Re: Larrabee details: Yes, it is based on the Pentium. :-)

Quadibloc wrote:
That I don't wonder.

The Pentium, like the IBM 360/195, combined the two performance-
enhancing features of aggressive pipelining and cache. The 486, in
contrast, could be compared to a 360/75 architecturally, so by going
back before the Pentium, you are taking a *really big* single-thread
performance hit.

The 486 had about the same pipeline depth and also cache. The advantage of
the Pentium was the second pipeline, i.e. it was the first superscalar x86
implementation. In contrast to the 486, it had branch prediction, and
separated the cache into I and D cache (the 486 had a combined cache). This
all can help, but remember, the 486 had a significant clock cycle
advantage, i.e. 100MHz 486 vs. 60MHz Pentium in the same process (and later
the Pentium II had a similar lead, last 450 vs. 300MHz in 0.25µm, where the
Pentium MMX was only used for mobile applications). For real world
application, this clock cycle advantage made it first difficult to sell the
Pentium - the 486 was cheaper and not really slower. AMD went further with
the 486, but in the end, the 486 was doomed - AMD replaced it with the
lackluster 29k-based K5, and Intel just stopped delivering faster 486er.

There's however another (management) reason to use the P55C as model: It was
designed in Haifa. Intel's Haifa team seems to be more successful than the
Oregon team, so giving such a task to Haifa isn't a bad idea. Haifa can
only go back to the P55C.

I'm not too happy with this idea of x86 in a GPGPU. x86 is not "easy to
program", it is a messy architecture. The current GPUs are difficult to
program since the vendors don't tell us how to program it (AMD just
recently released specs at http://www.x.org/docs/AMD/). Now, with Intel to
be releasing full spec of their GPU (by basing it on x86), this trend
continues. This will make GPGPU programming much simpler, just because we
now can do it as we like, not just with the vendor's toolkit. Whether r600
or Larrabee is easier to program for GPU or GP stuff remains to be seen -
you get the actual performance through the vector instructions, and we
don't know how those will work. I suspect not too much different from the
r600 ISA.

--
Bernd Paysan
"If you want it done right, you have to do it yourself"
http://www.jwdt.com/~paysan/
.

Relevant Pages

  • Re: Pentium vs Celeron
    ... >half the cache. ... Celeron processor than the usual shoe-scrapings plus "Pentium Tax". ... 128k full-speed L2 cache, 400MHz base, SIMD ...
    (microsoft.public.windowsxp.general)
  • Re: Need cache chips for 1995-vintage PC
    ... The Pentium OverDrive and the 233MHz Pentium MMX represent the ... levels of cache into the Socket 7 K6-3, which will not run in the Advanced/MN ... a 200MHz Pentium Pro CPUs ...
    (comp.sys.hp.hardware)
  • Re: Re: How does an 800Mhz celeron compare to lower speed Pentium II
    ... > cache as a P-II ... > The 370 socket Pentium 3?s had twice the cache and for it?s time was ... > eat for dinner a similar speed 370 socket Pentium 3 chip. ... > socket processors started appearing with the faster clock speeds ...
    (alt.comp.hardware.pc-homebuilt)
  • Re: Pentium vs Celeron
    ... >comparably equipped Celeron based system. ... Celeron and Pentium 4 are based on the same core, ... Zero Level 2 cache, 66MHz base ... games is your thing, a large and fast HD, and lots of RAM (or at least ...
    (microsoft.public.windowsxp.hardware)
  • Re: IBM 45nm -- new or licensed from Intel?
    ... Why do you think that Silverthorne's pipeline will have the same constraints as the original Pentium? ... I.e. use a pipeline which Intel knows how to compile for, and where they also know how to minimize gates/power needed. ... It's also worth remembering that there's also Diamondville, the SMT ... It seems like a far more general-purpose dual-issue design than the original Pentium, particularly since it needs SMT to get some useful work done during the very long branch miss shadows. ...
    (comp.arch)