Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uwm.edu!linac!unixhub!stanford.edu!leland.Stanford.EDU!leland.Stanford.EDU!fangchin From: fangchin@leland.Stanford.EDU (Chin Fang) Newsgroups: comp.unix.sysv386 Subject: Re: Performance of 80486 based machines running Unix Keywords: performance 80486 486 unix Message-ID: <1991Jun22.064103.15333@leland.Stanford.EDU> Date: 22 Jun 91 06:41:03 GMT References: <1587@caslon.cs.arizona.edu> <1991Jun20.233523.8184@leland.Stanford.EDU> <1991Jun22.015603.14718@virtech.uucp> Sender: news@leland.Stanford.EDU (Mr News) Organization: AIR, Stanford University, CA 94305 USA Lines: 71 In article <1991Jun22.015603.14718@virtech.uucp>, cpcahil@virtech.uucp (Conor P. Cahill) writes: |> fangchin@leland.Stanford.EDU (Chin Fang) writes: |> |> >As someone who takes care of RISC 6000s on a daily basis, I venture to |> >give you my own opinions, I post it instead of mailing because I believe |> >this is of general interest. As RISC 6000 low end model is on par with |> >high end 33Mhz 486s... I should have said that price-wise, after educational discount, a RISC 6000 320 is on par with high end name brand 33Mhz 486s (nowadays, for a 320H with 355 MB Maxtor SCSI II drive, a school can get it around 11K) |> One of our clients has a pair of 6000s. The low end model fees about |> the speed of our 33Mhz 386. The other model is about 30% faster. That's a bit an exaggeration I am afraid. My 320 is NFS mounted to a SUN 4, and I think it's definitely lively than a 33 Mhz 386, which I have seen many. BTW, your client may need to upgrade his/her machines OS to 3005/2006. The earlier ones are so terrible even IBMers hated 'em (I do too as I suffered from 3001 to current patch level, and in just three months! AIX 3.1 is definitely of pre-alpha quality, no joke here) This is a text book example that lousy software ruined superlative hardware. |> (They are now talking about getting a new disk drive and IBM has quoted |> them a price of $4,500 or so for a 330MB SCSI drive) I can understand this, IBM RISC binaries are FAT! X+unix need AT LEAST 16 Megs RAM to run without swapping. As most readers of this group know, you can run 386 Unix with X11R4 on top in 8 megs quite comfortably, at least when there is only one user. I once posted a joke in comp.unix.aix conjecturing that IBM is going to produce the first kernel that won't boot with less than 100 Megs RAM :-) [RISC 6000 won't boot with 8 megs, I know this for a fact] The drives are OEMed from Maxtor, I don't understand why BIG BLUE like to rip off people like this, shame on it. These drives are no different from Maxtors used in say, your favorite 3/486 boxes. |> >|> - Have enough power left over for me (the sysadmin) to handle |> >|> some heavy programming chores without having to step out for |> >|> a cup of coffee everytime I run a make. |> |> The 6000s do a fairly good job compiling, but when it comes to linking |> they are as slow as som 286s. If we run two links at the same time, |> the entire machine grinds to a hault. The xlc compiler generates pretty good error msgs, linking is slow but per AIX developers next door at IBM PALO ALTO, they are working on it (finger crossed). Executables generated in general is fairly fast executing, as far as my own apps are concerned. |> I'm not saying the 6000s are dogs, just that they aren't screamers either. |> They are screamers once you run floating point things in memory without touching hard disks. Definitely FOUR TIMES faster than your everyday 33Mhz 486s. CISC is woefully inadequte in this regard. When I was asked to enrich our RISC 6000s about three months ago, I hated 'em. I always rlogined into my RISC 6000 320 from my SUN SPARC 1+(I have both on my desk), then I started running my own research codes, floating point intensive ones, on my 320, gradually I developed liking to 'em. Now I sit in front of mine 320 all day long. I even have gone so far as to port so many things to it that it's /usr/local/bin is "richer" than our SUNs in some ways, to my head SA's dismay :-). I haven't touched my SPARC 2 for a month by now. Slow number crunchers don't get my attention :-) I am a mechanical engineer first, so most things I do have something to do with number crunching, I guess that's why even with it's awful OS/lousy R3 server (pretty fast with a 50 MHz TI34010 however), I now like my 320. I guess most UNIX programmers don't user floating point and if that's the case, AIX3.1+X suck. Don't get 'em (hardware is fine however, other than the flickering SONY Trinitron flat screen monitor) Sincerely, Chin Fang Mechanical Engineering Department Stanford University fangchin@leland.stanford.edu