In a world where convergence seems to
be the word of the moment it seems odd that the graphics
card industry seems so hell-bent on divergence. Different
architectures performing functions in increasingly
different ways with no real industry standard to glue
it all together is making life hell for the discerning
buyer and the hardware analyst who genuinely cares
about his or her results and the message they convey.
As the PC market has grown beyond anyone's wildest
expectations so it's taken increasingly bold measures
to get noticed and with these measures have come accusations
of unfair optimisations, benchmark rigging and generally
unsavory deeds driven more by political wranglings
than by the desire to further the industry towards
its ultimate aim of high resolution, real-time photo-realistic
3D rendering for the masses.
Well, you'll be pleased to hear that
I've decided to completely duck the issues of who's
cheating at what and just hit the ground running with
a simple, uncomplicated look at the GeForce 5600 Ultra.
There are plenty of forums around where you can read
about or make comment on who you think has sunk the
lowest in their quest for domination of the high end
graphics market, our own included, but for the next
few pages I'm going to pretend it never happened.
Neither of the big players, ATi and NVIDIA, are entirely
blameless any more than either of them are entirely
guilty, so let's put that issue aside for another
The graphics card on the bench for today's
review is the GeForceFX 5600 Ultra. This is a reference
card and is based on the newer Flip Chip GPU with
its increased core and memory frequencies. The core
has been increased from 350 to 400 MHz as has the
memory making for an 800MHz effective memory clock.
This results in an increase in memory bandwidth from
11.2 to 12.8 MB/Sec and an increase in pixel and texel
fillrate from 1400 to 1600 Mpixels/sec.
These increases in frequency were due
in part to the switch to a new Flip Chip design GPU.
So what's Flip Chip? Well, in very simple terms it
involves literally flipping the chip so the core is
on the bottom. This means that rather than having
to transfer connections down to the PCB with fine
wires arranged around the perimeter of the chip (known
as wire bonding) the core can be connected directly
to the PCB.
There are many benefits to this. One is that the shorter
connection paths leads to a reduction in resistance,
capacitance and inductance. Another advantage is that
chips with relatively small perimeters are not limited
in the number of connections available as they would
be using the wire bonding process.
Of course it's not all good news, there are some
disadvantages but fortunately none that affect the
user as they are mainly problems linked to testing
and handling at the manufacturing stage.
Anyway, enough small talk, let's get stuck in with
a look at the card we'll be testing.