MaximumPC 2008 09

46 | MAXIMUMPC | SEP 08 | http://www.maximumpc.com

ATI TO NVIDIA:

YOU’RE OBSOLETE

Under the Heat Spreader

ATI is kissing the giant GPU goodbye, preferring smaller, more

effi cient GPUs that can work in tandem on big workloads

We’ve walked this path before. When

Intel’s NetBurst architecture reached the

end of its life, we were seeing the largest,

hottest, most power-hungry CPUs ever,

but performance wasn’t scaling in line

with the power and heat increases. In or-

der to see a 10 percent performance boost,

the new CPU would generate 30 percent

more heat and require 30 percent more

power. This was an untenable situation, so

Intel and AMD quickly moved away from

monolithic cores to more effi cient multi-

core designs. If your applications can take

advantage of all the CPU cores in your

system, you should see signifi cantly better

performance with a much slower, cooler

multicore design than you would with a

similar-size single-core design running at

twice the speed.

The two main GPU manufactur-

ers are at a similar crossroads, but each

chose a diff erent direction with its new

generation of GPU. Nvidia has launched

its GTX 280 boards, which sport a mas-

sive, monolithic GPU design. These are

among the largest chips ever put into

mass production—a single GTX 280 chip

is 576mm^2 , features a 512-bit memory

interface, and draws 236W when running

full bore. By contrast, the RV770 chip that

ATI is using in its new line of GPUs is just

260mm^2 , features a 256-bit memory bus,

and draws about 170W when running at

full power. Yet despite a much smaller die,

a lower power draw, and a memory bus

that’s half the width of the GTX 280’s, the

Radeon 4870 delivers about 75 percent of

the speed of the GTX 280 in most of our

benchmarks.

THE GPU CORE

With this generation of GPU, ATI’s begin-

ning to see the payoff from its premature

move to a 55nm die size last generation.

While Nvidia languishes at 65nm, ATI is

packing more silicon into a smaller space

and increasing effi ciency at the same

time. But that’s not all ATI’s done. The new

RV770 series GPU features a complete

redesign with an astounding 800 stream

processors—the little silicon dynamos

that handle everything from rendering

soft shadows and bump maps to decoding

H.264 video from Blu-ray movies.

By integrating 16KB of cache with

bundles of 10 stream processors and a

dedicated texture unit into so-called SIMD

units (which combined make up the com-

plete shader core), ATI has juiced much

better shader performance out of the

overall package. The stream processors

in each SIMD unit can share information

using their shared memory, which makes

the new shader core more effi cient than

previous designs. And because the stream

processors pump their output directly into

a dedicated texture unit, there’s very little

time lost between writing the output to

texture memory.

The SIMD units themselves are each

integrated with four texture processors in

modular units, which minimizes latency

and improves the performance of the

RV770 design. Each SIMD connects to its

four dedicated texture processors with

480GB/s of bandwidth between them. This

was absolutely crucial to maintain perfor-

mance; otherwise, the texture processors,

which render the actual pixels that are

displayed, would remain a bottleneck.

The RV770 Unveiled

ATI packed the latency-sensitive silicon,

such as the stream processors and the basic

texturing units, in the center of the RV770

GPU. Surrounding it are the memory control-

lers and the L2 cache, and on the periphery of

the chip rests the memory interface (GDDR5

on the 4870 and GDDR3 on the 4850), the PCI

Express connection, the CrossFire controller,

and the various display controllers for DVI,

HDMI, VGA, and DisplayPort. And all of that is

packed on a 260mm^2 55nm die.

UNDER THE HOOD

SIMD UNITS

TEXTURE

PROCESSORS

UVD & DISPLAY

CONTROLLERS

GDDR MEMORY INTERFACE

PCI EXPRESS BUS INTERFACE