8514/A Adapter Experience

XGA - The Solution to the Video Subsystem Bottleneck
A Nifty OS/2 utility
Monitor Selection & Frustration
8514/A Explained... Sort of, Maybe
From the OS/2 Warp Unleashed Pages
A Note on Benchmarks

Content by Tim O'Connor (original archived HERE). Edited by Major Tom.


XGA - The Solution to the Video Subsystem Bottleneck

I confess - the best answer to curing a video system bottleneck in the 8570 was to use an XGA-2 card. I didn't want to toss the 8514/A info though, just because I wasn't using the adapter any longer. XGA offered me several benefits: It's a directly accessible card, unlike 8514/A, and thus doesn't have all the overhead involved with the older card. Screen resolution can be a nice, easy-to-read 640x480x256.


A Nifty OS/2 Utility

If (big if) you use OS/2 v 2.1x or above and you have 8514/A, there is a utility which might help you out with the WPS redrawing the entire screen:

FASTSS.ZIP

I've noticed good results with this.


Monitor Selection & Frustration

IBM at one time published the PC Assistant series of books in *.inf format (viewable with OS/2 or PC-DOS VIEW.EXE). The first monitor I tried with 8514/A was an IBM 8518 monitor. According to PC Assistant, this monitor was capable of 8514 1024x768x256 resolution. It's not. I'm currently using it with XGA and have found that with XGA the monitor only supports 640x480 and 640x450 modes. Rumor has it that the best monitor to use with the adapter is the 8514 display - although I've made do with a Packard Bell 1010 Multisync display that seems to do the trick.


8514/A Explained... Sort of, Maybe

Other people have explained this far better than I can (but I guess I'll try my shot at it, too). Keep in mind, my experiences with 8514/A are (so far) OS/2 experiences without WinOS/2 complicating things. So without further ado - The cheap & cheesy approach, extensively quoting from an e-mail with a friend who was used to more 'typical' video cards:

<begin 'way boring bit on 8514/A>
Ah... the mysteries of the 8514/A adapter card... 1024x768x256 (or else) Odd card, for most folks. I'm kinda stuck with the resolution/number of colors... it goes something like this: The base PS/2 VGA subsystem is on the planar itself, has 256KB of video memory. Now when you plug in an 8514/A (which can only go into the MCA slot with the Video Extension - one slot per machine) it adds to the VGA subsystem, and itself only handles 8514/A standard calls to the... 8514/A interface. VGA & text graphics on a machine with an 8514/A are actually handled by the machine's own VGA processor, passing the signals through the 8514 card and on to the monitor. Then number of colors the card generates is determined by how much memory it has - 512KB gets you 1024x768x16, and 1024KB gets you 1024x768x256. There are no other choices. On the Screen tab of my System Configuration Notebook object it lists one, exactly one resolution/color option: 1024x768x256.

My choices with this are, well, limited. If I apply VGA drivers & use the system as a VGA system... the 8514 card does nothing other than pass the signals on to the monitor... and I'm using the machine's rather limited 256KB of video memory. The only way I can drop the number of colors is to remove half the memory on the card... neither here nor there, other than risking damaging the card itself (it's a sandwich of the main adapter & daughterboard with memory). 1024x768 is the only resolution that it knows, all others are passed off to the system VGA... (OS/2 full screen session, for example)

Pretty funky, no? This thing actually gave me headaches trying to figure out how I could run VGA resolution (640x480x16) *and* take advantage of the memory on the card. Uh-uh, I was asking it to do something it's not designed to do...

As always, if I'm wrong, please let me know.


From the OS/2 Warp Unleashed Pages

IBM introduced the 8514/A the same time as the VGA to complement the VGA adapter. The 8514/A provides 1024 x 768 resolution in either 16 or 256 colors. Software support for the 8514/A got off to a slow start because the hardware interface specification was not published. This reason alone probably contributes to the wide number of SVGA designs rather than designs compatible with the 8514/A.

Despite the lack of information on the hardware interface, the introduction of the ATI 8514/Ultra and device drivers for Microsoft Windows and the OS/2 operating system make the 8514/A architecture a common choice for many users.

Regarding a choice of video adapter for use with OS/2, you should be aware of the following concerns:

  • IBM has withdrawn the 8514/A video adapter card. The ATI 8514/Ultra is still available, although it is being replaced with newer designs such as the MACH32 and MACH64.
  • The hardware architecture is not well suited to 32-bit operating systems.

This last point is the most critical for future compatibility. The 8514/A does not allow any operating system or application direct access to the video memory buffer. All access is through I/O ports (this is significantly slower than adressing memory directly). Additionally, I/O operations are priviledged in OS/2 Warp, and a process known as a ring transition must take place to switch the processor from user mode (with hardware protection) to kernel mode (when hardware access is permitted). This makes the 8514/A slower than designs that enable direct memory access.

From pp545-546 OS/2 Warp Unleashed, copyright 1995, Sams Publishing. Moskowitz, Kerr, et. al. This taken from The Video Subsystem chapter written by Bill Bodin.


A Note on Benchmarks

Interestingly - the Model 70 with the 8514/A card gets about 12 PM-Marks with Helmsley's SysBench tool, whereas my big 77 with XGA typically gets in the sevens. Why would I mention such a thing? Because benchmarking a thing can be taken out of context & raw scores alone say nothing: my 77 offers better performance overall, for a lot of different reasons. In short - Benchmark tools are great to see if a measure to tune a system has had any effect, but not necessarily valuable comparing machines... like judging machines by their processor speed alone, a misguided judgement.

Content created and/or collected by:
Louis F. Ohland, Peter H. Wendt, David L. Beem, William R. Walsh, Tatsuo Sunagawa, Tomáš Slavotínek, Jim Shorney, Tim N. Clarke, Kevin Bowling, and many others.

Ardent Tool of Capitalism is maintained by Tomáš Slavotínek.
Last update: 24 Mar 2024 - Changelog | About | Legal & Contact