27 Apr
2016
27 Apr
'16
12:47 a.m.
27 Apr
27 Apr
1:03 a.m.
Zitat von Dave Wade <dave.g4ugm at gmail.com>:
I actually wonder if an FPGA VAX chip could be made
that would run
faster than existing real VAXEN. That could perhaps form the basis
of a nice VaxStation...
... on browsing I found this...
If you get ~200VUPS on an i7, it should be possible to put it in an FPGA,
an do it right. So no real VAX, but if you compete with a 4000/90,
you should be fine on an FPGA ...
Cheers
5 a.m.
On Apr 27, 2016, at 3:47 AM, Dave Wade <dave.g4ugm
at gmail.com> wrote:
...
Digital is now a fond memory for most. Both VAX and Alpha are no longer manufactured.
I actually wonder if an FPGA VAX chip could be made that would run faster than existing
real VAXEN. That could perhaps form the basis of a nice VaxStation...
Clearly that's possible given that today's clock speeds are much higher than those
of any VAX, and you could put caches on-chip as well for additional points. The
difficulty would be to create an accurate enough implementation. Given the availability
of the VAX architecture manual, the odds are better than they would be for many other
processors.
For that matter, by the same reasoning it should be doable (and quite possibly easier) to
build an FPGA Alpha. Is the Alpha architecture manual ("SRM") online?
paul
6:07 a.m.
New subject: Digital VAX, Alpha (Was RE: Mac "Workgroup Server" ...
Zitat von Paul Koning <paulkoning at comcast.net>:
If "accurate" means to run VMS or Unix,
it shouldn't be to difficult.
On Apr 27, 2016, at 3:47 AM, Dave Wade
<dave.g4ugm at gmail.com> wrote:
...
Digital is now a fond memory for most. Both VAX and Alpha are no
longer manufactured.
I actually wonder if an FPGA VAX chip could be made that would run
faster than existing real VAXEN. That could perhaps form the basis
of a nice VaxStation...
Clearly that's possible given that today's clock speeds are much
higher than those of any VAX, and you could put caches on-chip as
well for additional points. The difficulty would be to create an
accurate enough implementation. For that matter, by the same reasoning it should be
doable (and
quite possibly easier) to build an FPGA Alpha. Is the Alpha
architecture manual ("SRM") online?
Is there any demand for it, besides of being fun?
Cheers
8:27 a.m.
New subject: Digital VAX, Alpha (Was RE: Mac "Workgroup Server" ...
On Apr 27, 2016, at 9:07 AM, emu at e-bbes.com wrote:
Zitat von Paul Koning <paulkoning at comcast.net>:
If
"accurate" means to run VMS or Unix, it shouldn't be to difficult.
You might be surprised. Getting a PDP-11 FPGA to be accurate enough to run standard
operating systems is hard enough (as I found out helping Sytze's "pdp2011"
project). And that's a much simpler CPU than VAX. In particular, the privileged
architecture tends to be critical for getting an OS to boot, and that part tends to be
poorly documented (as well as variable from one CPU model to the next).
On Apr 27, 2016, at 3:47 AM, Dave Wade
<dave.g4ugm at gmail.com> wrote:
...
Digital is now a fond memory for most. Both VAX and Alpha are no longer manufactured.
I actually wonder if an FPGA VAX chip could be made that would run faster than existing
real VAXEN. That could perhaps form the basis of a nice VaxStation...
Clearly that's possible given that today's clock speeds are much higher than
those of any VAX, and you could put caches on-chip as well for additional points. The
difficulty would be to create an accurate enough implementation. For that matter, by the same reasoning it should
be doable (and quite possibly easier) to build an FPGA Alpha. Is the Alpha architecture
manual ("SRM") online?
Is there any demand for it, besides of being fun?
28 Apr
28 Apr
1:02 a.m.
New subject: Digital VAX, Alpha (Was RE: Mac "Workgroup Server" ...
Zitat von Paul Koning <paulkoning at comcast.net>:
Probably not. I have both working here as Software emulations ;-)
> On Apr 27, 2016, at 9:07 AM, emu at e-bbes.com wrote:
If
"accurate" means to run VMS or Unix, it shouldn't be to difficult.
You might be surprised. Getting a PDP-11 FPGA to be accurate enough to run
standard
operating systems is hard enough (as I found out helping Sytze's
"pdp2011" project). And that's a much simpler CPU than VAX. In
particular, the privileged architecture tends to be critical for
getting an OS to boot, and that part tends to be poorly documented
(as well as variable from one CPU model to the next).
Funny as it Sound, the VAX was easier, but all the Floating Point
formats took a while to implement ...
I don't know. Fun is certainly a good reason for
many of us.
Yup, that's how it started ;-)
I don't know if there are any unexpired patents;
if not, then
implementing a machine from the published documentation seems fine,
though running the software might require answering some licensing
questions.
That's the real pain in the neck, to Transfer licenses from one
machine to the next ...
Cheers
27 Apr
27 Apr
8:34 a.m.
New subject: Digital VAX, Alpha (Was RE: Mac "Workgroup Server" ...
On Wed, 27 Apr 2016, emu at e-bbes.com wrote:
I would assert that there may well be. I could be wrong, but consider
these tidbits:
1. There are at least 3 commercial alpha emulators (Stromasys Charon, MSI
Avanti, and Alpha EmuVM). They seem to be staying in business.
2. I work with a lot of clients using Tru64. The current hardware support
companies are recycling parts off ebay. I believe this is the reason
you see way-to-high prices for categorically ancient Alphas. I saw
someone wanting nearly $1300 for an alphastation 4/266.
3. I know for a fact the US government and a few other folks are pretty
well stuck with using Alphas for "certain" things. If the vendor was
OK'd by the gubment, there might be some money to be made there, too.
4. If some guy in Poland (Lotharek) was able to make an FPGA box that
emulates more than 9 different platforms (albeit probably simpler ones)
for under $200, then making an FPGA alpha for anywhere near that cost
would certain raise some eyebrows of clients I'm dealing with. If it
would all fit in 1U, then so much the better.
-Swift
For that
matter, by the same reasoning it should be doable (and quite
possibly easier) to build an FPGA Alpha. Is the Alpha architecture manual
("SRM") online?
Is there any demand for it, besides of being fun?
8:44 a.m.
New subject: Digital VAX, Alpha (Was RE: Mac "Workgroup Server" ...
On Apr 27, 2016, at 11:34 AM, Swift Griggs
<swiftgriggs at gmail.com> wrote:
... If it
would all fit in 1U, then so much the better.
Ideal would be if it could fit in this:
https://www.element14.com/community/docs/DOC-69215/l/fpga-development-board…
paul
11:16 a.m.
New subject: Digital VAX, Alpha (Was RE: Mac "Workgroup Server" ...
On Wed, 27 Apr 2016, Swift Griggs wrote:
3. I know for a fact the US government and a few other folks are pretty
well stuck with using Alphas for "certain" things. If the vendor was
OK'd by the gubment, there might be some money to be made there, too.
That is interesting. So does Raytheon or someone still make a custom CPU
with the Alpha ISA? Or maybe foundryless, soft core, etc. I wasn't aware
of that.
I was browsing around a few months back and found a catalog from a Russian
electronics company. Current catalog ... and they appear to still be doing
PDP-11 ... I know for sure ... and possibly VAX ... might be confabulating
that ... CPUs. I thought that was neat. Assumedly for some mil-spec use.
Best,
Sean
28 Apr
28 Apr
1:15 a.m.
New subject: Digital VAX, Alpha (Was RE: Mac "Workgroup Server" ...
Zitat von Swift Griggs <swiftgriggs at gmail.com>:
3. I know for a fact the US government and a few other
folks are pretty
well stuck with using Alphas for "certain" things. If the vendor was
OK'd by the gubment, there might be some money to be made there, too.
But they don't Need an FPGA, for them Software Emulation of an Alpha
is sufficient ...
4. If some guy in Poland (Lotharek) was able to make
an FPGA box that
emulates more than 9 different platforms (albeit probably simpler ones)
for under $200, ...
just for clarification, Lotharek is a reseller not the developer of
it. About the development of it, ("MIST") you can read at
https://github.com/mist-devel/mist-board/wiki
cheers
27 Apr
27 Apr
11:46 a.m.
New subject: Digital VAX, Alpha (Was RE: Mac "Workgroup Server" ...
On 4/27/2016 7:07 AM, emu at e-bbes.com wrote:
1st build a front pannel, then worry about the inside.
For that
matter, by the same reasoning it should be doable (and quite
possibly easier) to build an FPGA Alpha. Is the Alpha architecture
manual ("SRM") online?
Is there any demand for it, besides of being fun? Cheers
Ben.
11:07 a.m.
On Wed, 27 Apr 2016, Dave Wade wrote:
Don?t fret, once OpenVMS v9.0 is released, on x86-64, there won?t be any doubt
as to who won. :-)
Sadly, no one won. I doubt any one (well perhaps not anyone) would consider OpenVMS for a
new deployment.
Upgrading existing environments, yes, but a new green field site. It would have to have
very good reasons.
(I know you will all come out with some, but perhaps one for every 10,000 Linux and/or
Windows Server deployments.)
Digital is now a fond memory for most. Both VAX and Alpha are no longer manufactured.
I actually wonder if an FPGA VAX chip could be made that would run faster than existing
real VAXEN. That could perhaps form the basis of a nice VaxStation...
... on browsing I found this...
http://h71000.www7.hp.com/openvms/journal/v7/vax_6000_emulator.pdf
Dave
I don't know ... I know the primary market is probably legacy environments
but I'm really, really excited that OpenVMS is making it to x86-64 and as
a working sys admin, I wouldn't hesitate to pitch OpenVMS for certain
situations if I felt it was right for the job ... I'm mostly a UNIX guy
but I've always thought OpenVMS was cool, too, no prejudice. I encourage
OS diversity ;)
Particularly in the HPC space ... I would love to put a cluster built on
VMS against a cluster built on Linux and compare availability, throughput
and uptime over the course of a year ... Results could be interesting.
I'm sure like a lot of folks on here, I have a lot of positive nostalgia
associated with DEC (and I'm pretty young) ... I will always be a fan for
sure. But versus Linux ... which has swept the market ... OpenVMS does
have some real technical merits to stand on. It doesn't just have to be
for legacy applications and hobbyists.
Best,
Sean
6:09 p.m.
On 04/27/2016 01:07 PM, Sean Caron wrote:
>
> I actually wonder if an FPGA VAX chip could be made that
> would run faster than existing real VAXEN.
Sure. A VAX 11/780 had a 5 MHz
clock! Would be hard for an
emulator to NOT beat that! Later models did run faster, but
not vastly faster, due to the technology of the time.
Jon
7:34 p.m.
On Wed, Apr 27, 2016 at 7:09 PM, Jon Elson <elson at pico-systems.com> wrote:
Sure. A VAX 11/780 had a 5 MHz clock! Would be hard
for an emulator to NOT
beat that! Later models did run faster, but not vastly faster, due to the
technology of the time.
I'm not sure what would qualify as "vastly faster", but I take it that
170.9 MHz doesn't?
(also required fewer clocks per instruction than the 11/780)
A well-tuned VAX design in a recent FPGA family (Xilinx 7-series or
newer, or Altera Stratix V or newer) might be able to outperform the
fastest "real" VAX, but perhaps not by a whole lot. FPGAs are
generally much more efficient at implementing RISC processors, but
it's difficult to get a whole lot more than 200 MHz for the
cost-optimized FPGAs, or 350 MHz for the (expensive)
performance-optimized FPGAs.
I've designed VHDL cores equivalent to microprocessors such as the RCA
1802, National Semiconductor PACE, and DEC/Western Digital LSI-11 (at
the microarchitecture level). I'd like to tackle something more
sophisticated, but it's hard to find enough time.
8:10 p.m.
On 04/27/2016 09:34 PM, Eric Smith wrote:
On Wed, Apr 27, 2016 at 7:09 PM, Jon Elson <elson
at pico-systems.com> wrote:
Well, running an
efficient simulator on a modern PC
processor with a 3+ GHz clock, with faster memory and larger
and faster cache, I think it would still beat that top/end
of the line VAX. But, I don't actually know for sure.
Sure. A VAX 11/780 had a 5 MHz clock! Would be
hard for an emulator to NOT
beat that! Later models did run faster, but not vastly faster, due to the
technology of the time.
I'm not sure what would qualify as "vastly
faster", but I take it that
170.9 MHz doesn't?
(also required fewer clocks per instruction than the 11/780)
A well-tuned VAX design in a recent FPGA family (Xilinx 7-series or
newer, or Altera Stratix V or newer) might be able to outperform the
fastest "real" VAX, but perhaps not by a whole lot. FPGAs are
generally much more efficient at implementing RISC processors, but
it's difficult to get a whole lot more than 200 MHz for the
cost-optimized FPGAs, or 350 MHz for the (expensive)
performance-optimized FPGAs.
I've designed VHDL cores equivalent to microprocessors such as the RCA
1802, National Semiconductor PACE, and DEC/Western Digital LSI-11 (at
the microarchitecture level). I'd like to tackle something more
sophisticated, but it's hard to find enough time.
Wow, yeah, designing a hardware emulator on an FPGA for the
VAX would be a big project. Doing the same for the Alpha
architecture would be REALLY daunting, since a lot less was
done by microcode.
Jon
2942
days inactive
2943
days old
14 comments
8 participants
participants (8)
-
bfranchuk＠jetnet.ab.ca -
dave.g4ugm＠gmail.com -
elson＠pico-systems.com -
emu＠e-bbes.com -
paulkoning＠comcast.net -
scaron＠diablonet.net -
spacewar＠gmail.com -
swiftgriggs＠gmail.com