On Fri, 17 Aug 2007 08:06:40 +0100, Shawk
wrote:

John Lewis wrote:


Seen the AMD stock price recently... slipping away?

You always sound to me John like someone that wants to see one graphics
card manufacturer with no competition?


Ummm, actually no.

But no point in being an ostrich and indulging in wishful thinking.
AMD is in serious financial trouble and developing new high end GPUs
(the largest die-size in mass-market computing-chips on the planet,
including CPUs...) on very small-geometry processes is exceptionally
expensive. nVidia spent $400million developing the 8xxx family. Where
is AMD/ATi going to get this sort of money for another full-fledged
turn at DX10 hardware? Seems as if they need to put every $$ they have
in fending off the Intel threat to their core CPU business.

In reply to your statement...

You always sound to me John like someone that wants to see one graphics
card manufacturer with no competition?

..... I will say--- Do you then want to see one CPU manufacturer with
no competition?? This starkly paints the AMD/ATI dilemma. Bleeding
re-ink... where should AMD focus its efforts? If I were AMD, I would
endeavor to dump ATi ASAP. Maybe they have privately tried already
with no decent bids? If I were AMD, I would focus most of what little
development money is still available on the CPU business and let ATi
take care of the volume sweet-spots in computer core-chips with
integrated graphics and graphics engines for high-volume
consumer-devices, cell-phones, media-players etc. I would cease any
further development in the high-end stand-alone graphics chip arena,
and just milk the 2xxx series designs as much as possible,
opportunistically adding low-cost peripheral functionality (such as
DisplayPort) to gain as much market volume traction as possible, at
least until the joint company turns a decent profi again.

Anyway, back to nVidia's current discrete-GPU exercises...

nVidia's next GPU turn is a complete re-run at their Dx10 family and
they have more than enough profits from their business to finance this
turn. Probably compatible with DX10.1; certainly double-precision
compute-paths, certainly optimized for the DUAL alternate roles of GPU
and GPGPU from the same silicon.

The expansion of Nvidia's efforts in the GPGPU arena is a major
strategic target for the upcoming turn of the nVidia GPU family.
Besides the potential uses in home computers ranging from accelerating
features of video processing to games, there is a huge professional
market for the massive hardware parallel-processing of the GPGPU in
tandem with nVidia's CUDA mid-level programming language specifically
targeted for the GPGPU.

John Lewis