Gadgetory

okay I me not to my opinion on it is I in terms of action items as a question should be media have a response to mantle and I think that's unequivocally they should know it would be I think a horrible mistake if some reason some how the video got panicked by this it made some lower level nvidia API you know you already have very little level access through the GPU extensions that nvidia is always rolled out of anywhere else give you as much performance as you want to trade the inconvenience of doing a video proprietary stuff now mantle is if amy has talked many times and fast about close to the metal type of architectures and it only ever became interesting because of their dual console if it was just a way to do on the pc be a little bit more I a lower level for them I couldn't have cared less I the landscape does matter that they have both the major console wins with a similar architecture we can get articles in all of us will be making systems architectural changes that neighbor those I so I think it's not a stupid thing for am need to be doing there are some benefits that they can breathe from it I could have implications for steam box type things there I but they're probably if Microsoft and Sony both embraced it that would be very very powerful for AMD but it doesn't look like they're going to at least Microsoft or though words that they're having and show you though I I mean if I was still doing all of them you know all of the major technology coding or I for next-gen game stuff I probably would not be embracing mantle right now but it would be there would be days when it was extremely tempting you know but what I would dispassionately step back and look at it I probably wouldn't come down to the side of savings obviously others huh conclusions sure there's something ideas in Maryland we really like that idea of heaven low level low overhead access to the GPU I mean if you look back at what to vote Direct X and OpenGL there's a lot of overhead and there's a pis and the multiple layers of them and the fact that they date back to the old SGI nah no partner rendering which is very very different modern shader based programming whether we have now you know where you have potentially unified memory and a lot of processing power male 1 the GPU and lots of ways of sharing data beyond just calling the ton of API functions for every little thing you want to do so they're good ideas there i hope that really helps the OpenGL committee and Microsoft shape the future api's I if you ask me what I want to prefer to have a low-level API for accessing the GPU the answer is yes but five of them each for different hardware architectures and vendor and operating systems absolutely not that is the wrong direction for the industry to go and to have yet another AP on the pc and it's no different than the standard pc Direct X avion it's different than the OpenGL OpenGL ES there's estan mac and android and iOS and it's different than the playstation 3 playstation 4 is little more enduring api and it's different than what Microsoft provides an Xbox I don't think it's a good idea yeah so did one piece things about it is and one of the main reasons for doing is that it's not the replacement this is not designed that way at all and the idea is that we can solve some of the motor problems that we have been having all of us performance consoles a robust stable performance those are things we can experiment and work but not we do that bit easier here we do that with geo lols of course this is another avenue you can see it's also a really interesting of this essentially opening up something that we're quite familiar with already we've been spending bollocks in the past two years Oregon next year also the architecture so I think we intimately familiar with me to put those architectures and it's a good fit there just as still described and I see even though we're not done with our work with it and it's definitely not not done yet either to be to be shipped out I still see it as a success even right now I should just because of these conversations and the amount of things that actually the amount of developments in the mountain after that both enthusiasm and even the opposite the verses yes oh all of the type of them and it's been a bit stale and you could say in especially in the pc graphics mark soft a switch focus for quite some time over to you other things which they probably really didn't need to do but I think now going forward it's a really exciting opportunity both on the on the piece of space in the stuff we're experimenting with mental and we'll be learning on the consoles but also mobile as we've discussed before okay what should be the future graphics pipeline sir it's a lot of movement in all these things and this is just good to have lots of different avenues to experiment and actually hopefully deliver a concrete value in these things but literally the most John and attended video we do their own API and then into the do na VI and then pop them with you an API that's not teaching one be that there would be a bad future that will be very very a lot of work that you myself and from us that it does play an important part of albinism I hope mentor will be quite influential in many aspects also and what we do by naz is specifically with with daily is is just a start yeah I can i yeah you give you stay aware we just read in his bed for the marketing videos um what would you say is the big question 20 thing is a cost of like in man years or whatever a plan mental support and healthy again what do you think it's the automatic pc that could you try to quantify that it's in this 2015 we've been very busy just to making sure our game works overall and actually and shipping it out when we're almost done with it so it's too early to say we'll have a lot more information in November mid-november talk more about it and then will be interesting to see and I think it's important also to get one thing that people Oh general people that about developers can be difficult to understand that because of all nights this actually it's not like people okay we solve this one with go home and game well as well and ball next is like whack-a-mole uu 561 and then you thought that that was the thing he made us four times faster and then the next wall itself after of course the biggest issue is that really you have to architect your orange ski ovation to take yep no need I just all of a sudden really makes a dramatic difference on Vegas because if you build a good game in 88 guy even if you magically made all that i will read vanish it doesn't think that much of difference it's a possibility of do I mean there's stuff one GC ends that I you're excited about some of the asynchronous pipeline cues for different things it would be great to make direct that right direct control over but thats you got your design for that really don't design yeah that the new google be a gradual approach also of sure we're doing well before it's not like that before us is only dedicated for a specific API it's a rich game there you want to work with something and deliberate concrete benefits and then valuate and then go forward and see where we are in the future we'll design games based on many of the concepts that we're proving out not earlier now if we can do otherwise and i think that that's especially gets changed for the entire industry even though people sometimes do get stocked up on unlike the double roles including again oh the same discussion as I had before okay its mobile replace all the pc thats i'm not saying that will replace all of the Equality of that then complementary i'll actually when you use that is complementary learner in the individual industry overall you want that competitiveness not necessary if i have 25 different aprs but having movement and i think that's really really i think i appreciate about cape town actually I can answer a couple questions that they wouldn't have to pay us anything to use it I've been fishing these ideas for for many many years for every single vendor at and it's actually talking just about my own egoistic view of solving my own specific problems is that we see a lot of stalls we see a lot of performance performance gone missing in many areas I creation get that performance we want to learn how to program pain a machine on our lower level it hasn't actually really really good architecture for and for that and I think that's something that will both sort of rolled out going forward also okay so it's good it is an R&D platform for golfers experiment well you're talking about the desired have front of us practice or something like that Laura only more likely think it's also that if it pushes microsoft and opengl falls to improve their drivers by realizing that hey you can actually achieve a lot more response on pc with some more low-level techniques i think it would only be bad if the outcome of this is that not only have to do is find vendors specificity is adjudicated effort and also the division for for mental well it can play up anyways you don't really come to the sacred for short and the division for mental is that it well it can become a maybe I also got something we're completely open for something like diesel also it's just that that's not the right way to go out and initially there are no abuse already bits Cronus and with these P this is because try something I wouldn't be a little battling boy well actually I don't think I don't forget sure you can make an API that's completely down to the level sort of what you like in this week we work to do a superior than abstract level it was good in specific without architecture what do you say the way modern GPUs work is quite different from from how thieves or leave absolutely kill the sign was wrong from the beginning during extensions of missing viola and we can access many things but there are commonalities between all our matrix like my this works out there's an awesome thing that the videos morning on the video that is really the future for all peoples and if you ask the different vendors even people in the madness cracks on mobile space that do not have it yet that is the future well now I'm talking for all the love we are probably not have been able it makes sense a service your modern design important and you see some some convergence towards things that are general design and then some people are some designs we will come a little late for that but when it's a great fit for porky hardware and is great that for developers then at last we might point about a unified memory is absolutely here we really need to get getting down to do GPS are just peers on the same februari well as the CPU and we really need like a documented textured layouts that's another thing that everything is coming associés also even the GPU is a combination of tons of different processors a significant amount the process you know if it's an interviewer is named a GPU there's quite a lot of quite independent engines there and that's something that's hidden from us now did you do is something that you from the cpu prepare your commands you just give it to them it has a driver that tries to figure out which goes where that's just generally good designed to expose those things of living like you're trying to do something pinky like using the hardware video to compression stuff or your recoding there you have a mentor for a new API to drive cavalry but you know even things like computer technology computers well one of the best in this case is compute is for graphics which is kind of yeah feels a little bit but but it's not like you want to use opposed tightly coupled with your visual graphics and you want to use it completely separately that the CPU dress for for loneliness is done and you want to have background tasks that have well can have really high latency and run in the background and just eat up whatever vector units are a freedom continue that's also general we can design the position you space like getting up these issues where one GPU task scheduling and we have no concept prioritization between GPU schemes where you know we need to be able to specify type slices in prior work yeah you know it's doubtful you seem to be scheduled report by metal resource and it will likely take quite a lot of time until you find the convergence to a definite all there and but in the meantime it's good to have a couple of models posing ha grab from software sort of competing or hard being tested there and see what sort of unsolvable because I do seem to booking quite interesting convergence