^/\^ PeaK /\^/\

May 2001 News

May 5, 2001

The strategy of focussing on X-box may be undone by focussing on a niche high end 3D solution generating both heat and significant fan noise that may limit it to being offered by OEMs as a add-on option. Grphic board vendors will take whatever they can get as the that market segment has no choice. The boards themselves will be expensive to address power and heat requirements.

Many are finding that even ATI's entry level Radeon VE has MX(all flavours)/TNT2 besting 3D performance offers home graphics that drive the new "digital" flat panels (versus analog flat panels that do not use the DVI/TMDS connector) and allow their kids to play games. With dual display, a game controller, and a VE...you can sit down with your two monitors ,pointed 90 degrees to each other,with you fiiling in your EXCEL TAX spreadsheet while your son/daughter figures out the intracasies of WarCraft on the second monitor....dual keyboards anyone ? To catch up on Radeon reviews recently published...click here

While you are at it take a look at the Anand's take on the Nvidia GeForce2 MX200 and MX400 (nee original GeForce2 MX clocked 25 Mhz higher in the core). The new chips are essentially a rebadging of technology that will confuse some and offer no real world imrovement to the knowledgeable...so that the unsuspecting consumer walking into a store asking for a GeForce2 may not get a GTS, or even a MX (original or mx400) but may get duped even more and walk away with a low-end MX (MX200) with only a 64 wide SDR interface. My suggestion to anyone considering the GTS/Original MX/MX400 to pick what others have known for the last 3 months

April 2001 News

April 22, 2001

0.3, 0.8, 1.2, 2.1 and 3.3 Megapixel sensors are needed to support resolutions of 640x480, 1024x768,1280x960, 1600x1200, and 2048x1536 resolutions respectively. Higher resolutions do not necessarily make for a better image, I have seen vast differences in images taken at 640x480 in terms of edge-to-edge sharpness, colour accuracy(vs washed out colours), and exposure latitude(details in shadowns). Not all lens, image sensors construction, and exposure metering systems are created equal. Note that for every doubling of the horizontal/vertical pixels (i.e. 640->1280, 800->1600,1024->2048) results in a 4x increase in memory requirement for a given level of compression. The Fuji cameras both offer 640x480(0.3 megapixel) and 1280x960(1.2 megapixel resolutions).

I am truly excited by excellent image quality and practical features offered in two modestly priced cameras from Fuji in the form of the MX-1200 and its replacement in the form of the FinePix 1300. The latter makes incremental improvements in adding USB, hiding less used previous features under menus, and about 20% less current/power draw and almost the same exact image quality (as the MX-1200) but a tad less.

What sets these two cameras apart from the field are the following:

1. Sharpness and quality of the lens construction playing its role in picutre quality. Equal to cameras costing two to three times as much.
2. Colour accuracy possibly due to the use of multi-coating antireflective lens coatings (pioneered by Fuji in the late 70's) and careful CCD design playig their role in picture quality. Top notch.
3. Included 8MB Smartmedia card allows for about 90 VGA/640x480 images with good picture accuracy. A upgrade to a 32MB SmartMedia card will allow for 360(10 rolls of 36 exposure film) pictures at a cost of about $30. This storage medium (for those of you familiar with film) can be uploaded to the computer and then re-used for again and again for more 360 photo sessions. Turn off the LCD display and flash to take picture after picture and you could survive two weeks away on a single set of 4 AA batteries.
4. Flash can be disabled most of the time as the camera can take well exposed photos by an available 40 watt bulb in an average sized room due to shutter range from 1/2 second to 1/1000 second.
5. LCD can be disabled for point and shoot operation. In this mode, the camera consumes about 40mA. Four AA batteries supplying 1300mA-hr capacity can operate for about 32 hours...that is 975 picture composing cycles. The camera is however has an auto-off feature so that it shuts off after two minutes so that you do not accidentally find your camera dead after 32 hours. Actual continuous-on time under normal operation is less when using the flash, LCD, and taking pictures/memory writing as average current/power consumption increases by factor of 10 for tens of seconds. Even in this high powered mode, the camera can run for about 2.5 hours. If you roughly take a picture every 12 seconds, you may be able to get off close to 800 pictures still. The smaller 1.6" (vs 1.8") LCD monitor is partly responsible for this long battery life.
6. Practical resolution modes include 640x480 mode with 1/8 JPG compression and a higher resoluton mode 1280x960 mode with selectable 1/16, 1/8, and 1/4 JPG compression. This results in roughly a 2x increase in memory footprint per setting change (90KB/image, 160KB/image, 320KB/image and 610KB/image).
7. Operation off four commonly available AA batteries. This allows for the use of advanced rechargeable Nickel Metal-Hydride batteries with alkaline capacity and no "memory recharge effect" common to Ni-Cads. The weight of these batteries adds a bit of useful heft to the camera weight to help stabilize the camera during picture taking.
8. An excellent menu system and camera manual.

1. Green LED near viewfinder could be drop down in brightness, made smaller, or moved to a the periphery of a user's vision.
2. The smooth plastic and small size makes the camera feel slippery to hold. Some of the expensive SLR digital cameras have a rubberized textured material that offers "tack" and some mechanical protection. How about a black rubberized textured version of the camera?
3. A more predictable and less spongy shutter release. And a more accurate viewfinder with a rubberized surround eyepiece.

April 5, 2001

1. Via Hardware's SBLive thread
2. Paul's Sound configuration hints
3. KX133/KA7 hardware workarounds

April 1, 2001

For now, I make due with a digital camera capable of only 8 VGA/640x480/0.3 Megapixel shots per 2MB of memory (no compression) or 250KB/image, a $6 3V lithium good for about 200 pictures, no frame counter, lens of questionable quality and construction, and built in flash in the form of the Relisys Dimera 3500. Now if Kodak could come up with a second generation EZ200 camera with incremental improvements in the form of a small modest zoom, selectable built-in flash, multi-element glass lens/better image quality and removeable memory...the who industry might be transformed. Update: Take a look at the Agfa CL20. This is the third generation of cameras from Agfa and these seem like winners. The unit offers a basic 1024x768 (0.8 Megapixel) resolution or an economy (512x384) near VGA mode for simple shots to mount on the internet. Unusual is a higer resolution 1280x960 mode(1.2 Megapixel mode). The flash mode can be disabled but the two AA akaline batteries are good for about about 36 pics. So the only question about this camera is really about the power efficiency and whether enough engineering effort was put into edge to edge sharpness. Beats the Kodak EZ200/Intel Pocket PC in terms of resolution, removeable memory expansion up to 16MB, and built-in flash...my new reference for the camera for the masses.

March 2001 News

March 22, 2001

Digital Interfaces offer the same advantages as the CD did to the record/phonograph/LP. Transistors were also allocated to relieve the CPU from processing the complex MPEG decode operations for viewing DVD and at the same time implement a more math accurate/better quality decoding algorithim. Architecturally, the use of a form of Hidden Surface Removal/Cache nick-named Hyper-Z allowed for RISC-like gains on near-code graphics image data. Careful attention to the feature list allowed for a low power consumption and low heat/quiet/fanless board design with no sacrifices on reliablity. The catch in all of this...it beats the Nvidia MX on a number of fronts in 3D performance and not to mention the TNT2 family of chips. You can thank John Carmack for taking the time to optimize his engine for the TNT engine, used in the MX, two years ago. Positive word of mouth about ATI has transformed the retail market for Radeon products due to improved drivers and regularly updated beta/special purpose drivers, breathtaking 3D rendering quality, sharp 2D text, and the aforementioned features above.

March 10, 2001
If you want to see the power of the Radeon with DX8, just take a look at some of the reported 3DMark 2001 scores reported. The 2x skew in the 3DMark 2000 scores for some hardware did not actually reflect real world gaming performance. Just as the Radeon took less of a hit in 32 bit that competing hardware, 3DMark 2001/DX8 is showing the strength of Radeon as the most future proof and hardware compatible DX8 that is available today.

February 2001 News

February 28, 2001
GameCube will probably be the first game console that I will purchase It is a reflection of what I call good design from a corporate, creative, and a personal view. The hardware will serve software (as it should) with a clean sheet of paper for both hardware and software with an set of software tools uncluttered by the legacy of the PC.

With Nvidia pushing the GF3 moniker(vs NV20)), the PClegacy of supporting multiple memory footprints and types, bandwidth limiting slot interfaces (instead of dedicated on board chip to chip high speed paths) will lead to large footprints and extra pins that will ulimately make is expensive in the consumer space.

Recent Mercury Research reports show the Radeon to be market favourite at the Retail level with clear market share and exponentially growing sales... it is an architecture with legs that will surpass Nvidia's two year stranglehold with the TNT2 product.

Stability, 3D quality, 2D quality and a performace delta can clearly be distinguished with the Radeon over the aging TNT2 product making up a large percentage of Nvidia's profit from last quarter.

People are beginning to appreciate the benefits of vertical integration for quality versus just selling chips based on a benchmark number alone. When framerates requriements extend beyond 30-45 fps in high quality game modes run at or beyond 1024x768...the 3D association with benchmarks is akin to the 2Dassociation with EXCEL/WORD 2D benchmarks...it just does not matter if it completes in a microsecond or a millisecond.

The software gurus at Nvidia have done an admirable job of rehashing/borrowed common architecture (from the TNT) and extending it (T and L). but things become significantly more complex with the DX8 re-architecture (new), new T&L mode calle pixel programmable shaders, new features(hype-Z and pixel shaders) that will put a significant driver burden into an already late project. The plethora of hype in pre-previews is evidnt by the lack of prototype silicon or prototype drivers.It is out of proportion and not really justified other that to keep the analysts bubble from bursting.

I think that most of the information has been out there for a while with the major sites and that the NDA signed back in December expired at the end of this February. They just decided to invite everyone to the same party. It seems interesting that NVIDIA are playing the same hand now with DOOM3/OpenGL that John Carmack played with "Quake3/OpenGL vs D3D stakes" in saying that there is a "reasonable" higher plane ia to what do hardware manufacturers need to do to service HQ mode under Quake3 and that D3D would not meet it.

OpenGL will do well by GF3/Apple and Carmack but X-box is supposed to Direct3D. Jobs and Gates have come to some sort of compromise. This might explain the recent licensing deal with Unreal to sub-license the UnReal Engine... they need one but it is going to be a pig in more ways that one. My opinion, is that X-box aimed high and late to what MS and Nvidia could comfortably/affordably build using just Windows serving/PC based components. A GF2 with no bump mapping and vertex skinning added to the works would have made a nice X-box platform. Hyper-Z would have bought the external memory requirements down. The problem there was that Radeon and Dreamcast covered some of that ground. Competition would also come from PCs (with the same or better graphics) or from someone else such as Nintendo's Game Cube. Like I said, I personally the position of GameCube at this point. The recent spat of hype from Nvidia is reminscent of previous early announcements from Nvidia based upon what they want they think the market needs/wants to hear at this point. Based upon the delays, I say there are probably fundamental design/technology issues in which key blocks are not functionaing properly with Hyper-Z due to different types (more central) of layout constraints related to memory distribution. All of Nvidia's to date have used one simplistic layout scheme comprised on one large memory instead of smaller distributed memories as in ATI's case.

A new market favourite in the PC Retail sector beginning to emerge called the Radeon. Perhaps it is all that the X-box really needed from a release point of view. Radeon 2 is a refinement of Radeon and probalby should have been what the real X-box chip required in terms of schedule requirements of games application development.

In addition to currently spanning all market segments from low-cost to high performance and I think it makes a better set of tradeoffs in the features/integration vs 3D performance in each category.

1. Extreme End Gaming + Vidon : 64MB DDR, lower cost, consumer grade DVD, Video-in, Video out, great 2D clarity and better image quality generation capable of playing latest generation games. Rock stable 7089 drivers. versus GF2 64MB(no Video).
2. High End Gaming: 32MB DDR, same great 2D and 3D accurate compute engine with 128 wide DDR memory
3. High End with Tuner above with 4th generation Tuner software. Asus has numerous teething problems keeping in step each time a new driver release comes out.
4. Budget/Performance: 32 MB SDR with same great 2D and 3D. 7089 drivers works with 98% of the software and increasing. . Easily beats MX.
5. Business/TNT2+ busting Performance : Matching the MX in performance but offering TMDS and dual monitor support. Same great 2D and 3D quality. Should be a hit. versus MX.

1. Limited Z-buffer accuracy
2. Unsharp 2D beyond 1280x1024...it might say 350MHz DAC on the package in that it can put out data at that speed...but what you get is something t measures well up to about 170MHz.
3. Speed up in drivers comes at the expense of using compressed textures in some games (Such as Quake) which are not supported correctly in the hardware. The problem is bad enough that a special version of Quake (1.27) was released that does not use compress textures by introducting a new flag called "r_ext_compressed" to disable it automatically. Older configuration files that have the old flag called "r_ext_compress" enabled will see no compression, lower performance and better quality(1.27) or higher performance and bad compressed(pre 1.27) sky textures. Your pick.
4. Non working TMDS in the graphics chip. Result is that the mass adoption of the the Digital Video Interface (TMDS and RGB in one connector) by bothe monitor manufacturers and graphics card manufacturers will require the board to implement an external ($$$) TMDS chip to drive a digital flat panel.

February 19, 2001

1. P4 runs slower than P3
2. Win2K performance is 50% less
3. Fury Maxx does not run under Win2K

February 15, 2001

February 10, 2001

The progression of improvements from 3056,3063,7020, WHQL 7041, 7062, 7068,7072,WHQL 7075 and 7078 drivers partly problems introduced in game application code development interpretation of hardware/software specifications by not beta testing on the Radeon when code was developed and have been fixed with a combination of patches to games and folding driver updates into new release driver CDs.

performance advantages (40%) in the Radeon architecture using new games such as MBTR (Mercedes Benz Truck Racing) or is that "mean time before failure" ? The new aspect to this is in 16-bit and the performance gain tradeoff in going from 32-bit to 16-bit. Clearly the SDR solutions beat the MX and the DDR solutions even more so. What is more interesting are the lack of comments on the clear superiority of the Radeon architecture in certain benchmarks when tested on new advance games beyond Quake where the playing field is more even. Despite this fact, guess what is dished up on the front page of this site...a review of five identically performing cards just so that the "free" hardware commitments are honoured. Still no mention of MX texture problems, 2D blurriness, and the significant performance improvements of the Radeon of late. Go figure. The users have and they are flocking to Radeon for smoother game play, sharper 2D, much better DVD, and now performance!!!

January 2001 News

January 25, 2001

I've not seen this posted anywhere for Windows 2000 so I thought I'd share. For those with the shimmering/text crawl problem, the fix that has been posted so far has been for Win 95/98/ME/NT.

For W2K go to the following in your registry:-


HKLM\SYSTEM\CurrentControlSet\Services\ati2mtag\Device0

Add a new DWORD:-


GCOOPTION_SlowerEdge

Modify the value to 1, then reboot. It's upped refresh rates by 1, ie instead of 85Hz you get 86, and so on. Worked for me anyway on my Iiyama VM Pro 400 17"

January 18, 2001

January 4, 2001

"...before I move on to the benchmarks I want to mention that after having used all of the current generation video cards, the ATI Radeon CLEARLY has the best image quality of all of them" Radeon Roundup, PCMonkey Jan, 2001

"...with the same 350MHz RAMDAC found on the other RADEON products, the All In Wonder delivers crystal clear test text at high resolutions. In fact, resolutions of up to 2048x1536 at 75Hz are supported. As far as PC usage is concerned, the RADEON offers 2D quality that matches the infamous clarity of Matrox's lineup. AIW Radeon review, Sharky Extreme/Chris Angelini, Jan 9, 2001

A related issue related to how AGP4x is implemented relates to driving strength. AGP4X has one of the most advanced "drive control" mechanisms found in a consumer product to ensure good signal integrity by having the output buffer drive correct for supply voltage, temperature and manufacturing variance. The goal is to keep the output drive fixed and constant at a very precise strength...not too strong and not too weak.

Many early adopters of the very early KX133 boards chose ABIT's KA7 due to the qualilty and stability of earlier generation BH6 products supporting Pentium II processors and AGP 2X. What is peculiar (in the case of the ABIT KA7) is that the driving strength is actually set in the BIOS and that various users have experimented to find out values are optimal for reliable booting and AGP2X operation of this interface. It seems that a value of "68" is optimal for these boards when used with Radeon products. As mentioned, differences in manufacture or environment will require slight tweaking of this value. For AGP4X operation one must first ensure that they have a CE revision of the chips to have any hope at all of running in AGP 4X mode. This is difficult to verify visually as the Northbridge chip is now covered by a heatsink. Secondly, a graphics driver that is aware of the Chip revision must exist(download any of recent WHQL drivers). Last, if the strength of the AGP I/O is set manually, then the value must be determined by trial and error for 4X. One suggestion would be to somehow clip on a fan onto the KX133 chip itself to stabilize the temperature as the "drive" will vary as the chip self heats. The last thing to do is to add in additonal bypass capacitors to the KX133 chip across the 1.5V AGP power supplies. Good luck.

I recently tried the drivers under Win95 with good success. The Win98 tested drivers operated correctly under Win95, but the OpenGL component was not accessible until the GLsetup OpenGL drivers were installed.

There was an increase in visual quality and frame rate under Quake. Hardware supported are Rage Pro, Rage 128 and Radeon. The most recent release of these drivers was Jan 1, 2000. Give them a try and report back here to see what you thought of them. Although there is an uninstall facility, it does not use Install Shield.

Monday January 1, 2001
Good Morning!!! It's a blanket of snow outside and the first day of the year 2001. So...what's the first post of the year going to look like? Well it is definitely not something new but a post about my aging ASUS P5A-B AT motherboard. My motherboard has served partly as a test bed to evaluate various chipset drivers and graphics drivers that form the basis of the "pipe to the eyeball" as someone nicely put it. As new hardware features and technologies come on to the market, graphics cards adopt an approach somewhat similar to car offerings narrowing the choices to one or two offerings. PCI graphics card volumes are well served by older chipsets and are rarities in new graphic chip offerings. I have a similar predicament with the Radeon AGP graphics cards on the software side. Putting in the graphics card will shift the bottleneck solely to the CPU. The costs of 700MHz CPUs and motherboards (getting more expensive due to expensive high amperage regulator designs) mean that very powerful and affordable sub $1,000 systems are pretty much the norm. I will try to support this platform for as long as it serves my needs by judicously upgrading drivers, operating systems, and hardware. So where is the hitch?

Basically the Radeon user's manual that states Win98 as the minimum operating system. It is a strong recommendation, I think, to circumvent a plethora of possible limitations with Win95 with regard to patches, workarounds and new hardware support. So is it possible to get the latest Radeon chipset up on the very popular ASUS P5A-B motherboard. I do not have first had experience yet until I assess how close the patched Win95 system is to Win98. In the meantime, here are a bunch of links on experiences with this path from the world wide web...product of the decade.

1. K63 and P5A success This user is running a new beta bios and no problems
2. Beta Bios post and some install hints from a FAQ. Partial confirmation on the need for the ALI driver update. Information is available on all driver files. The AGP169E readme covers the history of the driver. AMD users may wish to read the following post
3. 8 posts in www.rage3d.com forum. Particularly good is the MVP-3 user who outlines his install sequence and cleansing procedure for old drivers.
4. This is a catchall search for radeon posts concerning the Asus P5A and P5A-B motherboards.