AMD GPU14 Tech Day Event - Sept 25'th

[Love_In_Rio]

It instantly reminded me of this interview from early 2011:

He sounds as almost all devs are willing to program near to assembly level to get most out of the hardware. And however devs hated Cell, the most hardcore (and powerful if well coded) gaming cpu ever. Well, I supposse people like Repi liked it.
As stated by Dave the advantage here is that a few game engines rule the world now and only applying Mandle in those would bring its advantage to the majority of future games. But wouldn´t this same reasoning be against Sony ditching Cell 2 in PS4 after almost all engines have greatly adapted already to it?.

 

[rapso]

He sounds as almost all devs are willing to program near to assembly level to get most out of the hardware. And however devs hated Cell, the most hardcore (and powerful if well coded) gaming cpu ever. Well, I supposse people like Repi liked it.
As stated by Dave the advantage here is that a few game engines rule the world now and only applying Mandle in those would bring its advantage to the majority of future games. But wouldn´t this same reasoning be against Sony ditching Cell 2 in PS4 after almost all engines have greatly adapted already to it?.

I would expect it gives you just the same possibilities you have on consoles. and as there are already two similar platforms with different APIs, those APIs are probably wrapped to not duplicate most of the 'real' code. So, to get Mantle support, you'd just need to add another API-wrapper, but as the hardware behavior underneath is similar (or the same), you don't really have much work.

 

[Bouncing Zabaglione Bros.]

They optimized power consumption a lot, claiming that the arch can go as low as 1 Watts.

I'm amazed that this was passed over in one brief sentence during the presentation. I've been very impressed with the power saving abilities of my Haswell system, and it seems fitting that new GPUs should also be able to match thing kind of minimum performance. Most systems spend most of their time doing very little, and there's no reason for them to draw significant power during those times. I want my next GPU to run cool, quiet and with little power use when it's not really doing anything.

 

[liolio]

Memory model is quite different, GPU and CPU are closer in this regard than they were to Cell. Cell memory model has been widely rejected, even by IBM. End of the game, by the way it has been discussed some many times by lots knowledgeable people /OT

 

[DavidGraham]

The fact remains , so long AMD supports both DX and Mantle fully , then there won't be any side effect from an extra FPS bump even if made through an extra API. And AMD will have to fully support DX if they plan to keep customers of their legacy non-GCN products happy (HD 5xxx, 6xxxx) .

I really don't see much problem there.

 

[rapso]

I really don't see much problem there.

I wonder if they'll still be able to release drivers that boost some performance a month after launch. if it's console alike, then there won't be any of those magical 30% boosts unless the dev releases a patch.

 

[DavidGraham]

On the DX's side there is always a room for improvements via driver updates considering that most games will still rely on DX . Even on the Mantle's side , new optimizations could be created and implemented via game updates .

 

[loic]

What about https://developer.nvidia.com/nvapi ?

 

[entity279]

Doesn't the single most important difference between glide and mantle amount to the fact that mantle appears to have been requested by dice (the industry)? Thus this has a better chance of being adopted by other IHVs/ISVs as graphics engine writers may request it (or something similar).

So that;s why I'm interested in the future.

But note that I know very little about Glide; didn't even had a computer back then.

 

[Dominik D]

I'm guessing the assumption is that this is primarily going to be for "full screen gaming", but if those limitations go too deep, I'm going to be much less interested. Robustness/stability is also a concern, but only time will tell on that one

That was my thought as well. In order to do composition you'd have to do pre-emption which basically goes against the idea of low level, high performance access.

if it's still going through dxgk, then Mantle seems like a big waste of time--the kernel thunking overhead and allocation lists will still kill perf. guessing they're working around it somehow, but carving out memory and suballocating was (last I checked, which was admittedly 15 months ago) not a thing you could do, nor was reserving some amount of memory at boot for non-WDDM purposes.

You can allocate as much as you won't but AFAICT you won't pass WHQL with this kind of driver (then again: WHQL would never touch Mantle APIs so if you defer allocation, you should be able to do whatever you want).

What I'm curious about is why did AMD only mention draw call throughput. My completely unbacked theory is that Mantle - as "something that covers or surrounds something else"* - is mostly about efficient draw call batching and deferred flushing to the hardware to optimize SF<->HW chatter. There will be other benefits (direct access to memory perhaps?) that will be huge for game developers but numbers shown during the presentation were the only generic-PR-worthy. 5% or 10% improvements in real life scenarios would still be massive to developers but hardly quotable for news sites with the exception of the most technically-oriented ones.

* m-w.com

 

[itsmydamnation]

There will be other benefits (direct access to memory perhaps?)

they said there is, it was under the GPU benefits section. draw calls where in the CPU benefits section.

 

[rapso]

Doesn't the single most important difference between glide and mantle amount to the fact that mantle appears to have been requested by dice (the industry)? Thus this has a better chance of being adopted by other IHVs/ISVs as graphics engine writers may request it (or something similar).

I think the industry rather wanted to see DirectX fixed.

 

[entity279]

Absolutely. My point is that this initiative is somehow shared between AMD and Dice.

 

[Arwin]

I think the industry rather wanted to see DirectX fixed.

You can't say that DICE hasn't tried that first ...

 

[rapso]

You can't say that DICE hasn't tried that first ...

of course I can say that, I can "say nobody has tried to fix directX". simply for the reason that nobody can fix directx except microsoft, and if microsoft had wanted to fix it, it would have been done.

so... we all wanted DirectX to be fixed.
we all can do nothing about it (not even AMD and NVidia could make MS fix it).

so, I'm very happy with Mantle from a software point of view.
I'm sad about the market split and the divergence that is going to happen at some point.

But if AMD convince NV+Intel to use Mantle and all are equally compatible.... naah, world can't be that awesome

 

[DieH@rd]

Digital Foundry has an article about this, good read http://www.eurogamer.net/articles/digitalfoundry-could-amd-mantle-revolutionise-pc-gaming

 

[Psycho]

slide deck: https://amd.app.box.com/GPU14publicpreso

 

[DieH@rd]

Anandtech article about Mantle, even better read http://www.anandtech.com/show/7371/understanding-amds-mantle-a-lowlevel-graphics-api-for-gcn

 

[Love_In_Rio]

Anandtech article about Mantle, even better read http://www.anandtech.com/show/7371/understanding-amds-mantle-a-lowlevel-graphics-api-for-gcn

I see no sense in MS developing a low level API that would make the ports of its games see 10 times better in PC. If XB1 ( and PS4 ) had any advantage over PC is the low level API. If Jaguar cores were shit before with this move they could become even worse compared to the I5s and I7s and if XB1 gpu and Liverpool already had few shaders with this move is like they suddenly had 1/3 less against desktop gpus. I am with Carmack in this case. This Mandle thing won´t be loved by MS and Sony. And Steam is too hovering with SteamOS. I can even see them joining and making developers that want to publish in their consoles to renounce to a PC Mandel and/or SteamOS version. My prediction: if there is one console future it will be MS+Sony.

 

[lanek]

I see no sense in MS developing a low level API that would make the ports of its games see 10 times better in PC. If XB1 ( and PS4 ) had any advantage over PC is the low level API. I am with Carmack in this case. This Mandle thing won´t be loved by MS and Sony. And Steam is too hovering with SteamOS. I can even see them joining and making developers that want to publish in their consoles to renounce to a PC Mandel and/or SteamOS version.

If Mantle is so much linked to Xbox Low level API.. I believe MS is not completely stranger to it... And as it is said so far, it will permit to port Xbox games a lot more easy and faster to PC ( and vice versa ).

I dont know how it will turn, but its not an hazard if AMD push this right now, when they have all console hardware in hand.