Nvidia Corp. used
Steve Jobs’ keynote address at Macworld Expo Tokyo earlier this week to officially unveil GeForce3, its latest and greatest graphics processing unit (GPU). The company followed up that event late yesterday with a press conference. Nvidia’s conference with the Mac press was led by director of product management Tony Tamasi, who discussed the architecture and design philosophy behind the GeForce3.
GeForce3 represents the first chip in Nvidia’s next-generation architecture. Rumored about prior to its unveiling as the “NV20” chip, according to Tamasi, the GeForce 3 is being billed as “the Infinite Effects GPU.”
Most advanced processor ever
Tamasi explained that the GeForce3 isn’t just the most advanced graphics processor ever built — it’s the most advanced processor available, period. The unit is built using a .15 micron process and is comprised of more than 57 million transistors. With an operating speed of 800 billion operations per second, or 76 gigaflops, the GeForce3 is much faster than either the fastest Pentium or PowerPC chip available today.
Tamasi touted his company’s philosophy of “Moore’s Law Cubed.” Intel co-founder Gordon Moore once observed that the number of transistors on integrated circuits doubled approximately every eighteen months, thus giving rise to the industry term “Moore’s Law”. Tamasi said that since 1997, Nvidia has been able to double the density of transistors on their processors every six months.
GeForce3 brings the Mac and PC into parity with the same core technology as what will be used in Microsoft’s vaunted Xbox home video game console, which is expected to hit store shelves later this year. Tamasi emphasized that the chips are not exactly the same, although there are many similarities.
Tamasi said that the Xbox utilizes a different memory architecture that will result — at least initially — in higher polygon rendering performance than the GeForce3. Tamasi said that Mac and PC cards based on GeForce3 will run at higher resolutions than the Xbox will support, however, and that future generations of Nvidia products will probably outpace the Xbox’s graphics hardware, since the Xbox’s specifications are expected to remain static for several years.
‘Infinite’ flexibility for developers
Tamasi said that the GeForce3’s strength is in its programmability. The “nfinite FX Engine,” a core piece of the GeForce3’s hardware, makes it possible for developers to create their own hardware transform and lighting methods by manipulating vertex and pixel shading directly, rather than the old method, which relied upon preset functions built into the hardware.
For game developers and 3D application developers, this freedom should result in software that has completely unique characteristics and a unique look and feel. Many games that are out now, said Tamasi, look the same because the APIs — the core code base that such software is built upon — are too restrictive. GeForce3’s nfinite FX Engine promises to relieve developers of that restriction.
Tamasi said GeForce3 promises to offer more than just realistic environmental effects, as well. The new hardware can display more realistic facial and character animations, textures, and more, all in real time, as demonstrated by Id Software co-founder John Carmack in his presentation to Macworld Expo Tokyo keynote address attendees earlier this week.
Of course, these sophisticated new effects features are only useful if software developers build support for them into their products. And currently, the GeForce3’s development tools run only on Windows. Tamasi confirmed that efforts are being made to bring GeForce3’s software development kit to the Mac OS — he said that Nvidia and Apple’s engineers are collaborating on that effort now. When asked about a time frame for such tools to be released, Tamasi said “very soon now.”
Radically better performance than other chips
In addition to the nfinite FX Engine, the GeForce3 also sports greatly enhanced memory management features, which Nvidia calls the “Lightspeed Memory Architecture.” The Lightspeed Memory Architecture itself yields tremendously improved overall performance, even compared with Nvidia’s previous GPU offerings. The chip will be available in its initial Mac configuration equipped with 64MB of double data rate (DDR) memory on board.
The GeForce3 has also been designed to offer high-resolution anti aliasing (HRAA). The chip is capable of displaying anti-aliased real time 3D graphics at high resolution without a significant performance penalty, said Tamasi. HRAA provides users with smoother graphics absent of jagged images and artifacts created when textures and polygons are viewed at certain angles. Tamasi said that the GeForce3 also employs new sampling technology to yield higher quality images than previous products. The GeForce3 can achieve a fill rate of about 3.2 billion anti-aliased samples per second — four times faster than the GeForce2 Ultra, and seven times faster than any competitive consumer graphics product, according to Nvidia.
How is the GeForce3’s real-world performance, though? That question was answered during a product demonstration for journalist at Macworld Expo Tokyo. Scott Brodrick, Apple’s product line manager for graphics and displays, showed a chart comparing the four graphic card options now available for the Mac. They are the ATI Rage 128 Pro (with 16MB of SDRAM), the ATI Radeon (with 32 MB of DDR RAM), the GeForce2 MX (with 32 MB of SDRAM), and the GeForce3 (with 64 MB of DDR RAM).
Running Quake 3 Arena at 1280 x 1024 at 32 bits and with sound on saw the Rage 128 Pro delivering 10 frames per second and the Radeon 20 frames per second. The GeForce2 MX tallied 33 frames per second while the GeForce3 racked up 64 frames per second.
The Mac angle
Right now, Apple is the sole source for Mac compatible Nvidia-based graphics cards. Until April, the only way to get such a product is to buy a new Power Mac G4 with the card installed. Tamasi was asked about the GeForce3’s potential for the Macintosh graphics card aftermarket.
Nvidia does not manufacture graphics cards itself, but Tamasi says that the company is open to developing relationships with companies that manufacture Macintosh-compatible video cards.
During his keynote address earlier this week, Steve Jobs stressed that the GeForce3 will be available first on the Mac. When asked, Tamasi indicated that Nvidia is working to deploy GeForce3 with several PC vendors as well.
“Apple announced the GeForce3 first,” said Tamasi. Tamasi couldn’t say when its PC manufacturing partners will have GeForce3-based products ready, but he confirmed that Apple was “as aggressive or more aggressive” in getting a GeForce3-based product to market as its PC competitors.
“Apple will be among the first, if not the first, to ship the product,” said Tamasi.