERY useful ) it tells me "that extension is not available" but if I render on a standard PC screen it IS available.

Does it make any sense ? Surely "an output device" should not affect this ? Isn't Oculus "Just another HDMI display" ??

No idea. The multiview thing is a load of old cock if you ask me. We had a much better plan for VR in Navi10 (which was designed at the height of the VR boom) so you could send a single primitive to the rasteriser and get two primitives with the rendertarget scale adjusted to closely match the lens warp function. It's vastly better than forcing people to muck about with multiview, but VR dropped out of being the hot ticket in graphics so we couldn't get any traction with it and it's not currently accessible.

VR definitely isn't 'just another display' though - there's a private compositor involved, so the path by which the data gets from presented buffer to the eye is very different to the DWM path.

[gilesgoat]

Out of curiosity "VR is not the hot ticket ?" .. ok .. "what is it now ?" .. if it's not 4K vid what else is ?

[Dio]

Currently the hype is firstly RT, with strong contenders from streaming and content scaling algorithms. 4K also yes. ML lurks but nobody has yet found any real use for it.

On the technical rather than the marketing side, meshlets, VRS and task graphs are probably the most interesting topics.

VR died as a mass-market product the day Oculus stuck that price tag on the consumer Rift, unfortunately. It's still an interesting niche, but it's not much of a platform.

[gilesgoat]

Currently the hype is firstly RT, with strong contenders from streaming and content scaling algorithms. 4K also yes. ML lurks but nobody has yet found any real use for it.

On the technical rather than the marketing side, meshlets, VRS and task graphs are probably the most interesting topics.

VR died as a mass-market product the day Oculus stuck that price tag on the consumer Rift, unfortunately. It's still an interesting niche, but it's not much of a platform.

ML ? ... VRS ? ...

[edit] ah Machine Learning ? but .. VRS ?

[Dio]

Variable Rate Shading - drop the pixel shader to 1x2 or 2x2 rate.

[gravy]

FidelityFX was very competitive with the first version of DLSS - more efficient, a much more reliable result, works on all GPUs not just those with the funky ML hardware. There's really no need to use ML for this, but of course nvidia are desperate to do that to light up those otherwise dead transistors.

I don't think we're planning to sit on our arses on this one.

definitely looking forward to any announcements next month. You guys have been killing it recently.

[gilesgoat]

*CLICK*

VR died as a mass-market product the day Oculus stuck that price tag on the consumer Rift, unfortunately. It's still an interesting niche, but it's not much of a platform.

I am not so sure, to me "is more dead when they sticked the FB 'tag' on it".

I mean "how much price you wanted it to be ?" .. "100 quids ?" .. I think yes maybe the INITIAL price was a bit "for nerds with money only" but take the Quest and such ..yes it's still a bit high but "you get quite something" for it and you have to admit that "finally" VR is getting a bit better in SW, you begin to see some things that are really worth it.

Also with the increasing pandemic and such "VR could become more important than ever" and there are quite a number of "serious applications" where thankfully it seems it's starting to get a bit like "virtual properties tours" "virtual galleries" "historycal reconstructions" .. there are many possibilities.

If you think about it "even your graphic cards ARE NOT CHEAP AT ALL !" .. they are still "something for people with quite some money" and yet people buys them

I think VR "is goign to stay is going to be better" I would NOT "underestimate" it and likewise "call it even MORE NICHE/non-existent" but I DO LOVE that "Looking Glass" technology.

I would DARE to say that "companies like you that has a fuckton of money" SHOULD have "a small division working at loss" that "always keeps an eye on all those new techs and maybe makes plans/prototypes for 'when times will be' " ..

Ok SORRY to say this "maybe AMD has not that much money" but I'd expect NVIDIA to have "money at loads" they COULD be able to do that, and if they don't I'd say "you are idiots" because you coud do it.

If I'd be "in some position to be able to do it" at Nvidia/AMD I'd push VERY HARD to say "we MUST HAVE such a division working on such things ! Always !".

VR/AR is going to be at some point "something" and you don't want to be "just jumping in" waiting "for the markets to be bigger".

It's like the movie industry where they have "3 projects they are kinda sure they are going to make money and 1 'boh could do nothing' but they still keep working on 'just in case it could be really something else' " ..

This work-for-sure-profit-only is an EXTREMELY DAMAGING phylosophy, "you have pockets big enough to spend SOME money in things that yeah maybe for the next 5 years won't do any money at all but at some point could be something really big".

Or at least pay *ME* a TINY fraction of that to keep working on that .. in case you don't want to take the risks ( donations accepted please use Paypal, check Llamasoft donations page, if you want shit nobody else wants to do keep in touch with us, send us your HW prototypes we'll be happy to have a look, the strangest the better )

Ah this "big industry" .. "It was CRAZY .. THEY EXPECTED RESULTS THERE !!" - Ghostbusters

[Dio]

I am not so sure, to me "is more dead when they sticked the FB 'tag' on it".

I mean "how much price you wanted it to be ?" .. "100 quids ?" .. I think yes maybe the INITIAL price was a bit "for nerds with money only" but take the Quest and such ..yes it's still a bit high but "you get quite something" for it and you have to admit that "finally" VR is getting a bit better in SW, you begin to see some things that are really worth it.

Yeah, basically. VR had to build a platform and an ecosystem to be anything more than a niche interest. Which meant the backing of someone with Facebook's pockets, and a preparedness to sell the headsets on the console model, i.e. close to cost.

At $500+ it's a rare peripheral, maximum market size 1m units; at $300 it could have started building a platform, maybe heading towards 5m units; at $150 it would be an impulse purchase, 10m+ units. Only the latter allows for the possibility of a true VR ecosystem with many things being targetted very specifically at the platform. In the former category (which is where VR mostly is, even today) it's niche. Oculus had led the world to believe the latter was a possibility, with the price of the devkits being very much more reasonable than the actual consumer units were.

Sony understand this best, which is why PSVR is 3/4 of the VR market even though the platform is rather underpowered for VR.

I would DARE to say that "companies like you that has a fuckton of money" SHOULD have "a small division working at loss" that "always keeps an eye on all those new techs and maybe makes plans/prototypes for 'when times will be' " ..

Ok SORRY to say this "maybe AMD has not that much money" but I'd expect NVIDIA to have "money at loads" they COULD be able to do that, and if they don't I'd say "you are idiots" because you coud do it.

Yeah, AMD even now is not resource rich. We are, I dunno, about 10k people, growing again for the first time in a while but slowly - AFAIK nvidia and Intel are both over 50k. We are lean and efficient, punch WAY above our weight, but we aren't really resourcing research that can't draw a direct line to profit.

[gilesgoat]

*SNIP*
Yeah, AMD even now is not resource rich. We are, I dunno, about 10k people, growing again for the first time in a while but slowly - AFAIK nvidia and Intel are both over 50k. We are lean and efficient, punch WAY above our weight, but we aren't really resourcing research that can't draw a direct line to profit.

YOU REALLY WORRY ME !!! REALLY !!! If AMD can't do that then WHO should ?

50K .. 10K ... 20K .. "where are the days of Hewlett and Packard in a garage doing oscillators with an incandescent lightbulb (*)" ?

MAN this is "so wrong" .. if a company like AMD "can't do that" ? .. then really "who should do it ?" ..

"There's something wrong in the system" .. very wrong .. and/or maybe "this is really another chance for bedroom/garage companies ?" ..

There are "people living on ramen" SO enthusiastic about VR/light fields/strange tech "would spend all time to do those things" and companies that SHOULD "be part of this" seems to be in the condition "too big too fail but not big enough to have at least a bit disconnected from direct profit" ..

Like some "game industry" "too big to risk because they spend to much just to be what they are" ...

"Technological INvolution" ...

Ask me if I'd want to work for AMD/NVIDIA the answer would be NO then, "unless you put me in that experimental non-profit thing that you don't have".

Anyway if you'd pay me a tiny fraction of what you pay "to go on with the direct line of profit" maybe I can do it seriously if you can't have it "at home" you should consider "outsourcing" it .. but maybe you are already doing it ..

Growing growing growing .. " a cancer ".. just "stay cool man" .. why always growing ?

[Dio]

If you want product there's no choice but to grow. The days when one person could architect a piece of silicon or build the associated software are 30 years gone. As an example, just the ray tracing component of DX12 is a larger API than the entire of DX9.

[KlaxonCow]

If you want product there's no choice but to grow. The days when one person could architect a piece of silicon or build the associated software are 30 years gone. As an example, just the ray tracing component of DX12 is a larger API than the entire of DX9.

True, true.

Though giles has a point, that scale, unto itself, can be its own problem.

As we can see in the entertainment industries, as things get bigger, they become more and more "risk averse".

Understandably so, because when it costs, say, £500 to finance a Pink Floyd album then, fuck it, leave them to be as creative as they like. Do whatever weird shit they fancy. But when someone comes in with a "vision" and is asking for £5 million to make it real, then you can totally understand that, with such large sums of money on the line, people want assurances. That they might start messing with the creative process to insist that you've got to follow this or that "winning formula" that has proven itself successful before.

I mean, if it's personally your $500 million that's funding the next movie blockbuster, then you might well also insist that it's a sequel to something that's already proved very successful, or is based on a franchise - Marvel, Tolkien, Harry Potter, etc. - that's raked in the money elsewhere and already well-loved by a passionate fan base.

Basically, you want to ensure that you get your $500 million back, at least, much less hopefully make a nice big profit on it all.

When it's just two guys then you can do whatever mad shit you fancy, more or less. Experiment. Try out different genres and novel concepts.

But when it's half a billion on the line and there's 100 employees that need to be fed, it's understandable - at that scale - that creativity and experimentation, exploring uncharted territories, very much takes a back seat. Because, sure, that's nice and everything - but, first and foremost, I need a "return on investment" for my half billion.

And, rightly so, not just because of wanting greedy profits, but just so that an industry can continue to exist. If every movie cost $500 million, yet never once made that money back at any point, then movies would stop being made. You can't run at a loss forever. The money eventually will run out.

So all movies have become sequels, or are based on an already successful franchise - comic books, books, etc. - and the music industry "prints out" formulaic boy and girl bands, who mostly sing cover versions of songs that already reached the top of the hit parade back in the day. And, similarly, games are all sequels - Call of Duty, Tomb Raider, GTA - or they're copying an indie game success that proved that something else can work (e.g. Minecraft proving that folks like "survival games", PUBG proving that "Battle Royale" works, etc.), so it's tried-and-tested by the time they're doing it. Let the small indies do the experimentation and if they throw up something good, then we'll assimilate that into our "winning formula" going forward.

It all makes perfect sense. If it were my millions and billions on the line, then I'd totally be as equally worried that I'm not throwing it away on some weird "experimental" shit that's never going to work too. If I own a game studio with hundreds of employees, then I have a human responsibility to all those people not to bankrupt the company and put them all out of work on some mad whim that "VR yo-yo games are the future". Nope, we're making an FPS, with a "Battle Royale" mode, that involves resource gathering and crafting.

And this is the inherent problem with scale. Creativity withers, nuance dies, and we end up "risk averse" over those mad strokes of genius - which, you know, might, indeed, have been absolutely amazing. But I can't keep throwing half a billion at everyone's latest mad ideas, unless some of those ideas start paying back big time.

Scale has its own inherent problems. In that bigger ain't better. I mean, it helps some things flourish, no question, but it rots away other things.

[gilesgoat]

There HAS to be "a NEW system / DIFFERENT model" .. "in everything" .. from the "production" point of view to the "consumer" point of view.

We can not go on "increasing forever", we can't have production costs going up forever and/or resources necessaries going up forever because none of them is available "to infinity".

Now don't take me wrong I understand there's a fuckload of work but "ray tracing is older than all your graphic cards" in fact I am sure it did exist before AMD/Nvidia even existed ( not so sure about AMD ).

This story of "grow grow grow, buy buy buy" need "a cap" or "take more time" I mean "ok it's cool but" .. but in a way "it's not the ray tracing making better games" as in "better to play" it's a novelty but if you tell me for example "this is going to make Fortnite look better" I could not give less than a fuck, "yes is again the same game with ray tracing added" .. "fantastic" ..

I still think "you should have the 5 .. 10 persons team max" "for the innovation" .. you say "gone are the day ONE person ?" .. yes ONE I can believe it .. but .. 6 ... 10 people allow me to say "I see it differently there", I still think "if you find the right ones they can kick arses".

Then there's the one that comes out with the independent music album that is "far out" or the indie movie that's better than the latest Hollywood blockbuster or the video game that surpasses Nintendo at its own game, I am not saying you have to create "a new video card from scratch" ( however given things I've seen and tools available today I would not be 100% sure that 'this is not possible' ) but I am saying "You have 10K people, you should be able to invest 5 .. 10 people to keep an eye on the new stuff even if they don't make any money".

If things are SO BAD that "you ONLY can follow the direct instant money", so big you are, then I say "you are doomed" jeez that says many things about "the big industry" especially if you even don't have the money to invest in maybe "someone else" ( which makes some sense because if you could you could do it yourself ) for "innovation".

I mean fuck me let's say you are among the 3 in the world "that makes graphics" and you can't do that .. fuck .. that's so bad.

This "system" has to change there can't be just "immediate money or nothing" and "costs that never ends".

And I'd bet you have some CEOs that spend 10K a week in some posh bottle of wine "because they can" ( and they get conned by that guy selling fake wines bottles ) ?

On the other side I can't even stand those people "I cannot work for less than $1M a year" .. yeah .. sure .. that is wrong as well.

Mah ...

"If you want innovation, send donations here please, don't waste your money in 'the big industry' "

[Dio]

Now don't take me wrong I understand there's a fuckload of work but "ray tracing is older than all your graphic cards" in fact I am sure it did exist before AMD/Nvidia even existed ( not so sure about AMD ).

AMD is a vastly older company than nvidia. We just had our 50th anniversary last year. ATI was also around several years before nv were (and owned a significant part of the VGA patents).

But nobody has done it realtime before, and as anyone who's studied computing well knows, realtime techniques are not usually that related to 'slow' ways of doing things.

I mean fuck me let's say you are among the 3 in the world "that makes graphics" and you can't do that .. fuck .. that's so bad.

I mean, think about the driver codebase alone. The driver package is 300MB now - OK, 200+MB of that is different copies of LLVM but there's still tens of megabytes of actual object code there. It's tens of millions of lines of code and tens of thousands of hours to write it all.

[gilesgoat]

*SNIP*

I mean fuck me let's say you are among the 3 in the world "that makes graphics" and you can't do that .. fuck .. that's so bad.

I mean, think about the driver codebase alone. The driver package is 300MB now - OK, 200+MB of that is different copies of LLVM but there's still tens of megabytes of actual object code there. It's tens of millions of lines of code and tens of thousands of hours to write it all.

I meant again having "a small division of people" to keep an eye on all new techs and "have stuff prepared" in case one day you want to make a "step" into that.

I don't know if, suppose at some point VR "gets profiteable" you can say like "ok now let's do it" and within 6 months come out with "something incredible for VR new AMD Radeon something with special VR support" ?

But instead of "doing that in a rush", if you'd have "a group of people" always working on those things "to keep an eye on all that's going around and maybe putting up some crazy plans" what I am saying is "having something that does not necessarely works on immediate money" but "keeps an eye on all potential money and/or interesting tech".

Or are you going "the route" of "let's wait maybe there's some X startup out there doing that and if they do really good we buy them ?".

As in "let someone's else take the plunge if it's worth it we'll get it ?".

I still think given the situation ( Covid ) and such VR will become "more and more of a something" .. a bit "for necessity" ..