Today graphics cards based on NVIDIA chips boast the highest sales in all the price sectors. Until recently all the chips from GeForce2 family were positioned as follows: GeForce2 MX was intended for Low-End graphics cards; GeForce2, GeForce2 Pro and GeForce2 Ultra dominated among Mainstream and High-End solutions.

But not so long ago NVIDIA split its GeForce2 MX clan into three smaller types: GeForce2 MX 400, GeForce2 MX 200 and GeForce2 MX 100 for low-cost cards and the very Entry level ones . This move was taken to maintain the company's positions in the Low-End sector, since TNT2 chips got morally outdated and are not manufactured any longer. The only chips, which appeared more or less close to the former GeForce2 MX is GeForce2 MX 400. NVIDIA recommends equipping the graphics cards based on GeForce2 MX 400 with SDRAM chips featuring 128-bit bus and working at 183MHz. The core frequency should make 200MHz. GeForce2 MX 200 and GeForce2 MX 100 based graphics cards have lower working frequencies and cut-down memory bus bandwidth. They are supposed to stand for the outdated Low-End products.

However, the graphics card manufacturers sometimes tend to neglect NVIDIA's recommendations, especially when it goes about graphics memory type and frequency. Subsequently, graphics memory clocking on GeForce2 MX 400 based cards made by different manufacturers may vary from 143MHz to 183MHz. Occasionally we can even meet such monsters as the card based on NVIDIA GeForce2 GTS equipped with SDRAM graphics memory.

This article is devoted to another freak of the kind - a graphics card built on NVIDIA GeForce2 MX 400 with 64MB graphics memory named Gigabyte GV-GF1280RT.

Closer Look

Gigabyte GV-GF1280RT is designed on a traditional blue-green textolite so typical of Gigabyte:

The card features NVIDIA GeForce2 MX 400 chip:

To dissipate the heat from the graphics chip passive cooling system is used. It is actually a pin heatsink:

Most NVIDIA GeForce2 MX 400 based graphics cards are equipped with heatsinks only stuck to the chip and in fact, the absence of a fan doesn't tell on their performance. But the sample of Gigabyte GV-GF1280RT, which we had at our disposal, hung in the end of the half-an-hour 3DMark2001 benchmarks set.

The reason was an overheated core that couldn't be properly cooled down because of a thin aluminum foil covered with some sticky stuff on both sides and stuck between the heatsink and the chip lid:

When we removed the foil and installed a fan against the card's heatsink, 3DMark2001 was successfully passed. So, if Gigabyte doesn't take the trouble of changing the core cooling system as it starts releasing GV-GF1280RT cards in mass, we would recommend you to care about some extra cooling when purchasing this card.

Gigabyte GV-GF1280RT is provided with 64MB 128-bit graphics SDRAM. The 6ns chips are manufactured by Hyundai:

These memory chips are supposed to work at under-166MHz frequencies. However, the sample we got hold of showed 179MHz default graphics memory clocking. Increasing the frequency didn't affect the card's stability. We observed no irritating artifacts, which are so common for "over-overclocked" graphics memory.

This graphics card features two VGA-Outs and one S-Video Out. Fairchild FMS3815 chip (8-bit RAMDAC with 150MHz data conversion rate) is responsible for transferring the image to the second display:

TV signals are encoded by Chrontel CH7008 chip (a TV-coder without Macrovision technology support, 800x600 maximal resolution, supports NTSC/ NTSC EIA, PAL B/ D/ J/ H/ I/ M/ N display formats):

Unlike most NVIDIA GeForce2 MX or MX 400 based graphics cards bundled with a TV-Out, Gigabyte GV-GF1280RT is equipped with special jumpers allowing to set the display format for the TV-Out: NTSC or PAL:

Testbed

For our tests we assembled the following testbed:

• AMD Athlon 1.2GHz CPU (133MHz FSB);
• ABIT KT7A mainboard (VIA KT133A based);
• 256MB NCP PC133 SDRAM;
• Fujitsu MPE3084AE 8.4GB HDD.

We used the following software:

• Windows 98 SE build 4.10.2222 A;
• DirectX 8.0a;
• Quake 3 Arena v1.27;
• 3DMark2001 build 200.

For our tests we utilized Detonator 11.01 driver for our graphics card.

2D Image Quality, 3D Image Quality and Overclocking

Gigabyte GV-GF1280RT provided rather smart graphics. It gave no blurring in all resolutions below 1280x1024.

The card is built on NVIDIA GeForce2 MX 400 chip, so it represents nothing special in the terms of 3D graphics image quality and features if compared with other NVIDIA GeForce2 MX 400 based graphics cards. We guess there's no need highlighting these issues now in this review.

We managed to overclock the graphics memory up to the top of 190MHz absolutely painlessly, i.e. without any artifacts emerging in games. However, as long as the 6ns memory of Gigabyte GV-GF1280RT is clocked by default at a higher frequency (179MHz), we wouldn't say that this card features large overclocking potential.

The maximal core frequency where Gigabyte GV-GF1280RT still retains stable performance is 225MHz - quite a modest result restrained by the card's passive core cooling.

Testing Methods

All the settings were left by default except Vsync - graphics synchronization was disabled.

For Quake3 Arena in 16-bit color modes we chose 16-bit texture quality, in 32-bit modes we took 32-bit textures. Textures and the level of detail were set to the maximum value, tri-linear filtering was enabled, all the other settings stayed default.

Testing in 3DMark2001 was run with the following settings: in 16-bit color modes we enabled 16-bit textures and 16-bit Z-buffer, while for 32-bit textures we selected 32-bit textures and 24-bit Z-buffer. All the cards were tested in "D3D Hardware T&L" mode.

We found Gigabyte GV-GF1280RT worth looking at from the 3D graphics image quality and performance point of view because of its 64MB memory. We thought it would be interesting to compare it with 32MB graphics cards based on NVIDIA GeForce2 MX 400. Alongside with Gigabyte GV-GF1280RT we engaged such racers as:

• SUMA Platinum GeForce2 MX 400 built on NVIDIA GeForce2 MX 400 chip with 32MB 128-bit graphics SDRAM. The card worked at 200MHz core and 166MHz memory frequencies.
• ASUS V7700 GTS based on NVIDIA GeForce2 GTS chip with 32MB 128-bit graphics DDR SDRAM, the core and the memory were clocked at 200MHz and 333MHz (166MHz DDR) respectively.
• SUMA Platinum GeForce2 GTS based on NVIDIA GeForce2 GTS chip with 64MB 128-bit graphics DDR SDRAM, the core frequency was set to 200MHz, the memory frequency - to 333MHz (166MHz DDR).

Since the graphics memory of Gigabyte GV-GF1280RT is set by default to 179MHz, we tested it at 166MHz as well in order to compare with a 32MB NVIDIA GeForce2 MX 400 based card (166MHz is a typical value for NVIDIA GeForce2 MX 400 based graphics cards).

Performance

It's easy to notice that Gigabyte GV-GF1280RT outpaces the 32MB NVIDIA GeForce2 MX 400 due to its higher graphics memory frequency. When the memory frequency is set to 166MHz both the 64MB and 32MB cards come to equal results, that is, in Quake3 the additional 32MB of memory make no real advantage for Gigabyte GV-GF1280RT.

On the other hand, the 64MB NVIDIA GeForce2 GTS card in 1600x1200x32 mode looks tangibly faster than its 32MB counterpart. Why so?

The gap in performance between 64MB and 32MB NVIDIA GeForce2 MX 400 graphics cards exists as well, but it's really small because NVIDIA GeForce2 MX 400 cards mask insufficient graphics memory and the need to use AGP texturing with twice as small fillrate as that of GeForce2 GTS and twice as small graphics memory bus bandwidth. In other words, the shift from transferring the textures via graphics memory bus (with 2.5GB/sec bandwidth) to transferring them via AGP bus (with 1GB/sec) is not so obvious by NVIDIA GeForce2 MX 400 based cards as by GeForce2 GTS based cards (with 5.3GB/sec graphics memory bandwidth).

Now it's high time we checked the performance with full-screen anti-aliasing enabled. The modes with FSAA require more graphics memory and faster data transfer rate between the graphics core and the memory.

When full-screen anti-aliasing was enabled, 64MB of graphics memory provided a performance gain only in 1024x768x32 mode with 4x supersampling. However, this mode is not good for gaming on NVIDIA GeForce2 MX 400 anyway, because the performance is limited by the graphics memory bandwidth. In all the other modes Gigabyte GV-GF1280RT yields better performance only thanks to its higher graphics memory clocking.

32MB cards failed to cope with these tests in 1600x1200x32 mode. There is almost no difference in performance between 32MB and 64MB NVIDIA GeForce2 MX 400 based cards, while the difference between 32MB and 64MB NVIDIA GeForce2 GTS based cards did appear in 1024x768x32 and 1280x1024x32 modes, though it was not too large, not like in Quake3 at 1600x1200x32. The reason is that unlike Quake3, 3DMark2001 uses much more polygons and textures. The numerous polygons load heavily the T&L unit making the difference between 32MB and 64MB cards less obvious. At the same time, a greater amount of textures makes it necessary to download more the textures at higher resolutions starting with 1024x768x32.

As you have seen, the additional 32MB of graphics memory don't make the life of Gigabyte GV-GF1280RT easier either in 3DMark2001 or in Quake3.

Conclusion

Gigabyte GV-GF1280RT with its 64MB on-board graphics memory is doing no tremendous breakthrough: its 32MB rivals are no worse in many aspects. The only logical application for this graphics card are such programs as 3D Studio Max where its 64MB memory will come in handy to store the textures and the memory bus bandwidth is not so crucial as in the latest games. Dual-display support and a TV-Out turn this graphics card more flexible to work with.

As for the gamers, this card will bring scarcely any gain in comparison with the conventional 32MB NVIDIA GeForce2 MX 400 based graphics cards.

Highs:

• High performance (thanks to higher memory working frequency);
• TwinView and S-Video Out support;
• Supports a wide range of TV formats;
• Good 2D quality;
• High-quality mounting.

Lows:

• Relatively poor overclocking potential;
• Not the best graphics core cooling.