From: slavins@hearsay.demon.co.uk@localhost (Simon Slavin) Newsgroups: alt.folklore.computers Subject: Rough equivalence: early Cray with today's micro Date: Thu, 17 Jul 2003 00:42:58 +0100 Organization: None Lines: 10 Message-ID: NNTP-Posting-Host: hearsay.demon.co.uk X-Trace: news.demon.co.uk 1058398967 2248 194.222.24.177 (16 Jul 2003 23:42:47 GMT) X-Complaints-To: abuse@demon.net NNTP-Posting-Date: Wed, 16 Jul 2003 23:42:47 +0000 (UTC) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!switch.ch!news.belwue.de!newsfeed01.sul.t-online.de!newsfeed00.sul.t-online.de!t-online.de!kibo.news.demon.net!news.demon.co.uk!demon!user Xref: chonsp.franklin.ch alt.folklore.computers:146238 I want to give a ballpark figure in something I'm writing for non-technical people. There was a time when if you needed the fastest computer for scientific number-crunching you used a Cray. Can someone tell me what year that was likely to be and how that Cray would compare with one of today's Wintel/Linux/Mac machines in terms of speed and memory ? ###### From: "Glen Herrmannsfeldt" Newsgroups: alt.folklore.computers References: Subject: Re: Rough equivalence: early Cray with today's micro Lines: 25 X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2919.6700 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700 Message-ID: <6jqRa.68898$GL4.17073@rwcrnsc53> NNTP-Posting-Host: 12.207.204.17 X-Complaints-To: abuse@comcast.net X-Trace: rwcrnsc53 1058419010 12.207.204.17 (Thu, 17 Jul 2003 05:16:50 GMT) NNTP-Posting-Date: Thu, 17 Jul 2003 05:16:50 GMT Organization: Comcast Online Date: Thu, 17 Jul 2003 05:16:50 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!switch.ch!news.belwue.de!feed.news.nacamar.de!newsfeed.icl.net!newsfeed.fjserv.net!news-out1.nntp.be!propagator2-sterling!news-in-sterling.nuthinbutnews.com!cyclone1.gnilink.net!wn14feed!wn13feed!wn12feed!worldnet.att.net!204.127.198.203!attbi_feed3!attbi.com!rwcrnsc53.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:145970 "Simon Slavin" wrote in message news:BB3BA192966839C131@10.0.1.2... > I want to give a ballpark figure in something I'm writing for > non-technical people. > > There was a time when if you needed the fastest computer for > scientific number-crunching you used a Cray. Can someone tell > me what year that was likely to be and how that Cray would > compare with one of today's Wintel/Linux/Mac machines in terms > of speed and memory ? As close as I remember it, that was 1976. The Cray-1 was clocked at 12.5ns, that is 80 MHz in today's units. I am not so sure about memory size. It was interleaved, maybe 16 way, so that it could be accessed at one word per clock cycle in most cases. The closest I can remember is 256 MB, but I am not so sure on that. I believe it was measured in 64 bit words, and I am not sure which number I remember. The Cray 2 was 4.2ns clock cycle, about 238MHz. -- glen ###### From: Thomas Womack Newsgroups: alt.folklore.computers Subject: Re: Rough equivalence: early Cray with today's micro Date: 17 Jul 2003 10:06:12 +0100 (BST) Organization: Linux Unlimited Lines: 38 Message-ID: References: NNTP-Posting-Host: rapun.sel.cam.ac.uk Originator: twomack@chiark.greenend.org.uk ([193.201.200.170]) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!switch.ch!news.imp.ch!news.imp.ch!eusc.inter.net!news.tele.dk!news.tele.dk!small.news.tele.dk!lnewspeer00.lnd.ops.eu.uu.net!emea.uu.net!server1.netnews.ja.net!pegasus.csx.cam.ac.uk!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:146240 In article , Simon Slavin wrote: >I want to give a ballpark figure in something I'm writing for >non-technical people. > >There was a time when if you needed the fastest computer for >scientific number-crunching you used a Cray. Can someone tell >me what year that was likely to be and how that Cray would >compare with one of today's Wintel/Linux/Mac machines in terms >of speed and memory ? I think that was the case until the mid-nineties, though the Crays at that point were things like the T3E, large clusters of 21164 Alphas with custom interconnect. The first machine I had which unambiguously outpaced a Cray 1 was my first P4; other ones didn't have the sustained memory bandwidth. I believe the Cray 1 had eight megawords of memory, so 64MB. Memory bandwidth is the killer -- my current laptop (admittedly bought on Monday) has about six times the FLOPs of a single CPU on a T90-series Cray, which were the fastest Classical Vector Machines around. I can't find stream results for a P4 on the 875 chipset; for a two- year-old P4, the figures are around 2GB/second, which is a little more than a single J90-series Cray CPU could manage, and a factor three less than a single C90-series Cray CPU or a factor five less than a single T90-series one. Some Itanium2 machines manage 3.5 to 4 GB/second. Basically, a current desktop will beat one CPU of a Cray J90 on anything, and will beat a whole Cray J916 on jobs that don't need memory bandwidth. This is probably why J-series Crays show up on ebay. Tom ###### Newsgroups: alt.folklore.computers Subject: Re: Rough equivalence: early Cray with today's micro References: <6jqRa.68898$GL4.17073@rwcrnsc53> From: Morten Reistad X-Newsreader: trn 4.0-test76 (Apr 2, 2001) Originator: mrr@acer.reistad.priv.no (Morten Reistad) Message-ID: Lines: 34 Date: Thu, 17 Jul 2003 11:35:58 +0200 NNTP-Posting-Host: 80.111.136.96 X-Complaints-To: abuse@chello.no X-Trace: amstwist00 1058471102 80.111.136.96 (Thu, 17 Jul 2003 21:45:02 MET DST) NNTP-Posting-Date: Thu, 17 Jul 2003 21:45:02 MET DST Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!switch.ch!news.tele.dk!news.tele.dk!small.news.tele.dk!newsfeed1!bredband!amsnews01.chello.com!amstwist00.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:146302 In article <6jqRa.68898$GL4.17073@rwcrnsc53>, Glen Herrmannsfeldt wrote: > >"Simon Slavin" wrote in message >news:BB3BA192966839C131@10.0.1.2... >> I want to give a ballpark figure in something I'm writing for >> non-technical people. >> >> There was a time when if you needed the fastest computer for >> scientific number-crunching you used a Cray. Can someone tell >> me what year that was likely to be and how that Cray would >> compare with one of today's Wintel/Linux/Mac machines in terms >> of speed and memory ? > >As close as I remember it, that was 1976. The Cray-1 was clocked at >12.5ns, that is 80 MHz in today's units. Remember that it was a vector machine, and could do some hundred (was it 128 or 256?) operations in a streaming, fast mode. It was short of a billion operations per second though. So, in todays terms; unoptimized around the speed of a slow pentium-II; optimized around the speed of a fast PIII. >I am not so sure about memory size. It was interleaved, maybe 16 way, so >that it could be accessed at one word per clock cycle in most cases. The >closest I can remember is 256 MB, but I am not so sure on that. I believe >it was measured in 64 bit words, and I am not sure which number I remember. > >The Cray 2 was 4.2ns clock cycle, about 238MHz. Just barely reached a billion operations per second, AFAIR; or about at the midpoint of the P4 range in vector mode. -- mrr ###### From: slavins@hearsay.demon.co.uk@localhost (Simon Slavin) Newsgroups: alt.folklore.computers Subject: Re: Rough equivalence: early Cray with today's micro Date: Sun, 20 Jul 2003 00:26:58 +0100 Organization: None Lines: 17 Message-ID: References: <6jqRa.68898$GL4.17073@rwcrnsc53> NNTP-Posting-Host: hearsay.demon.co.uk X-Trace: news.demon.co.uk 1058657202 26598 194.222.24.177 (19 Jul 2003 23:26:42 GMT) X-Complaints-To: abuse@demon.net NNTP-Posting-Date: Sat, 19 Jul 2003 23:26:42 +0000 (UTC) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!switch.ch!newsfeed00.sul.t-online.de!t-online.de!kibo.news.demon.net!news.demon.co.uk!demon!user Xref: chonsp.franklin.ch alt.folklore.computers:146236 Lots of people wrote: >[stuff] Many thanks for the responses. Taking some averages and leaving out the vector-processing capabilities in consideration of my non-technical readers, I could say that Cray built the fastest and most impressive computers throughout the 1980s. By the end of the 1980s they were about equivalent to a Pentium 3 in speed/memory/processing but they had some advantages such as faster memory access and clustering ability. I'll then look up costs, size, power-consumpution, etc.. Thanks for your help. ###### From: jsavard@ecn.ab.ca () Subject: Re: Rough equivalence: early Cray with today's micro Newsgroups: alt.folklore.computers References: Organization: Edmonton Community Network X-Newsreader: TIN [version 1.2 PL2.6] Lines: 21 Message-ID: Date: Mon, 21 Jul 2003 17:22:15 GMT NNTP-Posting-Host: 198.161.206.2 X-Trace: localhost 1058808135 198.161.206.2 (Mon, 21 Jul 2003 11:22:15 MDT) NNTP-Posting-Date: Mon, 21 Jul 2003 11:22:15 MDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!switch.ch!solnet.ch!solnet.ch!news.imp.ch!news.imp.ch!snoopy.risq.qc.ca!news.uunet.ca!newsfeed.grouptelecom.net!localhost!jsavard Xref: chonsp.franklin.ch alt.folklore.computers:146452 Thomas Womack (twomack@chiark.greenend.org.uk) wrote: : Some Itanium2 machines manage 3.5 to 4 GB/second. : Basically, a current desktop will beat one CPU of a Cray J90 on : anything, and will beat a whole Cray J916 on jobs that don't need : memory bandwidth. This is probably why J-series Crays show up on : ebay. I remember noting, at the time the original Pentium came out, that it was roughly comparable to a 360/195, since, like that machine, it had hardware floating-point with multi-bit logic, cache, and pipelining. But usually, home PCs don't easily interface to dozens of terminals, dozens of disk drives, and so on, although specialized attachments are available. The Itanium architecture looked a bit "supercomputer-like" to me as well. But I didn't have any close comparisons of performance. This certainly shows how fast progress has gone! John Savard ###### From: Eric Sosman Newsgroups: alt.folklore.computers Subject: Re: Rough equivalence: early Cray with today's micro Date: Mon, 21 Jul 2003 15:08:01 -0400 Organization: Sun Microsystems Lines: 19 Message-ID: <3F1C3A11.4469C8C9@sun.com> References: Reply-To: Eric.Sosman@Sun.COM NNTP-Posting-Host: tardis.east.sun.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Trace: news1brm.Central.Sun.COM 1058814481 13599 129.148.168.113 (21 Jul 2003 19:08:01 GMT) X-Complaints-To: usenet@news1brm.central.sun.com NNTP-Posting-Date: Mon, 21 Jul 2003 19:08:01 +0000 (UTC) X-Mailer: Mozilla 4.79C-CCK-MCD [en] (X11; U; SunOS 5.8 sun4u) X-Accept-Language: en Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!switch.ch!news.tele.dk!news.tele.dk!small.news.tele.dk!priapus.visi.com!news-out.visi.com!petbe.visi.com!feed.news.qwest.net!namche.sun.com!news1brm.central.sun.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:146554 jsavard@ecn.ab.ca wrote: > > Thomas Womack (twomack@chiark.greenend.org.uk) wrote: > : Some Itanium2 machines manage 3.5 to 4 GB/second. > > : Basically, a current desktop will beat one CPU of a Cray J90 on > : anything, and will beat a whole Cray J916 on jobs that don't need > : memory bandwidth. This is probably why J-series Crays show up on > : ebay. > > I remember noting, at the time the original Pentium came out, that it was > roughly comparable to a 360/195, since, like that machine, it had hardware > floating-point with multi-bit logic, cache, and pipelining. "Roughly comparable," I suppose, means "Within the accuracy limits of original Pentium arithmetic?" -- Eric.Sosman@sun.com ###### From: "Glen Herrmannsfeldt" Newsgroups: alt.folklore.computers References: <3F1C3A11.4469C8C9@sun.com> Subject: Re: Rough equivalence: early Cray with today's micro Lines: 24 X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2919.6700 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700 Message-ID: NNTP-Posting-Host: 12.207.204.17 X-Complaints-To: abuse@comcast.net X-Trace: rwcrnsc54 1058828716 12.207.204.17 (Mon, 21 Jul 2003 23:05:16 GMT) NNTP-Posting-Date: Mon, 21 Jul 2003 23:05:16 GMT Organization: Comcast Online Date: Mon, 21 Jul 2003 23:05:16 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!switch.ch!solnet.ch!solnet.ch!newsfeed.freenet.de!216.166.71.118.MISMATCH!small1.nntp.aus1.giganews.com!nntp.giganews.com!cyclone1.gnilink.net!wn14feed!wn13feed!worldnet.att.net!204.127.198.203!attbi_feed3!attbi.com!rwcrnsc54.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:146601 "Eric Sosman" wrote in message news:3F1C3A11.4469C8C9@sun.com... (snip) > > I remember noting, at the time the original Pentium came out, that it was > > roughly comparable to a 360/195, since, like that machine, it had hardware > > floating-point with multi-bit logic, cache, and pipelining. > > "Roughly comparable," I suppose, means "Within the > accuracy limits of original Pentium arithmetic?" Actually, the 360/91 and /195 don't satisfy the Principles of Operation for the floating point divide instructions. The architecture requires a truncated quotient, and such machines provide a more accurate rounded quotient. This is pretty much a direct result of the algorithm used. There should be no error more than 1 LSB, though. -- glen ###### From: ab528@freenet.carleton.ca (Heinz W. Wiggeshoff) Newsgroups: alt.folklore.computers Subject: Re: Rough equivalence: early Cray with today's micro Date: Tue, 22 Jul 2003 04:21:15 -0400 Organization: National Capital Freenet, Ottawa, Ontario, Canada Lines: 11 Sender: ab528@freenet.carleton.ca (Heinz W. Wiggeshoff) Message-ID: References: <3F1C3A11.4469C8C9@sun.com> NNTP-Posting-Host: p45.tc1.std.dialup.ncf.ca X-Trace: freenet9.carleton.ca 1058862160 24781 134.117.137.122 (22 Jul 2003 08:22:40 GMT) X-Complaints-To: complaints@ncf.ca NNTP-Posting-Date: 22 Jul 2003 08:22:40 GMT X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 6.00.2800.1158 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1165 X-Given-From: "Heinz Wiggeshoff" Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!switch.ch!solnet.ch!solnet.ch!newsfeed.freenet.de!eusc.inter.net!priapus.visi.com!news-out.visi.com!petbe.visi.com!nntp1.roc.gblx.net!nntp.gblx.net!nntp.gblx.net!xcski.com!freenet-news!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:146663 "Eric Sosman" wrote in message news:3F1C3A11.4469C8C9@sun.com... > jsavard@ecn.ab.ca wrote: > "Roughly comparable," I suppose, means "Within the > accuracy limits of original Pentium arithmetic?" Are you suggesting that 2 + 2 is not 3.999999 ? Shame on you. And your accountant too!