The GeForce 3 is a DirectX 8 compliant 3D-card and brought new features and techniques to the market. It has programmable shaders, LMA (Lightspeed Memory Architecture) for better memory management, reducing overdraw en compression of the Z-buffer. In terms of image quality the GeForce 3 can do better anti-aliasing and 8-tap anistropic filtering. The GeForce 2 could only do 2-tap and instead of 'real' anti-aliasing the GeForce 2 rendered a high resolution and scaled it back to get less 'jaggies'.

Performance-wise the GeForce 3 is quite good but sometimes it's surpassed by the older GeForce 2 Ultra. The GeForce 2 Ultra has higher clock-frequencies and more memory bandwidth. Because both the GeForce 3 and GeForce two have four pixel pipelines which each can sample two textures per clock the GeForce 2 Ultra has an higher pixel and texel fill-rate. This causes the GeForce 2 Ultra to score higher in Quake 3 16-bits where the GeForce 3 excels (by quite a margin) in more advanced benchmark/games like Aquamark 3 or Unreal Tournament 2003. The effect of LMA clearly shows in Quake 3 on higher resolutions and 32-bit color as the GeForce 3 keeps pretty strong and the GeForce 2 Ultra pant for extra memory bandwidth.

Both the GeForce 3 Ti200 and Ti500 were released at a later date (October 2001). The original GeForce 3 Ti is positioned between them. > Read more

I bought this card new (unused) in 2009! Somewhere a few of these cards 'survived' in stock .

The GeForce 2 Pro was meant to be an high-end product and uses the GeForce 2 GTS core with faster RAM. My card, as seen on the photo's, has a GeForce 2 GTS chip with 'Pro' stamped onto it. The GTS was shipped with 166MHz RAM and Pro with 200MHz.

Leadtek made a nice card here. It has good cooling with a finishing touch. They also integrated a temperature sensor but I've never seen it working. It probably works only with the original Leadtek drivers and I used the nVidia reference driver. > Read more

The Ti4600 is, along with the Ti4400, released on 6 February 2002. A massive, fast and expensive 3D-card with good performance.

Leadtek's version of the Ti4600 comes with a large heatsink. It's working well with a low noise-level (just like GeForce2/3 cards from Leadtek). Unfortunately Leadtek started using lower quality parts on the GeForce 4 Ti cards. See the photo's and take a look at the green capacitors (cylinder shaped) that aren't flat anymore.

nVidia released the cheaper (but quite fast and popular) Ti 4200 in April 2002. Newer versions with AGP8x support and a NV28-core appeared months later. The GeForce 4 Ti didn't really need AGP8x as AGP4x offers enough bandwidth. However, the Ti 4200 AGP8x has faster memory so it does perform a tad better. > Read more

Leadtek's WinFast A280 is based on the nVidia GeForce 4 Ti4800-SE chip. This is an AGP8x version of the GeForce 4 Ti4400. Somehow the names of the GeForce 4 Ti aren't as expected. The first AGP4x cards were 4200, 4400 and 4600 and their AGP8x versions are 4200-8x, 4800SE and 4800. In my opinion nVidia should've continued with the -8x suffix and sold 4400-8x and 4600-8x's instead. > Read more

Very notable is the cooling design which seems and feels quite good. In reality it's not much high-tech as it's practically a box around the card with air flowing through. Nevertheless it looks good.

Leadtek's WinFast A350 TDH LX is based on the nVidia GeForce FX 5900 GPU. The 'LX' means it has lower clocked memory to reduce the price.

Unfortunately my card doesn't work so I can't get any benchmarks. > Read more

To get a better position in the market, nVidia launched the GeForce FX 5900XT GPU. This graphics processor was one of the slower parts for the high-end market. Leadtek build their WinFast A350XT using the 5900XT GPU. As the 5900XT is fitted on a high-end (but somewhat cut-down) PCB with high-end memory it has lot's of potential for overclocking. Back in 2003 many enthusiasts bought the 5900XT-cards and overclocked it to get good performance. > Read more

This card is based on nVidia biggest and fastest GeForce FX GPU; the 5950 Ultra.

Leadtek has done a good job on their custom cooling. It cools good and looks quite impressive. Many other companies released a GeForce FX 5950 Ultra so they had to come up with something special and attractive. Cooling, games/software-packs and things like that are often used to make a product more complete. > Read more

A budget card with an interesting twist: this card boasts all the ingredients for a budget card (single molex, DDR memory rather than GDDR3, lower clocks, less pipelines, D-Sub connector) but basically it's fitted with the same chip as the GeForce 6800 GT and Ultra (NV40 chips)! Of course this sounds better than it really is, as the 6800LE cards have partly defective NV40-chips which work fine when a few defective pipelines are disabled. Some cards had nearly perfect chips, however.

Back in the old days a buddy of mine bought an ASUS V9999TD/LE which is a GeForce 6800 LE with 1,2 volts on the core (the other brands had 1,1 volts and GT / Ultra cards work with 1,3 or 1,4 volts). Due to the increased voltage the card could overclock a bit more easily. All the pipelines were unlockable (from 8 to 16) and the GPU clocked amazingly well. Eventually we installed his card in my PC (Athlon XP 1800+ running at ~2,6GHz with nForce2 platform) which was not bad but slower than our competitors with 3GHz Athlon64's or Pentium-M's. Eventually we reached 6033 3DMark05's and 2689 3DMark06's which was good for #2 in 3DMark05 and #1 place in 3DMark06 (at that time, at least). Not bad considering we did not use volt-mods, did not use LN2 and used a slower CPU .

To compare the scores we got:
Standard 6800LE with 3DMark05: about 2500.
Standard 6800GT with 3DMark05: about 4800.
Our overclocked 6800LE with 3DMark05: 6033.

Unfortunately not every 6800 LE overclocks as well. Some can't open any pipelines at all or are stuck at 12. I also noticed, after a while when newer games were released, quite some reports of people who fully unlocked their card without problems but eventually got some/very few artifacts with new games.
> Read more

Just like the GeForce 6800 Ultra but a tad slower. Both GPU and RAM clock frequencies are lower which result in cheaper RAM and less heat output. Because of the latter this card only has one molex power connector. > Read more

A late GeForce FX 5200 from Leadtek. It comes with the box and has sort of acceptable entry-level performance for early 2003, which is two and a half years late as this card was made in 2005 . > Read more

This entry level GeForce 6 series card has an NV44-GPU. The first GeForce 6 6200 cards use a NV43-core which is meant for the GeForce 6 6600 series. To full-fill the demand of the budget versions nVidia used defective (or perfectly working cores if they used up the 'defective-reserve') cores which could be unlocked and thus allowing to transform a 6200 to a 6600. > Read more

nVidia's first card with TurboCache technology. Cards with TurboCache have very little memory on-board and use the computers' internal RAM (through the PCI-Express bus) in case more RAM is needed. This is cheaper, but also slower. > Read more

A decent middle-class card. Higher clocks but narrower bandwidth compared to the 6800. Due to the higher clock frequency it's sometimes faster than it's bigger brother, but in those cases an 6800 would have enough rendering muscle to get the job done as well. After all you won't notice a difference between 400 FPS and 500 FPS . > Read more

On Thursday 9 March, nVidia announced the GeForce 7600 and 7900 series. The 7600 GT was an ideal mid-range card with good performance, drivers and build quality. > Read more

> Read more