AMD Driver Support State For Radeon HD 7000 Series, Trinity

[Kano]

What i am wondering about is that when hd 7000 cards will be out then amd will not ship any pci-e 3.0 chipset. That means you test it with snb-e boards now or what

[bridgman]

Is John Bridgman referring to GCN for the integrated GPU within trinity?

No, I'm referring to Trinity GPU and GCN as two separate things. Putting both in the same sentence made it kind of ambiguous, sorry about that.

Trinity GPU initial support is done, including the same acceleration APIs as currently shipping parts. It has only been "developer tested" but seems to work. The GCN shader core is substantially different (as you can see from the AFDS presentations) so there's a lot of work required to implement acceleration support.

[chithanh]

I don't see anything in his statement which would infer that Trinity is GCN or not.

[Alchemist07]

???

Could somone clarify...

In terms of what is working and what isn't, Trinity is basically working and the focus now is building up the GCN acceleration stack. Kernel driver for GCN has been working for a while although changes to memory management are continuing, and work is ramping up on shader compiler and 3D acceleration.

Is John Bridgman referring to GCN for the integrated GPU within trinity?

If so...isnt this GPU supposed to be VLIW4?

http://www.pcper.com/news/Graphics-Cards/AFDS11-Upcoming-Trinity-APU-will-use-VLIW4-Cayman-Architecture

[bridgman]

Its funny Bridgman stop writing instantly every time i make a good point.

It's too hard to type and facepalm at the same time

Seriously, if you're talking about the *last* GPU/APU in the GCN family I guess I sort of agree with your numbers, although you are ignoring the contributions of non-AMD developers which IMO invalidates the numbers completely.

I was talking about the *first* chip in the GCN family not the *last* chip and there the difference is smaller, although still significant.

RadeonHD and radeon always shared the kernel and 3D drivers, so I wouldn't take time off for that, but I agree that we don't need to go through the classic=>Gallium3D transition again so that does save some time.

This is all academic until I see triangles though...

http://www.phoronix.com/scan.php?page=news_item&px=NjY3OQ

[duby229]

I'm pretty sure xenos is something much closer to x800.... I could be wrong, but it definitely does not have a unified shader architecture. It definitely has pixel shaders, and I'm almost positive it has vertex shaders too.

EDIT: Well, I looked into it, and I guesss it does use a unified shader architecture. So I was wrong.

[bridgman]

Writing English is hard work for me. Write a valid conditionally sentence is even harder for me. I do not have the skill to write in your Public Relations fashion way. Sorry if i conditionally hurt you.

I try to write more "simply", ie if I disagree with something I say that and nothing more. Your English is probably as good as mine these days -- but if I tried to write as colourfully as you do I would probably have trouble too

For me as a consumer there is no difference between different parts of the radeon driver in mesa and kernel because all parts must be completed.

Agreed, but you were disagreeing with someone who said that work on the *mesa* driver might not have started.

And in fact compared to an hd2900 amd start 5 year earlier.

It's more like a couple of years, isn't it ? The HD2900 came out in mid-2007 and by mid-2008 we had started working on r6xx acceleration. I agree that it is a really important change though... starting before launch rather than 12-18 months after launch makes a *big* difference.

[bridgman]

read the word "if" i know you don't do this "stupid" claim. he claim that by trying to quote you and trying to interpret your words.

That was how I interpreted your post as well, but could you find another way to deliver that explanation ? Preferably one which does not involve even "conditionally" calling me a liar

As I said earlier nearly all of the work has been on the kernel drivers. This discussion is about the userspace Gallium3D drivers.

for me the kernel part is a part of this driver.

I don't know how to respond to this

[bridgman]

Also Bridgman is a liar if he claim they don't have already started to code the GCN driver.

???

I said what the plan is. I didn't say if execution of the plan had started yet or not. Please don't call me a liar for something I didn't say.

I know they do have 2 devs coding on GCN for months.

As I said earlier nearly all of the work has been on the kernel drivers. This discussion is about the userspace Gallium3D drivers.

[AnonymousCoward]

You are just a ignorance person.
I know

...

Also Bridgman is a liar

...

so please stop to be naive.