Fabian@cup.portal.com (Fabian Fabe Ramirez) (01/20/90)
Copyright 1989 (c) Austin Macintosh Developers Association THE GREAT 32-BIT VIDEO CARD QUICKDRAW SHOOTOUT By Vincent Parsons Review by Frank Kurzawa [Editor's note: Vincent's last talk on Extending Excel has just been published in the January 1990 MacTutor. Congratulations! Vincent is currently extending his Hydra program and talking to a number of video card vendors to extend the results reported here into an article to be published in a Macintosh journal. Among some of the operations planned for the expanded article are CopyBits from an offscreen buffer to an onscreen window, short horizontal lines, and long vertical lines. Other suggestions for enhancements are welcomed. Keep tuned.] In the past several months, both SuperMac and RasterOps have come out with cheap (<$1000.00) 32-bit video cards capable of driving the Apple Color Monitor or its equivalent. In fact, SuperMac also offers an optional daughter-board QuickDraw accelerator to complement its color card. We were curious how well these boards performed. How do they compare with standard 8-bit performance? And how do they compare with each other? THE COMPETITION So at our last meeting we had a QuickDraw contest between five separate systems. Vincent Parsons wrote a small C++ MacApp program called Hydra which he designed to compare drawing performance on several monitors in sequence. At the meeting we had a Standard Apple 8-bit color card, a SuperMac ColorCard/24, a second SuperMac ColorCard/24 with the optional accelerator, a RasterOps ColorBoard 264, and an older model TrueVision card: the NuVista. All cards were driving their own Apple Color Monitor except the Truevision which was driving a Sony Monitor. The MacApp program simply ran a selected QuickDraw activity, such as CopyBits or FillRect, on each monitor in succession and displayed the time required on each monitor. The user could select the QuickDraw command from a menu. THE RESULTS For those who want to see all the gory numbers, the results are tabulated in the table on page 5. The raw numbers in the first section represent tick-counts. The second section show ratios based on the Apple Color Card being 1.0. The third section gives a few comments on the operations. WHAT IT MEANS A hardware junkie can have a field day with these numbers, making comparisons to determine what the boards are doing on the inside. But a few major points stand out. First, and most obvious, the Accelerated SuperMac card is an incredibly impressive performer; on many operations, it outperforms the Apple Video Card which runs in 8-bit mode. Second, while a lot of data is presented here, it's important to remember that actual performance as perceived by a user performing normal tasks is most affected by the speed of CopyBits and ScrollRect operations. These are the operations required when a user scrolls through a document or screen redraws are performed. Once again, the Accelerated SuperMac Card far and away outperforms the other 24-bit cards in ScrollRect and 1:1 CopyBits. In fact, it outperforms the Apple 8-bit card! Although the RasterOps card was slightly faster than the unaccelerated SuperMac card, the two are roughly equal performers. And while the TrueVision card came in at a very solid last place, it is an older card. TrueVision has come out with several newer cards which may be much better performers. [Editors Note: One disadvantage of the SuperMac card is its large power draw. The ColorCard/24 draws 2.5 amps and the accelerator draws 1.5 amps - well in excess of the 2 amps allocated per NuBus slot. this could cause problems in "low-power" Macs such as the Mac IIcx. The RasterOps ColorBoard 264 draws only a little over 1 amp.] WHAT THE MANUFACTURERS THINK The day after our meeting, Ron Parsons took our compiled results and sent them to both SuperMac and RasterOps. He received a very strong response. SuperMac, in particular, is interested in the nature of the test program we used and is interested in enhancements. Macintosh News has since called to obtain more information on the tests. RasterOps has taken the position with Macintosh News that acceleration is unimportant since CopyBits of offscreen bitmaps to the screen is the predominant operation that affects performance and this operations is already optimized. But this position ignores the fact that scrolling (ScrollRect) is a separate important operation that has a significant impact on performance. Quite aside from the benchmark tests, if a user simply tries each video card with a common program such as Microsoft Word, the difference is speed between the accelerated and unaccelerated SuperMac cards is quite substantial. Obviously, factors beyond raw performance enter into the purchase of a card. Price, reliability, warranty, etc. are all important factors which we haven't touched on. For example, the accelerator adds several hundred dollars to the price of the SuperMac outfit. However, these numbers may be a useful as a starting point towards a decision. AMDA P.O. Box 50447 Austin, TX 78763