So if there is no proof of its greatness, and no proof of its awefulness, what is the point of this thread exactly? We've talked about what it could be to death, but have yet to see any game that I can play with the Mantle API. We have yet to see any (p)reviews on its performance, enhancements, advantages... from any person that is not directly involved with the project. That says a lot to me.
And to think I was debating on getting an aftermarket 290(X) because of Mantle, thankfully I didn't wait.
There is plenty of evidence and example in this thread if you actually cared to read it. You dont. Once in a while some dumps the usual noise and piss in a page or two and then they leave. You are free to leave if you dont have anything of value to add.
There is plenty of evidence and example in this thread if you actually cared to read it. You dont. Once in a while some dumps the usual noise and piss in a page or two and then they leave. You are free to leave if you dont have anything of value to add.
Time have changed the last ten years so that today the devs build their own game engine on top of an engine eg fb3. It means that it is the experts that code to the api. Mantle is build on that change. And at the same time builds on pc and console portability. At the same time gfx evolves and what was once good abstractions is obviously not good abstractions any more.
It makes perfect sense for a new low level api in that new context. What you decribe is what was optimal 10-15 years ago. The new experts today is not the ones making the api but the devs making the engines. It places control where the competence and creative and artistic development is.
I think your arguments perfectly illustrates why mantle is meeting such enthusiasm from devs.
DX may have ways to improve things, but Mantle is not the way, as it is too hardware specific.
DX was fine 5-10 years ago but it's no longer good enough for performance. I believe this will be seen in ~20% performance gain, simply by switching to Mantle and doing nothing else with it. That's why I expect 20% as a baseline improvement.
DX was fine 5-10 years ago but it's no longer good enough for performance. I believe this will be seen in ~20% performance gain, simply by switching to Mantle and doing nothing else with it. That's why I expect 20% as a baseline improvement.
20% from what? Better core scaling? Lower driver and drawcall overhead? 20% overall improvement sounds rather low.
DX10 improved the API execution overhead by 20% of DX9. Yet it took years before we saw a result.
But 20% now, thats quite a lowering from your previous predictions.
I meant 20% minimum improvement.
Yes Mikk, better core scaling, lower driver and drawcall overhead. That's 20% for that alone, simply by switching to Mantle from DX will give those gains in the majority of games, in my opinion.
DX 10 didn't remove the core issues which are awful core scaling and the serial nature of the API.
Mantle basically removes all 3 (including draw calls with the 2 above) of these serious issues by default. That has to tell in better fps across the board - yes even in games like BF4 where CPU usage is actually very good, it can still get even better.
I expect to see large gains simply because the CPU and API aren't bottlenecking by forcing one thread.
You do know it's only AMD having a singelthreaded DX11 driver right?
NVIDIA's DX11 is multihtreaded....thus you just owned yourself.
You do know it's only AMD having a singelthreaded DX11 driver right?
NVIDIA's DX11 is multihtreaded....thus you just owned yourself.
Yes I clearly just owned myself. AMD not having multithreaded drivers and then suddenly replacing that situation with a far superior multithreaded API is going to show what kind of improvement?
Do you even stop to think before posting Lonbjerg? Ever?
Unless you are a graphics programmer yourself, you can't just claim that Nvidia's DX11 multithreaded rendering is solving the draw calls problem. Neither can the opposite be proven.
The actual programmers however, do seem to think DX11 multithreaded rendering is working badly, and unless you're better than them, you have to take them by their word.
It's just not being used much, not even for Nvidia's TWIMTBP titles. And if it were that good, you can be sure Nvidia would push for it in the games that's part of the TWIMTBP program at least.
Mantle introduces extra costs both for the studios and for AMD, the studios wouldn't want to touch Mantle if all AMD had to do was to enable multithreaded rendering, or providing an OpenGL extension.
And even if AMD did cover all the costs for supporting Mantle, the studios would be negatively affected by splitting the workforce.
64-bit only requires a different compiler. It hardly changed anything other than give much larger address capabilities, and larger numbers. It was not radically different. Besides, most games are still 32-bit. That was AMD's genius with 64-bit, it worked seamlessly with 32-bit.AMD64 was hardware specific, but it seems to have caught on. This is hardware specific in the same manor that only AMD currently has it implemented. Nothing will prevent Intel and NVidia from implementing the instructions.
NVIDIA's DX11 driver nullifies his empty argument...all it needs to do.
Unless of course he is soley talking about AMD's limitied DX11 driver...but I kinda doubt that.
64-bit only requires a different compiler. It hardly changed anything other than give much larger address capabilities, and larger numbers. It was not radically different. Besides, most games are still 32-bit. That was AMD's genius with 64-bit, it worked seamlessly with 32-bit.
I meant 20% minimum improvement.
Yes Mikk, better core scaling, lower driver and drawcall overhead. That's 20% for that alone, simply by switching to Mantle from DX will give those gains in the majority of games, in my opinion.
Only requires a different compiler? How about OS, drivers and shift in market? It's not like 64 bit popped up one afternoon and became the norm. It may not be Mantle, but very little in the PC world has stood the test of time and it seems many forget that. DirectX is not hear to stay.
AMD/Dice claimed vastly better core scaling and not to mention the lower overhead, therefore 20% sounds low for the extra work and limited clientage. 20% gains should be possible also for DX11MT, probably more with a proper implementation. And this is not exclusive to one IHV, AMD could bring a new driver anytime. I would assume 50% higher fps for Mantle in purely CPU bound scenarios on a 6+ core CPU. On APUs with 4 threads it might be more limited because there is less room for better CPU/thread scaling.
If Dx is not here to stay, what is going to run the 60+% of the market that runs on intel igps and the probably 50% or more of the remaining part of the market that runs on nVidia hardware? Dx has disadvantages of overhead, but advantages of working with a variety of platforms, unlike mantle.
Only requires a different compiler? How about OS, drivers and shift in market? It's not like 64 bit popped up one afternoon and became the norm. It may not be Mantle, but very little in the PC world has stood the test of time and it seems many forget that. DirectX is not hear to stay.