nVidia Optimus and ASUS UL50VF: An Invisible Game Changer

nVidia Optimus Technology

The main idea that guided the teams lead by Sasha Ostojic, Director GPU Software and Sridhar Pursai, Director Platform Engineering

When we first learned of the nVidia Optimus Technology back in November, it looked quite promising – even though we had no idea how nVidia would offer seamless switching. I admit I was wrong, but my idea was something in the lines of new NVIO chip that is usually bundled with high-end GPUs and that the new NVIO chip would accept signals from Intel or AMD integrated graphics part. This time around, I am really glad I was wrong. We received an invite to an Optimus Deep Dive event featuring us and 13 other journalists across the globe. The Optimus Deep Dive event took the better part of one cold Monday in January.


nVidia Optimus Technology - Good bye MUX, Software overlay controls the both GPUs

To answer immediately, Optimus is nothing else but a combination of software and internal hardware controlled graphics switching. Forget about expensive MUXes, this is an intelligent way of taking over the processing of DirectX/OpenGL calls the moment they’re made. Optimus works by leaving the Intel’s Display Driver to display image on the screen and actively monitoring everything that is happening in relation to displaying image. The library of applications inside nVidia’s driver will automatically react and switch to the GPU as soon as it detects application profile where nVidia’s GPU would do much better than integrated graphics.


How Optimus Technology works - a combination of hardware and software

The interesting bit about this detection is that nVidia will offer automatic profile updating - when an Optimus-powered notebook asks for updates, it should download new application profiles too, as well as some fixes. This is not an automated driver update - Optimus v1.0 will offer profile updates only, and the company said it is looking into the future variants of Optimus to bring as much seamlessness as possible. Naturally, in order to enable automatic driver component update, nVidia would have to change its internal driver stack. This is also the reason why ATI fell flat on their promise to offer automatic driver updates with Catalyst drivers [back in 2003].


Optimus Copy Engine is the secret to how Intel's IGP sees the finished frames rendered by nVidia GPU

The second part of the Optimus technology is actually a hardware one, but not the NVIO or anything similar. Rather, nVidia’s Optimus-capable GPUs feature "Optimus Copy Engine", a parallel pipeline next to the 3D Engine one. What Copy Engine does is takes the finalized rendered engine created by the 3D Engine and copies the contents from on-board memory to the system memory - which is then taken by Intel’s IGP and displayed on frame-by-frame basis. This approach removed a lot of unnecessary overhead and more importantly, killed latencies that would happen if the company didn’t put the afore mentioned Optimus Copy Engine.

Starting with GeForce G200 series, all nVidia GPUs have built-in support for Optimus technology. According to nVidia, even the upcoming GF100 architecture will support Optimus not just in notebook variants, but in desktop as well, on a top-to-bottom basis. The company is planning to stop Intel’s rising share in graphics by offering a complementary add-on, rather than a complicated discrete add-on to the design.


Launch Platforms

Naturally, in order to adopt the technology, you need a lot of partners behind your back. Starting with this week, ASUS launched five notebook designs on world-wide basis, while EU-oriented Medion launched a notebook for European markets. Six publicly announced designs does not seem like a great deal, but that is a natural consequence of launching the technology prior to China and Taiwan shutting down. As of today [Friday, February 12, 2010], almost all of China and Taiwan shut down for Lunar New Year festival, with offices and foundries opening up in two weeks time, around February 22, 2010. After that, a mad rush to complete the designs to CeBIT will commence, and we should see at least 20-30 design wins on CeBIT 2010 in Hannover, Germany [March 2-6]. For Back To School 2010, nVidia expects to ship around 50 design wins, including Acer, Dell, HP and the rest of top tier vendors.

The Apple Connection 
The relationship between Apple and nVidia has always been an interesting one. After all, Apple came on board and committed to OpenCL after nVidia sent engineers with a working demo of the OpenCL platform back in May 2007. A few months after, Apple launched second generation MacBook Pro and iMacs featuring nVidia GeForce 8-series GPUs and later, GeForce-9 series chipsets - much to the surprise of the analyst and media communities. According to our sources, that demo was the reason why nVidia ended up in Mac machines, as Steve himself was a true believer in the technology demonstrated to him.

After learning that, it was quite funny hearing "Apple is the driving force behind OpenCL" mantra and how some vendors criticized nVidia for forcing proprietary standards while at the same time, we might even say that OpenCL would not have happened if there wasn’t for the company exactly criticized for the lack of "openness". But that’s just how this game goes.

When Intel announced they’re integrating their graphics subsystem with their 32nm products, our sources at Apple were adamant they will not go with the Arrandale-based notebook unless a better graphics switching solution was found. I was personally told "Steve doesn’t trust them [Intel]. We got burned with mediocre graphics on Apple TV and MacBooks and that will never happen again." To make the matters worse for Intel, there seems to be a lot of tension between Apple and Intel as far as patents go. It isn’t widely known, but Intel patents a lot of non-core business related patents, with one of them being keyboard air intake. Apple used this technology on a lot of their products and when Taiwanese vendors started to utilize the same technology, the giant from Cupertino started to pressure Intel not to allow this to happen and got owned in the process.

Getting back to the subject, on the subject of whether or not we are going to see an Apple refresh using Arrandale thanks to the Optimus technology, your guess is as good as mine. If it does happen, remember where you read it first. For now, Apple’s official stance is that Arrandale, just as any other part with Intel integrated graphics, will rot in hell. Who is to blame for that, given that Intel’s integrated graphics can normally run Windows Vista and Windows 7 Aero interfaces, as well as Project Looking Glass and its "offspring" Compiz on Linux. Thus, we believe nVidia Optimus technology has a really good chance of winning the Apple deal all over again.