> Compare with Sparc 20, Sparc 10 and  Sparc 5 which one is > > a new model and which is more powerful ?
 
> They're all reletively old, discontinued models, but you've > listed them in power (potential) order.

    There are a variety of optional features available for these computers. The oldest is the SparcStation 10. The SparcStation 20 is the new improved version of the 10, while the SparcStation 5 is the simplified low cost version of the 20. There is also the SparcStation 4 which is a not too popular cheap version of the 5. 

    I guess by "power" you mean CPU speed and cache size. This is not an easy question to answer since this power is set by plug-in modules in the 10 and 20. The SS5 comes in old and new versions with different speed CPU's. The CPU modules for the 10 and 20 are called mbus modules. There are many versions of these. Not all of them will work in the 10 or 20 SparcStations. The first SS10 had slow mbus modules that were supposed to be replaced by Sun when they got their production line for over 30 MHz CPU's running well. The fastest mbus modules for the SS20 is 200 MHz; the fastest for the SS10 is 180 MHz. These speeds are still pretty good so they are not available for sale at bargain prices. It is much more economical to get a CPU in the 50 to 100 MHz range.

    Another factor is the video processor. The SS10 originally came with an 8 bit color framebuffer. Just after Sun developed a good fast 24 bit color framebuffer, the SS10 was replaced by the SS20. So if you want 24 bit color, then look for a SparcStation 20. 

    There are people who understand the variations and performance of all the different possibilities of these computers. Unfortunately none of these people has written a short clear summary of this information for the education of the used Sun computer buyer. So you'll have to study a lot of comp.sys.sun.hardware to learn what is available and what you are willing to pay for. Sun is very spotty in providing info on their old computers. Things like the SS10 service manual http://docs.sun.com/ab2/coll.208.1/SMSS10/@Ab2PageView/ or the SparcEngine 5 OEM manual(for the SS5 motherboard) http://www.sun.com/microelectronics/manuals/801-7104.pdf  are found only with lots of luck.

--
Michael Coughlin  m-coughlin@ne.mediaone.net  Cambridge,MA USA



In article <Pine.GSO.4.21.9912201451220.19582-100000@zen>,
  Rich Teer <rich@rite-group.com> wrote:
> Depending on how fast the original poster's CPU is, and what work > they're
> doing, there may not be much difference between the two: as a rule of > thumb,
> SpuerSPARCS can do 2 ro 3 times more work per MHz than a HyperSPARC, > due
> to the increased cache and context switch efficiency.

That's not quite the understand that I have.  Here's a quote from the SunHelp mailing list, by James Lockwood (who seems to know a whole heck of a lot about these, compared to me).

===========QUOTE===========

On Fri, 26 Nov 1999, Gregory Leblanc wrote:

> Not to claim to be an expert...   It depends.  For some things, the SM81
> will be faster because of it's larger cache, while for others the HS150 (I
> think that's the number) will take the race.  150MHz with .5MB of cache, or
> 85MHz with 1 MB of cache.  I would expect the HS to be faster most of the
> time, and always cooler.  I just spent yesterday trying to find a fan with
> enough oomph to add to my 20 for cooling, and I only have a pair of SM41s.

The SuperSparc CPU's can do 3-way issue, compared with 2-way in the HyperSparcs.  This means that the effective clockspeed advantage of the HyperSparc is only 17% instead of 76%.

SPEC is not a perfect benchmark but it usually gets you in the ballpark:

System            CPU        ClkMHz  Cache      SPECint SPECfp  Info
Name              (NUMx)Type ext/in  Ext+I/D      92      92    Date
================= ========== ======= ========== ======= ======= =====
Sun SS20/71       SuprSP2    50/75   1M+20/16    125.8   121.2  Jan95
Sun SS20/81       SuprSP2    50/85   1M+20/16    148.6   141.4  Jan95
Sun SS20/151      HyperSP    50/150  512+8/0     169.4   208.2  Nov95

Note that there are several results for the SM81 floating around, this seems to be about the average.  The SM81 actually exhibits superlinear speedup from the SM71 as it supports multiple command queueing on the mbus, allowing it to saturate it more effectively.

Generally, a 150+MHz HyperSparc will be the fastest.  If your code fits in
the 1MB cache of an SM81 or does a lot of context switches, the SM81 may well be faster (as the SM81 uses physical cache vs the virtual cache of the HyperSparc which has to be dumped whenever the physical/virtual mappings change).

The HyperSparcs generally run hotter than the SuperSparcs.  The CPU generates a lot more heat than the cache SRAMs, and the cache controller is integrated into the same package as the CPU.

To the original poster I ask you this: why do you want to upgrade a SS20 to this point?  Unless you must run SunOS 4 (or NeXTstep, I suppose) a low
end Ultra will be a better deal.  Even the Ultra 1/140 has far greater memory bandwidth than the SS20 does, and you strain that bandwidth more with each additional processor.

The SS20 is a great machine, one of the best "desktops" that Sun ever made.  But it's not an Ultra.

-James

===========END QUOTE===========

Not quite doubled, but the 150MHz hyperSPARCs should be faster for almost EVERY operation than the SM81s, and should be faster than any other SuperSPARC.  There are rumors of some 200MHz SuperSPARCs, which should smoke the pants of the SM81s.  A few more years, and the prices should come down enough for normal people to buy them.
	Greg

--
It's pronounced "sexy" not "scuzzy"!


Sent via Deja.com http://www.deja.com/
Before you buy.


cjf wrote:
 
> I recently picked up a sparcstation 1+, and I'm wondering if > I can run Solaris 2.6 on it (not recommended, I know).

     Solaris 2.6 will work just fine on the SparcStation 1+. Do not believe impatient people who say it will run too slow. You just have to learn to relax and be less frantic. You can also use Sparc verions of NetBSD and Linux which will be faster since they have removed some of the clutter that has accumulated over the decades in propriority Unix.

> When I try and boot the CD,
> I get a 'bad magic number ~blah~ not executable', > HOWEVER, I can partially boot a RedHat/SPARC disc, so I know > my pioneer CD is working (it has a 2048/512 jumper). I say > partially, because my PROM needs to be replaced/reprogrammed, > but if I'm not mistaken, I should still be able to start the > install process (right?).
 
> I'm also wondering if updated firmware is available. It > currently has rev 1.3, that does not appear to 'directly' > support a cdrom, ie through an 'sr' device type. I have > found updated versions for the 3/80, implementing that > particular support, but I've been hard pressed to find > documentation and/or a source for the 1+. Er, so to > backtrack, my boot command is 'b sd(0,6,0)', which > 'partially' works. Curiously, the Sun CD FAQ mentioned that > the boot command for a sun4c was 'b sd(,30,3)', which gave > me nothing (nor did any of the other examples). Am I doing > something wrong?

     Yes, you are doing something wrong. There is a key technical term that describes this situation. It is "FUBAR", which means (in polite terms) "foulded up beyond all recognition". The process of numbering disk drives (or even different verions of Unix) is much too difficult for programmers. So they are always getting this wrong and trying to fix it -- but just making things worse. The CD ROM drive can be set to any number. Sun has decided that it should be number 6 (but you can change that if you want). At one time on Sun computers, you would refer to SCSI disk number 6 as number 30. Hence the command 'b sd(,30,3)'. After many years of otherwise intellegent people being unable to boot and install the operating system, Sun changed their boot ROM so the CD ROM was called number 6, the same as the disk jumper setting. But there are three numbers in the b command. First number is 0 for the controller for the internal drives. Of course you could make that something else, or have a workstation with more than one SCSI controller, but that doesn't happen very offen. The third number is the partition on the disk. I've lost count of how many different versions of Sun Sparc processors there are, but each different type needs to boot a different kernel. For the Sun SparcStation 1+ the necessary code is on partition 2 of the Solaris CD ROM. The correct boot command for your machine is  boot sd(0,6,2)   or   b sd(0,6,2)
I've tried this myself on two different versions of Solaris so I know it works. For the Red Hat or NetBSD CD ROM, check the install instructions carefully for the right numbers.

     The latest versions of Sun's Open Boot prom have made a major advance in programming. All you have to do is type 'boot cdrom' and the computer will figure out where everything is. Amazing!

--
Michael Coughlin  m-coughlin@ne.mediaone.net  Cambridge,MA USA