facebook rss twitter

NVIDIA GeForce GTX 470 1,280MB review. Fermi done right?

by Tarinder Sandhu on 29 March 2010, 20:28

Tags: GeForce GTX 470 1,280MB, NVIDIA (NASDAQ:NVDA)

Quick Link: HEXUS.net/qaxoy

Add to My Vault: x

Snip, snip, snip

GeForce GTX 480 and GTX 470 are both high-end desktop iterations of the GF100 architecture that we took a good look at on Friday. We encourage you to have a read and find out what makes them tick along.

Shamelessly taking the specification-comparison table out of that review and pasting it below.

Graphics cards NVIDIA GeForce GTX 480 1,536MB NVIDIA GeForce GTX 470 1,280MB NVIDIA GeForce GTX 295 1,792MB NVIDIA GeForce GTX 285 1,024MB ATI Radeon HD 5970 2,048MB ATI Radeon HD 5870 1,048MB ATI Radeon HD 5850 1,024MB
General clock 700MHz 607MHz 576MHz 648MHz 725MHz 850MHz 725MHz
Shader clock 1,401MHz 1,215MHz 1,242MHz 1,476MHz 725MHz 850MHz 725MHz
Memory clock (effective) 3,696MHz 3,348MHz 1,998MHz 2,484MHz 4,000MHz 4,800MHz 4,000MHz
Memory interface and size 384-bit, 1,536MB GDDR5 320-bit, 1,280MB GDDR5 896-bit (2 x 448-bit), 1,792MB, GDDR3 512-bit, 1,024MB, GDDR3 512-bit (2 x 256-bit), 2,048MB 256-bit, 1,024MB, GDDR5 256-bit, 1,024MB, GDDR5
Memory bandwidth 177.4GB/s 133.9GB/s 2 x 111.9GB/s 159GB/s 2 x 128GB/s 153.6GB/s 128GB/s
Manufacturing process TSMC, 40nm TSMC, 40nm TSMC, 55nm TSMC, 55nm TSMC, 40nm TSMC, 40nm TSMC, 40nm
DirectX/ Shader Model DX11, 5.0 DX11, 5.0 DX10, 4.0 DX10, 4.0 DX11, 5.0 DX11, 5.0 DX11, 5.0
Vertex, fragment, geometry shading (shared) 480 FP32 scalar ALUs, MADD + MUL  448 FP32 scalar ALUs, MADD  + MUL  480 FP32 scalar ALUs, MADD + MUL  240 FP32 scalar ALUs, MADD + MUL  3,200 FP32 scalar ALUs, MADD + MUL 1,600 FP32 scalar ALUs, MADD + MUL  1,440 FP32 scalar ALUs, MADD + MUL 
Single-precision GFLOPS (single-issue) 1,345 1,088 1,192
4,176 2,720
Texturing 60ppc bilinear
30ppc FP16
15ppc FP32
56ppc bilinear
28ppc FP16
14ppc FP32
160ppc bilinear
80ppc FP16
40ppc FP32
80ppc bilinear
40ppc FP16
20ppc FP32
160ppc bilinear
80ppc FP16
40ppc FP32
80ppc bilinear
40ppc FP16
20ppc FP32
72ppc bilinear
36ppc FP16
18ppc FP32
ROPs 48
64 32
GPixels/s throughput 33.6
27.2 23.2
GTexel/s bilinear 42
Board power (max) 250W 215W 289W 183W 294W 188W 170W
Multi-GPU Three-way SLI Three-way SLI Two-way SLI Three-way SLI Two-way XFire Four-way XFire Four-way XFire
Board length 10.5in 9.5in 10.5in 10.5in 12in 11in 9.5in
Connectors (native) 2x dual-link DVI
2x dual-link DVI
2 x dual-link DVI, HDTV-out,HDMI 2x dual-link DVI
2x dual-link DVI
2x dual-link DVI
HDMI, DisplayPort, 
2x dual-link DVI
HDMI, DisplayPort, 
Etail price  £440+ £299+
£550 £310 £225

GeForce GTX 470 is differentiated and hamstrung when compared to GTX 480 in several ways. It loses what NVIDIA terms a stream-processing unit, composed of 32 cores, which bring the total count down from 480 to 448.

The SM units also carry four texture-unit blocks and, consequently, GTX 470's complement drops from a GPU-wide 60 to 56. The snip in architecture extends to the ROPs at the bottom, where one partition is deactivated. What's more, NVIDIA further reduces the GTX 470's potency by switching off a memory-controller partition, resulting in a 320-bit pathway. The loss in memory-bus width also results in removal of two memory chips, bringing the frame-buffer down from 1,536MB to 1,280MB, assuming same-density modules are used. Last, but not least, the clock-speeds see a downward bump across the board.

NVIDIA's GF100 uses a big, expensive die. Any GPU that doesn't make the GTX 480 grade - and we suspect that to be many in the initial runs at TSMC - is binned, hopefully, at GTX 470 grade. What this means is that it's still the 3bn-transistor GPU which has proven rather difficult to cool.

The numbers tell us that GTX 470 will have difficulty in matching the peak power of the Radeon HD 5870 GPU, yet we don't think NVIDIA can make a huge loss by selling the GPUs and memory to their partners at a cost low enough to compete with the Radeon HD 5850's street price of £225, including  VAT.


In an altered block diagram which the author's five-year-old niece could have done a better job on, here's how GTX 470 stacks up against the ideal GF100 architecture, based on a 512-core design. It's not quite that simple, of course, but you get the idea....