AMD’s DirectX 12 Advantage Explained – GCN Architecture More Friendly To Parallelism Than Maxwell

Harlequin · 30 Aug 2015 at 19:09

layte said:
So has Mahigan. At least we know where one of them is coming from. The other has appeared out of nowhere posting hugely pro AMD pieces (with a nice sprinking of incorrect NV info thrown in to stoke the flames) and appears to have direct access to Oxide dev's. It's all very fishy.

as opposed to razor1 who is anti amd , with a lot of incorrect information and was proven very wrong even about NVidia topics.

they both have the good grace now to accept they both have learnt things - except ofc where Mahigan hasn't posted on a forum , so razor1 is twisiting his owns words to make himself look like the good guy; they both have made errors and both learnt thing.

so when I get tim sweeny to reply to me personally it means I have direct access to devs? no - its down to how a request is made , the why , and relevant links showing the extent of the discussion.

razor1 himself has shown his knowledge on cuda isn't as good as he thought it was.

layte · 30 Aug 2015 at 19:21

It's pretty unusual (read non existent) to get that level of technical information regarding an unreleased engine given out by a developer to a random person on the internet.

Harlequin · 30 Aug 2015 at 19:23

layte said:
It's pretty unusual (read non existent) to get that level of technical information regarding an unreleased engine given out by a developer to a random person on the internet.

you haven't read much from tim sweeny then have you.

again , devs are proud of their offerings , and when approached in the best way they will openly talk about it.

edit:

note - they are correct - other than day 1 Nv PR saying the rubbish that MSAA is broken (it isn't) , they haven't said anything else , because Oxide are in fact correct.

layte · 30 Aug 2015 at 19:28

Nope, no devs give out that kind of info on unreleased engines to people who randomly email them. Now, giving that kind of information to a partner who is sponsoring said game. Well that I wonder.

Something about this whole situation stinks. I really cannot wait to see how other DX12 titles perform, Fable being DX12 from the ground up should give us a better picture.

Harlequin · 30 Aug 2015 at 19:31

layte said:
Nope, no devs give out that kind of info on unreleased engines to people who randomly email them. Now, giving that kind of information to a partner who is sponsoring said game. Well that I wonder.

its not unreleased - anyone can go and buy it right now. why are you intentionally lying?

http://www.ashesofthesingularity.com/store

notice oxide have said NVidia have had just as much access, in house, as amd have . Again , why are you making stuff up?

layte · 30 Aug 2015 at 19:45

It's an early release alpha (their own words). The engine was created as a Mantle showpiece and has been touted by AMD's PR (including our very own now outed embedded AMD PR) as something that could only be managed in Mantle, I doubt Nvidia were having much involvement then. Nvidia have had to try and work around that. It would seem that they have only had since recently to do so.

Harlequin · 30 Aug 2015 at 20:08

you are aware that mantle is vulkan , which has been built upon and is now the defacto opengl standard??

so saying mantle as if its been found under your shoe is unbelievable rubbish - I'm sure carmack would happily and vocally correct you;

and carry on about the game engine

http://www.anandtech.com/show/8962/the-directx-12-performance-preview-amd-nvidia-star-swarm/7

the only difference between those tests and now is the introduction of async compute - want to carry on looking stupid?

layte · 30 Aug 2015 at 20:14

None of which has anything to do with the posts in question. AOTS was built as a Mantle showpiece and NV have had to try and work around that.

It is strange that the SW DX12 benchmark was never seen again when it showed an advantage to Nvidia, almost as if the IHV sponsor was non too pleased.

Stay classy with the insults though, that's always a sure fire way to make people listen to what you are saying. Let me know if you want to have an adult level discussion.. Actually don't bother, it's obvious that you have swallowed this PR hook line and sinker, I wonder what the excuses will be when the next DX12 title we have doesnt have such a positive AMD bias.

Harlequin · 30 Aug 2015 at 20:16

I`ll pass you the JCB to carry on digging that hole you are in.

so you ignore anandtech when it pleases you ,ignore the StarSwarm demo using the very same engine as ashes when it proves you wrong? really?

layte · 30 Aug 2015 at 20:29

Exactly the same engine? I'd wager their are some differences between the two. It's a real shame we never got a further look once it got buried for not towing the sponsors line (or perhaps somebody forgot to tick the extra AS box).

Harlequin · 30 Aug 2015 at 20:40

http://www.oxidegames.com/star-swarm/

and

http://store.steampowered.com/app/267130

apparently being open and free on steam equates to being `buried` to you...

somewhat `interesting` theory

layte · 30 Aug 2015 at 20:45

You should do your research before being so aggressive and rude fella, you shouldn't take it so personally.

None of those have the buried DX12 engine. A real shame as I guess a large number of people would have wanted to run it for themselves.

Harlequin · 30 Aug 2015 at 21:12

layte said:
You should do your research before being so aggressive and rude fella, you shouldn't take it so personally.

None of those have the buried DX12 engine. A real shame as I guess a large number of people would have wanted to run it for themselves.

erm , nitrous engine runs starswarm - I think you actually need to understand ` fella` and what you said and the actual programs , you asked for , which I gave links to.

I`ll get you the JCB.

r7slayer · 30 Aug 2015 at 21:27

layte said:
You should do your research before being so aggressive and rude fella, you shouldn't take it so personally.

None of those have the buried DX12 engine. A real shame as I guess a large number of people would have wanted to run it for themselves.

I thought DX12 was a API not engine? It's not hard to adapt an engine to DX12. Many devs have demonstrated a couple of devs can do it in 2 weeks to a month. Just porting a game to another API.

Stanners · 30 Aug 2015 at 21:35

Harlequin said:
erm , nitrous engine runs starswarm - I think you actually need to understand ` fella` and what you said and the actual programs , you asked for , which I gave links to.

I`ll get you the JCB.

Why do you get so angry ? Why not put your point across and engage in a conversation or debate?

layte · 30 Aug 2015 at 21:39

Harlequin said:
erm , nitrous engine runs starswarm - I think you actually need to understand ` fella` and what you said and the actual programs , you asked for , which I gave links to.

I`ll get you the JCB.

I think you will find that the DX12 version (which is what we are discussing) was never released, or spoken about since.

You gave me links to something different, then tried to play semantecs when called on it.

Harlequin · 30 Aug 2015 at 21:44

layte said:
I think you will find that the DX12 version (which is what we are discussing) was never released, or spoken about since.

You gave me links to something different, then tried to play semantecs when called on it.

Porting a game engine when the underlaying API is similar to another API is relatively easy - given that Mantle formed the basis of Vulkan , and the history of OGL follows DX quite closely - and shader calls are quite similar in functionality ; its not a giant leap of faith to understand that porting Mantle to DX12 wouldn't take long at all.

anyway - it shows you didn't even click the anandtech link:

again

http://www.anandtech.com/show/8962/the-directx-12-performance-preview-amd-nvidia-star-swarm/7

they test starswarm on D3D12

andybird123 · 30 Aug 2015 at 22:01

Harlequin said:
Porting a game engine when the underlaying API is similar to another API is relatively easy - given that Mantle formed the basis of Vulkan , and the history of OGL follows DX quite closely - and shader calls are quite similar in functionality ; its not a giant leap of faith to understand that porting Mantle to DX12 wouldn't take long at all.

anyway - it shows you didn't even click the anandtech link:

again

http://www.anandtech.com/show/8962/the-directx-12-performance-preview-amd-nvidia-star-swarm/7

they test starswarm on D3D12

Id forgotten all about that, so isn't it interesting that the starswarm demo on dx12 shows nvidia hardware getting a big boost from DX12...

Harlequin · 30 Aug 2015 at 22:12

andybird123 said:
Id forgotten all about that, so isn't it interesting that the starswarm demo on dx12 shows nvidia hardware getting a big boost from DX12...

and the change to Ashes - async compute , since , even on my i7 with 6000 units my gtx 980 is getting canned under DX12 (14fps at times) async compute allows more units which any aspiring RTS wants.

whats funny - my 980 gets an average of 70fps under the default `extyreme` option - but D3D11 deferred context is OFF and small batch optimisations (mantle) is enabled , meaning its running under D3D12.

why not *everyone8 download starswarm from steam and run it undere DX12 or mantle ; drivers have come along way since jan 2014 - and windows 10 will help both along!

andybird123 · 30 Aug 2015 at 22:21

But the dev already said that they disabled async on ashes for NVIDIA hardware, so what else is causing the drop in performance from 11 to 12, when starswarm, using the same engine shows a boost from 11 to 12

I think posting up the starswarm on dx12 anand article is a massive own goal as far as trying to prove ashes isn't being deliberately optimised for AMD hardware, as starswarm shows NVIDIA can and should get a performance improvement going from dx11 to dx12

Whatever it is about ashes that causes a drop in perf from 11 to 12 should be optional for NVIDIA hardware, which is probably what nvidia asked for, options, not disabled entirely or removed, but the option to turn it down, like turning down tesselation in other games