Re(2): [SunRescue] What is difference between IPX and IPC sun machines?

James Lockwood lockwood at ISI.EDU
Fri Aug 13 16:04:45 CDT 1999


On Fri, 13 Aug 1999, Tim Hauber wrote:

> I agree, if video is what you do.  However, the faster bus on the IPC, and
> the total lack of color if you are running it with the built in FB  make
> it about as effective for command line work.  I have always preferred
> busses running at chip speed, my favorite intel chip is still the

In the sun4c family this is true, as memory accesses are locked to the
sbus.  In other Sun families where memory has an independant path the sbus
speed is less important, as it is only hit for peripheral access.

> 486DX/50, not to be confused with the DX/2.  I'd really like to see AMD
> come our with a 200Mhz K7 running on a 200Mhz bus, with 200Mhz memory
> clocking.  I bet it could give the 600/200/100 Mhz one  a run for it's
> money on bus intensive applications.

Busses running above 150MHz or so are very difficult to design at this
point, due to their tendancy to act like transmission lines.  200MHz
memory clocking is blue-sky at this point without some wicked interleaving
that the PC crowd wouldn't go for.  Large fast caches are not economical
to put on-chip.

My favorite Intel chip is the i860XP, actually.  :-)

> These new multi-pipeline chips running twice the speed of their L1 cache
> and 6 times the speed of the PCI bus must spend an awful lot of time
> twiddling their thumbs.

This all depends on how fast the cache is emptied by the CPU.  The Intel
Celeron is a terrible design, but it works well regardless.

I like the architecture of the AXi quite a lot: a fast CPU with L1 cache
running at core speed, L2 cache running at bus speed (typically 1/3 CPU
speed for Sabre modules) and a 66MHz PCI bus feeding dual 33MHz busses
which actually go to the slots.  Far better structured than those x86
boards with slower and smaller L2 and a single 33MHz PCI bus spread out
over 5 slots.  Unfortunately in the PC world all the money goes to the CPU
makers and high-quality motherboards are shunned in favor of cheap ones. 
The public has been brainwashed by MHz numbers as better motherboard
performance is not easily numerically quantifiable. 

-James







More information about the rescue mailing list