DirectX 12 news unfurls a bit...
It seems as though Microsoft unveiled some more news about DirectX12 the other day. Which means I can clean up some of my assumptions from my last post. Normally, I would just edit the last post, but there is actually quite a bit here to dissect that I felt it deserved a new post.
What I have to say on the topic is this; the news is odd, the timing odder and what was not mentioned oddest.
As I said in my last post, the expectations that it would make a significant boost, in my opinion, would only be justified by support for hardware level optimizations. But, it doesn't sound like any such optimizations exist. DX12 will run on most recent gen video cards with no mention at all of new hardware instruction sets.
And that is what makes the announcement odd. The announcement sounds like it could mean one of two things. Either, code level optimizations or a total rewrite of the underlying code. It seems odd that Microsoft would rewrite such a huge, critical framework. But it is probably even more odd to assume that they would simply be able to tweak the existing code and get the results they are advertising. And if they could just tweak the code, it makes you feel like "wow, how did they miss THAT much room for improvement all of these years". So, the assumption that they are fundamentally rewriting DirectX seems much more logical.
If it really is a complete rewrite, things get weirder. The odder part of the announcement was availability in holiday season 2015. That's over a year and a half away. That is one long ass chunk of time off in the future. That means that they are nowhere near complete on this. Which begs the question... if they are that far out, how accurate are todays performance numbers and in which direction will they change as we approach that timeframe.
The oddest bit however was what wasn't mentioned. Tiling. This was, for me, a potentially huge area for performance gains. I thought it was supposed to be part of DX12. And with the potential it has, it is remarkable to hear it not mentioned at all if it is actually going to be a part of this. Is it scrapped? Is it something unrelated to DX12? Or was it simply ignored?
Circling back to the release date. The other open question is what they mean by availability in holiday 2015. Was that just for the PC? Or all platforms? Lets face it, Xbox One could use it now or as soon as possible. If it helps boost performance up to or above what the PS4 is capable of, pushing it out for over a year and half just to have a synchronized release across all platforms means more bad PR over console performance which may affect not only sales over that period but ongoing sales as well. And, DX12 is fundamentally a framework for games developer. If you can focus all effort on that one platform which is also fundamentally about gaming and get it out faster there, it makes sense. In fact, I would target Xbox One, Windows Phone and then Windows. Yes, Windows is probably more of a traditional DX game platform than WP8, but it is phones that would benefit the most from such improvements.
And finally, with it still being over a year out before this thing lands, it is just super weird that Microsoft would talk about it all. I swear that company would love to just release a new product into the market without telling anyone and then act surprised when no one buys it. I would attribute this new attitude to having a new CEO, but I don't think Satya has been at the helm long enough for his decisions on these sorts of things to have taken effect.
So, with all that being said... I'm skeptical. I read the Microsoft PR on it. And it sounds like they are making some smart improvements to the way that DirectX handles certain things. It's just a little hard to believe that software only level improvements will yield such results across the board. Certainly, it is possible. If the approach they used historically was flawed enough for newer GPU's then, yeah software level changes can also yield huge gains. Just very rare that it is actually the case.
What I have to say on the topic is this; the news is odd, the timing odder and what was not mentioned oddest.
As I said in my last post, the expectations that it would make a significant boost, in my opinion, would only be justified by support for hardware level optimizations. But, it doesn't sound like any such optimizations exist. DX12 will run on most recent gen video cards with no mention at all of new hardware instruction sets.
And that is what makes the announcement odd. The announcement sounds like it could mean one of two things. Either, code level optimizations or a total rewrite of the underlying code. It seems odd that Microsoft would rewrite such a huge, critical framework. But it is probably even more odd to assume that they would simply be able to tweak the existing code and get the results they are advertising. And if they could just tweak the code, it makes you feel like "wow, how did they miss THAT much room for improvement all of these years". So, the assumption that they are fundamentally rewriting DirectX seems much more logical.
If it really is a complete rewrite, things get weirder. The odder part of the announcement was availability in holiday season 2015. That's over a year and a half away. That is one long ass chunk of time off in the future. That means that they are nowhere near complete on this. Which begs the question... if they are that far out, how accurate are todays performance numbers and in which direction will they change as we approach that timeframe.
The oddest bit however was what wasn't mentioned. Tiling. This was, for me, a potentially huge area for performance gains. I thought it was supposed to be part of DX12. And with the potential it has, it is remarkable to hear it not mentioned at all if it is actually going to be a part of this. Is it scrapped? Is it something unrelated to DX12? Or was it simply ignored?
Circling back to the release date. The other open question is what they mean by availability in holiday 2015. Was that just for the PC? Or all platforms? Lets face it, Xbox One could use it now or as soon as possible. If it helps boost performance up to or above what the PS4 is capable of, pushing it out for over a year and half just to have a synchronized release across all platforms means more bad PR over console performance which may affect not only sales over that period but ongoing sales as well. And, DX12 is fundamentally a framework for games developer. If you can focus all effort on that one platform which is also fundamentally about gaming and get it out faster there, it makes sense. In fact, I would target Xbox One, Windows Phone and then Windows. Yes, Windows is probably more of a traditional DX game platform than WP8, but it is phones that would benefit the most from such improvements.
And finally, with it still being over a year out before this thing lands, it is just super weird that Microsoft would talk about it all. I swear that company would love to just release a new product into the market without telling anyone and then act surprised when no one buys it. I would attribute this new attitude to having a new CEO, but I don't think Satya has been at the helm long enough for his decisions on these sorts of things to have taken effect.
So, with all that being said... I'm skeptical. I read the Microsoft PR on it. And it sounds like they are making some smart improvements to the way that DirectX handles certain things. It's just a little hard to believe that software only level improvements will yield such results across the board. Certainly, it is possible. If the approach they used historically was flawed enough for newer GPU's then, yeah software level changes can also yield huge gains. Just very rare that it is actually the case.
Comments
Post a Comment