| Title: | Hudson VLSI |
| Notice: | For Digital Chip Data - CHIPBZ::PRODUCTION$:[DS_INFO...] |
| Moderator: | RICKS::PHIPPS |
| Created: | Wed Feb 12 1986 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 701 |
| Total number of notes: | 4658 |
Is the 21164 a braindead 21164? What's been done, has the cache been
dropped, clock turned down, transistors shaved? Will this in turn make
the chips much slower than the 21164... What are the target specmarks,
as we are getting about 20-21 SpecFP and 14-15 SpecINT from the current
500 Mhz parts? (i know specs.. are an average of many things and not a
very good measuring stick for application performance, but.. as a
watermark they are OK) Has any comparison preliminary testing been done?
I remember a low cost 21064 which was way slower than an 21064 at same
clock speeds... ie Multia Alpha...
MP
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 662.1 | NETRIX::thomas | The Code Warrior | Sat Mar 22 1997 19:34 | 4 | |
No. What a 21164PC is has been covered in this conference man times before. Please do a bit of digging... | |||||
| 662.2 | 21164PC announced SPECint95=14.3e and SPECfp95=17e | PERFOM::LICEA_KANE | when it's comin' from the left | Mon Mar 24 1997 10:13 | 16 |
| (i know specs.. are an average of many things and not a very good
| measuring stick for application performance, but.. as a watermark they
| are OK)
You've been reading comp.sys.sgi.hardware too much.
Take an Alpha in a real workstation, and while holding everything
constant, increase its clock rate by 20%. You can safely guess that
"application performance" might go up by as much as 20%. You
can also safely guess that the "application performance" might go
up as little as almost nothing.
SPEC CPU95 ratios will go up as much as 20%, and as as little as almost
nothing. Just like "real applications."
-mr. bill
| |||||