GeForce 8800 GT

Written by - on 05 November 2009. Posted in GeForce 8 Series

There was a real hole in the NVIDIA lineup - between the GeForce 8600 GTS (upper representative of the middle class) and the GeForce 8800 GTS 320 MB (the initial video card of the high-end sector). At the same time, the empty price niche of 0-300 is very attractive for the manufacturer in terms of profit, because the volume of sales in quantitative terms here, in contrast to top solutions, is quite significant. One way or another, NVIDIA decided to correct the situation by releasing the GeForce 8800 GT based on the G92 GPU.

As for the rest, the GeForce 8800 GT does not fundamentally differ from the G80 - the same unified architecture, the same principles of building a 3D scene. But there are also significant changes. This video card was the first on the market to support PCI Express 2.0, a new version of this interface that allows you to increase the throughput from 2.5 to 5 Gb / s per channel. At the same time, the permissible power consumption of the device has been increased, which should in the future lead (again) to the appearance of even powerful video cards without an additional power connector. In addition, the PCI Express 2.0 specification brought some more useful innovations, but for us the most important thing is that the new version of the protocol is top-down and bottom-up compatible with the previous standard and makes it possible to use any combination of motherboards and video cards that support PCI Express 2.0 or without it.

The G92 contains 754 million transistors, more than the G80. At the same time, the area of the core has decreased due to a thinner process technology. The additional transistor budget was spent on the integration of the NVIO chip responsible for RAMDAC (in the G80, NVIO was soldered separately from the GPU), as well as on a number of architectural changes in the core.

The GeForce 8800 GT has once again returned to the technical characteristics of the usual digital expression - the memory bus width is 256 bits, and the graphics RAM itself is set to 512 MB. Note that it was the 512-MB version of the board that went on sale first, and then the 256 MB version, which, in addition to half the amount of RAM, has a memory clock frequency reduced to 1400 MHz.

641821

Of all the changes, only a reduced number of blocks and some changes in TMU. G92 consists of seven universal computing units (shader processors), and of 112 processors. Each of the blocks, in which 8 TMUs and 16 ALUs are grouped, can execute part of a vertex, pixel or geometry shader over a block of 32 pixels, vertices or other primitives, can be engaged in other (non-graphical) calculations. Each processor has its own first level cache, which stores textures and other data. In addition to the control unit and compute shader processors, there are four ROPs that perform visibility detection, framebuffer writing, and MSAA, grouped with memory controllers, write queues, and a second-level cache.

Specifications NVIDIA GeForce 8800 GT

Name	GeForce 8800 GT
Core	G92 (D8P)
Process technology (µm)	0.065
Transistors (million)	754
Core frequency	600
Memory frequency (DDR)	900 (1800)
Bus and memory type	GDDR3 256 Bit
Bandwidth (Gb/s)	57.6
Unified shader blocks	112
Frequency of unified shader units	1500
TMU per conveyor	56 (total)
ROP	16
textures per clock	16
textures per pass	16
Shader Model	4.0
Fill Rate (Mpix/s)	9600
Fill Rate (Mtex/s)	33600
DirectX	10.0
Anti-Aliasing (Max)	SS&MS - 16x
Anisotropic Filtering (Max)	16x
Memory	256/512
Interface	PCI-E 2.0
RAMDAC	2x400

GeForce 8800 GT, conceived as a mid-range product, unexpectedly turned out to be so good that it eclipsed everyone, except, perhaps, only the GeForce 8800 GTX/Ultra. As a result, users got a really worthwhile video card, which is a worthy candidate for purchase.