NVIDIA GeForce 8-series

Overview

This TekSpek constructs an overview of NVIDIA's 8th GeForce series of graphics products.

The Technology

NVIDIA's GeForce 8-series graphics products, powered by G8x GPUs, sees it usher in the first truly radical architecture change since the Riva TNT, and that includes the jump to fully programmable shading in the GeForce FX. G8x GPUs implement a fully-threaded, fully-unified, fully-scalar shading architecture that's targetted at Direct3D10, the 3D component of DirectX 10.

Input data hits the thread setup units for organisation, before hitting the global scheduler. Pre-shading optimisations are performed and data is sent to the local cluster schedulers for execution on a particular cluster. The global and local cluster scheduler's implement G80's hardware load balancing effectively, organising data per cluster, the local schedulers sleeping and waking the threads in their control as needed and for free in any given cycle. The L1 store per cluster is connected to the shared L2 store and finished data from a cluster flows to the ROP for further processing, or back to the global scheduler for further assignment to other upcoming threads.

Shading takes place on 128 scalar ALUs, grouped in clusters of 16, each cluster capable of 16 scalar MADD+MULs per cycle, 16 scalar interpolations and 8 bilerps of data filtering, all in one clock. That shading, interpolation/SF and data fetch and filtering setup is the core of G80.

It's therefore in strict contrast to every non-unified, fixed-ratio vector-based GPU ever unleashed into the consumer space, and the target is clearly D3D10.

After the shading core, the rest of the chip is setup for new levels of image quality via a true 8x antialiasing mode that works with 'HDR', and a further coverage sample AA mode that can potentially offer even higher IQ than that. Coupled with the best AF level selection of any consumer hardware yet made and 8-series GeForce hardware has the ability to push the best looking pixels ever seen on the PC.

Current GeForce 8-series boards

Current 8-series GeForce products are the GeForce 8800 GTX and 8800 GTS. The GTX is the flagship and uses a full G80 with 128 SP units, 768MB of memory on a 384-bit memory bus, and clocks of 575/1350/900 (1800MHz effective). The GTS has 96 SPs, 640MB on a 320-bit memory bus, and clocks of 500/1200/800 (core/shader/memory), creating a clear performance gap between itself and the GTX, leaving room for a possible extra SKU at some point.

NVIDIA, however, has now introduced a 320MB version of the GTS that's priced at around £200 for a full-retail model. The only difference between it and the 640MB SKU is the size of the onboard memory.

The GTX is over 10 inches long and requires two 6-pin PCIe power connectors, running double-slot cooling and announcing itself as the biggest and most power hungry (in terms of connectors needed) consumer graphics board ever made. Make sure your chassis has the room and you have the connectors on your PSU for it (natively or via adaptors).

Both variants of the GTS are over an inch shorter and just need one power connector, but retain dual-slot cooling. They're almost an identical size to a 7900 GTX or something like an ATI Radeon X1950 XTX, and outperform them both.

Performance is outstanding, the GTX especially outrunning anything that's come before in PC graphics by some margin, and the GTS stands a cut above the likes of the 7900 GTX and with better IQ to boot. Current pricing is ~£375 for GTX and ~£275 for GTS 640MB and ~£200 for the GTS 320MB.

Summary

Check out HEXUS for the full lowdown, but be rest assured that 8-series GeForce is the fastest, best-looking and most forward-thinking graphics series on the market today, with 8800 GTX stomping on everything else from a high height. All three SKUs support SLI, should you want to run two, and none will disappoint. Microsoft Windows Vista and DirectX 10 drivers are now available, for users looking to D3D10 as the real use of their new 8-series GeForce product.