Introducing the GeForce4: nVidia strikes back

Introducing the GeForce4: nVidia strikes back

Models and specifications

In contrast to the now known procedure, first skim off the most affluent/willing to buy layer of hardware freaks and first the top Bringing out models in order to later make the euro with the masses with the mainstream offerings seems to be the other way around this time. According to manufacturers, the cheaper MX cards will be available soon, while you will have to wait a little longer for the more powerful Ti models.

GeForce 4 MX
GeForce4 MX 420 chip
GeForce4 MX 440 chip
GeForce4 MX 460 Chip
GeForce4 Ti card with VGA and DVI
GeForce 4 Ti
GeForce4Ti 4400 chip
GeForce4Ti 4600 Chip

Attention: Also in contrast to previous practice, this time there will be a clear difference in terms of the supported 3D features between high-end and mainstream. The MX models will not contain DirectX 8 pixel and vertex shaders. The Ti models, on the other hand, have to do without the Video Processing Engine (VPE).

Enough said, let's take a look at a table that shows the planned models of both the MX and the Ti series . In the GeForce4 MX series in particular, there are significant mathematical differences in performance. In the following tables it was calculated with the simplification that 1 GByte is exactly 1000 Mbytes, or 1 million Kbytes and so on.

GeForce4 MX compared to GeForce2 MX
Segment Low-End Mainstream Mainstream Mainstream Architecture nV11 nV17 nV17 nV17 Chip clock 200MHz 250MHz 275MHz 300MHz Memory clock 166MHz 166MHz 200MHz 275MHz Connection 128Bit-SDR 128Bit-SDR 128Bit-DDR 128Bit-DDR memory bandwidth 2.7 GB/s 2.7 GB/s 6.4 GB/s 8.8 GB/s Max memory size 128MB 64MB 64MB 64MB T&L performance 25 MT/s 31.25 MT/s 34.375 MT/s 37.5 MT/s Shader Hardware nVidia Shading Rasterizer, 2nd Gen. Hardware T&L Rendering Pipelines 2 2 2 2 Max. Theo. Fill rate 0.4 GPix/s 0.5 GPix/s 0.55 GPix/s 0.6 GPix/s 0.8 GTex/s 1GTex/s 1.1 GTex/s 1.2 GTex/s AGP interface 4x 4x 4x 4x transistors 20M? ? ? FSAA technology Supersampling Multisampling (Accuview) MultiMonitor Support TwinView nView nView nView Other PU, LMA-II PU, LMA-II PU, LMA-II

As you can see, the memory bandwidth is the GeForce4 MX420 clearly below that of the other GeForce4 MX cards. This shortcoming in particular should contribute to a rather mediocre performance of the smallest GeForce4 MX. The following table shows the top models of the GeForce4 Ti series compared to the older GeForce3 and GeForce3 Ti 500.

GeForce4 Ti compared to the GeForce3

Here we can only appeal to nVidia to reconsider the naming, if you don't want to accept greater confusion on the part of the paying customers and maybe the chips with a prefixed number à la ATi with regard to their DirectX compliance.

The GPUs themselves are still manufactured in the meanwhile mature 0.15µ manufacturing process. But since the GeForce4 Ti4600 in particular, with its high number of transistors of 63 million and a clock rate of 300MHz, produces a good portion of waste heat despite the PBGA (plastic ball grid array) package, nVidia has now come up with an integrated cooling solution. This just seems to be a cooler permanently connected to the chip, which is at least optically characterized by a very low design. The usual packaging of the chip may be dispensed with and, similar to Intel's Coppermine and AMD's Athlon/Duron series, the cooler is attached directly to the chip for better heat dissipation. The memory chips are also an innovation, as they are also manufactured using the BGA process in order to keep the signal paths as short as possible and thus enable the high clock rates.

On the next page: Features