Something we all know, and can say without doubt is that, NVIDIA has some hotshot 3D chip engineers. But, what most folk don’t know is that they have an equally smart marketing team. Now why would I say that? If you just take a glance at there product range you’ll realize that there is a product offering from NVIDIA targeted at each different level of PC buyer. They’ve segmented the market so well, and offered each segment a product that is so ideally priced and by giving the optimum bang for the money spent they’ve basically taken over the 3D graphics market from the competition. And they don’t stop there, no no no!!! NVIDIA wants to push every pixel to the maximum in every market segment, let it be the TNT buyers or the high end Ultra; their new products keep popping up like mushrooms. This kind of market segmentation and expansion is not only adopted by NVIDIA, but also the processor manufacturing ‘giants’ INTEL and AMD have grouped there processors in features and price to cater to such individual markets. This way they can have a market presence in each of the small markets instead of having to settle for just one large group of PC buyers. The benefits work both ways cause then the consumer also has the choice of selecting a processor or graphics card to suit his ideal requirements and price. So where does the GeForce2 MX line of cards belong? Well it’s the value and business PC market that the GF2 MX is aimed at. Basically what NVIDIA has used as the recipe for the GF2 MX is pretty simple.
Due to the dual pipeline in the MX, it theoretically has half the power of the GF2 GTS. Though the original GF 256 had a quad-pipeline, it could only handle one texture per pipeline, per clock. So the MX with the dual-pipeline that can handle dual textures per pipeline, per clock, should be an in-expensive way of matching or passing the original GF 256 SDR. The standard GF2 MX which is clocked at 175MHz is well behind the High-end 200MHz GF2 GTS or the 250MHz Ultra, which both have quad-pipeline and dual-textures per pipeline, per clock. In single texture games the GF256 clocked at 120MHz default, will be faster than the GF2 MX with 480 megatexels/sec over the 350 megatexels/sec of the MX. But when it comes to multi-texture games the theoretical specification and core speeds aim the GF2 MX to be faster than the GF 256, with 700 megapixels/sec for the MX versus 480 megapixels/sec for the GF256. (Megapixels/sec=clock rate* No. Of pipe lines
Megatexels/sec=megapixels/sec*No. Of textures per pipeline)
Though having 2 pipelines less than the GTS the rest of the 3D features of the chip are unchanged. But since the MX is clocked lower than the GTS the T&L engine can only handle up to 20 million triangles/sec. (GF2 GTS 25 million/sec)
The dreaded section when it comes to the GTS range of cards. We all know the memory bandwidth problems these very fast 3D chips suffer. The limiting factor of the GTS chip is the memory technology, which is unable to keep up with the fast 3D chips of NVIDIA. Again, it's this very same area that the GF2 MX enters rough seas. As we know memory does not come cheap, and as a value card the price of the card had to be kept to the bare minimum. The GF2 MX is designed to only support standard SDR memory using the 128-bit memory bus, or SDR/DDR using a 64-bit memory bus. So here we have a faster chip, with a higher fill rate, put into the GF 256 SDR league due to slower memory. The memory is clocked at 166MHz, which is the same as the GTS's. For a chip that's already suffering from memory bandwidth problems, cutting the bandwidth in half will hurt real bad in high resolution gaming especially with all bells & whistles on. Patience now!!. We'll soon find out at the benchmarks.
GeForce2 MX 3D Processor
· GeForce2 MX engine architecture
· 2nd generation hardware transform engine
· 100% hardware triangle setup
· 2 dual-texturing rendering engines
· 4 texels-per-clock pipeline engines generate 20 million triangles per second
· Incorporates NVIDIA's Shading Rasterizer, delivering per-pixel shading, seven pixels operations in a single pass
· Supports OpenGL and DirectX 7: environmental bump mapping, vertex blending and projective textures · 350MHz RAMDAC
· 175MHz Core Clock
· 32 MB SDR SDRAM
· 183MHz Memory Clock (OEM 166MHz)
· 700 MegaTexels and 350 MegaPixels per second
· 4X AGP with Fast Writes / AGP 2X compatible
· AGP Support Only
· Resolutions of up to 2048x1536 in 16 million colors
Driver Version Included
· Windows 95/98 5.30
· Windows NT 5.30
· Windows 2000 5.30
· Pentium II and higher or compatible
· Available AGP slot, AGP 2.0 compliant
· 32MB RAM
· 10MB hard disk space
· CD-ROM or DVD-ROM drive
· Microsoft Windows 95 OSR2, Windows 98 or 2000
The Prophet II MX is very similar to the NVIDIA reference design. The PCB is very similar to the reference card in shape and size but has a blue PCB instead of the usual green. The Heatsink also comes in blue, to blend-in with the rest of the board.The memory speed of the Hercules is clocked at 183Mhz, which is a jump from NVIDIA's 166MHz reference board, so that memory bandwidth is increased from 2.7GB/s to about 2.9GB/s. The board comes with 32MB's of Hyundai memory (I used to drive a Hyundai a while back!), which is rated at 5.5ns. (6 ns on the reference board). Hercules has used the faster memory to support the higher clock speeds of their board. One important note to make on the Prophet is that it does not support the TwinView technology of the GF2 MX. ( TwinView is very similar to Matrox's DualHead ,which gives two independent displays from one card.). So essentially the 3D Prophet II MX is more of a 3D card meant solely for gamers rather than for a business PC,and incorporates all the main 3D features of the GF2 MX and comes a bit cheaper than other GF2 MX cards. The Pack:
The complete box comes with the card, a 48-page manual (in several languages), and the driver CD. The drivers are not very different to the NVIDIA reference drivers and just has a Hercules feel towards it. The driver CD also has a few 3D games but didn't impress me much.
We tested the card on two systems, a low end Celeron 600 and a mid range P III 700. The benchmarks will show what to expect from this card. So go on reading!
Intel Celeron 600, Pentium III 700
Pentium III 550E Micro-Star International BXMaster Motherboard
Elsa Gladiac Geforce 2 GTS(200/333)
Hercules 3D Prophet II MX(175/183)
MSI Starforce GF 256(120/183)
ELSA Erazor X2 GF256 DDR(120/333)
128MB of Mushkin PC133 RAM
It's quite clear that the GF2 GTS is leading the pack. The GF2 256 DDR takes a comfortable second place at the HQ and MAX settings. It boils down to a battle between the GF 256 SDR and the MX especially when above the Normal settings. It's interesting to note that the SDR falls behind the MX at lower resolutions and then suddenly surges ahead of the MX at 1600x1200.
In the much more demanding quaver demo it's still the GF2 GTS, which is the star performer. The GF 256 DDR keeps pace with the GF2 GTS to a respectable margin. Though the MX performs better than the GF 256 boards at the normal settings, it starts falling back at higher detail settings. Sad but true!
With a mid range processor like the P III 700 it becomes evident how well the GF 2 GTS can perform in the Quake III arena. At the lower detail levels and resolutions the MX still leads the GF 256 boards. But at the high details level the DDR comes in at an easy second place.
It's again evident that the GF2 SDR keeps well ahead of the MX at the high detail, color depth situations. So it makes us realize that the 3D Prophet II MX, though a good performer overall, is a step behind the GF2 SDR in these more demanding situations.
MDK 2 uses the OpenGL API and is a 3rd-person action game and also features T&L support to enhance performance. So it is a somewhat different testing to the Quake 3 arena. The lower resolutions bring pretty much identical scores for all cards. But at 1600x1200 the GF2 GTS is almost 30 FPS faster than the MX.
MDK 2 Performance- Pentium III 700
The GeForce 256 DDR maintains its second place standing, while the GeForce 256 SDR is just a bit faster than the 3D Prophet II MX at the normal setting.
At the MAX setting the GeForce2 GTS and GeForce 256 DDR are still the leaders, while the 3D Prophet II MX passes the GeForce 256 SDR at almost all resolutions.
Overclocking Overclocking the 3D Prophet II MX is pretty straightforward. The Hercules drivers have an Additional Properties section where Hardware Options could be changed. Here you can change both core and memory clock speeds. We managed to get the core right up to 200MHz and the memory up to 192Mhz. BUT there are issues I have to straighten out here. Yes, the core would go up to 200MHz and be very stable, but the heat is a major problem. So the sanest overclocked speed is around 194MHz where the temperature is not too high. But put in an additional cooling fan and 200MHz won't be too far fetched. The memory, which is rated at 5.5ns, is already running at its maximum specified speed of 183MHz. And since the memory chips don't have any special cooling heat sinks, overclocking the memory beyond 192MHz proved impossible without having texture problems and lock ups. The most stable speed for the memory we determined to be at 192MHz. Anyway, if extreme overclocking is what you're after then adding a cooling fan to the 3D Prophet 2 MX is highly recommended. Overall, large performance increases in frame rates are only visible at the higher resolutions.
The GeForce2 GTS and Ultra boards are still going to be the choice of the performance crowd, and if you currently own any GeForce board, it is not likely that a GeForce2 MX card will be a worthwhile upgrade. But if you have a low end 3D card, then the Hercules 3D Prophet II MX will give you GF 256 performance at a price much lower than before. Unlike in the early years of 3D gaming, gamers with limited budgets don't have to settle for a card that has frame rates as slow as your grandma's rickety old Morris Minor. The Hercules 3D Prophet II MX paints a whole new picture for the budget conscious 3D gamer!
· Best performance at the lowest cost.
· GeForce 2 GTS features
· The same old memory bandwidth problems.
· No software bundle