[I'm posting this to several newsgroups to address people in
different areas. My apologies to those who see this more than
I humbly present this mini-FAQ regarding Sun's new line of
low-cost/end workstations, the SparcStation 4, and the new TCX
graphics framebuffer that's associated with it. It is addressed to
those who are considering or have recently purchased one or more of
them, or those who are just plain curious. They were found to be
unsuitable to our environment, but may work quite well for others. I
am therefore being (mostly) objective about this presentation. This
mini-FAQ is dated May 10th, 1995 with the latest information available
to me (but I doubt is likely to change in the very near future). It is
derived from our experiences and exchanges with regional area Sun
Microsystems SE's (ie, I do not represent Sun); I also grabbed some
numbers from Vikas Jha's article floating around usenet during this
time (Subject: _SS5 vs. SS4 (esp. graphics performance)_). This is a
one-time effort for me with no intention of further verifying/updating
this information with reasons explained below.
We shall begin....
1 - What is the SparcStation 4 ?
2 - What is the TCX framebuffer ?
3 - Will SunOS 4 run on the SparcStation 4 ?
4 - How is the performance of the TCX rate to other framebuffers ?
5 - Will the MIT X server run under the SparcStation 4 ?
6 - Will "Doom" run on the SparcStation 4 ?
1: What is the SparcStation 4 ?
In short, it is Sun's new line of low-cost/end workstations. It is
modeled after the SparcStation 5 (same architecture), with slight
performance improvements (see below) but sporting only a single
Sbus slot and a special port for an optional 16-bit audio
controller (using the same CS4231 chipset found on the SS5). With
a 20" monitor (well, Sun's definition of 20"), it was priced
a couple hundred dollars cheaper than the SS5 (even after adding
in the audio controller) making it a seemingly attractive line.
CPU PERFORMANCE [vj]
Performance numbers from SUN:
(1).SPECint92 59.6 59.6
(2).SPECfp92 46.8 46.8
(3).SPECrate_int92 1414 1352
(4).SPECrate_fp92 1100 1122
(5).MIPS 105.4 100.3
(6).MFLOPS 15.8 13.1
[mileage may vary above; best to assume them being equal]
POTENTIAL FOR UPGRADES [vj]
(1).MainMemory: 16-160M 16-256M
(2).Max: 535M(*) 2.1G
(3).Max.Ext.disk: 28G 56G
(4).Packageslots: 1 SBus conn. 3 SBus connectors
(5).Serialports: 1 sync 1 async 2 synchronous
(5).Audio 16bit(optional) 16bit
[the cabinet in the SS4 is almost identical to that of the SS5,
with four drive bays (incl floppy/cdrom); however, in the SS4,
the one harddrive sits immediately above the motherboard with
no connector plane for the other 2 bays; one could always hack,
i'm sure a larger drive could be used, with the correct
2: What is the TCX framebuffer ?
The "TCX" is the standard 8-bit graphics controller hardware
that is built-in to the SS4 motherboard. By default, it runs
in the standard 1152x900 resolution mode, but can be expanded
to support 1280x1024 resolution by adding VSIMM memory to the
special socket on the motherboard, alledgedly with no performance
degradation. Summarized below :
GRAPHICS CONFIGURATION [vj]
graph. 8-bit pixel-accelerated (i) Turbo GX 8-bit
1152x900 res standard (8bit 2-D/3-D wireframe,
1280x1024 res optional 1152 x 900).
(video memory expansion
(ii) Turbo GX 8-bit plus
(8bit 2-D/3-D wireframe,
1152 x 900 or 1280 x 1024,
(iii)24-bit true color graphics,
1152 x 900.
Pixel Accelerator [sun]
o Feature o Benefit
- On-board pixel accelerator - Speeds screen functions: fills,
text, and scrolling
- More than twice the screen
performance of the CG3 frame
buffer used in the SPARCclassic
- 1280 x 1024 video memory expansion - Tailors the system to apps
supporting high resolutions
- No degradation in performance
- Low cost
- Fast video memory interface - Further speeds execution of
- State-less, single-instruction - Limits graphics circuitry for
architecture high performance at a low cost
The capabilities of the TCX are only accessible under Solaris 2 (2.4+),
however. This is a new frame buffer that will not be recognized by
current applications that directly access the graphics hardware. THIS
INCLUDES THE STOCK MIT X SERVER. [there is no plans at this time from
the X consortium to expand support for the TCX]. Another interesting
note regarding performance from an SE, "(Even when fully-exploited,
this device's performance will generally not equal that of higher-end
accelerators such as the TGX/TGX+ - particularly for vector-intensive
Running Solaris 1.1.2 (SunOS 4.1.4), "this display presents itself as
a "CG3"-type device - an unaccelerated 8-bit frame buffer." [sun].
This is the same controller used by the SparcClassic systems and
in typical color IPC configurations. For comparison, the
performance resembles a Tektronix XP18 color X terminal and not
far off from a Sun 3/50 running Xkernel. This is the case running
either X or OpenWindows.
3: Will SunOS 4 run on the SparcStation 4 ?
SunOS 4 will run on the SS4, but you MUST run SunOS 4.1.4 .
Earlier versions, including SunOS 4.1.3_U1, will NOT support the
Note: there is a caveat running SunOS 4 in that it will only
support the TCX running as a CG3 (unaccelarated). (See #2)
4: How is the performance of the TCX rate to other framebuffers ?
I do not have any actual numbers but performance will vary depending
on the type of application, and whether you run SunOS 4.1.4 or
Solaris 2 (see #2). The following is quoted from an (unnamed)
The SPARCstation 4 built-in display is a low-cost 8-bit
frame buffer with hardware support for a class of common
windowing operations including text, fills, stipples, and
scrolling/"blit"ing. The capabilities of this display are
fully exploited by Solaris 2.4 and its X server (Xsun).
(Even when fully-exploited, this device's performance will
generally not equal that of higher-end accelerators such as
the TGX/TGX+ - particularly for vector-intensive rendering.)
5: Will my MIT X server run under the SparcStation 4 ?
The short, direct answer is "NO", the MIT X window server does not
support the TCX framebuffer, with no indication at this point from
the X consortium that it will be supported anywhere in the near
future (partly due to non-disclosure of the TCX intrinsics).
There is an exception to this, however. Since under SunOS 4.1.4,
the driver will driver the framebuffer as a CG3, which *is*
supported by the MIT X server, one can therefore run the server
on an SS4 under SunOS 4. The R5 server requires the JPL "multiscreen"
patch to work properly (see #80 in the X FAQ found in the usenet
newsgroup, comp.windows.x for availability info; also found via
ftp:ftp.x.org/R5contrib/Xsun.multi-screen). As reported by Sun :
All the "R5.Xsun.multi-screen" patches do is clean up
Xsun's initialization code so that it recognizes that
"/dev/fb" is a "CG3"-type device and it can use its
"CG3" ddx code to drive it. This improved initialization
code is included in R6 (no patch necessary).
I have not actually tried R6 under SunOS 4 to validate this.
This route will NOT work under Solaris 2.4, since the driver
does not present a "CG3" :
The patch doesn't add Solaris 2.X native TCX support.
R6 doesn't have Solaris 2.X native TCX support. Solaris
2.4 doesn't provide CG3 emulation of TCX (doesn't need it).
Customers will be expected to simply run the OpenWindows server
under Solaris 2 to get the performance and applications they need:
The MIT sample server is just not always going to be an
option. That's why vendors ship X server products. If you
want to get a feel for what the full performance of this
display device is, just run the Solaris 2.4 X server (Xsun).
It's very standard and very directly based on MIT X11R5 -
with the added value of *full* support for the range of Sun
display devices including that of the SS4 (not to mention QA,
vendor support, Display PostScript, etc.).
And for those adventurous to want to write their own drivers, our
response was simply, "The low-level display hardware programming
interfaces are not public."
I'll let this sit as is....
6: Will "Doom" run on the SparcStation 4 ?
The all important question. Running Solaris 2.4 with the audiocs
patch (102125-01) [rev 02 is now available] and OpenWindows, the
X version of "Doom" was able to run with good performance. The
"DGA" version, however, dumped core for us (as though it was
running under a straight X server). I'll leave this to someone
else to file a service order.
Regrettably, our labs are still running SunOS 4, given that alot of
software won't run and/or is not presently available for Solaris 2.
We also run the MIT X server under both SunOS 4 [R5] and Solaris 2
[R6] as numerous software that we have deployed do not properly work
under OpenWindows (either 3.0 or 3.4, the X11R5 based server). We
will subsequently be returning our entire lot [15+] (and thus, I
won't be tracking progress on future SS4 developments).
I realize changes are often necessary, but I would advise Sun to take
a lighter attitude towards simply expecting customers to adopt their
applications/work-environments just to make any use of new hardware,
as this is not always practical or feasible for those customers. We
also ran into major mis-communications (or lack of) regarding new
products and local representatives who deal directly with customers.
Having this improved would save many people time and headaches with
this information available up front.
I have concluded....
-- Eric Berggren | "Parts of this product may be derived from Portland State University | from UNIX and Berkeley 4.3 BSD systems..." email@example.com | -- Label on Solaris 2.x disk
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:24 CDT