November 22 - 30, 1998
Archive

Matrox regrets to announce that it has pushed back the release of  its OpenGL ICD to Wednesday, December 9, 1998. On this date, Matrox will release a pre-beta version with support for Windows 95/98 only. Windows NT 4.0 support is expected during Q1'99.

While the pre-beta ICD will not exceed the performance of the currently available Matrox Direct3D Wrapper, Matrox wanted to assure customers that, despite the unfortunate delays, we are aggressively working on the ICD development under Windows 95/98 and NT in order to deliver the maximum performance possible for MGA-G200 based graphics accelerators.

This pre-beta ICD released on December 9, 1998 will support a limited number of OpenGL games and applications including, Quake II and Half Life.

 

Issues Addressed/Features Added

DirectX 6 support added
Fixed AOL lockup issue in modes other than 8 bit color depths
Macrovision support added for improved DVD quality
Updated Glide to version 2.54
Added 200MHz modes (640x480, 800x600)

• Diamond Monster Fusion Driver (Win9X - version 4.10.01.0211 - 3.54 MB)
• Diamond Monster Fusion BIOS Update (AGP - version 1.20b4803 - 240 KB)
• Diamond Monster Fusion BIOS Update (PCI - version 1.20b4002 - 240 KB)

So, apart from the TNT, the Banshee is pretty much without competition. So, good job that the drivers and software seem reasonable too. The control panel element of the product too seeming even better than that prevalent in the Maxi Gamer2. Of course, there are other Banshee boards, hell, some are even cheaper than Guillemot's offering, but the Phoenix has one thing going for it that not even Creative's cut-price answer can beat. Half Life Day One, what else? Yes ladies and gentlemen it's in there, and even if I do say so myself, this board is worth buying because of it's inclusion alone! Not only as a showcase for it's power and Voodoo2-esque image quality (not to mention that 1024x768 resolution), but also because it could well turn out to be the best first-person shooter of the year.

 

Ease of installation, mature software drivers, full optimization for DirectX 6, full 3D feature set, AGP 2x compliance and mind-blowing performance at high resolution and colour bit-depth speak volumes for the Viper. Combine this with the fact that this is a board that provides both Matrox-quality 2D and greater-than 3dfx 3D capability for a little more than the cost of a Voodoo2 3D-only board, this would be a wise hardware purchase. Economy apart (for hardware buyers that are unconcerned with cost) the Diamond Viper V550 is a very well designed and manufactured component that provides the ultimate in 2D and 3D quality and performance.

 

-HALF LIFE
-V3 vs. 32-BIT
-D3 & THIEF DEMOS
-SIN REVIEW
-FLIGHT SIMS COMPARISON
-TOMB3 DEMO
-HAPPY ANNIVERSARY!
-...and the usual ramblings.

 

The most significant change in the current crop's makeup (aside from the new core chips) is that it includes much more on-board memory than in the past. Whereas 2MB and 4MB of memory were the norm at this price level a year or two ago, plummeting memory prices now let vendors equip their boards with 8MB, 16MB, or even (as with the $200 Number Nine Revolution IV 32MB) 32MB of dedicated graphics RAM and still deliver affordable products. This extra memory is necessary for memory-hungry 3-D rendering and also enables the cards to deliver higher resolutions and color depths for the 17- and 19-inch monitors that have become more commonplace.

The glasses did exactly what they're supposed to: make everything more 3D. To sum it up:

I was actually flinching when objects were traveling at the screen in Unreal, Final Fantasy VII's map seemed a lot more interactive and the game seemed more fun, Quake II and GLQuake were unbelievable; almost like VR, and the trucks in Monster Truck Madness were just plain scary!

The only problem that I had was the strain my eyes felt after an hour or so of Quake II.

Joystick, a french game magazine, scored some benchmark on ATI Rage 128 in the December issue.

Here are the score (ATI Rage Fury / STB Velocity 4400)
D3D 3D Mark Pro 99 640*480 16 bpp   2909 / 2729
D3D 3D Mark Pro 99 640*480 32 bpp  2529 / 2236
D3D 3D Mark Pro 99 1024*768 16 bpp  1834 / 1495
D3D 3D Mark Pro 99 1024*768 32 bpp  1493 / 1108

------------

OGL Quake II 640*480 16 bpp  65.5 / 76.2
OGL Quake II 640*480 32 bpp  67.8 / 61.6
OGL Quake II 1024*768 16 bpp  34.9 / 38.1
OGL Quake II 1024*768 32 bpp  34 / 22

 

To our suprise this processor made it to 558MHz (4.5 X 124) first then later successfully booted at 598MHz (4.5 X 133). Windows 98 ran but it was very very buggy and lots of crashes occoured. We suspect the RAM to be the blamed piece of hardware because the maximum bus the ram was designed to run was at 125MHz. It is American Megatrand's CAS2 8ns 125MHz RAM.

 

When it comes to FPU power, it looks like Intel won't be alone in the x86 market. The K7 features 3 floating point units (Load, Add, Multiply) each fully pipelined and superscalar. The latencies on the Add and Multiply are both 4 cycles, higher add latency than the PII (PII has 3 cycle), but lower multiply (PII has 5 cycle). The pipelined FPU should help give AMD a high throughput rate in intense FPU applications, especially when combined with the already advanced K7 core. (Out-of-order execution especially helps maintain high throughput (it tries to keep the pipelines full by executing instructions "out of order") Will the K7's FPU be faster than Intel's? It's hard to tell right now, but find out what I think later in the article.

 

• Matrox Unified BIOS (version 3.45 - 722 KB)

 

Sharky Extreme has put together a very good Intel Celeron overclocking guide.  As I'm sure you all know, getting a Celeron 300A up to 450MHz is a very easy task.  This guide gives you all the details and guides you step by step to achieve 450MHz speeds.

The K7 and the K6-3 will be produced in parallel. Just about everyone is in agreement that the K7 will be the most powerful architecture available in 1999, but an inordinate number of them are not aware of the fact that the K7 is not going to be a "25% less than Pentium II" priced product. Indeed, the factor that allows the K6-2 to be far cheaper than the Pentium II is that it costs far less to make (with a die size of about 70sqmm compared to the Celeron at something like 130sqmm). The K7, on a quarter micron process, will have a die size of 184sqmm, far above even Intel's offerings. At this point (around midyear, perhaps a little earlier), the K7 will be an expensive product, suitable for high end uses like servers, workstations, and insane gamer boxes. The chip will not be made in very massive quantities until about three months after when it migrates to a 0.18um process. At this point, the K7 will be comparable to a Katmai/Coppermine in die size (and likely somewhere similar in production price), and will be buyable as a mid-to-high end product. During the first few months of the K7, the K6-3 will be the middle end product line competing directly against the Katmai (the K6-2 might still be around as a competitor to Intel's Celeron line). When the K7 makes the jump to 0.18, so will the K6-3. At this point, the K6-2 will likely be gone and the K6-3 will take its place as the low end solution, placed directly against Intel's mid end Coppermine processor. AMD has been confirmed to say that there will be a budget version of the K7; however, the news stories that passed along this info ignored the fact that AMD also said that this would likely not be the case until the year 2000.

 

The results show that the TNT chip set uses about 90%-100% of the CPU when running these tests.

The Permedia2 and GMX chip sets uses less CPU than the TNT on all three of the tests.  The GMX on the CDRS test used about 5% on the 5 poly tests and 20% on the 2 wireframe tests.  The Permedia2 used about 40% on the CDRS.    On the DRV test the the Permedia2 and GMX are at about 80%.  On the Fog City test the Permedia2 and GMX have about the same level of CPU use  ~55% .

Clearly the TNT is a CPU hog but is that bad?    With the current cpu's ( i.e. Pentium II ) running  CAD applications the TNT's CPU load may cause performance problems.  With current games it is not so clear. However the next generation of games will be more poly intensive and thus a little more like CAD apps.

A faster CPU would make the TNT perform much better;   the GMX and Permedia2 performance would not be improved as much.  The current P2 FPU is designed for double precision ( 64 ) floating point and in the main, graphics computations need only 32 bit floating point.  The next generation P2 ( Katmai ) has a revised FPU to speed up these types of geometric calculations.  Thus the TNT might perform significantly better with this type of CPU. A description of the Katmai is at Tom's Hardware

 

Compared to a Voodoo2, the Banshee fares quite well. In games that take advantage of multitexturing, such as Quake 2 and Unreal, the Banshee takes a performance hit relative to the Voodoo2. In all single texturing games (most other games), the Banshee is slightly faster because its core and memory clock are faster than that of a Voodoo2.

 

• Guillemot MAXI Gamer Pheonix Driver (Win9X - version 1.03 - 2.0 MB)

How can you take advantage of the improvements found in the new CXT core?  Unfortunately it isn't as simple as popping in a new chip, you will have to make an update to your BIOS to recognize the processor as well as to enable the enhanced WHCR and the Write Merge Buffer.  As you'll be able to see from the tests AnandTech conducted, the K6-2 400 using the CXT core exhibited a 7% increase in overall performance in comparison to a K6-2 400 without the modifications enabled in the WHCR and with the Write Merge Buffer.  As far as gaming performance goes, the K6-2 400 doesn't really benefit all that much from the Write Merge Buffer, although you may see a few fps increase in performance in a few of your games, nothing too major.

 

The last big roundup on this site has been the one from July 98, so it was about time to get all cards lining up once again. Since then new chipsets have emerged and drivers from existing chipsets been optimized for better performance. This roundup covers about all major chipsets including Riva ZX, Riva TNT, Voodoo I, Voodoo II, Voodoo Banshee, Intel 740, Ticket To Ride 4, Savage3D and G200. So, let the cards come out into the open and let the truth be known.. Results from the K6-2 test system will be added in a few days as they aren't finished yet. So, here's the list of cards that I've looked at:

 

The Forzando Plus speakers are really designed for people on a budget, and for that they fit the bill quite nicely. These are not supposed to be the "Balls-to-the-wall" speakers, like the 4100’s. No, they are hitting a price point that most people can justify to themselves, because these are usually going to be the same people that spend less than $1000.00 on their whole computer.

"The primary focus of these drivers were to fix the various freezing problems some I740 owners were plagued with. Third party Chipset support was finally added for those not on an Intel motherboard, the Intel740 as Primary adapter in multimonitor systems is enabled, and a whole bunch of other fixes are included"

• Intel I740 Reference Driver (Win9X - version 3.1 - 871 KB)

Midiland's S2 series has been widely acclaimed for its amazing sound quality. However, one of the biggest complaints about their speakers is that they are just a little too expensive. I have to agree in this case - there are other systems that cost a lot less but sound almost as good. Of course, if sound quality is extremely important to you, you may just decide to take the plunge with the S2/4030 set because that is where they excel - especially if you don't listen to much bass-heavy music.

 

• 3DFX MiniGL (210 KB)

 

New features added:

1. Frequency Shifter Effect - allows you to create bizarre distortions, phaser, and rotating speaker effects.
2. Number of software voices can now be adjusted.

• Sound Blaster Live! Driver (Win9X - Revision 3 - 2.6 MB)

 

By now you're probably getting the clue that we really thought the ASUS P2B-LS was a fantastic product. It is. We test a lot of products each week for Sharky Extreme and there aren't very many that achieve a 4.0 or 4.5 "sharkfin" rating from us. We're tougher than most online publications when doling out high praise, our feeling is that it takes a superior product to get us to spout accolades. (We don't want to be like some of the less scrupulous movie reviewers who give wonderful sound bites to movies that you know are crap….Think of us as the online Siskel and Eberts of PC hardware) The P2B-LS deserves very high praise for its fantastic support, its pain free installation, its impressive combination of LAN and U2W SCSI and its wonderful instructional guides.

 

Overview 
-Samples: Nov '98, Production: March '99
-Supports AGP 4X (Savage 2+) and 2X (Savage 2)
-Major features:  AGP 4X, 32MB memory, 32/64-bit memory i/f
-Improved 143 MHz SGRAM i/f (up from 125MHz with Savage3D), continues to use block-write feature
-Video capability: unchanged from Savage 3D (i.e. h/w subpicture blending, motion comp, 60Hz VIP port for HDTV)
-270MHz DAC:  1600x1200@100Hz, 1920x1440@72Hz

 

Even after so many successes throughout the year, profits continued to elude the company. Either the volume of products sold had to increase, or the average selling price had to rise, or both. Volume was not a problem, yields were doing remarkably well and demand was strong, but AMD needed to earn more for each processor that went out the door. To do that, they needed a product with performance that was not merely comparable, or equivalent. If AMD was to command a higher price, the company would require a processor that performed better than Intel's line when running the exact same programs.

All of these features together made quite an impact on the SOHO community when the Epson Stylus 600 and 800 series printers hit the market, so why on earth would it be necessary to revise the plan yet again and offer the Stylus Color 850?  The answer is speed.  In order to really knock the socks off the business community, Epson had to make their printers fast on top of all of the printing advances that were already present in the Stylus 600 and 800 printers.  It was one thing to have brilliant printing, but business customers needed it to be fast as well.  The Stylus Color 850 boasts 9ppm (pages per minute) for text printing and 7ppm for color, even at the higher resolutions.

 

Electronic Arts really put a lot of detail into the game, and it shows even more when you have a Force Feedback device. They have all sorts of textures in the road that start to hassle you when you have Force Feedback on. The aforementioned covered bridge is a very good example of it. I turned the road grip down a little bit and ran into a few problems, namely losing control on some curves as the wheel decided it wanted to go left when I wanted to be right. There is a lot of good employment of "The Force" in NFS3. Slamming into walls, or getting put there by Smith in his Diablo, or driving in poor weather conditions brings The Force into play in a serious way.

 

OK, enough rants. Running on identical machines, I first started by running a standard Quake II timedemo. Over three consecutive Timedemos with nothing in the background, the SBLive clocked in an average of 102.5 fps. The MX300 stayed a rock-solid 100.7 for all three tests. Timedemo2 showed nearly the same result, with the SBLive at 102.4 and the MX300 at 100.1. Unreal Timedemo yielded the similar results, with both cards clocking in between 45 and 47 fps through three passes of the intro flyby.

• Quantum3D Raven Driver (WinNT4 - version 1.2)

 

Tell us a little about the Voodoo3 and how it managed to pull off the performance lead that it has.

The Voodoo3 is really an integration effort that has taken the Voodoo2 SLI technology, which is 6 chips, and integrated the Voodoo Banshee technology. The end result is a single chip that is basically 2-3 times faster than our existing products. What we are seeing right now is Quake 2 is 1600x1200 at 30fps with multi-texturing. We are really giving the technology to beat over the next few years and we thing we have really set the bar for performance.

Do you think that we will see a hold on the 180MHz V3 3000? I do not think that there will be lot of that 183MHz memory on the market when the card comes out.

The chip is going to be going into production by 2^nd quarter of next year, and we believe by that time there will definitely be memory to support it. The key here is timing your technology and your speed with the memory speed that is out there, and we work closely with the vendors to make sure that the memory will be out there.

And that's your lot. The Voodoo3 is basically a 3Dfx Interactive based product which will perform a little bit better than a dual Voodoo2 SLI configuration, except for the fact that it's on a single board. Although we've seen Quantum 3D do the 'single board SLI' configuration before with their Obsidian2 X-24, the price was way too high for most gamers. In addition there was no 2D part or support for DVD full motion video playback either. Can't be bad eh? Well as long as the competition doesn't do more than release 'Voodoo2 SLI Killer' press releases and actually deliver a product that outperforms the Voodoo3, then 3Dfx looks to be on to another money maker with Voodoo3. Or does it? Until 32-bit rendering can be achieved at an acceptable level, speed is still king and the Voodoo3 will be the fastest 3D performer out there. Comdex '98 didn't indicate that 3D chip vendors had anything that could scale well against the Voodoo3 in terms of speed. The rest of the 3D chip manufacturing pack is banking on improved image quality being the dish of the day. The specs of the other next-gen chipsets didn't really threaten the Voodoo3 in the speed stakes.

 

• Matrox G200 Driver (Win9X - version 4.50 Beta - 1.7 MB)

 

When I first received this pair of samples from AOpen, I was told they were capable of booting at the 133 MHz bus speed. Naturally, that was the first test I tried, a complete failure. Over the course of the past month or two, I dug up any information I could on the chips used on these modules. Sure enough, they're not 7 or 8ns, as can be seen on AOpen's page (screenshot below), but a 10ns cycle time. It's no mystery why 133 MHz, or even 124 and 112 MHz wouldn't even POST. A cycle time of 10ns is the bare minimum to meet the PC100 specification, meaning these chips cannot operate at over 100 MHz.

 

Adaptec’s Threadmark benchmark was used to test the maximum throughput of the hard drives, working them to their full potential. Threadmark testing is where the DiamondMax really flexes its muscles, showing the largest performance increase of all the tests, 37%. Why is there such a drastic difference in performance improvement between Quake II and Threadmark (22% vs. 37%)? This is related to the fact that Quake II doesn’t rely as heavily on the hard drive (for example: data is loaded, something is executed, and then more data is loaded). This is similar to the reason why AMD’s 3Dnow! Instruction set doesn’t offer four times as much performance even though it supports four times as much bandwidth.

 

• DirectControl Version 2.0.8:
• Added support for DirectX 6.1.
• Added support for DirectX Viewer and DirectX Texture Tool
• Added 'Set DirectX Tools Path' option.
• Added support for DirectPlay settings (Break on Assert & Debug Levels)
• Added support for DirectSound3D
• New DirectX Driver Version dialog box
• Added 'Refresh' option (this allows DirectControl to re-read registry settings if another application has made changes, i.e. DirectX Control Panel) NOTE: Not all visuals (checkmarks) are refreshed at this time.
• Corrected 'Remove Prefetch Settings' checkmarks
• Corrected known problem : If VCACHE settings are NOT detected or present, DirectControl will display checkmarks for all memory settings.

• DirectControl (Win9X - version 2.0.8 -112 KB)

• Matrox Overclock Version 2.0.8a:
• Added 'Command Line' text box. This displays the command line options that MILLCLK/MYSTCLK is using
• Added 'Reset' option for Powerdesk settings. This allows the user to reset all the Powerdesk settings to their installation defaults.
• Added 'Remove' option for Powerdesk settings. This will remove the Powerdesk settings from the registry. This option is for troubleshooting and may require you to reinstall your latest Matrox drivers.
• Added 'Default Clock' button. This displays information on how to obtain your video cards' default clock settings
• MOC now displays the clock setting in the tray for the custom configurations. For example, Overclocked Setting (55)
• Additional clock settings for G200 cards (126, 127, 128, 129, 130)
• Additional clock settings for Mystique/Mystique 220 (supports all three factory clocked Mystique170's, 150Mhz, 165Mhz & 180Mhz). Added clocks speeds: 170 - 220.
• MOC.BAT file support removed (Windows startup options now saved/removed directly to/from the registry)
• Added support for 5 custom configurations. These custom configurations allow the user to store / recall different card and clock settings. Users can save clock settings for use with specific games, overclocked and non-overclocked settings, different clock settings for different resolutions, etc. Currently, this option only supports card settings and clock settings, future versions will support additional options including Powerdesk and Advanced settings.

• Matrox Overclock (Win9X - version 2.0.8a - 170 KB)