Wow, what a response to this one! I think I touched a nerve somewhere
in the nmr community. I'd like to thank everyone personnally and I started
to do just that but there are just too many so I'll summarize the
responses for you to look at (names removed to protect the innocent).
For what its worth, my opinion is that on a system that is used by lots
of users, NT is a bit of a dangerous choice. My experience with Windows
and NT computers that are used by many people is that they tend to get
jammed up with all sorts of junk software. Its easy to download stuff for
Microsoft OS PC's and 'try it out'. I know that on NT normal users can't
alter the system software but I also know that they can still install
stuff. On a multiuser nmr computer I wouldn't allow *anything* but the nmr
software ... period. This would include Microsoft Word and any other word
processing software. This is easy to do under Unix. MS Word isn't
available for IRIX but xwinnmr is. Also, it is really easy to get unix
machines talking to each other ... file transfers, clock settings etc all
without human supervision. Makes management of the system that much
easier. My two cents worth. Anyway, here is the response summary:
-------------------------------------------------------------------
> I'm wondering what kinds of workstations folks are using these days with
>their spectrometers. Are you using SGI's under IRIX or PC's running NT? If
>you're using a mixture of the two how well do they complement each other
>and how do they communicate (if they communicate) with each other? Is
>there a Linux version of Xwinnmr available yet? I've heard disturbing
>rumors about the possible demise of SGI and wouldn't particularly want to
>be stuck without support but on the other hand ... I'm a unix kind of
>person.
>
> I recall a vote being taken a year or so ago on www.nmr.de concerning
>unix verus NT and the overwhelming response was unix. Is this still the
>prevailing opinion?
>
> Oh yes, now that I think of it, one more question. Is the Bruker file
>structure under NT the same as under IRIX? That is,
>"filename/expno/procno" so that data generated on one system can be put
>onto another.
--------------------------------------------------------------------
I just went to a Bruker users meeting where they said that by August,
xwinnmr 3 will be available for spectrometer control. There is a free test
version for processing available now that you can order from their
web-site. The problem is that xwinnmr 3 will support only Avances. Those
of us with ARX's are out of luck.
Please summarize any relevant information you get on this topic.
---------------------------------------------------------------------
I realize that what I left out of my previous message is that the
version of xwinnmr3 that will be available in August runs under Linux. (I
must be losing it; that was the crucial point.) I have only SGI's, even on
Avances that had a choice. I worry a lot about SGI but the fact that they
use unix is the reason I chose them. I don't personally have a great love
of unix, but we have for many years done all our off-line processing by
simply telnetting to the spectrometer and running another session of unix
while the spectrometer is being used by a different user. We have no
problem with this with SGI's. Presumably one could also do this with a
linux version of xwinnmr but of course I don't know that for a fact. And
since I have 2 ARX's I'm out of luck there.
---------------------------------------------------------------------
The Linux version of Xwinnmr has been announced last November. Runs
very well for processing. We are using it quite a bit for off-line
processing (NFS mounted). We still use an SGI for processing, some
things (DECO, Bruknet) are not yet functional under Linux.
Acquisition has not been released yet (but most likely is working...)
Never tried NT, and (hopefully) never will.
---------------------------------------------------------------------
I have XwinNMR running on SGI, NT and Linux. All three are more or
less the same, with each having strengths and weaknesses. The SGI
version has the swiftest graphics (thanks to SGI) while the NT version
of XwinPlot is definitely the best in that it is a native Microsoft app and
interfaces nicely to Word, PowerPoint and chemical drawing programs.
The Linux version is O.K. and is essentially just a port of the SGI version.
I do have some Colourmap problems but this may be with my particular
Linux Xserver or graphics card. My Linux XwinNMR is an "alpha" version.
Linux comes with its own set of bizarre wierdnesses though, mostly in
the X desktop environment.
File format is compatible between all versions of XwinNMR (and X32 UXNMR
as well). Byte order differences between the processors is handled internally
and is transparent to the user.
I typically acquire data on the SGI (Avance) or X32 (AMX) and then process
off-line with the NT or Linux versions.
---------------------------------------------------------------------
With regard to your questions, I can answer only one. The data structure
under NT is the same as UNIX ("filename/expno/procno" ) and can be
transferred. We do it regularly. I have no insight on SGI versus NT. All
our spectrometers are SGI (UNIX) and the user workstations on the network
are NT.
---------------------------------------------------------------------
I'm sticking with SGI/Irix for 3 spectrometers and 3 workstations.
There is a processing only version of XWIN-NMR for Linux, but if SGIs
should become non-viable I'm sure this would be completed. PC
hardware is relatively cheap and so if we ever have to move that way
it would be affordable. In the mean time I intend to stick with real
computers for as lonmg as possible.
---------------------------------------------------------------------
A linux version is available (I got mine from our Bruker rep). A few little
issues over colours, but it works great. I just NFS linked a directory from
my LINUX box onto the sgi, and copy data directories over to it. Very nice.
---------------------------------------------------------------------
I have made the change to the PC format. After some problems with being
one of the first to switch over, it seems to work just as well as the Unix
version. I have used XWIN-NMR on both platforms and don't really see much
difference. My deciding factor was the Indy I was using was becoming
obsolete(has trouble running the higher Irix versions of Unix) and the
alternative PC version was much much cheaper. They do communicate with
each other just fine. I still use my Indy as a processing station and the
two are linked via our intranet. Hope this helps.
---------------------------------------------------------------------
We have many systems out with either computer system running them. The
tendency of late, in my personal experience installing and supporting them,
is that there may be a slight majority of Windows systems over unix ones. I
have been personally assured by a SGI rep that they will continue to support
their OS and hardware well into the future- they're just not doing any R&D
for newer software and hardware technology as far as the O2 is concerned.
For you unix types, we d/will officially support a linux version- see our
www.bruker.de web site for the latest going on there.
The reason why unix is/should be preferred is because of the more intimate
relationship between the XWIN-NMR software (unix native) and the unix (IRIX)
OS. There are fewer layers of software to go through with a unix box whereas
in a Windows box, XWIN-NMR has to go through the X-windows (i.e. unix)
emulator, Hummingbird Exceed. This extra layer gives rise to potentially
added interfacing problems like "ghosting" and the like. However, the
Windows boxes have their benefits over the unix boxes.
Lastly, the file structure for where the data is stored is virtually the
same as you have shown. The difference as you may know is the location
specificity dependent upon the OS: /u/ vs c:\
Feel free to contact me if you have other questions.
---------------------------------------------------------------------
We're still using an Indy to control the spectrometer. Hope that
it lasts until/if Bruker does make a Linux version available.
OTOH used O2s are cheap and the spectrometer doesn't need
speed, just reliability. I think Vic said that there should
be a Linux data processing version available.
---------------------------------------------------------------------
we have both UNIX and NT installations and I really
love my UNIX environment. Although there are some
benefits of the NT version for processing (especially
cut & paste), there is no equivalent for the easy
remote administration of the UNIX machines.
Yes, a linux version is available for data processing
and this works very well. Until now this is not available
for spectrometer control and their is onelarge problem:
the color mode of XWIN-NMR. Remember, that this
program requires 8 bit pseudocolor. Using SGI
machines you can run different color modi parallel.
Simply ask for the available modes using xdpyinfo
Do the same using a Linux environment and you
will see a very restricted number of modes only.
Especially there are large problems to run 8 bit
pseudocolor and 24 bit true color parallel.
There are some helpful tips available from Rud Nunlist
http://www.tsoft.com/~rnunlist/geek/linuxnmr/XlinNmr.txt
Hope this helps.
---------------------------------------------------------------------
Well, there are some rumours, but, as Andy pointed out, nothing authorative.
For the time being, we support SGI, Windows NT and Windows 2000.
> I recall a vote being taken a year or so ago on www.nmr.de concerning
> unix verus NT and the overwhelming response was unix. Is this still the
> prevailing opinion?
We have seen a slight shift towards PCs from the beginning of this year on.
The distribution was appr. 55% SGI and 45% PC last year. In 2001, this has
changed to appr. 60% PC and 40% SGI. But whether this trend holds for the
rest of the year remains to be seen. With our Linux version coming up some
time later this year, people might turn back to the Unix based version. But
they will ALL stay with PC's (is my personal guess), as this is the cheaper
hardware solution.
> Oh yes, now that I think of it, one more question. Is the Bruker file
> structure under NT the same as under IRIX? That is,
> "filename/expno/procno" so that data generated on one system can be put
> onto another.
Yes, XWIN-NMR has the same data structure on all hardware platforms. You can
cross-mount directories and XWIN-NMR knows how to handle the data. That
saves you copying the data over the network.
---------------------------------------------------------------------
Regarding your query about workstations and Linux I can tell you that
Bruker have a Beta release of XWinNMR for Linux and the commercial
release is, I believe scheduled for late or at any rate later this
year.
Regarding SGI I have heard similar stories but nothing authoritative.
We are using the NT version and find it fine and reliable. I am
running the Linux Beta version and all I can say at this point is
that Linux is a fast and stable platform for the software. I have
had little time to evaluate it.
---------------------------------------------------------------------
I use both SGI and NT (pc) platforms to run my demo instrument at Bruker in
Billerica. I can ftp files and data from one to another with no problem,
via a command prompt or through a pc Explorer tool. The Bruker Suite
installation on the NT does have a GNU shell if you miss Unix terribly, so
you can tar and untar, etc., files on the NT. It works brilliantly.
Yes the file structure is the same, exactly, and the files are compatible.
If I have collected data (for some reason) on the SGI, I pull it over on the
NT using an Explorer tool and put it into my data dir and process it however
I want, then I can cut and paste directly to a MS Word or Powerpoint
document (I also have MS Office on my NT). This is a powerful feature that
simply is not available with an SGI.
The only reason I had both SGI and NT was for testing purposes, but now I
have to choose one and leave the other. Without hesitation I chose the NT.
Far more versatile: you can do absolutely everything on the NT that you can
do with an SGI, plus a whole lot more. I've seen rumours now and again on
the BUM list about the NT platform being unreliable or crashing and I've not
seen one instance of that with my system. That may have something to do
with memory size? I only use computers, I don't understand them
particularly well.
We do have a Linux demo version of the software available. If someone else
doesn't offer to send you one, let me know and I'll ask the right people.
---------------------------------------------------------------------
I will share some of my experiences with you, having used both. From the
outset, I will say that if the money was not an option, I would buy an
SGI next time. Comparitively, an SGI O2 is approx. AU$13,000 while a PC
equivalent is AU$2,500. This is quite an influencial difference in an
annual operating budget of AU$51,000.
I needed to replace my aging SGI Indy (1993 vintage) which couldn't
handle the latest versions of Irix and Xwinnmr and was excruciatingly
slow. I have also had systems elsewhere with SGI O2s and these were
fantastic. The assumption here is that you know enough Unix to cope. I
should also say that up until I moved to CSIRO, I had used Unix and
Macintosh systems exclusively, and I had to learn to cope with PCs.
When I changed (Nov 2000) I was advised by the Bruker Australia rep that
I should get Windows2000 and not NT as Xwinnmr v3.0 was about to be
released. Mistake. There were a number of incompatibilities between
Xwinnmr v2.6 (which I was running at the time) and Win2000 which meant
that I couldn't: wobb; use edte; rga, expt or gs when using echo/antiecho
2D; gs killed graphics server every time; graphics autorestart always
strated two Xwinnmr windows. There were probably others but I couldn't
be bothered finding them. I went to the b-release of Xwinnmr v3.0 in Jan
2001 which solved all of these problems, but left me with two which Mike
Engelhart at Bruker Karlsruhe still can't solve: I cannot gs for td>15k,
and there is a 60 sec delay from the start of wobb until the wobb curve
is displayed (since described as a 'design feature' - time to get a cup
of coffee). I haven't got the official release of v3.0 which is
supposedly the same as the prerelease version I have, and quite frankly,
I don't want to pay AU$1,200 for it if isn't going to work.
There are other things which are annoying with PC based systems including
not being able to telnet spect (you have to use hyperterm instead, but
this only works if there are no problems, which is why you want to use
telnet!?). The file systems are the same, apart from case-specific file
names which aren't supported by Windows, but are needed for some of the
Unix operations from within Xwinnmr. Permissions can be set to the Unix
equivalents. There are problems with using Win2K as an FTP server for
outside (non house network) access to data which I have resolved, but not
entirely to my satisfaction.
All in all, I am not impressed with the PC running the system and would
change back if the PC, still with all its faults, wasn't better than the
ancient Indy. Next time an O2 for me.
-----------------------------------------------------------------------
Keith Brown
University of Saskatchewan
Saskatoon, Saskatchewan
-----------------------------------------------------------------------
100 buckets of bits on the bus
100 buckets of bits
Take one down, short it to ground
FF buckets of bits on the bus
FF buckets of bits on the bus
FF buckets of bits
Take one down, short it to ground
FE buckets of bits on the bus
ad infinitum...
-----------------------------------------------------------------------
This archive was generated by hypermail 2b29 : Mon Jan 21 2002 - 18:08:05 PST