AlphaServer 8400 for Trade
vaxorcist
This is the AlphaServer 8400 current configuration:
8 CPUs = 4*Dual-CPU 300 MHz E2056-DA
4GB Memory = 2*2GB MS7CC-FA
2 KFTIA I/O Modules (4*SCSI / 2*Network) E2054-AA
1 Bus-Terminator E2034-AA
3 48V DC Power Supplies 303379602
3 BA356 StorageWorks SCSI-Boxes
12 DEC SCSI-Drives (RZs)
1 AlphaServer 8400 PCI Box
There are some AlphaServer 8400 spares that will go together with the
computer:
1 48V DC Power Supply 303379602
2 CPUs = 1*Dual-CPU 300 MHz E2056-DA
2GB Memory = 1*2GB MS7CC-FA
1 Bus-Terminator E2034-AA
1 Box BA655-AA
The AlphaServer 8400 is in working condition, but without software and
licenses.
The only known deficiency is that the console CDROM drive (RRD45) does
not open properly, but is nevertheless usable. The condition of the
spares is unknown.
The AlphaServer 8400 system is located in Northern Germany.
What do I expect in return:
- A VAX-11/750 in working order
- 1MB VAX-11/750 memory
- A working VAX-11/750 disk subsystem (controller & drive, any model)
- A DEUNA or DELUA UNIBUS ethernet adapter
All items are negotiable.
The VAX-11/750 will become part of my private collection of VAX
computers and will be used to run VMS version 2, 3 & 4, ULTRIX-32 and
some old UNIX versions.
Regards,
Ulli
(The VAXorcist)
Phillip Helbig---remove CLOTHES to reply
<d5ad96c2-cf55-4151...@v56g2000hsf.googlegroups.com>,
vaxorcist <vaxo...@googlemail.com> writes:
> I will trade off my AlphaServer 8400 for a VAX-11/750.
I guess the power consumption will stay about the same.
The VAX is 1 VUP, right? The 8400 must be about 1500. Not a judgement,
just an observation. While I don't have such big machines at home, I
can appreciate the joy of running a VAX (of which I have several, all
small and relatively fast (for VAXes).
Peter 'EPLAN' LANGSTOeGER
>The VAX is 1 VUP, right?
Nope. The 11/780 was (per definition) 1 VUP. The 11/750 was well below (0,7)
--
Peter "EPLAN" LANGSTÖGER
Network and OpenVMS system specialist
E-mail Pe...@LANGSTOeGER.at
A-1030 VIENNA AUSTRIA I'm not a pessimist, I'm a realist
Wilm Boerhout
23:23:
Alpha's don't count VUPs, see (a.o.)
http://www.compaq.com/alphaserver/performance/vups_297.html
If you still want to compare, from Table 1 on that page, I infer that
the (SPECxxx) performance of an Alphaserver 8400 5/300 is roughly 10
times a VAX 7000-610, or 350 VUPs. YMMV.
/Wilm
Dennis Grevenstein
> Alpha's don't count VUPs, see (a.o.)
> http://www.compaq.com/alphaserver/performance/vups_297.html
>
> If you still want to compare, from Table 1 on that page, I infer that
> the (SPECxxx) performance of an Alphaserver 8400 5/300 is roughly 10
> times a VAX 7000-610, or 350 VUPs. YMMV.
Then again, you can't be sure how those benchmark numbers really
scale, so you can't infer how an 11/750 might come out on SPEC.
Personally, if you want to compare performance, I would advice to
get those old dhrystone sources out of the crypt and get *proper*
results as a VAX MIPS rating for both. :-)
There's nothing better than collecting data yourself. Much more fun.
Dennis
--
Don't suffer from insanity...
Enjoy every minute of it.
Bob Willard
The basis for the VUP was the VAX-11/780. IIRC, the VAX-11/750 was ~0.7 VUP.
--
Cheers, Bob
Galen
>
> Personally, if you want to compare performance, I would advice to
> get those old dhrystone sources out of the crypt and get *proper*
> results as a VAX MIPS rating for both. :-)
If I recall correctly--and it's been a long time, so I may be
inaccurate about the details--good modern compilers (since, maybe, the
early 1990's?) may optimize away much or all of a simple benchmark
like dhrystone or whetstone, rendering them, in effect, a great big no-
op. If so you don't really get a measurement that will help predict
the performance of real applications. Also, modern processors may fit
most or all of these small benchmark applications into high speed
cache or use other hardware features that render the old benchmarks
less than useless. And how do multiple processors
I am not an expert on any of this, and I hope you'll pardon any
blunders or oversights I've made. I also don't know to what extent the
most recent OpenVMS VAX compilers are capable of this kind of
optimization. But I do recall a lot of debate about the usefulness of
dhrystone and other benchmarks way back in the late 1980's or early
1990's when we were evaluating early workstations from Apollo, HP, and
other companies.
Of course, if all you're looking for is some cheap fun then don't let
any of this stand in your way.
:-)
Dennis Grevenstein
>
> If I recall correctly--and it's been a long time, so I may be
> inaccurate about the details--good modern compilers (since, maybe, the
> early 1990's?) may optimize away much or all of a simple benchmark
> like dhrystone or whetstone, rendering them, in effect, a great big no-
> op. If so you don't really get a measurement that will help predict
> the performance of real applications. Also, modern processors may fit
> most or all of these small benchmark applications into high speed
> cache or use other hardware features that render the old benchmarks
> less than useless. And how do multiple processors
That's just my point. You can't predict how an 11/750 might
score on SPEC. It wouldn't be a fair comparison, because modern
benchmarks are made to show differences between modern computers.
If an old benchmark doesn't move enough data to go beyond the
CPU cache on modern machines, that's fine, because old machines
don't have a cache like that and that's one reason why they are
slower. It all depends on what kind of an application you are
thinking about. You can't say that an AS8400 is X times faster
than an 11/750. For whom? For a single user editing a text file
with vi or for a programmer who can use the entire machine for
serious number crunching?
In the end, it's only a useless number, but you could say that
the newer machine can do the same work much faster than the old
machine. If you want to make that comparision, stick to the old
type of work that the old machine was built to do.
H Vlems
wrote:
All these tests tried to do is come up with a performance statistic
for a new system that customers could compare to what they already had
running in their computer rooms. If your workload was as "average" as
the test then you could predict performance gains and possibly get
more work done and send more bills to customers.
Unfortunately we don't run "average workloads". And I recall
applications that ran faster on certain VAX models than on the first
Alpha's even when the latter "ran faster'.
I wrote a small DCL program that 'computes" VUPS ratings for VAXes. Of
course it runs on Alpha's and I ran it on the test Itaniums once. The
DCL code produces a number for ALpha and I64 but the significance
ought to be rated low. It does not make use of the advantages of these
64 bit systems.
A microVAX II, a VAX 11/780 and a VAX 8200 have approximately the same
rating (0.9, 1.0 and 1,0 VUPS). For heavy network traffic (10BASE5
remember) I'd certainly prefer the 8200 with its much faster BI bus.
The 11/780 was faster than the microVAX II in number crunching but the
microVAX had a faster Qbus and could handle graphics. Price (purchase
and maintenance) and application decided what was "best" and certainly
not VUPS. Possibly why the 10000 never became a success and the 4705A
is one of the most underrated VAX models (it was fast).
Hans
Glen Herrmannsfeldt
(snip)
> If I recall correctly--and it's been a long time, so I may be
> inaccurate about the details--good modern compilers (since, maybe, the
> early 1990's?) may optimize away much or all of a simple benchmark
> like dhrystone or whetstone, rendering them, in effect, a great big no-
> op.
That problem is much older. There is a story from the OS/360 days
of a Fortran benchmark that was a very complicated set of statement
functions doing complicated calculations. The OS/360 Fortran H
compiler expands statement functions inline, and also does constant
expression evaluation. The particular benchmark compiled very slow
and ran very fast. The compiler evaluated the whole thing down to
one constant and a WRITE statement.
> If so you don't really get a measurement that will help predict
> the performance of real applications. Also, modern processors may fit
> most or all of these small benchmark applications into high speed
> cache or use other hardware features that render the old benchmarks
> less than useless. And how do multiple processors
But if you want to compare old processors running old programs the
old benchmarks might be the best choice.
-- glen
Bob Koehler
>
> just an observation.
Actually 1 VUP was defined by the 11/780. Your 11/750 will be
somewhat slower, most of the time.
Rick Jones
> That's just my point. You can't predict how an 11/750 might
> score on SPEC.
Doesn't help that there are what, four or five of them from which to
choose (off the top of my head - I was doing SPECsfs in '91 and wasn't
paying _that_ much attention to the CPU guys ):
SPECcpu89 (specmark iirc)
SPECcpu92
SPECcpu95
SPECcpu2000
SPECcpu2006
If the 750 is going to have 1MB of memory, unless one wanted to wait a
_really_ long time for paging/whatnot you'd probably have to go all
the way back to the '89 version. '08 is designed to "fit" in a system
with 1GB of RAM, I forget the specifics of the prior versions but
would guess that there may be some information on www.spec.org
rick jones
--
Process shall set you free from the need for rational thought.
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...