The fastest GPU simple graphics card
nowadays
It is the first time that AMD has launched
the new graphics card architecture before Nvidiia. It is Southern Island AMD
Radeon HD 7970 that is the fastest graphics card currently.
It is AMD’s bravery to launch a completely
different graphics card design compared to its previous vector processors,
right after bringing us a new CPU architecture, especially basing on the
serious failure of the FX chip series.
We saw little information from AMD about
its previous new architecture – simply named Graphicss Core Next – but at that
time it was running on our test desk and processed plenty of polygons and
pixels on high-resolution screens without any concern about the world.
AMD Radeon HD 7970 may have reason to worry
if considering the fact that Nvidia is going to launch the latest knowledge of
graphics, code-named Kepler, in the spring.
It can also explain the reason why choosing
this launch time is so odd. AMD decided to launch Radeon HD 7970 only 3 days
before Chrismast, a notoriously dead time in technology.
However, talking to AMD’s partners, we
recognize that they are not allowed to launch overclocked cards until a long
time later, so what we see are these reference Radeon HD 7970 cards with new
stickers.
That means most of their echo was
dispossessed by the launch before Christmas, and even the remained echo quickly
became silent by holiday time.
So, like a Minister in Cabinets in crisis
time, AMD is trying to bury Radeon HD 7970’s launch, is not it?
Maybe the firm recognizes the market for
quite small $750 graphics cards. The upcoming Radeon HD 7950 may be the card
that it really wants to focus on. Perhaps it just wants to make sure that
technology journalists have to work even in Christmas time.
Despite the reason for launching this
unequal launch, it is time for us to consider what creates AMD Radeon HD 7970.
Architecture
About real architecture, it is a monster
quite different to HD 6970, which is provided power by the previous Cayman GPU.
This card presents a new tendency in GPU
technology for AMD as well as a production process and new GPU technologies.
As the first, AMD has almost all first-time
things coming with Radeon HD 7970. It is the first GPU built with extremely
tiny 28nm transistor, the first 11.1 DirectX graphics card, and the first
official PCI Express 3.0 component.
One of the really interesting things is
minimizing 28nm die shrink coming with this new Tahiti. Tahiti GPU will be the
chip providing power for high-end cards and built on the new Graphicss Core
Next architecture.
The whole HD 7000 series also comes with
the name Southern Islands, thus having Tahiti mark. Minimizing die shrink means
that it can pack more components into a smaller chip form, and this means more
power for you.
Compared to the number of 40nm transistors
in Radeon HD 6970 – 2.6 billions – AMD works well in packing more transistors
into the Radeon HD 7970. This card has 4.3 billion transistors in GPU. It is a
terrible number – and AMD also has more transistors for HD 7970.
AMD’s announced computing performance for
Tahiti XT GPU in HD 7970 is such number. With 3.79 TFLOP, that is impressive
computing ability deeply buried in the chip. All it needs is the software to
take advantage of.
Computing performance is an impetus fact in
changing from Nothern Islands to Southern Islands, and traditional vector
processors to scalar processors creating this new GPU.
Before, AMD focused most of its silicons to
create a card dedicated to graphics card processing, which leaves the
definition of ephemerous multi-functional GPU computing for Nvidia and its CUDA
cores.
So, the firm lays its all stake in 4-way
vector processor architecture. Basically, this means sorting simple commands to
batches before running them along GPU duct. It is a subtler dolution, and with
graphics processing, it is incredibly effective. 4xxx HD to 6xxx HD card series
are wonderful products with impressively low power requirement compared to
peering products.
However, at this time, AMD wants a slice of
the GPGPU cake, and thus changing to scalar architecture similar to the
architecture used by Nvidia, its hereditary rival.
This means it will use a more vilolent
approach including having a large number of simple processors in an array,
giving them each work to carry out individually until all commands are
finished.
|
Important statistics
|
|
Price
|
$720
|
|
Producer
|
AMD
|
|
Web
|
www.amd.com
|
|
GPU
|
Tahiti XT
|
|
Production process
|
28nm
|
|
Core clock speed
|
925MHz
|
|
Memory clock speed
|
1375MHz
|
|
Caculating units
|
32
|
|
Thread processors
|
2048
|
|
ROP
|
32
|
