Announced in Aug. 1998, shrink-wrapped versions of the Rage Fury graphics card finally hit the shelves in February '99. Here's what's inside....
Touted as "the world's fastest graphics accelerator" when it was announced in August 1998, the Rage 128 chip at the heart of the Rage Fury created a stir in the graphics world, with its promise of unsurpassed 2D performance and 3D acceleration. But beta versions suffered from buggy drivers and troublesome configuration routines. We ran several different driver iterations through our usual battery of tests to find out whether these prerelease glitches were ironed out before shipping or whether Bennett's "never buy v1.0 of anything" Law still holds true.
The Rage Fury, as shipped, includes drivers for Windows 95, 98 and NT, as ATI promised it would. Also included were four CDs: a driver disc (version 100 - uh-oh!) and three games: Half-Life Day One, MotoRacer2 and Expendable "Lite." Our box also included a slim installation manual, a US$30 rebate offer against the purchase of certain other ATI products and both S-VHS and RCA cables for use with the card's TV out connectors. Missing, however, from our box was the DVD driver that the package promised would "quickly and easily" have us watching DVD movies. Instead, we received a coupon, to be returned with proof of purchase, in exchange for the DVD player. Quickly and easily, but please allow 3 weeks for delivery. Whether this is a ploy to get owners to register or an issue related to a not-quite-ready DVD Player, we were not sure, so we asked ATI.
A company spokesperson explained that it elected not to ship the DVD Player in the Rage Fury box because it has to pay a royalty on every copy shipped. The company subsequently reconsidered this and Rage Fury retail packs are now shipping with the DVD Player missing from current boxes. (It's still not available for download, however.) Of course, there have been numerous driver revisions since then, fixing a number of bugs, including the widely reported spurious dithering problem we noted in our Rage 128GL image quality tests (caused, claims company spokesperson Brian Hentschel, by a missing "green bit") and adding driver certification for 3D Studio MAX 2.5 -- which misbehaved badly under the 1.01 driver release, along with Caligari trueSpace 4, Accolade Test Drive 5, and several other well-known games, including Microsoft's Monster Truck Madness 2 and Rainbow Six. Even Quake II required some manual tweaking to work properly.
The retail card has a heatsink -- reportedly added primarily to appease some reviewers who complained that the original version ran hot. Also, the heavily posterized 16-bit display problem we noted in the Jan. 25th beta is gone -- replaced, unfortunately, by what appears to be the Dec. 98 beta's "heavily dithered" algorithm. Why ATI can't make a 16-bit screen as smooth as that provided by Nvidia, Matrox or 3Dfx cards (etc.), we can't understand. At any rate, 16-bit 3D displays look rough, but 32-bit 3D displays look incredible and 32-bit mode performance is, as ATI claims, just as fast. For more on this topic, see our previous reports, listed below. In 2D modes, the fidelity of the images produced by the Rage Fury was nothing short of excellent.
However, since installing our first Rage 128-based card in late 1998, we upgraded drivers perhaps a half dozen times before we achieved a reasonable degree of stability and bug-free operation. It's safe to say that at least a year went by before the card worked properly -- and early adopters of Windows 2000 are still waiting for final drivers for the All in Wonder 128.
Indeed, back in Jan. 1999, we expressed some serious concerns about the stability of the shipped "1.00" drivers (dated Jan. 27th). We tested the card in two otherwise stable machines. One was a Socket 7 (Pentium) system based on an Asus P5A-B motherboard; the other, an Abit BX6 running a Pentium II/266. In both cases, the machines began misbehaving as soon as we inserted the Rage Fury card and stopped malfunctioning immediately after replacing it with a different graphics card. Hmmm!
The Pentium-based system was particularly troublesome. With the CD-based drivers, it would spontaneously crash and/or reset every five minutes or so; with the updated drivers from ATI's website, it didn't work at all -- the best we could get was a blank screen. This behavior stopped as soon as we swapped out the Rage Fury in favour of a different card, be it Permedia 2, Voodoo Banshee or TNT-based (etc). We even tried updating the Asus board's BIOS to a version that Asus claimed addressed a "crash problem with the Rage 128," to no avail.
The Pentium II system had its own quirks. Initially, the card appeared to be working reasonably well, but while attempting to resolve a conflict between the Rage Fury and a network card -- both of which insisted on using IRQ 11 in our system -- we used the tried-and-true trick of reserving that IRQ in the Device Manager, by clicking on the Resources tab, typing in the number 11 and then restarting. This produced an unusual effect: the Rage 128 display went entirely blank each time Windows subsequently loaded. We had to boot to Safe mode and choose a generic VGA driver to recover. Again, pulling the Rage Fury and inserting a different graphics card (a TNT-based Spectra 2500, this time) solved the problem -- and, using the "Reserve IRQ" trick mentioned here, we quickly re-allocated the TNT card's IRQ and easily resolved our network card conflict.
Thus, it is clear that the Rage 128 drivers, as shipped, were not yet ready for prime time. Not surprisingly, it didn't take ATI long to release an update and, on Feb. 23rd, the ATI support website posted a new set of drivers for Win95/98; followed on Mar. 22nd by another update known as Win9x 6.10-A4V (CD89), which the company said addresses Socket7 board compatibility problems. Then, on March 29th, another update appeared (6.11-A1E), adding compatibility with the ATI TV Wonder. Then, on March 31st, yet another update appeared: 6.10.1-A5F (CD93). This one contained an entirely new Open GL ICD, D3D and OpenGL property tabs and claimed 100% compatibility with Quake3, according to information on the Rage 3D website. You certainly have to give them points for trying. According to ATI's Brian Hentschel, the earlier release versions of the Rage 128 drivers did not fully support video overlays. For example, TV tuner and video capture boards would not work properly. The latest drivers have reportedly corrected this. (For more info, see www.rage3d.com)
Our test results of the CD89 drivers revealed a crash that occurs when a 3D program such as trueSpace 4's or 3D Max's Open GL display option is selected -- a problem we also noted with the beta releases. Also as noted in previous releases was an apparent glitch that causes Windows' dialog boxes to display occasionally incorrect colours (the shadow behind the little red "application has crashed" symbol is yellow instead of gray, for example) when the Rage Fury is in 32-bit colour mode.
Several changes from the Jan. driver and 02-23-99 update we tested were apparent in CD93. Most obviously, the Mar. 22 release displayed a different icon in the Win9x taskbar, and did away with the annoying "Game Engine Performance" options for Quake II and Direct 3D games that you have to, in the boxed retail drivers, manually turn on and off via the Taskbar, ignoring at your peril the warning that leaving these options on all the time can cause system stability problems. The TV feature also works. (ATI media relations rep Brian Hentschel seemed surprised to hear that it wasn't working in the 02-23-99 drivers, as plainly noted in the readme file on ATI's own website.) For reasons we can't begin to fathom, the heavily posterized 16-bit display problem we noted in the Jan. 25th beta was back, too. (A registry tweak noted in the LightSpeed FAQ can turn 16-bit dithering back on, if you prefer.)
ATI was busily releasing beta drivers throughout April; the most recent versions finally addressed the incompatibility with our Asus motherboard. As recently as version 6.11.1, the drivers still displayed some anomalies. In a dual-display configuration, with a Permedia2 configured as the primary display and the Rage Fury set as the secondary display, the Windows 98 message on the Rage Fury startup screen indicated that it had "successfully initialized the display adapter." However, the Rage Fury didn't show up in the Display control panel's Settings. Instead, a message warned us that we needed to remove EMM386 from our Config.sys and restart. Unfortunately, we weren't running it, or any other such memory manager. Both adapters did, however, work properly with the Rage Fury configured as the primary display. More recently, ATI released, in late Oct. 1999, "special purpose" drivers for Windows 95b+98 and Windows NT4.0, with optimizations for K6/K7 3DNow!, Pentium III SSE, Quake3 Test. These drivers finally implement a full OpenGL ICD. As mentioned earlier, subsequent driver releases finally started to shape up into something resembling a stable product.
While it's safe to say these won't be the last updates for the Rage Fury, they are, in general, an improvement. But we have to wonder: would the average user have put up with this much hassle -- or packed the thing up after the second or third upgrade attempt and asked for a refund? At any rate, we'd certainly recommend installing the latest updates from support.atitech.ca right off the bat.
Sidebar:
ATI now says that its Xclaim VR 128 for the Mac, originally expected to ship over half a year ago, is, as of Oct. 28, 1999, finally shipping. Also available: the 32 MB version of the All in Wonder 128 and the Rage 128 Maxx, released in Dec. 1999. Notably, the MAXX requires two IRQs and isn't currently compatible with Windows 2000 or Linux. PC Insight has details.
For further Reading
External Links:
Post new comment