NVIDIA Quadro vs GeForce - Head to Head #33

NVIDIA's latest release shows Ivon Smith the company still knows what it's doing a both ends of the workstation graphics card market.

The new GPU lineup
After NVIDIA ramped up it's GPU line by moving-on from the FX5800 to the 5900, it was only a matter of time before these changes made their way into the professional graphics card line of products. Hence the stunning 128MB Quadro FX2000 has recently been shunted down NVIDIA's hierarchy of uber-cards, humbling itself to the new 256MB Quadro FX3000, (based on the later and greater FX5900 GPU), which will undoubtedly be a serious contender for the workstation graphics card champion over coming months as the other manufacturers release their own updated high-end products.

Although NVIDIA, with this latest release, are taking-on the world again, challenging the boundaries of what is possible in terms of 3D application display features and OpenGL (or even DirectX) speed and performance in 3D design work, they have not forgotten the little guy who perhaps doesn't have the need or budget to purchase a four or five thousand dollar graphics card.

The important factors are?
Hardware
When it comes to working with 3D models and animation and purchasing a graphics card for design, TV, film, game development or industrial visualisation, a key word is practicality. Does the card work reliably, stably and have the display quality and performance in your chosen applications for you to be able to be productively creative and have the turnaround on 3D jobs that you (or your employers) want?

Obviously, a key factor in this criteria is the hardware architecture. Memory size, memory clock speed and core GPU clock speed are three of the most powerful influences over general card performance. The larger the memory, the larger the polygon count and texture maps on models that can be moved around at the speed that the RAM can push it. Vertex processing engines and caching dictate how easily large geometry models can be worked with, and the number and speed of pixel pipelines is important when it comes to texture handling within 3D applications as well many of the newer lighting and real-time material effects available now for designers and artists.

Software
Much of the latest hardware evolution has had to be developed in close conjunction with the software language and driver codes. NVIDIA have been a leading developer in this area, working closely with Microsoft on the progression of DirectX implementation and support within their graphics cards. OpenGL coding is of prime importance to workstation cards as the majority of 3D applications achieve most of their high performance results when geared towards OpenGL drivers rather than DirectX. This is partly because the development and design process involves much 3D work in simple texture-absent shaded display mode, or even 'simple' wireframe mode, essential when trying to create accurate industrial models of mechanical devices, engines, vehicles (CAD) or when creating and editing organic looking 3D models of faces, heads and character bodies in the world of TV or film entertainment (DCC).

Game developers do need to be able to utilise the most accurate and advanced DirectX display features to best visualise immediately what their target audiences are going to see when the full-blown version of a game is played, likely using the consumer versions of the workstation graphics cards. Supplementary to this, is the development and integration of NVIDIA's own Cg graphics language, which enhances both OpenGL and DirectX display modes when working in 3D applications. Leading 3D software developers, such as AVID have worked with NVIDIA to enable the use of Cg within their high-end non-linear 3D animation production application Softimage/XSI. Advanced texture and shader formats are allowing game developers to plug their game engine directly into XSI to display almost exactly what the PC or console platform user will see in terms of lighting, shaders and effects within the 3D viewports of Softimage/XSI. Not only is this just cool for the clients and 3D artists working on a game title, but it really speeds and improves the production process cutting down the necessity to export models, scenes or animations to try out in the game engine. This WYSIWYG approach to game creation has largely been made possible via the advanced coding languages that are implemented on today's workstation graphics cards.
 
So, hardware and advanced coding languages on workstation cards allow performance and special features to be used. The other and vitally important part of workstation 3D card design, is the development of drivers. Well-written and reliable drivers can be the key to really unlock the ultimate performance from both consumer and professional 3D graphics cards. In fact it is this, rather than radically different hardware architecture, that is often attributed to the vast chasm between investment costs of the top-of-the-line games cards compared to workstation cards. Pricing is typically 300-500% greater for the equivalent model GPU workstation card.

Is this justified in terms of real world reliability, performance and feature sets available on these professional cards or are these price hikes more likely influenced by the market place? It's true that the 3D industry sector has larger and more expendable resources of cash than your (not so) average Joe Blow gaming fanatic to throw after 3D accelerator hardware, and that they are chasing return capital on their investments, but surely the manufacturers haven't just been exploiting this factor for so long have they?

A mention of marketing vs. market readiness
I would never deny that to date NVIDIA have produced some fantastic 3D graphics products, both in terms of hardware and GPU development as well as software and driver production pipelines. They broke the mould years ago with their GeForce 256 gaming card and then bravely ventured into the more aloof and particular workstation market, and their line of Quadro cards has always been successful and truly powerful within the 3D production environment. However, one thing that has managed to keep pace with their famous and voracious development cycles is their marketing machine. It would seem that in a bid to retain their marketplace image, in the face of stiff competition from new products by ATI and 3dlabs, NVIDIA has pushed out their new workstation hardware before proper and full software support has been finalised.

Both the Quadro cards used in these tests were completely unable to run on the now familiar, reliable and high-performing certified 3dsmax v.43.03 and MaxTreme v.4.00.29 drivers.

ABOVE: A scientific particle-based fluid flow visualisation test used for data analysis and the 3D display represents flow tracking with coloured tubes based on particle velocity.

With previous workstation cards, these drivers, created alongside the 3D software developers, invariably produced performance results in tests and general working environments that were 25-30% faster. To get the FX500 and 3000 cards to even function, the generic Quadro drivers v.44.71 had to be used, which undoubtedly may produce a few slightly miffed professionals out there who invest in these new products expecting great leaps and bounds in performance and dedicated driver sets within their 3D production software of choice. Admittedly, these are brand new products, drivers will undoubtedly become available so