AMD's Tonga - R9 285 (Specs) and R9 285X (Partial Specs)

[krumme]

This is not reflected by AMD's statements from the article: In explaining the situation, AMD tells us that this is an application level issue due to these games not being familiar with Tonga, and that this can be fixed through further patches.

Nowhere does AMD state it will be resolved through driver updates. And again, doing so would completely defeat the purpose of Mantle given the driver would be doing per-game optimizations, as it happens now on DirectX.

Not to mention that in the earlier days, one of the things we were told about Mantle being superior to DirectX was that we wouldn't have to wait for driver optimizations to get good performance.

But according to you we have to wait for new drivers, and according to AMD we need to wait for application patches - all the while being locked to certain GCN hardware. I say certain because even now some versions don't work properly.

So exactly how is this "superior" to DirectX again?

Again, this is not reflected by reality: If you have ever used some of the lower end GCN products (e.g. Cape Verde) then you’ve seen first-hand that these games already are hit & miss depending on the GPU in use, so Tonga is an extension to that limitation.

Again, Mantle can't even do what it's supposed to on existing hardware. Basically only the flagship 290X has a chance, and even then the BF4 path still has issues such as memory leaks.

Eventually there's a point where saying "it's beta" no longer cuts it as an excuse. The reality is that the paradigm around the design of this API is extremely brittle from a software engineering standpoint.

Whats all this raging about?

As AMD says this is solved through game patches. As you say yourself thats the purpose of Mantle.
Its a new arch.
Ofcource it needs new driver, no matter how thin it is, and how minor the changes to arch is.
And then in game adjustments needs to take place.
Thats how Mantle works. And how DX12 will work.
Its bound to take time after the cards is released.

Thats probably the reason why we see this 285 now paving the way for the new arch, as 7790 did. No surprises. Everything as planned and expected. And thats the pattern we will see for all cards in the future when dx12 comes online.

 

[ShintaiDK]

The issue is game patches. There is zero guarantee.

Also its against all the false marketing we heard about Mantle.

Mantle would fix everything and anything. But when reality hit, it was a whole other matter. And now we dont just need driver updates, but also game updates.

 

[MathMan]

Can we just agree that 285 not performing well on Mantle is a big fail for AMD? If it's a problem with the application, then it raises some major concerns about game devs having to target individual chips. If it's a driver issue, then AMD was insane to not make Mantle a priority for the launch of this GPU, because it directly undercuts their Mantle story.
The former would be a fundamental problem. The latter just AMD screwing up a launch, the way we've come to expect them to do.

 

[wand3r3r]

The reality is that the paradigm around the design of this API is extremely brittle from a software engineering standpoint.

Which paradigm and what API design are you talking about? I assume you have valid sources or API documentation to make this vague rant about the alleged poor engineering standpoint?

 

[PPB]

What viral marketers cant grasp here is that no one said Mantle wont need optimizations for each GCN itineration, what it was said is that Mantle will always work as long the GPU is GCN. And optimizations with Mantle and GCN are really easy because:

- The dev has a lot of control because the API is transparent and can predict performance much better than the Black Box called DX.
- The dev knows GCN like the palm of their hand. Cant say the same for NV's uarchs tho, like some dev said:

NVIDIA kept moving away from its devs; while AMD kept moving closer. I know so much about the GCN architecture that I can even predict the next AMD-specific GL extensions (like GL_AMD_vertex_shader_viewport_index) or lift of useless restrictions (like the 64kb UBO limit). Why? Because AMD keeps being open, with lots of documentation, GCN performance tips, and more. Not to mention their full spec docs are open for Open Source driver implementations.

What do I know about the Kepler architecture? Nothing. Zero. Zip. Nada. It’s like it’s all a secret. I better not find out how it works.

So this begs the question, how much of a nightmare will be optimizing for DirectX12 with NVidia not telling ANYTHING about their Uarch to the game developers, and DirectX12, being the carbon copy of Mantle it is, needing optimizations for specific uarchs/cards?

 

[AtenRa]

This is not reflected by AMD's statements from the article: In explaining the situation, AMD tells us that this is an application level issue due to these games not being familiar with Tonga, and that this can be fixed through further patches.

What article is that ??? link ??

Again, this is not reflected by reality: If you have ever used some of the lower end GCN products (e.g. Cape Verde) then you’ve seen first-hand that these games already are hit & miss depending on the GPU in use, so Tonga is an extension to that limitation.

I have used Mantle in BF4 and Thief with GCN Tahiti(HD7950), GCN 1.1 Banaire(HD7790) and with Kaveri APUs. They all work fine.

Again, Mantle can't even do what it's supposed to on existing hardware. Basically only the flagship 290X has a chance, and even then the BF4 path still has issues such as memory leaks.

See above

 

[ShintaiDK]

What article is that ??? link ??

You must have missed Anandtechs article:
http://www.anandtech.com/show/8460/amd-radeon-r9-285-review/6

 

[AtenRa]

You must have missed Anandtechs article:
http://www.anandtech.com/show/8460/amd-radeon-r9-285-review/6

ahh thx, couldn't know what article exactly he was talking about without a link.

People forget that both BF4 and Thief required a Mantle Patch AND GPU driver. Also, i would like Ryan or anyone else, to see if 285 with Mantle has better gameplay (lower frametimes) than with DX11 even with those Beta drivers.

 

[wand3r3r]

Ryan did bring up some good points.