For those of you who haven't read our previous ELSA
reviews or have never heard of them before, here is quick synopsis for
you. ELSA is becoming a bigger and bigger name when it comes
to 3D consumer graphics cards. Just over a year ago, most of you
would have been puzzled if I asked you who ELSA was and what products
they offered. That's because they were primarily in the
Workstation class graphics business and have established quite a name
for themselves in that market. Then just over a year ago, they
decided to jump into the consumer 3D graphics board market and primarily based
their cards on NVIDIA chips. Now, if you're a person interested in 3D
graphics boards and follow the news, you'll be pretty familiar with
ELSA. This is a perfect testament of how well the Internet works
to the word out there.
Today, we are going to take a look at ELSA's top of the
line consumer graphics card, the ELSA ERAZOR X2. As you
may already know, this board is based on NVIDIA's extremely powerful 23
million transistor 0.22µ chip, the GeForce 256. The X2
board also has the luxury of carrying 150MHz DDR Infineon 6ns SGRAM
whereas the regular ERAZOR X has 166MHz SDR SDRAM. The difference
between DDR (Double Data Rate) and SDR (Single Data Rate) is that DDR
RAM operates on the rising and falling edge of a clock pulse whereas SDR
RAM only does operations on the falling edge of a clock pulse.
Don't let the 150MHz of DDR RAM fool you. Since it transfers on
the rising and falling edge of a clock pulse, to compare it to SDRAM we
would have to multiply that clock speed by two and we would effectively
have 300MHz SDR RAM.
We're going to tell you straight out, this card rocks
and performance wise you can't get much better than this. This
statement applies to just about all other GeForce DDR cards on the
market as well. The performance difference between two GeForce DDR
cards will be very minimal.
Let's take a closer look at the GeForce 256 before
diving into the review, feel free to skip this part if you've already
read about the GeForce 256.
First and foremost, the main difference between the
GeForce and previous NVIDIA chips as well as all other chips currently
on the market (other than then Savage2000) is that the GeForce is the
first "GPU" Graphics Processing Unit as NVIDIA calls it.
What this means is that the GeForce is a single-chip with integrated
transform, lighting, triangle setup/clipping and rendering engines that
is capable of processing a minimum of 10 million polygons/s. To
give you a visual example of what the GeForce is capable of, here are
some screenshots of one of NVIDIA tech demos:
Without GPU |
With GPU
|
As you can see, the image on the right is much more
detailed because the GeForce can handle a lot more polygons because of
its built-in transform and lighting engine. In essence a GPU
offloads much of the work from the system processor so that it can work
on other things like AI and physics.
The other major improvement over previous generations
that the GeForce offers is the 256-bit architecture. This does not
mean it's memory bus is 256-bits wide, in fact the memory bus is still
128-bits. Instead, the 256 refers to its quad rendering pipeline which
renders for 64-bit pipelines at the same time (4 x 64 =
256). Therefore, the GeForce can pump out four 32-bit pixels
per clock cycle compared to the two pixels per clock of the TNT and
TNT2.
This brings us to next point of clock speeds. All
GeForce cards available as of this moment come preset at a core speed of
120MHz. Since the chip renders four pixels per clock, a clock
speed of those greater than the TNT2 Ultra's is not needed. For
example, a TNT2 Ultra at 175MHz has a fillrate of 350 MPixels/s where
the GeForce has a fillrate of 480M Pixels/s at 120MHz.
Furthermore, when you increase the clock speed of the GeForce, you
increasing the performance 2 times more than you would with the TNT2
with the same clock speed increase. For instance, if you increase
the GeForce's clock speed to 135MHz, you'll have a theoretical fillrate
of 540M Pixels/s. If you increase the TNT2 Ultra from 175 to
190MHz (same 15MHz increase), you'll only increase the theoretical
fillrate to 380M Pixels/s. Keep in mind that these values are only
theoretical and you won't even get close to them with SDR ram because of
the limited memory bandwidth, which is why NVIDIA developed the chip to
work with DDR RAM as well. DDR RAM will allow the card to reach
those maximum fillrates.
Next Page: Features
In Depth
Hardware Pros Home Page
|
