HowToFixComputers.com

R420 · Guest

disclaimer: how does this post have anything to do with Xbox and
Nintendo groups? simple, ATI is designing the graphic cores for both
next generation consoles. Dave Orton is ATI's commander in chief. what
he does, and has been doing with ATI is critical for those two
upcoming consoles 'nuff said on that.

the Dave Orton ATI interview from Beyond3D

quote:
-------------------------------------------------------------------------------

Prior to the launch of Radeon X800 ATI held an "ATI Technology Days"
seminar for press over in Toronto. Whilst there we had the opportunity
to talk with the man in charge of steering ATI's operations at the
moment. We had a short time to sit down with Dave Orton and quiz him
about the directions ATI is taking and to look at their engineering
operations for the PC graphics side of ATI's business.

-------------------------------------------------------------------------------

Hi Dave. I guess the first thing is congratulations are in order (it
was recently announced that as of June '04 Dave will relinquish his
role of COO and move to the CEO position within ATI, with KY Ho, the
current CEO, moving on to Chairman of the board).

Oh, I guess – some people think condolences are in order!

No, what's interesting is that we've operated the company this way for
some time anyway, so it was really a case of aligning the titles with
the jobs and there are no organisational structural changes in the
company. It was a nice gesture by KY and the board, so I appreciate
it.

Now that you're fully heading up the company are there going to be any
significant focus changes for ATI or is it still a case of "stay the
course" for the moment?

No, its stay the course as it is.

I think what KY found was that he wasn't really doing anything
operationally within the company and he's been spending a lot of time
talking to partners and trying to develop a lot of our strategic
relationships as we look our handheld business and DTV in Korea and
Japan, and then continuing to work the relationships with the
foundries, and TSMC in particular, our key partner there. This is more
representative of the role of Chairman anyhow and he doesn't really
deal with the financial community at all, which has really been one of
the roles of the CEO with the CFO, so we've just realigned internally.

Advertisement

One of the things we're here at the ATI Technology Seminar event for
is to learn about your latest high-end desktop part (Radeon X800), but
this part of ATI's business appears to be becoming an increasingly
smaller string to ATI's bow, isn't it?

Well, it is and it isn't.

To date, still about 70% of our revenue is discrete PC standalone
business, and 70% is a big number! And if you look at NVIDIA's
revenues its even higher as they don't have the notebook business that
we do to go along with that – our notebook business is running about
35% of our total business and 40% of our PC business every quarter,
and a third of that is integrated. So the discrete business is still
extremely important, but you're right, when you look at the horizon
there's a lot of other business out there.

I've been very frustrated with the HDTV business for a couple of years
here, when you look from the standpoint of our investment verses the
market. When you look at where the market is today and you then think
about winding the clock on two years from today where every TV in
North America is going to ship HD ready and today there is only 3
Million sets, which could grow to over 30 Million sets in this period
representing a 10X growth in two years. You have to think that ATI's
potential in that market is fantastic. In the short-term the wins
we've had recently with handhelds and the growth of that platform is
great.

So we're excited about these new markets but I think we're set up well
as a company to not defocus from one and defragment and destroy the
other.[quote]

[quote]
Well, that leads into another question because the biggest growth area
of the company appears to be in the handheld segment. It appears that
you've had an explosion of employees recently as it seemed that you
stood at 1900 employees for a long time and now its standing at 2500.

We're currently at 2600 employees, although I think what happened is
we weren't being diligent about keeping that 1900 number in the
external communications current! 100 of that increase was converting
our rep firm in Taiwan to a direct team.

....but the growth does appear to be in the other areas rather than the
PC desktop part of the business...

We've grown some there, but you're right, we're at 300+ people in the
consumer group and its growing much faster than the rest of the
company and we expect it to.

At the moment the business model operates along the lines of taking
the high end and trickling the technology down to the low end, even as
far down as the handhelds. Is there ever a point where you look at the
amount of R&D invested in the high end products – especially when
you're now looking at 160 Million transistor parts from yourselves and
222 Million from NVIDIA, and you've already previously acknowledged
that your margins will drop from the cost of these chips – and
conclude that this is getting a little ludicrous at the high end?

No.

In everyone of our segments, including handheld and DTV, we're coming
out at the high end because what we find is that the high end drives
the innovative thinking and it drives the customers desire from a
branding standpoint as they like to have the halo around their product
line – a Sony or a Mitsubishi or a Qualcomm are going to want that.
So, you have to drive at the high end.

I think you've got a valid point though, because there's the high end
and then there's the lunatic fringe and you cross a point there of
going to high of the high end. What we've looked at from a technical
standpoint, in order to achieve that ultra high end is whether its
better to take a performance part and figure out a way to scale it up
through two parts, as an example, or whether its better to design at
the high end and then figure out how to slice it down to meet the
lower end markets. We do keep challenging ourselves at an
architectural level as to whether you end up creating inefficiencies
in the low end because you are designing so high up which end up
making you less competitive at the low end. And what we are leaning is
that the answer is yes, there are features that we're designing for
the high end and as you move down its very hard to pull them out so
we're looking at features like Hi-Z [Hierarchical-Z Buffer], some of
the floating point precision questions and the 3.0 shader model and
figuring out how to implement them at the high end and also how to
pull some out for the low end, and even integrated, effectively.
That's a challenge.

However, the big thing that I feel, and I came in with this
conviction, was that watching us trying to target the mainstream and
win was a dying model and so we said we are going to have to arc up.
R300 was really the first part where we really opened up the thinking
to what you can do to hit performance and hit schedule and relaxing on
die size to an extent and I think it helped ATI get back in the game.

At the end of the day though, is there really the desire to continue
with that – is the drive there to keep pushing that type of model?

There's always the debate of who steps down first. I think what's
going to happen is we're going to hit a power limit, so through other
innovations and technologies we have to manage efficiencies. And then
I've heard some, tongue in cheek, talk that NVIDIA isn't counting
die-per-wafer, but wafers-per-die, and whenever this is the case
you've certainly crossed a threshold!
--------------------------------------------------------------------------------

quote:
--------------------------------------------------------------------------------

There will be the temptation for people to look at R420 and NV40 and
compare the differences between the two. Clearly NVIDIA have a
compelling part in terms of features and the performance – while R420
is unlikely to disappoint much in terms of performance, in comparison
it looks a little behind the curve in terms of feature set.

I think the main feature that people are looking at is the 3.0 shader
model and I think that's a valid question. What we felt was that in
order to really appeal to the developers who are shipping volume games
in '04 Shader 2.0 would be the volume shader model of use. We do think
it will be important down the road.

How much of this comes down to engineering resource? Would it have
actually been possible for you to have had a Shader 3.0 part available
now if you'd wanted to, at a reasonable performance level?

As you say, there's always trade-off's. There's the trade-off of
performance and die size. The answer is yes we could – the die size
would have been fairly large to the point where we weren't sure how
produce-able it would be in 130nm and we didn't think that 90nm was
really going to be there for '04 production. Now, NVIDIA has put
something in 130nm that's die size is 10-15% bigger and there's still
some understanding we have to get on their architecture.

In comparison to NV40 do you think you undershot the expectations for
die size this time around, or do you feel that larger die sizes than
R420 are not really feasible at this point in time?

We focused on performance, schedule, features and cost. Our trade-off
was that we wanted to maintain our performance leadership and hit a
die size that we felt could be produced in volume. ATI is very
confident that we picked the best path for the enthusiast market in
2004.

With respect to engineering resources its been suggested to us that
the "West Coast Team" (Santa Clara - Silicon Valley) has become the
main focus for all the PC parts coming from ATI and that now even
R500, which we initially understood to be an "East Coast Team"
(Marlborough) product, is being designed at Santa Clara. Is it the
case that Santa Clara will mainly produce the PC parts now, while
Marlborough will be active with "special projects" such at the next
X-Box technologies?

We had this concept of the "ping-pong" development between the west
and east coast design centres. On paper this looked great, but in
practice it didn't work very well. It doesn't work well for a variety
of reasons, but one of them is the PC architecture, at the graphics
level, has targeted innovation and clean sheet innovation and whenever
you have separate development teams you are going to, by nature, have
a clean sheet development on every generation of product. For one, we
can't afford that and its not clear that it's the right thing to do
for our customers from a stability standpoint. Its also the case
that's there's no leverage from what the other development team has
done, so in some cases you are actually taking a step backwards
instead of forwards.

What we are now moving towards is actually a unified design team of
both east and west coast, that will develop our next generations of
platforms, from R300 to R400 to R500 to R600 to R700, instead of a
ping-pong ball between them both. Within that one organisation we need
to think about where do we architecturally innovate and where do we
not in order to hit the right development cycles to keep the
leadership, but it will be one organisation.

If you dissect in, for example, to the R600 product, with is our next,
next generation, that development team is all three sites - Orlando,
Silicon Valley, Marlborough – but the architectural centre team is in
the Valley, as you point out, but all three are part of that
organisation.

Would I be correct in suggesting that mainly Marlborough and Orlando
would be the R&D centres – with the design of various algorithms for
new 3D parts – while the Santa Clara team would be primarily
responsible for implementing them in silicon?

No, because the architecture of the R300 and R500 is all coming from
the Valley, but we've got great architects in all three sites.

Bob Drebin in the Valley is in charge of the architecture team and so
he's in charge of the development of all the subsequent architectures
but he goes out to the other teams key leaders and that forms the
basis of the unified architectural team. At an implementation level,
you're right – Marlborough is mainly focused on the "special projects"
and that will probably be another 18 to 24 months for them. So the
R600 family will mainly be centred primarily in the Valley and Orlando
with a little bit from Marlborough, and then the R800 would be more
unified.

--------------------------------------------------------------------------------

quote:
--------------------------------------------------------------------------------

As we go through each architectural change we begin to see that an
increasing number of chips are being produced by the IHV's – first
only one ASIC was produced to meet the entire market, then two, and
last year it was three. Does ATI have the engineering resources to
produce that many parts in a timely manner?

Well, I think four is needed now.

What's interesting is that you could argue there are more offerings
down below, but that's not the case. We used to design to a 10x10 die,
then a 12x12 and then the R300 was 14x14 and now we're at 16x16, so
you see what's really happened is we've added more SKU's in the high
end as time has gone by, so if you're at 16x16, what's the next one
down?

We think that there still is a market for a $10 part. Integrated isn't
taking over the world, so there's a huge opportunity for discrete
parts out there for about a $10-$12 selling price. So, with that at
the bottom end and you're high end part you ask if one will fill the
gap, and the answer is probably not, you might need two. So, how you
implement those total 4 SKU's is the question – is it 3 design centres
and some migrant technology or is it 4 design centres?

However, as you'll see this fall from us, and our RV lines, a total of
4 parts too. But the key is, from an architectural level, if you
architect it properly it becomes more of a system level design
question, and that's what we're looking to do.

Just after the introduction of R300 you talked about R400, however
shortly afterwards that appeared to go off the roadmap and R420
appeared – what happened during that period? And would that be related
to upcoming contracts with console vendors?

We changed the roadmap.

You can look at cause and effect. That was not the cause, internal
changes was the cause and the outcome was that we decided the best way
to go forward was to do "this" with the PC roadmap and take "that" and
use it with X-Box. But that wasn't the cause of the roadmap change, we
had to make this change anyhow due to execution issues.

-------------------------------------------------------------------------------

We' like to thank Dave for his frank and detailed answers to our
questions. you can see that it he has highlighted a number of changes
in ATI's engineering structure over the past year or so and it should
be interesting to watch their development and execution over their
coming architectures.
-------------------------------------------------------------------------------

my comments:

No mention of the Nintendo graphics processor for N5 / GCNext but I
wouldn't worry. It's still coming from what I understand.

It is very interesting that R600, ATI's next-next generation product,
is a joint-effort between ATI's 3 main design centers: Orlando,
Silicon Valley, Marlborough. Xbox 2 is supposed to be R600-based, if
not a hybrid of R500 and R600. so this is interesting news for Xbox 2.

Hopefully some solid info will surface on the Nintendo console soon.
then again, if Nintendo has its way, it probably won't until Nintendo
themselves shows it.

I still believe the Nintendo graphics processor will be custom.