HP/DEC/Compaq

I've heard that NextStep on PA-RISC is much faster than on comparable hardware of the time. I've run NextStep on everything but PA-RISC would really love to hear from folks with firsthand experience what it was like.

Anyone who like me has run it on multiple platforms and can really speak to the comparison would be much appreciated.

I once ran a lab of 70x Hp 712s that I could have had my pick of when we decommissioned it. At the time I had no idea they were NextStep compatible so I didn't take one. *self kick* Though I did take a scsi drive out of one to use with my Indy. I'm thinking of picking one up again.

It would be nice to finally use that last of the binaries in those quad-fats I've built over the years.

I've got NS3.3 on an HP 712/60 with 128mb and it is definitely peppy, when compared to my TC slab. I'm not sure if it runs better than my SS20 at 75mhz and SX graphics.

antagonizt wrote: I've got NS3.3 on an HP 712/60 with 128mb and it is definitely peppy, when compared to my TC slab. I'm not sure if it runs better than my SS20 at 75mhz and SX graphics.

and consider that the /60 is the slowest of all 712s, because it is not only clocked the least, but also has a smaller cache than the /80 and /100. The /100 has the two additional memory slots, but I'm not sure 192mb will make that much of a difference on NeXTSTEP.

I didn't run NeXTSTEP on my 712 yet (because I run it headless and NeXTSTEP cannot be installed via serial console), but from what I've heard the 712/100 is one of the fastest non-intel NeXTSTEP machines (after the 715/100XC)

oreissig wrote: I didn't run NeXTSTEP on my 712 yet (because I run it headless and NeXTSTEP cannot be installed via serial console), but from what I've heard the 712/100 is one of the fastest non-intel NeXTSTEP machines (after the 715/100XC)

Just for the sake of argument, here are some actual NEXTSTEP 3.3 benchmark results.

'make all' -- ISC BIND 8.2.2
715/80: 21:29
P5/133: 18:35

'make all' (non-fat) -- perl 5.6.0
SS5/70: 26:41
SS5/110: 21:20

NWbench results:
SS5/110 TCX: 157068[dhrystone] 99.725716[MIPS] 0.784[gfx v/v] 6.384[gfx d/v] 24.678793s[webster] 45.900536[compile]
712/100: 126582[dhrystone] 80.369522[MIPS] 9.44[gfx v/v] 9.52[gfx d/v] 18.327635s[webster] 35.44598s[compile]
715/75 CRX24: 105263[dhrystone] 66.833649[MIPS] 5.936[gfx v/v] 6.432[gfx d/v] 20.026428s[webster] 38.612015s[compile]

Seems to indicate that the HPs have slower graphics and processors, but turn in better overall application performance.

Just for the sake of argument, here's the NWbench result for VMware Fusion on my Mac Pro. The dhrystone and MIPS ratings are not shown because they complete faster than the timer resolution of the dhrystone benchmark (resulting in scores of zero).

Fusion: 0.048[gfx v/v] 0.016[gfx d/v] 6.15974s[webster] 4.819834s[compile]

The TCX in the SS5 is clearly a pretty nice framebuffer. I never tested the SS20, so I can't comment on how the SuperSPARC or cgfourteen stack up. NEXTSTEP won't run on the 170 MHz SS5 (as the turboSPARC processor is not supported). I would expect the HP 735/135 to be faster than the 712/100 in all respects but I never tested it, either.

I have both 712/100 and sm81 ss20. The hp fells snappier, but that's pretty much always been the case with these workstations, even running hpux/Solaris.

PA-RISC is a nice architecture all the way around and was a particularly performant RISC implementation for the time. It's a shame they just drink watts of power and run rather hot. Part of this is the insane amount of cache later models shipped with.

I've always had a soft spot for PA-RISC because my first job out of college was working on an HP K-class.

I had a B&W '030 Slab for many years. The king gecko I have is noticably faster processing wise. Disk wise it's much of a muchness with Fast-5.

uridium wrote: I had a B&W '030 Slab for many years. The king gecko I have is noticably faster processing wise.

Nothing to be surprised of. Integer performance of the early PA-RISC processors was uncanny - too bad the worse than average I/O made it useless in most cases. While Sun systems were the opposite. And the 68030's lack of any real cache will always make it lose against a 040 or any RISC processor of that era.

I've never seen a 68030 NeXT slab. Do they really exist?

No.

ClassicHasClass wrote: I've always had a soft spot for PA-RISC

me too, simply because they were good as so was/is hpux

simply because they were good as so was/is hpux

Well, I don't know about that ...

(having suffered through HP-sUX 8.0, 10.20 and 11i)