<!–#set var="article_header" value="The GeForce2 MX400:
NVIDIA’s Hold On The Mainstream” –>
Let’s Be All We Can Be: GeForce2 MX200 and MX400
On March 6, NVIDIA released two new versions of their mainstream GPU, the GeForce2 MX. Many people did not even notice as there were no special features to be revealed.
Have you ever noticed how many systems are sold that still use a TNT2 or Vanta based graphics card? Although many home users prefer to specify their individual hardware configuration rather than purchasing a system off the shelve, wanting better 3D performance for gaming, the demands of the typical business computer are not as stringent. A decent amount of RAM, a modern processor and sometimes a big hard drive are the dominating factors for office machines. However, if you, for example, had to specify the graphics card for a corporate division, you would definitely favour a technically more advanced product if it was not any more expensive than the standard issue graphics subsystem.
NVIDIA is using its newest versions of the GeForce2 MX with that thought in mind. Sooner or later, the TNT2 family will be obsolete. The GeForce2 MX is supposed to be the successor, but it is facing stiff competition these days. ATI’s Radeon VE is an attractive alternative and the new Kyro II chip from ST Microelectronics delivers good performance at favourable prices. By creating two new MX versions of the GeForce2 NVIDIA is adding to its product portfolio in order to make the GeForce2 family attractive to as broad a range of mainstream users as possible. Let’s now take a look at the two new chips.
The Newcomers: MX200 and MX400
As I already mentioned, there is no technical innovation to announce this time. The GeForce2 MX will remain unchanged. Instead the two new versions will simply be added to the MX lineup.
The GeForce2 MX200 will be the low-cost version. Running at the same core and memory clock speed as the MX (175 MHz core, 166 MHz memory), and the memory interface has been cut down to 64 Bit (Single Data memory only). That makes graphics cards with only two memory chips possible – this is a very good approach to saving costs for price-conscious buyers. Budget buyers will get a graphics card that carries a GeForce2 chip including desired features like T&L and NVIDIA’s Shading Rasterizer, as well as 32 MB memory.
Here I can see some confusion arising, a fact which will certainly be taken advantage by computer marketing people. You could possibly run across a high-end computer with an ultra-fast CPUs and GeForce2 chips, but it might only be high-end in marketing speak. Actually, lots of these types of systems could be equipped with a low-end MX200.
There is better news from the MX400. I guess the strong competition from the Kyro II is the main reason NVIDIA has opted to release this faster version of the MX. The memory interface remains 64 or 128 Bit wide and can operate Single Data or Double Data SDRAM memory. Unfortunately, DDR memory can only be operated at 64 Bit, making it no faster than a 128 Bit Single Data rate setup. ‘MX400’ is an allusion to the maximum pixel fill rate of 400 MPixels/s. Some of you might have already realized that the chip has to run at 200 MHz core clock to achieve that.
As a result, this accelerated MX has one disadvantage which limits its attraction to some OEMs: besides a heat sink, it definitely requires a fan to keep the chip temperature down.
For more information about the features of the GeForce2 MX, please read the initial review published in June 2000.
Chip | Memory Bandwidth | Core Clock | Fill Rate | Memory Setups |
GeForce2 MX 400 | 2.7GB/sec | 200 MHz | 400M pixels/sec 800M texels/sec |
64/32MB |
GeForce2 MX | 2.7GB/sec | 175 MHz | 350M pixels/sec 700M texels/sec |
64/32MB |
GeForce2 MX200 | 1.3GB/sec | 175 MHz | 350M pixels/sec 700M texels/sec |
32MB |
The Competition: Kyro II from ST Microelectronics
Guillemot/Hercules is the first manufacturer to offer a graphics board based on the Kyro II. As you may know, this little chip is definitely able to play in the league of NVIDIA’s GeForce2 MX and ATI’s Radeon SDR. As the Kyro II does not offer either AGP 4x or wide memory bandwidths (5.5 ns SDR SDRAM at 180 MHz), the benchmark results are quite diversified. However, in most benchmarks the Kyro II is able to beat both the GeForce2 MX and ATI’s Radeon DDR board. Depending on the resolution and colour depth, Kyro II may even be able to beat the GeForce2 GTS.
ST Microelectronics played it smart as they did not even try to boost memory bandwidth by using fast and expensive memory chips. Instead they decided to avoid unnecessary data movements and pixel fills using TBR technology (Tile Based Rendering), this being the main technology selling point for the Kyro II technology.
Typical 3D scenes are populated with an array of 3D objects. TBR simply checks which objects are visible to the spectator and only fills the required polygons. ATI’s HyperZ goes a similar way. In contrast, the NVIDIA engine always fills everything and then sorts out the visible polygons using Z-buffering – resulting in a big burden on memory bandwidth. The impact on memory bandwidth is tremendous when there is a lot going on in a scene, which is the case with most 3D shooters and action games. NVIDIA has opted for brute force to overcome this problem; TBR finesses an alltogether different approach to bypass potential bottlenecks.
Test Setup
System Information | |
CPU | Intel Pentium III 866 MHz |
Motherboard | Asus CUSL2, BIOS 1006a |
Memory | Crucial PC133 CAS2 setting 2-2-2-5/7 |
Network | 3Com Etherlink 3C905TX |
Hard Drive | IBM DTLA 307030, 7200 rpm, 30 GB |
Graphics Cards | GeForce2 MX: MSI StarMaxx32, 32 MB SDR GeForce2 MX400: GeForce2 GTS: |
Software | |
Operating System | Windows 98 SE 4.10.2222 A Windows 2000 Pro 5.10.2195 |
DirectX Version | 8.0a |
Quake 3 Arena | Retail version command line = +set cd_nocd 1 +set s_initsound 0 |
Mercedes Benz Truck Racing | Default Features |
3DMark 2001 | Downloadable Version |
Aquamark 2.1 | 1024x768x16 FSAA Disabled 24 MB Textures Pixelshader: Off |
BAPCo SYSmark 2000 | Patch 5 |
SYSmark 2000
The first benchmark is not very useful for graphics benchmarks, as most graphics boards do not differ in terms of 2D performance and have not done so for some years now. However, I decided not to remove it as many of our readers would like to to see these results anyhow.
Aquamark
The Aquamark benchmark looks pretty amazing and shows awesome realism under water. To achieve this, the processor is always under quite a heavy load, making the graphics board less important at the resolutions of 640×480 and 800×600. As soon as it comes to 1024×768 or above, you can see that only the GeForce2 GTS is able to sustain the frame rate.
Mercedes Benz Truck Racing
At 16 Bit per pixel, there is hardly a difference between the three GeForce2 models. At 1024×768 and above, the GeForce2 is still able to produce high frame rates thanks to its fast DDR memory. If it was possible to equip a GeForce2 MX or MX400 with DDR memory over a 128 Bit interface, I’m quite sure that it would not be too far behind.
A similar result at 32 Bit per pixel. The MX400 is slightly faster than the standard MX board; still it lags the results of a GeForce2 GTS by some distance.
Quake III Arena
Quake III Arena is still one of the most important benchmarks – and likely will stay until another 3D shooter will surpass its popularity. The 16 bit run is very theoretical, since the frame rates are of course fast enough for smooth game play even at 1280×1024. However, it is also the best way to show the real world difference between graphics cards.
Basically, your desired resolution should run at 30 fps or faster. A demand which even the standard GeForce2 MX is able to fulfil. I did not include the results at 1600×1200 but all three cards drop below those 30 fps than. After checking those Quake III results I cannot see any reason why I should spend more money for a MX400 board. The standard MX will do it.
3DMark 2001
Although I am not a strong supporter of the 3DMark because of its inexpressive results, I decided to use it this time for a number of reasons. The latest 2001 version has some very impressive tests and also supports DirectX 8, making it a good choice technically. Again the MX400 is obviously faster than the MX, but still quite far behind the scores of a GTS.
Conclusion
NVIDIA has proven once again why they are market leader by executing very strongly on these product line additions. This time, it’s not a technological revolution, but a clever modification to an existing chip. From a business perspective, NVIDIA now can offer a GeForce2 chip across the broadest spectrum of PC users (including the mobile market with the GeForce2 Go). The GeForce2 MX currently is the undisputed #1 choice if you are looking for a powerful solution which is not too expensive. The MX200 will slowly replace the ageing TNT2 and Vanta chips, GeForce2 MX remains unchanged, and the MX400 is meant to make you feel superior to users with an ordinary MX.
Most of us will only get in contact with the MX200 if you opt for an off-the-shelve system – but we hope you try and keep your hands of that kind of stuff! The ‘new’ MX400 is likely to be in good supply at computer stores soon; of course at prices above the standard MX. I’m quite sure that the yield is good enough to validate a 200 MHz core clock with every chip produced. So, if you think of the MX200 and 400 as a positioning statement by NVIDIA, and a way for the company to round out its product offerings in light of the possible threat of Kyro II then, you can’t go wrong.
Any other company in NVIDIA’s position would certainly have gone the same route. As soon as a product is mature, chip vendors start harvesting the design. Intel did this for quite some time until a company called AMD was finally able to gain enough attention to give it some competition.
I’m really looking forward to seeing how the acceptance of the Kyro II goes in the market, as it is a better product than the GeForce2 MX, and at a better price point. NVIDIA needs a little more competition, and a little less domination.