lm@cottage.UUCP (06/05/87)
Hi. Does anyone out there know about the ETA machine? It's supposed to be a supercomputer made by a spinoff from CDC. What can you tell me about this machine? Will it run normal applications? What's the memory management look like? Etc. I'm interested in information from an OS point of view. Thanks in advance. Larry McVoy lm@cottage.wisc.edu or uwvax!mcvoy
rchrd@well.UUCP (06/05/87)
The ETA-10 is (will be? was? might be?) an 8 processor system.
Each processor is essentially a Cyber-205.
This means that each processor has a scalar and a vector
processor. THe scalar processor is a 200 mips machine with
a 64 word instruction stack, and a 256 word (64 bit) register
file, with independent segmented functional units.
The CPU memory is 64k mos sram, 4 million 64 bit words.
It is secded for each 32 bits with 75 billion bit/sec
bandwidth. It is a virtual memory system (just like the 205).
There is also a shared memory with 256k mos dram, 65 to 256
million words secded, 8 cpu ports at 12 billion bit/sec
bandwidth per port.
Apparently, there will also be a 4 processor version.
CPU Clock cycle is supposed to be 5 to 9 nanoseconds.
Like the cyber-205, it will have all the problems of that
machine. CDC has a very arrogant attitude towards programming,
and to attain the top speed of the machine, you have to
write in 205-FORTRAN with many non-standard syntax constructs.
Unlike the CRAY and other vector processors, the 205 is a memory to
memory pipeline, and can only attain high speeds for long
vector lengths. Virtual memory also degrades vector performance
everytime there is a reference out of in-core memory.
Still, the ETA-10 is not a real machine, with no real working
systems in place. It is also perhaps significant that rumor
has it that CDC has pulled out some of its staff on loan to
ETA.
Time will tell... ETA is going to have a hard time selling this
machine.
--
...Richard Friedman [rchrd] Pacific-Sierra Research
2855 Telegraph #415, Berkeley CA 94705 415 540 5216
uucp: {ucbvax,lll-lcc,ptsfa,hplabs}!well!rchrd
- or - rchrd@well.uucpbaum@apple.UUCP (06/08/87)
-------- [] >In article <3238@well.UUCP> rchrd@well.UUCP (Richard Friedman) writes: >Still, the ETA-10 is not a real machine, with no real working >systems in place. Do you know something we don't? ETA claims to have delivered a working machine to U. of Florida. -- {decwrl,hplabs,ihnp4}!nsc!apple!baum (408)973-3385
grunwald@uiucdcsm.UUCP (06/10/87)
I think that the ETA systems are cyber-205 work-alikes. THe ETA-10 is a box containing 10 of the CPUS. I haven't heard much about the memory heirarchy (if any) or the performance.
utterback@husc4.UUCP (06/11/87)
In article <3300007@uiucdcsm> grunwald@uiucdcsm.cs.uiuc.edu writes: > >I think that the ETA systems are cyber-205 work-alikes. THe ETA-10 is a >box containing 10 of the CPUS. I haven't heard much about the memory >heirarchy (if any) or the performance. The ETA-10 Can have from 1 to 8 CPUs. Since there has been so much interest recently, I will give some facts(?) that I know about the ETA-10. This is from a product presentation that they gave, and I am working for the competition, so take everything with a grain of salt. 8-) 1st system to FSU 2nd system to Princten, JVNC. 240 Chips on each PC board (16"x22") Each chip os 1 cm2 Each CPU holds 4MW of local memory Memory is air-cooled. CMOS Gate Arry [ALSI - 20k] Cyber-205 instruction compatibility. Factor of 10 improvment over 205 on short vectors. G Series is Liquid Nitorgen cooled, 45 Minutes to cool down, 90 Minutes to warm back up. 7 ns clock with Liq. N. 14ns clock in air. OPTIONS: 2,4,6,8 CPUs with 4MW memory each 64, 128, 192, 256 MW Shared Memory 2 to 18 I/O Units Optional Redunancy on 2 and 4 processors. Redundancy standard on 6 and 8 processors. Memory: Local: 64K MOS SRAM SECDED for each 32 bits 75 billion bits/sec bandwidth virtually addressed. Shared: 256K MOS DRAM 32-256 MW with SECDED 8 hi-speed cpu ports 18 lo-speed ports O.K., there it is. Like I said, this was from a marketing product presentation, so I can't vouch for the accuracy of the claims. Also, don't ask me for more details, I don't have them, and I don't even understand all that I wrote above. Brian Utterback >Everything in this message is the opinion of myself and not my employer. < >I am representing only myself, and if you can, forget that you saw my < >name on it, O.K.?
eugene@pioneer.arpa (Eugene Miya N.) (06/11/87)
Brian-- (and others)
Since you noted you work for the "competition."
Just a word of note: Neil Lincoln, the chief architect of the 10 does
read this news group (although he tells me that he spends more time on
his Amiga reading comp.*amiga [interesting contrast Cray builts Crays on
Apples, Neil builds ETAs on Amiga, (will Commodore buy an ETA?) think
of those lines .....;-)]). Seymour does not read news, but I heard
he gave a chuckle when people (like Patterson) started crediting him
with RISCs.
From the Rock of Ages Home for Retired Hackers:
--eugene miya
NASA Ames Research Center
eugene@ames-aurora.ARPA
"You trust the `reply' command with all those different mailers out there?"
{hplabs,hao,ihnp4,decwrl,allegra,tektronix,menlo70}!ames!aurora!eugenetrb@stag.UUCP ( Todd Burkey ) (06/12/87)
Slight Correction...Neil Lincoln uses Atari ST's, not Amigas...But he laughed when he saw the message... -Todd Burkey (YAAA-Yet Another Atari Addict) ..ihnp4!meccts!stag!trb