Graphics : The New Era Beckons

The next wave of computing is going to be dominated with rich visuals. You

don't need to go too far to realize that. Social networking sites, apps to

enhance Internet experience like Silverlight, a lot more powerful GPUs and

integrated graphics on board, all these point to the fact that high-end graphics

are not going to remain the preserve of 3-D or seismic modeling for buildings,

or medical laboratories and of course gamers; but in today's world of 3-D web,

rich graphics shall be the must-have for every machine.

Let's do some number crunching first. According John Peddie Research, the

number of GPUs shipped for Q3 2008 saw a 17.8 percent growth rate, the highest

quarter-over-quarter growth rate for the last six years; with more than 111

million GPUs shipped as compared to 91 million shipped during Q3 last year.

Thanks to all the visually intensive apps that we are used to these days,

having a good GPU can improve user experience tremendously as opposed to just

having the latest and fastest CPU. Even a somewhat unlikely area such as a data

center is witnessing an increase in GPU usage with some of the resource hungry

tasks moving from commodity CPU boxes to GPUs. This is good news for GPU vendors

as this would ensure a steady demand for their GPUs from large enterprises!

Intel is soon coming up with its Larrabee GPU chip that can be considered a

hybrid between multi-core CPUs and GPUs. The architecture similarities with a

CPU with the added dose of GPU functionality means the chip has been primarily

designed to compete in the General Purpose GPU (GPGPU) and high-end computing

markets. It is slated for release in late 2008 and video cards based on this

chip can be expected by late 2009 or early 2010. This has been a rather late

response by Intel to NVIDIA's Tesla and AMD's Firestream, the GPGPU products

that have in some places already displaced Intel's CPUs in supercomputing.

Not looking too far, you'll see increased graphics processor usage right in

your desktop. Apart from the known resource hungry games and animation apps,

most of the other specialized apps such as Google Earth, GPS navigation systems

and office apps starting Office 2007, require the power of a GPU. Even on Macs,

the flip amongst photos and movies is made easy by the visual interface that is

handled through the GPU. The rise in the usage of innovative visual interfaces

such as multi-touch for faster browsing, is also attributed to a growing

reliance on GPUs.

GPUs can provide a lot of value for money for enterprises when it comes to

rapid information processing. Quite handy as they are for applications that

require a lot of parallel data processing, while the CPUs take care of processes

that require decision-making at each step.

GPUs are in no way ready to kill CPUs right now, as they would shoulder some

of the load in corporate data centers to run compute-intensive apps or database

queries. The increasing interest in delivering visual rich content over IP and

the amount of data stored in corporate databases all present potential usage for

GPUs, which means these chips are bound to move out of desktops toward the more

complex enterprise IT equipment.

Let's take a quick peek at some of the recent developments in this enchanting

domain.

Vapor Chamber-based cooling



The increase in performance of GPUs or CPUs comes at a cost: increased power
consumption. This calls for more advancement in the design of thermal

dissipation systems.

ATI, graphics product group of Advanced Micro Devices, and Celsia

Technologies have joined hands to develop higher-performance cooling solutions

for GPUs. The new coolers will used vapor chamber pipes instead of traditional

heat-pipes.

The system, called a NanoSpreader, uses a copper encased two-phase vapor

chamber into which pure water is vacuum sealed. The liquid is absorbed by a

copper-mesh wick and passed as vapor through a micro-perforated copper sheet

where it cools and returns as liquid to the wick. However, just like heat pipe

based coolers, this system also requires space consuming heat spreader fins.

Where they score is that they are half the weight of solid copper, yet can

transfer heat at roughly ten times the rate. Also, in this configuration, the

thermal resistance from the device to the cooling surface, ie heat sink is

reduced by minimizing or eliminating conduction and the associated interfaces.

ATI Hybrid Graphics



ATI Hybrid Graphics is a conglomeration of the multi-GPU connectivity
technology, CrossFireX, that lets you connect upto 4 discrete GPUs and other

technologies that provide power state adjustment and multi-display

functionality. The pre-requisite being that you need an AMD 7-series chipset

such as 780 that has integrated graphics and an ATI Radeon HD 3000 and above

Series GPU. Two monitors can be connected to each discrete GPU card and the

integrated graphics accelerator can work in parallel with a discrete GPU to give

enhanced visual experience. The same technology can be used to enhance battery

life on notebooks. This is done through dynamic switching between integrated

graphics and discrete graphics processors which depends on what app you're using

at a particular instance. For routine tasks such as checking email, web surfing,

etc the system works on integrated graphics accelerator but for graphically

intense apps such as games, animation software and videos, it switches over to

discrete GPU mode.

NVIDIA Hybrid SLI



The next version of SLI, the Hybrid SLI is NVIDIA's solution for hybrid
graphics. The older SLI technology enables two or more discrete GPUs to work

together and provide increased graphics performance. The Hybrid SLI technology

is similar to SLI in allowing multi-GPU collaboration but with two added

benefits: apart from connecting two discrete GPUs (dGPUs), the

NVIDIA motherboard GPU (mGPU) can be made to work with an NVIDIA dGPU to enhance

the overall visual experience. Here, the mGPU and dGPU share the rendering load

by rendering different frames of an image; and secondly, the HybridPower feature

enables the user to switch off the dGPU when the maximum performance of the dGPU

is not required and use the mGPU for non-intensive graphics applications, such

as high definition DVD playback, Web surfing and office productivity

applications. Switching off the dGPU not only lowers the total system power

consumption but also lowers total system noise and heat.

The GeForce Boost mode (part of the technology) allows up to two displays to

be connected to either the mGPU or dGPU. When displays are connected to both

adapters, Hybrid SLI mode is disabled and multi adapter extended desktop mode is

enabled. In this mode, up to four displays may be connected (two driven by the

mGPU and the other two by the dGPU).

Both ATI Hybrid Graphics and NVIDIA Hybrid SLI have been designed to take

full advantage of DirectX 10.1 and the Unified Shader Model 4.1, and so they

work only on Vista.

What's new in DirectX 10.1



No discussion on the future of graphics can ever be complete without discussing
the Microsoft API for graphics developers: Direct X. The latest version DirectX

10.1 comes with a host of changes over the revolutionary DirectX 10, released in

parallel with Vista. One of the main improvements is the improved access to

shader resources. In particular, better control when reading back samples from

multi-sample anti-aliasing. Along with this, the ability to create customized

downsampling filters is also available.

New formats for render targets which support blending mean that render

targets can now be blended independently of one another.

Direct3D 10.1 has improvements in shadow filtering capabilities within the

API but we shall have to wait and see how well developers can use this in their

future apps.

More importantly, the API has been modified keeping in tune the needs of

multi-core processors. So, it will allow for higher performance in multi-core

systems. The number of calls to the API when drawing and rendering reflections

and refractions (two commonly used features in modern game titles) have been

reduced in Direct3D 10.1, which should make for some nice performance boosts.

Finally, another oft-used feature, cube mapping, now comes with the ability to

use an indexable array for handling cube maps.

One of the major additions that would impact image quality in DirectX 10.1is

the introduction of 32-bit floating-point filtering. Currently, 16-bit filtering

is used in both DirectX 9 and 10. This would impact the quality of High Dynamic

Range rendering in particular. On top of this, the overall precision throughout

the rendering pipeline will also be increased. These increases in precision

could make for an interesting challenge for the graphics independent hardware

vendors (IHVs), for they would need to spend major resources in upgrading their

equipment to match these new requirements.

Looking at improvements on the image quality front, DirectX 10.1 introduces

full application control over anti-aliasing. This allows applications to control

the usage of both multi-sample and super-sample anti-aliasing, as well as give

them the ability to choose sample patterns to best suit the rendering scenario

in a particular scene or title. These changes in DirectX 10.1 give the

application control over the pixel coverage mask, which is used to help to

quickly approximate sampling for an area of pixels. This in particular should

prove to be a boon when anti-aliasing particles, vegetation, scenes with motion

blur and the like. All of this additional control handed to the application

could allow for anti-aliasing to be used much more effectively controlled by

game developers.