Date: Wed Nov 29 2000 - 10:47:13 PST
Gregory Alan Bolcer writes:
> Sony disagrees & is why they are launching their "Tool Creative Workstation"--a Linux
You sure about this? I can't google anything about it. URI?
If the thing is any good and affordable, I'll buy it. I'm rather sick
of PC architecture, to be frank. Otoh, Sony's technology chrome is by
now getting rather flaky about the edges. And their policy re open
> based workstation based on the PSX architecture. Last time I posted
> this article I got hate mail from the CPU guys saying that there will never
> be a .01 micron fab of a chip in the next decade.
10 nm structure size is not awfully large. Something like less than
100 atoms across. While this might still be doable with semiconductor
photolitho (and lots of advanced voodoo), this is about the threshold
or already slightly beyond where you'd better switch over to molecular
circuitry. I'm not sure that we'll have molecular circuitry by ~2010,
but I won't rule it out. Maybe 2d crystal memories. Processors and
all-purpose logics not likely (but they should arrive soon after).
> "In the past, workstations and PCs had more power than
> home game consoles, so we could use them as development tools. But when
> the power [of Playstation 2] matches or exceeds their power, it becomes
> difficult to use them for development." Ken Kutaragi on the roadmap of
> the emotion engine and graphics synthesizer chips found in the EE Times.
I dunno, 1.2 GHz Athlons with good code should kick PSX2 CPU's ass
right now. The embedded RAM engine is probably still superior to a
modern GeForce 256 derivate, but not by wide margin. Also see the
stuff Nvidia let leak out (kindly rescale to accomodate for hype
factor). Still no embedded memory, but I don't see how you can avoid
this on the long run.
> He should know what he's talking about, he has the benefit of a billion
> dollar investment in two new chip fab facilities.
You've surely already seen this:
Nvidia GPU is seven times faster than NV15
By Kai Schmerer, ZDNet Germany
Fri, 24 Nov 2000 06:35:07 GMT
Secret documents reveal Nvidia's next screaming chip design
Pioneer of the first ever GPU (Graphics Processing Unit), Nvidia is now
introducing a programmable GPU, seven times faster than the original
Geforce 2 Ultra, the NV15.
Confidential documents seen by ZDNet from the graphics chip manufacturer
state the new chip will outperform previous processors even in complex 3D
According to the document, the chip's performance will depend on the
application it is running on. In environments where there are low detail
scenes (large triangles, simple geometry, hardly any depth) the NV20 is
only twice as fast as the previous Geforce 2 Ultra. The difference in
performance grows however with the complexity of the 3D scene (multitude
of small triangles, very complex geometry, a lot of depth).
The anti-aliasing performance of the chip is also said to have increased
substantially. Nvidia gives the NV20 a 300 percent increase in
performance compared to the NV15. The performance of the chip doubles
when handling geometrical data. According to Nvidia the NV20 is able to
handle one trillion operations and 100 gigaflops (floating-point
operations) per second.
The performance increase is achieved through better design, according to
the documents: the storage interface is a completely revised version
compared with the Geforce 2. Downloading large quantities of geometrical
data should now flow substantially faster, according to the documents.
In addition, the NV20 is said to have a ramdac speed of over 500MHz as
well as an faster clock rate.
There is a new pixel shading experience for 3D-textures and curved
surfaces will cater for a 3D visual quality. The NV20's gradual fogging
effects and per-pixel bump mapping will primarily enable the presentation
of even more realistic characters as well as natural effects.
If the specifications and performance statistics highlighted in these
confidential documents are proven to be true, Nvidia position as market
leader looks set.
This archive was generated by hypermail 2b29 : Wed Nov 29 2000 - 11:13:37 PST