interview of Sony's Ken Kutaragi on: PS3, Nvidia, graphics, eDram, more

the following is all quote from the message board thread:
http://www.ga-forum.com/showthread.php?t=51518

quote:

"This is actually part 3 of the Watch Impress Kutaragi interview."
http://pc.watch.impress.co.jp/docs/...3/kaigai189.htm

"He talks about why a Cell-based GPU wasn't use, the close bidirectional
Cell<->RSX relationship, eDram, unified shaders vs dedicated shaders, why
they went with NVidia, backwards compatability and more. Unfortunately I
can't see any proper translation yet, so we'll have to make do with the mess
that is babelfish's for now."

Things I took from it:

- PS3 backwards compatibility is through software on Cell (perhaps Cell
alone). Cell has been made bi-endian to emulate the PSX and PS2 CPUs.

- A 2-Cell PS3 was considered - one acting as a GPU, one as a CPU. But
Cell's nature as a CPU eats away at its efficiency for graphics relative to
a dedicated GPU. So they've married Cell with a specialised GPU.

- Cell, however, can be used for graphics tasks. Cell and RSX have direct
access to the results of each other's operations. Kutaragi specifically
mentions displacement mapping as a sample workload the SPEs could take on. I
think he mentions something about harmonising floating point rounding errors
between the CPU and GPU to further ease data exchange between the two. Seems
like they focussed quite a lot on harmonising communication between the GPU
and CPU, and binding them quite tightly together.

- Originally there was no VRAM in PS3. The graphics chip accessed only the
XDR ram.

- There is no eDram in PS3, because they couldn't put enough on the die to
support a 1920x1080 frame let alone 2. Kutaragi thinks the transistor cost
of eDram is too high when you consider what the same amount of transistors
buys you elsewhere in terms of more shading power.

- He touches on unified shading and mentions something about problems with
stalls..I don't know, hard to make out exactly what he's saying there.




Quote:
The reason which does not adopt the Cell based graphic chip

With PS3 architecture, those where last year you are surprised
graphics are to apply with Cell based architecture. Why, not making Cell
based GPU?

Kutaragi: 7 SPE of Cell (Synergistic Processor Element) it can use in
graphics. Actually, several of the demonstration with E3, when it has not
been the graphic processor, has done everything of graphics to rendering
with just Cell. But, such using is wasted. There are times when you do more
in other things in Cell.

2 placing Cell, (you use Cell of one side) with also the plan which is
said it was on the graphic center, but as Cell and Shader as the computer
thinking that weight is different from function you stopped. We would like
to make the architecture which Shader with, all things thoroughly (in
graphics) can specialized Shader. Though, for example, displacing mapping
(it does) with it can also say with SPE.

So far (real time 3D graphics), that it seems and shows, (3D graphics)
in the space is different really. Even then, with present resolution it was
good. The game which has come out with Xbox 360 the majority, is 3D such as
that even at present point in time.

But, I would like to make 3D where (metamorphosis and the like is
reflected securely in the 3D space). For that, there is the thought as much
as possible of liking to share the data (with CPU and GPU). Therefore, the
latest architecture was taken. Originally if as for the floating point unit
of GPU and Cell, from precision round (making round) until error we would
like to make entirely the same. This time, it has become close, almost
rather simultaneous. Therefore, each other, it can use (the operational
result) bidirectionally.

- EDRAM is stopped because of full HDTV

EDRAM (it installed graphic memory and DRAM) expected, but the HDTV 2
picture hearing, you could agree upon the reason which it does not make
eDRAM.

Kutaragi: Originally, there is no graphic memory in GPU, Redwood (the
high-speed interface which connects Cell and RSX) with YDRAM (the code name
of XDR DRAM) with also correspondence is achieved. Because YDRAM (memory)
has become unified.

But, so entering, that there is a problem which you say whether ()
definite form processing and Shader of graphics (computing) when it does,
the distant place (accessing memory) with, it is possible to make (zone and
cycle time) wasteful. It is not necessary to use (memory zone), with
definite form task of the side of with special care Cell. Because Shader
with Shader, calculates the tremendous quantity, there is a necessity of
memory even to there. Especially, when with full HDTV, 2k×1k (the
1,920×1,080 dot) being progressive, when 2 pictures or more we would like to
handle, it becomes, mass VRAM becomes necessary.

When that happens, eDRAM is unreasonable. As for using eDRAM, at the
time of PS2 it is good. Are not enough in so or this time just 2 pictures.
We assume that eDRAM of the quantity which (can support HDTV) 200 squares mm
or 300 squares mm was inserted in the stone. So, when it does, (it can load
onto the tip/chip logic because of the area of eDRAM) at once to decrease,
the quantity of Shader decreases. That compared to, in full using in logic,
the method which places large quantity Shader is better.

- NVIDIA vision of the processor of the ideal which is shared

In the first place, why uniting with NVIDIA in the GPU vendor?

Kutaragi: So far, we, Toshiba and together did the graphics for
computer entertainment, personally. Including also process technology, the
finish you did. And, this time, it united with NVIDIA in order to do the
computer itself.

The finish pursuing PC graphics, Intel to the processor doing, it is
about NVIDIA probably to do with Programable Shader. NVIDIA pursuing
function and efficiency as a processor, the ???? kirk (Chief Scientist of
David B. Kirk and NVIDIA) to include the person, because the developer is
graduate person of various computer enterprises such as SGI. There is a
character that they, size and the like of the tip/chip do in the air, resign
the fact that we would like to do and would like to pursue. Occasionally,
there are also times when you do too much, but culture has been similar to
me.

The approach of NVIDIA my approach, agrees in the point that finally
the finish it will pursue the full programmable processor. ????? (president
and CEO of Jen-Hsun Huang and NVIDIA) and ???? to be good there is an
opportunity which does story, but a story that the processor of ideal it
probably will do that it appears in that time. Ideal, naturally, present PC,
well, is the processor which exceeds the present all processor.

They are opposite to the direction steadily, in that sense, us and
vision share. It shares also the road map. In addition, they have received
influence even from our architecture. Mutual temper having known, because
you think, that we would like to do the same thing, it united with NVIDIA.

As for another element, display fixed pixel system (the liquid crystal
and the like) the point which is moving. When it becomes fixed pixel system,
TV and PC, are times when everything fuses. Therefore, we would like to
support everything perfectly.

So it puts out also the downward compatibility of PS, dirty ????
(graphics) from to up-to-date Shader would like to support entirely. As for
resolution ??? we would like to put out those above WSXGA. When such as
that, rather than we making scratch (build) with, the method which the van
has all is quicker (from NVIDIA).

As for Microsoft, with Xbox 360 GPU of ATI took Unified-Shader type
architecture. With program characteristic Unified-Shader with advanced.

Kutaragi: As for the architecture of ATI, as for Shader however Vertex
Shader and Pixel Shader equality (architecture) with, are visible at first
glance well with joint ownership, you think that it is difficult. For
example, whether the result of doing apex processing is made somewhere, how
doing that, in Shader letting flow (for pixel processing) again? When
somewhere is plugged, the stall it does entirely. Really it is different
from those which are drawn in the picture. If of realistic performance is
thought, you think that method of approaching NVIDIA is superior.

PLAYSTATION 3
- As for maintenance of compatibility with combination of hardware and
software

As for the compatibility of past PlayStation actualizing with the
hardware?

Kutaragi: You take with the combination of the hardware and the
software. You try if probably to do (with just the software) how it becomes,
but just which drives to perfectly close compatibility, it is important.

As for the person who develops the software unexpected, the fact that
you cannot imagine is done. For example, however it is not logical as a
program, you said that it moved accidentally. It is moving however, with,
that is moving with, is a kind of case which is said in completely another
reason. Passing through also our tests, "as for this cord/code which is
what! "There are times when the cord/code which we would like to see passes.

We do not take either the compatibility for the cord/code such as that
and the ? is not good. ) Just a little it is painful but, because there is
no logic, (to take compatibility with just the software. There are also
times when hard becomes necessary. If so, this time (PS3) there is a power
of extent, as for a certain place the correspondence such as with the
software can do the place where it is hard.


When the cord/code of CPU side is emulated with the software, as for
the endian of CPU.

Kutaragi: Therefore as for Cell bi- endian, how it becomes.

Xbox 360 takes compatibility almost with just the software. Because
they have not produced the tip/chip at the respective company, it is the
case that it is not the choices in other things, but how seeing?

Kutaragi: As for Xbox, when the new generation comes to November of
this year, as for current Xbox you become the old generation. So when it
does, Xbox means to kill by your by your. The only method of rescuing that
takes 100% compatibility from first day. So or that it probably cannot
commit (Microsoft), technically it is painful.


<Watch Impress Comments below>

- As for SCEI and NVIDIA those which are similar

Hisashi ? well, being the element, directivity of enterprise above
transacting the device development and agreement of culture simply is
transmitted to the relation of SCEI and NVIDIA, from word of the wooden
person. Also both corporations, like original idea, with ???????, pursue to
cost function and efficiency last. Is not the case, always, but the
propensity such as that is strong. In addition, also both corporations,
presently agree with the conception which pursues the processor.

In the GPU vendor, NVIDIA directivity to especially programmable
conversion is strong. Speaking accurately, it has the direction whose also
ATI Technologies and 3Dlabs are strong in programmable conversion, but it
was most aggressive in NVIDIA raising general purpose. As for NVIDIA,
because of that, the die/di size of GPU (the area of the semiconductor
itself) it enlarges, it does not leave either the fact that production cost
soars. As SCEI, that the directivity of such NVIDIA, it seems that you
thought is faced as a partner.

Presently, GPU stream processing (with the small program piece keeps
processing the mass data) has dashed forward to the programmable processor
which specializes in stream system. By the fact that the general purpose of
Programable Shader which is the operational core is raised, it is the case
that it tries to be able to do, also the general-purpose processing other
than graphics. It is the idea of the general-purpose processor which places
the sub processor which on the one hand, also the basic idea of Cell,
optimizes in stream processing. Evolving the general-purpose processor, it
makes the structure which faces to the stream type processing which in the
future would become important. If it tries saying, SCEI and NVIDIA are
approaching to the same goal from another direction. So when you think, SCEI
and NVIDIA agreed, are not strange thing with vision. You can understand
also the fact that it is the agreement point which is called to the
directivity both, the processor of ideal.

Hisashi ? well, with the graphic architecture of PLAYSTATION 3 it can
support the fact that several choices were examined from explanation of the
wooden person. First, with 1 Cell processor, the plan which can let do
graphic processing. SPE which is the data processing processor core of Cell,
SIMD (Single Instruction and Multiple Data) has the operational unit of
type, can designate the same thing as Programable Shader of SIMD structure
similarly basically. But, to be able to let do graphics to Cell, proper
thing, it is not realistic because the efficiency of Cell as CPU is shaved.

Next, 2 loading Cell, the plan which uses Cell of one side in graphic
exclusive use. Expanding the architecture for graphics Cell, it is presumed
in this plan that also the plan which makes SPE for graphic processing was
included. In that case, it is presumed that also loading and the like the
operational unit for specification processing of graphics probably is done.
Though, plan of the Cell based graphic tip/chip, is said that it went out
rather at early stage.

By the way, even with present PS3 architecture, it can use Cell in
graphic processing. As for Kirk of NVIDIA, with the combination of Cell and
RSX, it has made clear that pre- processing and post processing of 3D
graphics can be done with SPE of Cell. For example, metamorphosis is done
Displacement Mapping which (displacement mapping) and the like also to do on
SPE side it is possible the apex data.

SCEI eDRAM (installs in RSX and DRAM) the reason which is not placed,
is clear from the picture resolution which as written even in the past, is
supported. In addition, in order to actualize high Shader processing
performance, thinking that it is not possible to consume the die/di area
with eDRAM it is recognized. This took the special graphic architecture
which utilizes the wide band of eDRAM, conception differs from Graphics
Synthesizer of PS2 fundamentally. If you look at the information which is
open with RSX, architecture, NVIDIA color quite is strong.

SCEI with PlayStation 2 solved the problem of compatibility by the
fact that it loads the chip set of old PS as the sub processor with the
hardware. Because this, unless it makes the hardware base, cannot guarantee
almost 100% compatibility. When the hardware emulation is done completely
with the software, enormous CPU power becomes necessary. This, like PS2
releasing the content of the hardware, in the machine which the developer
that tries can access the resource freely, especially is critical.

Those where presently it is clear are basically "perfectly to have
been about to actualize close" compatibility even with PS3. Because of that,
as still compatibility in the hardware base the direction, continues even
with PS3. However, this time, utilizing the high processing power of Cell,
the interchangeability in the software base (emulator) it is taken. That
specially Cell was designated as bi- endian the fact that compatibility of
the CPU side is taken with Cell it means. At early stage of the cooperation
development with IBM, SCEI is conveyed that it requested that bi- endian is
necessary because of compatibility. By the way, this time, the compatibility
of PS, PS2 and 2 generations is actualized. The both of PS and PS2 has
loaded CPU of MIPS architecture.



As always, it's probably best not to read too much into anything until we've
a proper translation..the above is just a rough idea of what was being said.
A better translation should hopefully be on the way soon..