They said they wrote their game engine specifically for DX12, to meet all the requirements/QA, even MS looked at their code and approved it. Playing by the book.
Whether you believe it or not, is your prerogative.
And this gives him access to secret nvidia insider info how?
(hardware plans for gpus)
If they wrote async especially for GCN (they are working with amd so why shouldn't they) and just run the exact same code for maxwell then of course it must be maxwell's fault if it runs slower right?
No way that you have to code differently for a different architecture right?
And GCN is undoubtedly a less efficient architecture than Maxwell. Fury X has more transistors than GTX 980 Ti, but is slower and uses more energy even with the benefit of liquid cooling. Same for the 390x and GTX 980..
Im sorry, but that is not correct. Fury X has 8.6 TFLOPs of compute power, whereas GTX 980 TI has only 6.1.
That is the reason why R9 290X is so close in DirectX12 performance to GTX 980 Ti. R9 290X has 5.6 TFLOPs of compute power.
Times have changed. Better get used to this situation, guys.
And GCN is undoubtedly a less efficient architecture than Maxwell. Fury X has more transistors than GTX 980 Ti, but is slower and uses more energy even with the benefit of liquid cooling. Same for the 390x and GTX 980..
They said they wrote their game engine specifically for DX12, to meet all the requirements/QA, even MS looked at their code and approved it. Playing by the book.
Whether you believe it or not, is your prerogative.
Theoretical flops numbers doesnt mean anything. And 1 flop isnt equal to 1 flop. Also if its unused, it means power consumption will go up with usage as well.
Just to nitpick... the only flop here is nvidia dx12 performance and support...
https://en.wikipedia.org/wiki/FLOPS
Now oxide programmers arent experienced...... Yeah right.
They said they wrote their game engine specifically for DX12, to meet all the requirements/QA, even MS looked at their code and approved it. Playing by the book.
Whether you believe it or not, is your prerogative.
Yes I know who Dan Baker is, which is why I am surprised that they managed to get such poor results with Maxwell's Asynchronous Compute function. It's as though they only optimized it for AMD, but not for NVidia..
GCN might REQUIRE dedicated ACEs to achieve maximum throughput as it has a much bigger problem with underutilization, but Maxwell likely does not..
Willing to bet on that with other DX12 games?
Crypto mining is a classic example of 1 flop isnt equal to 1 flop.
Also if flops was the only parameter. It would make the IGPs a lot faster than they are.
Sure Fiji has 4 times as many samplers as the PS4 does but games today are COMPUTE bound and not graphics bound ...
If nVidia supported it why go through all of the effort they did with the DX11 path (doubling performance the last month) and then supply code to disable AC on their hardware?
Once again, something doesn't run well on nVidia and it's the Devs fault.
And this gives him access to secret nvidia insider info how?
(hardware plans for gpus)
If they wrote async especially for GCN (they are working with amd so why shouldn't they) and just run the exact same code for maxwell then of course it must be maxwell's fault if it runs slower right?
No way that you have to code differently for a different architecture right?
Being fair to all the graphics vendors
Often we get asked about fairness, that is, usually if in regards to treating Nvidia and AMD equally? Are we working closer with one vendor then another? The answer is that we have an open access policy. Our goal is to make our game run as fast as possible on everyones machine, regardless of what hardware our players have.
To this end, we have made our source code available to Microsoft, Nvidia, AMD and Intel for over a year. We have received a huge amount of feedback. For example, when Nvidia noticed that a specific shader was taking a particularly long time on their hardware, they offered an optimized shader that made things faster which we integrated into our code.
We only have two requirements for implementing vendor optimizations: We require that it not be a loss for other hardware implementations, and we require that it doesnt move the engine architecture backward (that is, we are not jeopardizing the future for the present).
DX11 games were graphics bound.
DX12 games are more compute bound and will be.
What evidence do you have that games today are compute bound and not graphics bound? Take a popular game like BF4 for instance. BF4 made use of computer shaders for lighting and post processing, but Kepler was actually faster than Hawaii (even with the advantage of Mantle), and Kepler isn't exactly a compute monster..
Same thing for Bioshock Infinite, which also used DirectCompute for post processing.. Here's Dragon Age Inquisition, another FB3 engine game that makes use of DirectCompute. A GTX 680 with feeble compute capabilities can still match the first generation GCN with much stronger compute capabilities:
Directcompute isn't asynchronous compute though. They're two different things. The benefit of async comes from the GPU being able to do the compute work in parallel with the graphics work. Directcompute is a serial compute function (afaik).
Particularly Dan Baker. Even Johan sat back at the Mantle presentations that Dan was at and let him do 90% of the speaking.
This sounds like Project Cars :sneaky:The OP that I responded to said games today, and not games tomorrow. I don't know of a single game today that is actually compute bound..
And compute is still compute. A game can still be compute bound whether the compute shaders are executed in serial or in parallel.
That said, I think it would be bad game design if a game was compute bound rather than graphics bound..
That and at the recent SIGGRAPH 2015, at the next-gen API summit (covering Vulkan & DX12), there was only a select few presentations. Guess from who?
NV, AMD MS, Dan Baker (Oxide), Valve.
People need to lay off on the ad hominem whenever a messenger is delivering news that they don't like to hear.
Get to the bottom of it before slinging mud.
You know what's hilarious? Dan Baker was educating the conference go-ers ABOUT ASYNC COMPUTE and how BEST to employ it for Vulkan/DX12! Oh man that cracks me up.
So you are proving my point,they don't have the knowledge of the hardware and are forced to ask,well not forced but they are only doing what they want to do and tell PR that whoever wants something better come and code our game for yourselves because we sure as hell will not.NVIDIA DID gave them MAXWELL OPTIMIZED code they integrated in to the engine, AotS is not GCN optimized only. Now if Async Compute is working better with GCN, It is not the software fault that NVIDIA Maxwell doesnt work as good as AMD GCN but the Hardware, simple as that.
Directly from OXIDE and Dan Baker.
http://www.oxidegames.com/2015/08/16/the-birth-of-a-new-api/
What evidence do you have that games today are compute bound and not graphics bound? Take a popular game like BF4 for instance. BF4 made use of computer shaders for lighting and post processing, but Kepler was actually faster than Hawaii (even with the advantage of Mantle), and Kepler isn't exactly a compute monster..
Same thing for Bioshock Infinite, which also used DirectCompute for post processing.. Here's Dragon Age Inquisition, another FB3 engine game that makes use of DirectCompute. A GTX 680 with feeble compute capabilities can still match the first generation GCN with much stronger compute capabilities:
The whole thing is funny as heck.... Well, not really. So, from reading through some of this stuff, Oxide made an engine to take advantage of asynchronous compute. Supposedly, based on Nvidia's spec sheet, asynchronous compute is supported on Maxwell. Based on this assumption, Oxide went ahead and built a engine based on asynchronous compute. However, right after they finished the engine, Oxide found out that Maxwell ran worst with this DX12 feature than it should. People question the results, picked and pried and found out that Maxwell actually lacks the proper asynchronous compute feature set. Am I getting this correct?
If this is true, Nvidia could have avoided all of this if it was more forthright with asynchronous compute. I'm pretty sure Oxide wouldn't build an engine that is not optimized for the majoirty of the GPUs out there. It would be foolish of Oxide to build an engine that isn't optimized for the majority of its potential customers.
That means, I really think Nvidia was less truthful about asynchronous compute. It wasn't until after the fact that Oxide found out about Maxwell's shortcoming. Now, Oxide is left scrambling to rework the engine that is better optimized for Maxwell. Now, Nvidia is scrambling to help Oxide patch weak points in Maxwell's uarch. I think that's what we're at right now.