burzio@mmlai.UUCP (Tony Burzio) (12/10/90)
>I've posted this so amny other places, I might as well post it here >too. We have $150,000 to spend on hardware for a computer >graphics visualization system. Currently, we have an Apollo netwrok >with about 25 nodes. We're looking to do animation of finite >element analysis results and CFD stuff. Obviously, at this >dollar level we'll get a lot of CPU speed. > >What would you buy? HP, Apollo, SGI, SUN, STARDENT? How would >you configure your dream system. Why, HP or course! Regardless of the abysmal marketing ploys by HP (i.e. none :-), the 835 TurboVRX is a real corker! Running SDRC IDEAS in X Window System mode, you should have your socks blown off! Nice animations too, and they are VERY fast. If you can wait, the "snake" replacement for the 800 series, around the spring(ish), should give you alarming speed (50+ MIPS?)! If you really want mucho bang-for-the-buck, get a 400 TurboVRX. As fast as a Sparc (until the $2K 68040 upgrade arrives, then twice as fast) but more capable because most of the processing of graphics is offloaded to the TurboVRX. Some notes on the 835: This processor is rated around 12 HP MIPS, or 15 the way Sun counts them. The difference between a Sparc and the 835 is multi-user access (or one user doing two things, like the 37 processes that SDRC spawns off while X is running :-) The Sparc will die at process two, because of hardware limitations (see old .arch postings for details), while the 835 will chug along unhindered while around 8 CPU burdened jobs are running. Another marketing secret at HP is the stability of HP-UX, HPs' UNIX variant. Very easy to use, and it works great (as long as you use Suns' manuals to figure out UUCP :-) The local HP support (at least for the moment, the recession may change things) is SUPERB!!!! Buy from HPs' remarketing or demo sales. You usually get 50% off on currently produced systems. A computer that has run for a while is much more reliable than a "new" one. We got 2 GB of disk for the price of one. All our "used" parts are under contract and are still running after three years. ********************************************************************* Tony Burzio * Time to SKI!!! Martin Marietta Labs * mmlab!burzio@uunet.uu.net * *********************************************************************
bb@reef.cis.ufl.edu (Brian Bartholomew) (12/11/90)
In article <54@gauss.mmlai.UUCP> you write: > I've posted this so amny other places, I might as well post it here > too. We have $150,000 to spend on hardware for a computer graphics > visualization system. Currently, we have an Apollo netwrok with about > 25 nodes. We're looking to do animation of finite element analysis > results and CFD stuff. Obviously, at this dollar level we'll get a lot > of CPU speed. > What would you buy? HP, Apollo, SGI, SUN, STARDENT? How would you > configure your dream system. Usually I agree with what Tony says, but in this case I almost completely disagree. Here's why: The *first* rule of computer purchasing is to select the software you want to run, then find a computer that runs it. Fast hardware is useless without powerful and flexible software to use it. *Useless*! For a start, compare the size of Sun's SPARCware catalog (about 1200 pages, three applications per page) with HP's equivalent software catalog. If you can find a close enough match and are happy with the HP selections, fine. Just remember that you are choosing from a substantially smaller pool of applications than you would with Sun. [Tony paraphrased: The HP 800 series with graphics accelleration is the fastest machine this week] Let's see SPECmarks, not MIPS! What do MIPS mean when comparing between RISC (SPARC) and CISC (800?) architectures? Let's see performance comparisons between equally-priced Sun and HP configurations. [Tony paraphrased: The 835 deals better with multiple CPU-intensive users on the same machine] I have no data to indicate if this is true or not, and in fact sounds like a tuning problem (i.e. don't run an animation machine on 8 megs of RAM). Assuming that it is real, is this capability useful, given that this guy needs a head for each seat, to display animation on? > Another marketing secret at HP is the stability of HP-UX, HPs' UNIX > variant. Very easy to use, and it works great (as long as you use > Suns' manuals to figure out UUCP :-) The local HP support (at least > for the moment, the recession may change things) is SUPERB!!!! As long-time readers of this group know, this is one of my favorite complaints about HP workstations. HP-UX is consistantly behind SunOS in terms of features, most notably networking features. Within the first week of getting our new HP 9000/345 workstations with the pre-installed software (including X Release 2 - yum-yum), we had managed to crash or hang 3 of the 7 machines, at least once. At the time, we were taking particular care not to do anything unreasonable on them, as we didn't have backups of the disk (and the OS tapes failed upon installation). I don't know about Tony's area, but at the University of Florida (the premier Florida Engineering University, perhaps one of the big 3 or 4 in the Southeast) the HP reps are incredibly uninformed about the workstation product line. It is extremely difficult to impossible to get such things as patch tapes out of our sales reps. The telephone-based tech support is much better, but usually we have questions that go over their heads. And,they keep referring us to the local reps for patch tapes (!) The local Sun reps are rather more clued-in, but perhaps not overwhelmingly so. However, the existance of Sun-manager mailing lists, informal ftp sites for patches, and other such net resources makes expert support just 8 hours and a mail message away. I know that HP is attempting to start participating in this milieu, but they are very far from there yet. Besides, it seems that a very large percentage of the software posted on the net has been developed on Sun 3 or 4 computers. That saves an awful lot of porting time when that hot new piece of free data- visualization software comes NNTP'ing down the net. Examples: KHOROS (~100 Megs of source), NCSA-something. [Tony paraphrased: buy cheap, slightly used or demo] I agree, as long as you get the warranty, like he suggests. One last note: anyone who cares to dispute these points should attack my facts, not my attitude. -- "Any sufficiently advanced technology is indistinguishable from a rigged demo." ------------------------------------------------------------------------------- Brian Bartholomew UUCP: ...gatech!uflorida!mathlab.math.ufl.edu!bb University of Florida Internet: bb@math.ufl.edu
jlol@REMUS.EE.BYU.EDU (Jay Lawlor) (12/11/90)
>>>>> On 11 Dec 90 03:35:47 GMT, bb@reef.cis.ufl.edu (Brian Bartholomew) said: Brian> In article <54@gauss.mmlai.UUCP> you write: > I've posted this so amny other places, I might as well post it here > too. We have $150,000 to spend on hardware for a computer graphics > visualization system. Currently, we have an Apollo netwrok with about > 25 nodes. We're looking to do animation of finite element analysis > results and CFD stuff. Obviously, at this dollar level we'll get a lot > of CPU speed. > What would you buy? HP, Apollo, SGI, SUN, STARDENT? How would you > configure your dream system. Brian> Usually I agree with what Tony says, but in this case I almost Brian> completely disagree. Here's why: Brian> The *first* rule of computer purchasing is to select the software you Brian> want to run, then find a computer that runs it. Fast hardware is Brian> useless without powerful and flexible software to use it. *Useless*! True. Brian> For a start, compare the size of Sun's SPARCware catalog (about 1200 Brian> pages, three applications per page) with HP's equivalent software Brian> catalog. If you can find a close enough match and are happy with the Brian> HP selections, fine. Just remember that you are choosing from a Brian> substantially smaller pool of applications than you would with Sun. Sun's only saving grace. Brian> [Tony paraphrased: The HP 800 series with graphics accelleration is Brian> the fastest machine this week] Brian> Let's see SPECmarks, not MIPS! What do MIPS mean when comparing Brian> between RISC (SPARC) and CISC (800?) architectures? Let's see Brian> performance comparisons between equally-priced Sun and HP Brian> configurations. Since when was the 800 series CISC? Sounds like you are a little less informed than you could be. > Another marketing secret at HP is the stability of HP-UX, HPs' UNIX > variant. Very easy to use, and it works great (as long as you use > Suns' manuals to figure out UUCP :-) The local HP support (at least > for the moment, the recession may change things) is SUPERB!!!! Brian> As long-time readers of this group know, this is one of my favorite Brian> complaints about HP workstations. HP-UX is consistantly behind SunOS Brian> in terms of features, most notably networking features. HP's getting better in this area. You can't catch up all at once. Otherwise you'd have bug distribution lists the size of Sun's lists. Brian> Within the first week of getting our new HP 9000/345 workstations with Brian> the pre-installed software (including X Release 2 - yum-yum), we had Brian> managed to crash or hang 3 of the 7 machines, at least once. At the Brian> time, we were taking particular care not to do anything unreasonable Brian> on them, as we didn't have backups of the disk (and the OS tapes Brian> failed upon installation). Well, we have R3 and are running an R4 server from HP that works great. Brian> I don't know about Tony's area, but at the University of Florida (the Brian> premier Florida Engineering University, perhaps one of the big 3 or 4 Brian> in the Southeast) the HP reps are incredibly uninformed about the Brian> workstation product line. It is extremely difficult to impossible to Brian> get such things as patch tapes out of our sales reps. The Brian> telephone-based tech support is much better, but usually we have Brian> questions that go over their heads. And,they keep referring us to the Brian> local reps for patch tapes (!) The local Sun reps are rather more Brian> clued-in, but perhaps not overwhelmingly so. However, the existance Brian> of Sun-manager mailing lists, informal ftp sites for patches, and Brian> other such net resources makes expert support just 8 hours and a mail Brian> message away. I know that HP is attempting to start participating in Brian> this milieu, but they are very far from there yet. I guess it depends on where you live. The ee dept here won't buy any more Suns because of the "quality" of support they received. Our HP support has been much better. Brian> One last note: anyone who cares to dispute these points should attack Brian> my facts, not my attitude. Of course. -Jay
paul@actrix.gen.nz (Paul Gillingwater) (12/12/90)
In article <25849@uflorida.cis.ufl.EDU> bb@reef.cis.ufl.edu (Brian Bartholomew) writes: > One last note: anyone who cares to dispute these points should attack > my facts, not my attitude. OK here goes... > Let's see SPECmarks, not MIPS! What do MIPS mean when comparing > between RISC (SPARC) and CISC (800?) architectures? Let's see The Series 800 is based on HP-PA, which is RISC. The series 300 and 400 are 680x0, which is CISC. I guess the keywords *are* correct. -- Paul Gillingwater, paul@actrix.gen.nz
bb@sandbar.cis.ufl.edu (Brian Bartholomew) (12/12/90)
jlol@REMUS.EE.BYU.EDU (Jay Lawlor) & paul@actrix.gen.nz (Paul Gillingwater) > The Series 800 is based on HP-PA, which is RISC. The series 300 and > 400 are 680x0, which is CISC. Oops. Foot-in-newsfeed syndrome. The comparison I drew was between Sun 4 (RISC) and 800 (RISC, I now learn). I still want to see SPECmarks between equally-priced systems. jlol@REMUS.EE.BYU.EDU (Jay Lawlor) > HP's getting better in this area. You can't catch up all at once. > Otherwise you'd have bug distribution lists the size of Sun's lists. I have heard reports that HP is funding some academic institution in Utah to port true BSD to the 400 line. Jay, could this perhaps be your institution? If so, are you able (legally and properly) to make a progress report to this newsgroup? > Well, we have R3 and are running an R4 server from HP that works > great. All of our (CIS, Math) Suns are completely running R4. This would be an empty point save that I am told the R3->R4 upgrade mainly consisted of bug fixes. The (non-)reliability of R2 on the 345's seems to suggest that as well. > I guess it depends on where you live. The ee dept here won't buy any > more Suns because of the "quality" of support they received. Our HP > support has been much better. Uh, what was your rep's name again? :-) -- "Any sufficiently advanced technology is indistinguishable from a rigged demo." ------------------------------------------------------------------------------- Brian Bartholomew UUCP: ...gatech!uflorida!mathlab.math.ufl.edu!bb University of Florida Internet: bb@math.ufl.edu
fmonaldo@aplcomm.JHUAPL.EDU (Monaldo Francis M. S1R x8648) (12/12/90)
Arguments between buying SUN vs HP computers are interesting, because we were faced with this same dilemma 2 years ago. In the past, for technical computing we had purchased HP computers. We were given significant funds to set up a small network of workstations and could choose between SUN's and HP's. At the time we were working with researchers at Harvard, who used SUNs. The questions at the time was do we jump aboard the SUN bandwagon or do we maintain our long satisfying relationship with HP. We did both. We purchased an HP9000 Series 360 as combination fileserver/workstation an Seres 319 workstation. About 1 Gbyte of hard disc space was connected to the HP server. On the SUN side, we purchased 2 SUN 3/60's with enough disc space to boot up and swap with, and little else. The HP disc was mounted onto the SUN's file system via NFS. After a couple of years, here is our experience. (1) It is amazing to me that the systems work together well. The HP file server and SUN workstations communicate with no problems. (2) Although significantly more software is available for the SUN's, most software we use is available on both. It is, however, a pain to maintain two sets of software. (3)The SUN's came with more tools such as an editor. In time, we were able to accumulate the necessary software for HP. For example, we bought HP's visual editor, which works great. (4) In general the CPU speeds have not been a problem, we have generally been RAM and I/O limited in our processing. We are just now getting to the point that the systems have enough RAM. This problem was particularly acute for the HP's because X-window is more of a RAM hog than Sun View. (5) In the long term, I believe our hardware investment in the HP equipment has better supported. SUn is, in effect, dumping its 68000 support and going entirely to SPARC. This means we either radically upgrade our hardware or live with our current systems. HP is suppoting both the 68000 machines and the HP-PA RISC systems with an essentially unifie operating system.d Upgrades can be made more incremently. Since HP follows the standards, instead of trying to drive them as much, with the HP equipment, we have always had X-windows. Any comments to the above would be appreciated. Let it be said that my bias has been with the HP machines, but in retrospect, I think we would have been better off with one vendor (either), less sysstem administration headaches. Frank Monaldo fmonaldo@aplcomm.jhuapl.edu
jbb@hpcvlx.cv.hp.com (Jim B. Byers) (12/13/90)
Regarding the X server, HP shipped the following servers as products: HP-UX 6.2 server - R2 based (As far as we can tell the first major Un*x system vendor to ship an X11 product) - sold as separate product as it came out after 6.2 shipped HP-UX 6.5 server - R2 based (as far as we can tell the first server to ship as a product that allowed 3D graghics programs to utilize 3D h/w within an X window) - bundled with Series 300 HP-UX HP-UX 7.0 server - R3 based with many R4 fixes (included the first shipments of Motif by any major vendor) - bundled HP-UX 7.03 server - R4 based - bundled Post 7.03 server available via anonymous ftp on hpcvaaz (15.255.72.15). This is the 7.03 server made to run fast on the 98550 (1280x1024) graphics card. This allows this card to run as fast as the Series 400 color VRX. This server is supported on 7.0 and 7.03. Get it if you have this card. Do not confuse the vendor release number (you get this with some kind of query to the server) with the version of the software (R2, R3 etc.) For MIT these match because their "vendor release number" *is* the MIT release number. But for any other vendor these can have no relation. Jim Byers Interface Technology Operation Marketing/Lab Team "The X11/Motif/Vue/Architect folks in Corvallis Or."
mike%jaguar.utah.edu@cs.utah.edu (Mike Hibler) (12/13/90)
> jlol@REMUS.EE.BYU.EDU (Jay Lawlor) > > > HP's getting better in this area. You can't catch up all at once. > > Otherwise you'd have bug distribution lists the size of Sun's lists. > > bb@sandbar.cis.ufl.edu (Brian Bartholomew) > > I have heard reports that HP is funding some academic institution in > Utah to port true BSD to the 400 line. Jay, could this perhaps be > your institution? If so, are you able (legally and properly) to make > a progress report to this newsgroup? At the University of Utah (you know, the "other" football team in Utah) we have been involved in BSD work on the HPs for quite some time and, more recently, Mach work as well. A synopsis and then a detailed list follows. It should be noted that all of this work has been generously supported (or at least condoned :-) by HP. Here is the Reader's Digest version for those in a hurry, a more detailed account follows: 300/400 4.3 BSD in production use, available from Utah or Mt Xinu 300/400 4.4 BSD under development, support will be on 4.4 tape 300/400 Mach 2.5 in use at Utah, no distribution (yet) 300/400 Mach 3.0 will probably happen Real Soon Now 800 4.3 BSD in use at Utah, cannot distribute right now 800 Mach 3.0 just starting to run, hope to distribute and a couple of other items of interest: 800 GCC robust, fairly good code, ftp from jaguar.utah.edu 800 GAS/GDB running, will be distributed when ready [ details follow ] First, the 300/400 680x0 series work. Even though our emphasis has shifted to the 800 PA-RISC machines, we still do a significant amount of work here since we have over 100 320/350/340/360/370/345/375/425/433 machines in the department. 1. 4.3 BSD. This started in 1987 with a 4.3 port to the 200 series done at Berkeley. We added 300 support, more device drivers, HP-UX binary compatibility, and integrated the 4.3bsd+NFS (Sun's NFS 3.0 in 4.3) work done at Wisconsin. We have continued to evolve this system with significant code contributions from other Universities and individuals and it is the operating system run on the majority of our 300/400 systems. It runs on all 68020 and 68030 based machines except the 332 and support is planned for the 68040 machines. It contains no HP proprietary code but does contain both AT&T and Sun code, the latter being NFS. GCC and GAS are used to build all kernel and user code. We do distributions, but the latest (called HPBSD 1.7) is out-of-date with respect to what we run internally. Mt Xinu also has a distribution of their More/BSD product that is based on this work but with significant added features (e.g. NFS 4.0, diskless support). 2. 4.4 BSD. This is work in progress and is a joint effort of ourselves and CSRG at Berkeley. This will be, by definition, "pure BSD" unencumbered by any commercial entity other than possibly AT&T. 300 series support did go out on the 4.3-Reno tape but only in source form, making it rather difficult to bootstrap if you don't already have Utah or Mt Xinu BSD running. The Reno tape does not include some late changes necessary to run on 400 series machines though we can provide them. HP 300s have been the development platform (currently still the only platform) for the "new VM" (mmap and friends) that will appear in 4.4bsd (note: it is not in 4.3-Reno). 4.4+NEWVM is running on one machine at Utah (mine) and a couple at Berkeley. 3. Mach 2.5. Funded by the OSF, this has turned out to be mostly an experiment as the focus of our work has shifted to Mach 3.0. It is a spin off from the 4.3 and NEWVM (which is Mach based) work with some influence by an earlier 2.0 port done internally at HP Labs. This was primarily a kernel port, the Mach kernel runs in our normal BSD user environment (says a lot for the Mach BSD emulation). We have it running on a couple of machines here at Utah. If there is sufficient interest, we may put together a "wizard" (i.e. very rough) distribution. 4. Mach 3.0. This hasn't started yet, but probably will soon. (funded by the OSF?) Should be straight forward given our long involvement with the 300 series and our recent 3.0 work on the 800 (described later). This should result in an unencumbered pure-kernel that might be useful to various people (among them, the FSF which has HPs and could use them to develop the GNU OS). Now for the 800 PA-RISC series machines. This is where we have focused our attention lately. In addition to OS work described here, we have also done a lot of work on GCC, GAS and GDB for the PA-RISC. 1. 4.3 BSD. Has been ported to the 835 (should run on the 834 as well). It is robust, however it's far less complete than the 300 BSD port. It currently uses modified HP-UX drivers and can not be distributed as such. It supports only the 6-port MUX, LAN and HP-IB (on the CIO bus). Besides writing new, distributable drivers, there is plenty of general clean-up work yet to be done. The kernel is still built with the HP-UX compiler running in compatibility mode though it has been built using GCC/GAS (the resulting kernel runs "for awhile"). The majority of the user code is built using GCC/GAS. User-level utility sources are common between the 300 and 800, the kernel trees are still separate. We have 4 835s, including our major 800 development machine, running BSD. As a result of our continuing Mach work, we should be able to replace the device drivers and old BSD VM to complete 835 support suitable for 4.4 BSD. We would like to get this support into the official Berkeley release. We have an 832, but are not anxious to take on another bus architecture (NIO) with its associated drivers, hence no work has been done. 2. Mach 3.0. Our latest, and by far most involved, effort. The current work is with the pure-kernel and UN*X single server. The kernel first ran (well, "idled") at the end of October, we got our first shell prompt about two weeks ago, and we could finally execute interesting binaries (BSD not HP-UX) last week (two days before the HP site visit, what timing! :-) Lots and lots remains to be done. We plan to use no HP-UX code, at least in the kernel/server(s)/emulator. We have already redone the VM code (pmap module, TLB miss handling) and MUX driver, are working on a disk driver (current Mach testing is done using a memory disk) but still need LAN and display drivers. We do plan to make a release of this, but I have no idea when this will be. We would also like to see this (as well as the 68k work) get back into the main Mach distribution at CMU. A couple of notes on the 800 Mach work. There was a project at HP to experiment with replacing HP-UX's VM with Mach's as well as adding Mach threads and ports to HP-UX. We have used this work, known as "Tut", as a reference (and for drivers). Also, our work is unrelated to what is being done at the OSF and HP/Apollo for OSF/1. Mike Hibler mike@cs.utah.edu ...!utah-cs!mike
mjs@hpfcso.HP.COM (Marc Sabatella) (12/14/90)
Another "fact" that needs clarification is the inferrence that somehow SPEC numbers are more "meaningful" than MIPS numbers when comparing RISC vs. CISC. This would be true only if MIPS really did mean "Millions of Instructions Per Second". However, no one actually calculates MIPS ratings that way any more. They are calculated *exactly* the same way as SPEC numbers are - ie, you run a few benchmarks, divide by the time the same benchmarks take on a VAX, and produce some sort of average. RISC/CISC is entirely orthogonal to this. SPEC ratings differ from MIPS rating in three important respects. The MIPS benchmarks are almost entirely integer, almost entirely C, and almost entirely "toys" (ie, Dhrystone). The SPEC suite is more floating point intensive and uses a lot of Fortran, and includes more "real" applications. Neither suite tells you much about overall system performance - except for gcc in SPEC, none do much I/O, or use much VM; none attempt any graphics, networking, etc. In theory, because of the way the numbers are generated, SPECmarks would always equal MIPS ratings. This is almost never true in practice, though - MIPS numbers are artificially inflated because several of the benchmarks in the suite are unrealistic "meatballs" for an optimizer to speed up. -------------- Marc Sabatella (marc@hpmonk.fc.hp.com) Disclaimers: 2 + 2 = 3, for suitably small values of 2 Bill and Dave may not always agree with me
mjb%hoosier.utah.edu@cs.utah.edu (Mark Bradakis) (12/15/90)
As one who works at the U of Utah, I may be a bit biased in terms of
comparing Sun workstations to HP units. I will, however, offer this one
observation.
We have a student lab with 40 machines in it. Half are HP 400t machines
and half are Suns, a mix of Sparcstation 1 and SLCs or some such. Usually
when I go over there and the lab is only half used, nearly all the students
are sitting in front of the Hewlett Packard machines, only one or two are
using Suns. Remember that TV commercial where the two people are asking
"Which is the best computer?" It is indeed the one that people use.
But to get back to the original poster's question, about a high end graphics
workstation, have you looked into the Evans & Sutherland ESV?
mjb.
"Come on Dad, gimme the car tonight..."
mjb@hoosier.utah.edujbb@hpcvlx.cv.hp.com (Jim B. Byers) (12/15/90)
I am reposting this because I understand that it got garbled by the time it reached some sites. Also It seems to have gotten posted in two spots. Please excuse if this is redundant or due to an error I made. ---------------------- Regarding the X server, HP shipped the following servers as products: HP-UX 6.2 server - R2 based (As far as we can tell the first major Un*x system vendor to ship an X11 product) - sold as separate product as it came out after 6.2 shipped HP-UX 6.5 server - R2 based (as far as we can tell the first server to ship as a product that allowed 3D graghics programs to utilize 3D h/w within an X window) - bundled with Series 300 HP-UX HP-UX 7.0 server - R3 based with many R4 fixes (included the first shipments of Motif by any major vendor) - bundled HP-UX 7.03 server - R4 based - bundled Post 7.03 server available via anonymous ftp on hpcvaaz (15.255.72.15). This is the 7.03 server made to run fast on the 98550 (1280x1024) graphics card. This allows this card to run as fast as the Series 400 color VRX. This server is supported on 7.0 and 7.03. Get it if you have this card. Do not confuse the vendor release number (you get this with some kind of query to the server) with the version of the software (R2, R3 etc.) For MIT these match because their "vendor release number" *is* the MIT release number. But for any other vendor these can have no relation. Jim Byers Interface Technology Operation Marketing/Lab Team "The X11/Motif/Vue/Architect folks in Corvallis Or."
bb@beach.cis.ufl.edu (Brian Bartholomew) (12/15/90)
In article <1990Dec14.095519.14818@hellgate.utah.edu> mjb%hoosier.utah.edu@cs.utah.edu (Mark Bradakis) writes: > We have a student lab with 40 machines in it. Half are HP 400t > machines and half are Suns, a mix of Sparcstation 1 and SLCs or some > such. Usually when I go over there and the lab is only half used, > nearly all the students are sitting in front of the Hewlett Packard > machines, only one or two are using Suns. Remember that TV commercial > where the two people are asking "Which is the best computer?" It is > indeed the one that people use. Well, this about a content-free posting as you can get. Not a word about features, performance, reliability, or price; just sitting there with a big grin and quoting television ads at me. At this rate I should be using Windows, because that's what Microsoft reccommends. Here are some easy questions to ask, that might clarify the situation: Are the HP's color and the Sun's mono (SLC's are)? Is there software used locally that only runs on the HP's? Is the local support and system defaults concentrated on the HP's? "Any sufficiently advanced technology is indistinguishable from a rigged demo." ------------------------------------------------------------------------------- Brian Bartholomew UUCP: ...gatech!uflorida!mathlab.math.ufl.edu!bb University of Florida Internet: bb@math.ufl.edu -- "Any sufficiently advanced technology is indistinguishable from a rigged demo." ------------------------------------------------------------------------------- Brian Bartholomew UUCP: ...gatech!uflorida!mathlab.math.ufl.edu!bb University of Florida Internet: bb@math.ufl.edu
mikeg@dali.gatech.edu (Mike Gourlay) (12/18/90)
In article <BB.90Dec15033453@beach.cis.ufl.edu>, bb@beach.cis.ufl.edu (Brian Bartholomew) writes: |> In article <1990Dec14.095519.14818@hellgate.utah.edu> |> mjb%hoosier.utah.edu@cs.utah.edu (Mark Bradakis) writes: |> |> > We have a student lab with 40 machines in it. Half are HP 400t |> > machines and half are Suns, a mix of Sparcstation 1 and SLCs or some |> > such. Usually when I go over there and the lab is only half used, |> > nearly all the students are sitting in front of the Hewlett Packard |> > machines, only one or two are using Suns. Remember that TV commercial |> > where the two people are asking "Which is the best computer?" It is |> > indeed the one that people use. |> |> Here are some easy questions to ask, that might clarify the situation: |> |> Are the HP's color and the Sun's mono (SLC's are)? |> Is there software used locally that only runs on the HP's? |> Is the local support and system defaults concentrated on the HP's? Some more such questions: Does speed not matter? The Sparcs are much (10x?) faster. Are the HP's the only machines with local drives? If the Suns are using remote-mounted drives (diskless Suns), then they'll seem slower. Do you do anything with multi-threaded processing? Do you use/program UNIX as part of the projects that use the machines? Do the students program X? Do they use X?
bijal@hpopd.pwd.hp.com (Bijal Shah) (12/18/90)
So a Sun Sparc 1 runs at over 80 MIPS then? Are you sure? A Sparc 1 runs at between 8 and 12.5 MIPS. (I'm not certain) A HP 400t runs over 8 MIPS. Compare. Please do not make referances to speed differences. There is virtually no difference. If you use sparcs, fine, be happy with them. If you use HPs, fine enjoy them too. There are too many squabbles like this betweem users of any computer (Mine's bigger than your's - So? Mine's faster then your's - So? Mine's got better graphics than your's). It just gets a bit ridiculous after a while does it not? BJ
chan@hpfcmgw.HP.COM (Chan Benson) (12/19/90)
> Does speed not matter? The Sparcs are much (10x?) faster.
I'm willing to admit some advantages to Sun, but the Sparcstation I is
about the same performance as the 400t. At the worst the multiplier is
some number greater than one and less than two. It's certainly nowhere
near ten.
-- Chan
If all the false statistics presented on usenet in a year were placed
end to end they would circle the equator 50,000 times.chan@hpfcmgw.HP.COM (Chan Benson) (12/19/90)
Previously, I posted this and then tried to delete too late... >> Does speed not matter? The Sparcs are much (10x?) faster. > >I'm willing to admit some advantages to Sun, but the Sparcstation I is >about the same performance as the 400t. At the worst the multiplier is >some number greater than one and less than two. It's certainly nowhere >near ten. > > -- Chan > >If all the false statistics presented on usenet in a year were placed >end to end they would circle the equator 50,000 times. OOOOOOOOOPSSSSSSSSSSSS! And I'm doing my part to make that happen. I should have qualified my performance statement with the word "integer". The SparcStation I is 3 to 4 times faster than the 400t on floating point benchmarks. The 68040 upgrade will pretty much close that gap. -- Chan
chan@hpfcmgw.HP.COM (Chan Benson) (12/20/90)
Well, I seem to be having some problems with getting my "remove foot from mouth" posting posted. So if you see this multiple times, I apologize. In a previous note, I wrote... >> Does speed not matter? The Sparcs are much (10x?) faster. > >I'm willing to admit some advantages to Sun, but the Sparcstation I is >about the same performance as the 400t. At the worst the multiplier is >some number greater than one and less than two. It's certainly nowhere >near ten. > > -- Chan > >If all the false statistics presented on usenet in a year were placed >end to end they would circle the equator 50,000 times. I inadvertantly added a few yards of misinformation of my own. I should have qualified the above statement with the word "integer". Floating point performance on the SparcStation is about 3-4 times faster than a 68030 based 400t. The 68040 upgrade narrows that gap substantially. -- Chan
irf@kuling.UUCP (Bo Thide') (12/23/90)
In article <17780001@hpfcmgw.HP.COM> chan@hpfcmgw.HP.COM (Chan Benson) writes: >Previously, I posted this and then tried to delete too late... > >>> Does speed not matter? The Sparcs are much (10x?) faster. >> >>I'm willing to admit some advantages to Sun, but the Sparcstation I is >>about the same performance as the 400t. At the worst the multiplier is >>some number greater than one and less than two. It's certainly nowhere >>near ten. >> >> -- Chan >> >>If all the false statistics presented on usenet in a year were placed >>end to end they would circle the equator 50,000 times. > >OOOOOOOOOPSSSSSSSSSSSS! > >And I'm doing my part to make that happen. I should have qualified >my performance statement with the word "integer". The SparcStation I >is 3 to 4 times faster than the 400t on floating point benchmarks. > >The 68040 upgrade will pretty much close that gap. > > -- Chan To put this discussion into perspective, here are the MIPs and MFlops ratings as published by HP: Model SPU/Clock speed MIPs MFlops -------------------------------------------- 425t MC68040/25 MHz 20 3.5 433s MC68040/33 MHz 26 4.5 Both the 425 and the 433 were initially shipped with an 68040 emulator card, based on the MC68030/50 MHz delivering 12 MIPs and 0.5 MFlops. Sun are known to sometimes overestimate their MIPs rating by 10-20% and to use the inflated arithmetic mean of the Single and Double Precision LINPACK MFlops rating whereas HP always has used the uninflated Double Precision LINPACK values. I hope HP keeps their more honest policy. Bo ^ Bo Thide'-------------------------------------------------------------- |I| Swedish Institute of Space Physics, S-755 91 Uppsala, Sweden |R| Phone: (+46) 18-303671. Telex: 76036 (IRFUPP S). Fax: (+46) 18-403100 /|F|\ INTERNET: bt@irfu.se UUCP: ...!uunet!sunic!irfu!bt ~~U~~ -----------------------------------------------------------------sm5dfw
jan@bagend.uucp (Jan Isley) (12/24/90)
irf@kuling.UUCP (Bo Thide') writes: >Sun are known to sometimes overestimate their MIPs rating by 10-20% and to >use the inflated arithmetic mean of the Single and Double Precision >LINPACK MFlops rating whereas HP always has used the uninflated Double >Precision LINPACK values. I hope HP keeps their more honest policy. Sun also uses compilers that cost *major* dollars more than the standard ones to get these numbers. -- Jan Isley jan@bagend {known universe}!gatech!bagend!jan (404)434-1335