Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
I presume the guys who absolutely attacked mercilessly AMD/Mantle for daring to talk about Mantle all the way back in was it November, about 2 mins worth in October and brought it to the first game in Jan/Feb, second game in March....... who complained about lateness, talking about things they "can't deliver", taking ages, massive delays, the ridiculousness of "talking" up an API that isn't available will all be harsher on DX12/MS because they first spoke about DX12 about a month ago and have no plans to get games delivered on it before the end of next year?
I won't complain about MS doing this because it's standard. Industry changes take time, new API's/software/hardware changes take time and I have precisely no problem.
DX12 WILL have faster adoption than any previous DX api purely based on the hardware compatibility situation.
For say DX10 they would be working on the API for 2-3 years before released, but there was only hardware available basically around the time it launched and the install base for DX10 hardware was very small comparatively.
Interesting but as yet I've seen very little about how low level and key features that devs actually want, their own memory management. I suspect they've conceded control of such stuff. but we'll see. Their graphs are interesting if accurate, finally going properly multi threaded and well, it's important to look at the numbers there. It's not always about how much CPU you have, you could only be at 50% cpu load but that one thread taking 75% of the time means the gpu is waiting 5ms extra before getting data. even on a not fully loaded cpu you can still gain ms which is almost certainly why both the minimums with Mantle have been phenomenal and why the games feel smoother, and why console games have often felt smooth despite lower frame rate. Consistency in frame rate is almost as important as the frame rate itself.
They pointed out that it was "lower than ever before" rather than just calling it low level, and considering this is the highest level of DX/openGL(with extensions)/Mantle by quite some distance it doesn't fill me with massive confidence, nor does their 50% reduction in CPU performance. IT's an improvement and I'd never refuse it, but Mantle does seem to offer a significantly higher reduction in cpu ultilisation. It shouldn't be surprising but you get the impression it will be a kind of midway point between DX11/Mantle.
But it's a good step and one MS hasn't taken in years, hasn't been forced to take in years and maybe bodes well for DX13. DX12 gets broad hardware support and gets the industry used to coding at a lower level, then deliver DX13 which is a further step forward, maybe even closer to Mantle but like a more normal DX is limited to newer hardware and so can go much lower without needing such wide support.
I won't complain about MS doing this because it's standard. Industry changes take time, new API's/software/hardware changes take time and I have precisely no problem.
DX12 WILL have faster adoption than any previous DX api purely based on the hardware compatibility situation.
For say DX10 they would be working on the API for 2-3 years before released, but there was only hardware available basically around the time it launched and the install base for DX10 hardware was very small comparatively.
Interesting but as yet I've seen very little about how low level and key features that devs actually want, their own memory management. I suspect they've conceded control of such stuff. but we'll see. Their graphs are interesting if accurate, finally going properly multi threaded and well, it's important to look at the numbers there. It's not always about how much CPU you have, you could only be at 50% cpu load but that one thread taking 75% of the time means the gpu is waiting 5ms extra before getting data. even on a not fully loaded cpu you can still gain ms which is almost certainly why both the minimums with Mantle have been phenomenal and why the games feel smoother, and why console games have often felt smooth despite lower frame rate. Consistency in frame rate is almost as important as the frame rate itself.
They pointed out that it was "lower than ever before" rather than just calling it low level, and considering this is the highest level of DX/openGL(with extensions)/Mantle by quite some distance it doesn't fill me with massive confidence, nor does their 50% reduction in CPU performance. IT's an improvement and I'd never refuse it, but Mantle does seem to offer a significantly higher reduction in cpu ultilisation. It shouldn't be surprising but you get the impression it will be a kind of midway point between DX11/Mantle.
But it's a good step and one MS hasn't taken in years, hasn't been forced to take in years and maybe bodes well for DX13. DX12 gets broad hardware support and gets the industry used to coding at a lower level, then deliver DX13 which is a further step forward, maybe even closer to Mantle but like a more normal DX is limited to newer hardware and so can go much lower without needing such wide support.
), so a lot of the stuff is doing what Mantle is doing, but it also stated that this is targeting the best dev's, performance when you need it. I wonder if they have some scaling options, like smaller dev doesn't want to write memory management, you can use DX memory management but you get an increase in overhead? Is DX12 only aimed at game dev's that want to use it like Mantle, for pure performance and complete control but it's not aimed at everyone.
