As Tom Petersen, Nvidia's director of technical management, explains in the video, the GTX 480 shines when it's tessellation time. During the tessellation-intensive parts of the benchmark, Nvidia's card outpaces ATI's considerably, though at other points they're neck and neck.

It seems obvious that Nvidia would choose something that really played to the GTX 480's strengths for its video debut, so we're hoping that the card lives up to our expectations for insane speeds when it shows up in the wild and people start running their own tests. [YouTube - Thanks Doug]

Sure, it would've been funny if something had gone wrong at that 1:40 mark, but the fact that this little demonstration worked just the way it should is great news. By switching between integrated and discrete graphics, Optimus could one day double our laptop batteries' lives. And that's more exciting than any bit of sparks and smoke could've been. [nTersect Blog]

According to Jon Stokes' well-trusted sources, there is no Cortex A9 dual-core inside Apple's A4, but just a single Cortex A8 and a PowerVR SGX GPU. According to Jon, it's probable that Apple has taken out a lot of the communication, image, and video blocks that the Cortex A8 uses, making it an extremely lean and efficient processor.

He also says that the people from PA Semi—which Apple acquired in 2008—are probably building a custom ARM-based core for future products.

Do you have to care about this? No. As Jon says, the important thing about the iPad is not its specs and how many polygons it can push or how many cores it has. That doesn't matter to normal consumers—that just matters to feature-list-obsessed geeks.

What is important for consumers is that the iPad provides an smooth experience, using a new user interface paradigm. As Jon points out, what's important about the iPad is that it may be the new Wii. Just like Nintendo's console opened gaming to a huge amount of normal consumers who don't have their brains wired to a game pad, the iPad may do exactly the same, simplifying computing for that giant and silent majority of users who dislike or don't really understand the current user experience of Windows and Mac OS X. [Ars Technica]

The GeForce 310 is the same as the GeForce 210, using the GT218 core; the GeForce 315 uses the same GT216 core as the GT 220; and the GT 340 is basically a GT 240. And then there's the GT 320 and GT 330, which use the G92b core, which has roots going back to the GeForce 8800 GT (a card that debuted in 2007).

Man, those new Fermi-powered chips can't come fast enough. [PC Perspective]

Until recently, foundries which manufactured chips on contract stuck to simpler chip designs because that's what their tech was suited for. But now smartphones, and the chips inside of them, are a BFD, so competition's ramping up, with $3 billion plants. GlobalFoundries, which was spun out of AMD, is one of the hot-and-heavy new guys, and about to open a massively advanced (and expensive) new plant in Germany. The first chips they're making? For mobile devices.

Also expensive? Designing chips. The NYT pegs the cost of simply designing a chip at a billion dollars. (Exactly just how much "from scratch" they mean is debatable, since Apple's A4 chip and Nvidia's Tegra use off-the-shelf designs from ARM and others.)

Where things will get interesting is when these mobile chips, mostly ARM-flavored, finally start crossing the same line as Intel's, since ARM chips are scaling up as Intel scales down, and the intersection's not too far away. And that's where Intel's got a chance to really show what it's made of, since they're the last game in town that still designs and makes its own chips. [NYT]

The benchmarking was done by Netbook News while at MWC, who managed to run a 3DMark03 graphics benchmark test during their hands-on time. The resulting score of 3,049 is lower than some Ion LE netbooks achieve; the Samsung N510, for example, scores a 3,593.

There are possible explanations, of course, and we won't know for sure how the Ion 2 stacks up until we're able to test a production-ready unit. As Netbook Choice points out, it's possible that Acer and Nvidia purposefully hamstrung the machine to keep it stable during demonstrations. Then again, it's also possible that Nvidia has had a difficult time working with the Aspire One 532G's Pine Trail processor. Let's hope it's the former: the last thing netbooks need is even more limited graphics capabilities.

[Netbook News via Netbook Choice]

If you remember, the unibody MacBooks from Oct. 2008 were the first to use Nvidia's GeForce 9400M, a combination chipset/integrated GPU that ridiculously outperformed Intel's own integrated graphics, and the Pro models at the time could switch between power-saving integrated and beefier discrete graphics, though it requires a logout.

Nvidia got cockblocked from making chipsets for Intel's latest-gen chips (read: Core i3, Core i5, Core i7), ruling out using Nvidia's better integrated graphics; Nvidia canned their chipset business entirely. So! Nvidia's Optimus tech works with Intel's crappier integrated graphics that are built onto the same die as the newer Core processors, but even more seamlessly than before—the computer automatically switches between Intel's power-sipping integrated and Nvidia's monster discrete graphics, depending on what you're doing. No logouts required.

In the demo we saw on a Windows machine, the automatic part of the tech has some drawbacks—namely, it's entirely dependent on software to tell your computer which graphics card to run. But Apple's obviously worked pretty closely with Nvidia on graphics before, so it seems logical they're doing so here, too. Though I don't expect we'll know until the new MacBooks finally arrive, whenever that happens (hopefully, soon). [AppleInsider]

Optimus is a new technology that allows Intel's integrated GPU to coexist with a discrete Nvidia GPU in a new way, seamlessly alternating between the two depending on the task at hand. It works with GeForce 200M series, GeForce 300M series, next-gen GeForce M, and next-gen Ion GPUs, as well as Intel's Core 2 Duo, Core i3/i5/i7, and Atom N450 processors.

At the most basic level, Optimus is similar to the switchable graphics that Nvidia pioneered a few years ago. But switchable graphics as currently conceived are a pain in the neck: You have to switch manually, there's generally a 5-10 second delay, your screen flickers, you have to shut down certain applications. Optimus still includes a manual option, but otherwise automatically decides what can run on integrated graphics (regular web browsing) and what needs an extra boost (games, Flash video, etc), making the switch for you behind the scenes. It makes the change so quickly by letting the Nvidia GPU handle the processing duties while still employing Intel's IGP as a display controller, as in the diagram above.

One drawback is that Optimus relies on the software to tell it which graphics to employ. That'll be done through an Nvidia verification process and accessed through automatic online updates, but there may be instances where your PC doesn't recognize a software and won't know whether to use the IGP or the Nvidia GPU. You can still switch manually in those cases, but it's an unfortunate extra step.

Performance hasn't been confirmed yet by a third party, but Nvidia posits that Optimus is up to an 8x improvement across apps and games than Intel's integrated solution. And because it only kicks in when needed, there's purportedly not much of a battery drain, depending on how often you use intensive graphics.

Nvidia says there will be more than 50 notebooks packing Optimus by this summer, although today it's launching only on a few Asus notebooks like the UL50Vf, an ultraportable which houses both a Core2Duo SU7300 and a GeForce G210M.

What we don't know—but what we strongly suspect—is if Optimus is the solution Apple has been waiting for before refreshing its Macbook line. The hiccup: currently, Nvidia says that Optimus is only compatible with Windows 7. But with the Macworld just around the corner, we may be seeing Optimus Macbook Pros sooner than later. [Nvidia]

AMD running with the "Application Processor Unit" name for the guts of their Fusion processors isn't as gimmicky as it sounds, because the Llano is genuinely unique: It's four processing cores and a DX11-capable GPU on a single processor die. In simpler terms, this means that AMD has created a tidy little system on a chip, aimed at a few portable markets. In the simplest terms, they've shrunk laptop graphics and processing into a single chip, which saves power and space.

So! Not much is known about the Llano right now, but we can pick out some broad themes. The chip's power regulation is novel, monitoring specific chip functions to gauge power draw rather than sensors. The graphics capabilities, though still generally a mystery, wouldn't have to be very good at all to trump Intel's lame integrated graphics. In other words, as Ars notes, this could be the first real baby of the still torrid AMD/ATI marriage, and the start of an ATI comeback, at least in laptops.

Or, given that we're not expected to see these processors in products until 2011, when everything could be completely different, it could be none of these things. [Ars Technica]

Tom's Hardware, Anandtech, HotHardware and others go pretty deep on the new architecture, which is now eminently scalable. Here's the overall structure of the GF100, which should give you an idea of the scalability—the GF100 is made up of four graphics processing clusters (GPC), themselves composed of four streaming multiprocessors (which are made up of 32 CUDA cores and texture units) and a raster engine:

To go deeper on architecture, you're better off reading the 10-page reports from any of the sites linked above, but bottom line, Tom's Hardware is predicting something like double the performance of Nvidia's current GTX 285. Anandtech also points out that Nvidia's geometry performance only went 3x between the NV30 engine in the ancient GeForce FX 5800 and current GT200 in the GTX 280, but the Fermi-based GF100 has 8x the geometry performance of the GT200. The endgame being that " it allows them to take the same assets from the same games as AMD and generate something that will look better. With more geometry power, NVIDIA can use tessellation and displacement mapping to generate more complex characters, objects, and scenery than AMD can at the same level of performance."

There is a cost. Even though it's at the 40nm process, those 3 billion transistors are going to run hot, and the GF100 maybe the hottest single-card GPU ever. It's also not going to be cheap. At all. [Tom's Hardware, Anandtech]

Says Sascha Segan:

According to a Marvell press release, the ARMADA 610 can render 45 million 3D triangles per second and control four displays at 2k x 2k pixel resolution each. That makes it more powerful than the PowerVR SGX 530 core used in the Motorola Droid and the Qualcomm graphics core in the QSD8250 Snapdragon, which handle 14m and 22m triangles respectively.

Noted.

[Anand via PCMag]

Or, even worse than a delay: New MacBooks could have worse graphic than older MacBooks.

Here's the story: Starting with the unibody MacBook Pros in Oct. 2008, Apple dumped Intel's own chipset and integrated graphics for Nvidia's GeForce 9400M as a combined GPU/chipset, since it wildly outperforms the Intel's integrated garbage, which had hampered previous MacBooks. Since the 9400M is in basically every Mac now, there's a baseline of graphics performance across every Mac—nothing has crappier graphics than the 9400M. Important, because the OpenCL tech in Snow Leopard leverages your graphics card for extra processing power.

Since Oct. 2008, Intel's introduced its blazing fast Core i7 and i5 processors, which use the Nehalem microarchitecture. The problem is that Nvidia can't make compatible chipsets for it. Intel claims that Nvidia's license to make chipsets for its processors doesn't apply to any current or future processor with an integrated memory controller, which all Nehalem and Westmere—the 32nm die shrink of Nehalem—processors do. Nvidia sued and is pulling out of chipsets entirely, at least at the desktop level. (Intel's also cut them off at the Atom level, making what the Ion 2 will look like something of a mystery as well.)

Which produces a question: What are the next set of MacBook guts going to be? The Arrandale Core i5 mobile processors Intel is expected to announce at CES don't just have integrated memory controllers, they have integrated graphics, built right onto the die, too. If the MacBooks were upgraded to off-the-shelf Arrandale processors, it can't, on the face of it, use an Nvidia chipset or more to the point, Nvidia's superior integrated graphics. Intel's integrated graphics still suck. So there are a couple of possibilities from here, it looks like.

• Possibility 1: Some kind of discrete or separate graphics cards for all MacBook Pros. Pre-unibody MacBook Pros, and even the 12-inch PowerBook G4, had discrete graphics cards only. The problem is that it's more expensive, and that now-famed 6-8 hour battery life would take a hit. It's how the latest iMac got away with using a Core i7 on with an Intel chipset, though.

Or maybe Apple will put discrete graphics cards in every MacBook Pro, but use Intel integrated graphics as a battery-saving fallback. Which is sort of the way all but the low-end MacBook Pros work now, with both integrated and discrete graphics. (Though the Nvidia integrated graphics are good enough to be the default option on current MacBooks.) It would rock the boat the least.

• Possibility 2: Suffer the crappier graphics on lower end models. A problem, given that any machine using Intel integrated graphics would result in worse graphics performance than the current MacBook or MacBook Pros. Which sounds counterproductive, given Apple's obvious bet on graphics cards for processing juice with OpenCL.

• Possibility 3: A customized set of hardware of some kind from Intel, either on the processor or chipset level that would let the next MacBooks match the power consumption and graphics capabilities of current models. It wouldn't be unprecedented: Apple asked for and received essentially custom chips from Intel before, for the MacBook Air. (Though Intel later let everybody else play ball with other chips meant for really skinny laptops.)

• Possibility 4: Apple's gonna wait on something else before upgrading from Core 2 Duos. Will people have to wait longer for blazing new silicon in MacBooks than in PC notebooks? Sometimes they do, yes, but sometimes Apple gets Intel's latest first—Nehalem Xeons in Mac Pros, and the ultramobile chip in the MacBook Air.

Something else to consider is that for the first time in a long time, if Apple wants to push new guts soon, it could switch to ATI graphics (which it's using in the iMac) for notebooks because of delays in Nvidia's Fermi architecture that push their truly new graphics cards out until Spring 2010. ATI's got a solid 4 months where it's got the newest graphics silicon around.

Whatever happens, it's a mystery for now. Which is kind of a fascinating point, actually, given that Macs run on PC guts now, yet it's still trying to do something different on the hardware level.

And you have to figure that's pretty far behind, since the Larrabee launch timeframe was 2009/2010. The only way you'll be able to touch Larrabee now is as a development platform for graphics engines or high-performance computing, in order to develop for future Intel products.

Intel says they're going to announce new plans for discrete cards some time in 2010—mayyybe CES, where we talked to former Intel Chairman Craig Barrett about Larrabee last year? But, more likely at the Intel Developer Forum later in the year. [Intel]

MaximumPC put the 5970 to the test (check out their review here), and found that it lives up to its promise. The 2GB dual-GPU card is the first to support DirectX 11, and basically doubles its wholly respectable predecessor (the 5870) in specs, capable of delivering nearly 5 teraflops of raw processing power. It's a massive card, about a foot long, designed mostly for heat dissipation, at which aim it apparently succeeds. It's also got easy access to overclocking via AMD's OverDrive, and can drive up to three displays simultaneously with a maximum resolution of 7680x1600. So it's pretty much the greatest thing ever, and it's got a pricetag to match: $600 upon its undisclosed release. Yow. [MaximumPC]

The architecture really is a huge leap forward, according to people who've gone through it in-depth. Interestingly, the huge focus for Fermi is GPU computing. The first actual goods coming out using Fermi should be the GT300 series cards, which, besides the 512 cores sorted into 16 streaming processors with 32 cores each, uses a brand new GDDR5 memory setup.

PC Perspective has an epic write-up breaking down Fermi in detail that's worth a whirl, and of course Nvidia's got lots of fluff themselves all about Fermi. Strangely, they don't explain the name, which sounds like a sad little poodle. [Nvidia, PC Perspective, Anandtech]

Specifics on the technology are being kept close to the vest, but a recent demonstration revealed, amazingly, that it runs on only one GPU. it also features several DisplayPort connectors—In this case, six 30-inch Dell displays were configured to run as a single 7680x4800 monitor.

Eyefinity is enabled through a combination of hardware and software being developed by AMD. On the hardware front, AMD's upcoming Radeons will sport between 3 and 6 display outputs of various types, DisplayPort, DVI, HDMI, etc. And those outputs will be managed by software currently dubbed SLS, or Single Large Surface. Using the SLS tool, users are able to configure a group of monitors to work with Eyefinity and essentially act as a single, large display.

Maximum PC witnessed XPlane 9 and Far Cry 2 running at full resolution on Eyefinity at 12-20 frames per second. HotHardware notes that an upcoming DX11 racing game, Dirt 2, was played at 7680 x 3200 with "perfectly acceptable frame rates" (although 12 fps is not what many would consider "acceptable"). They also claim that there are plans to integrate CrossFire support down the line and that AMD has partnered with manufacturers to create ultra-thin bezel displays specifically designed for use with Eyefinity. How long we will have to wait and how insanely expensive all this will be has yet to be determined. [Hot Hardware and Maximum PC]

They Call It "Open" For a Reason
One of the more excellent aspects of Snow Leopard, actually, is its full-scale deployment of OpenCL 1.0—Open Computing Language—a framework that allows programmers to more easily utilize the full power of mixes of different kinds of processors like GPUs and multi-core CPUs. (Much of the excitement for that is in leveraging the GPU for non-graphical applications.)

OpenCL lives up to its name: It is a royalty-free open standard managed by the Khronos Group, and supported by AMD/ATI, Apple, ARM, IBM, Intel, Nvidia, among others. Interesting thing about this open industry standard is that it was developed and proposed by... Apple.

What Is a Standard?
By "standard," we're talking about a format, interface or programming framework that a bunch of companies or people or organizations agree is the way something's going to get done, whether it's how a movie is encoded or the way websites are programmed. Otherwise, nothing works. A video that plays on one computer won't play on another, web sites that work in one browser don't work in another, etc. With increased connectedness between different machines and different platforms, standards are increasingly vital to progress.

Standards can range from open (anybody can use them, for free) to open with conditions (anybody can use them as long they follow conditions X, Y and Z) to closed (you gotta have permission, and most likely, pay for it). Some companies view standards strictly as royalty machines; others don't make much money on them, instead using them to make sure developers do things the way they want them to. Apple falls into this latter category, by choice or possibly just by fate.

Kicking the Big Guy in the Shins
Of course, OpenCL isn't the only open standard that Apple's had a hand in creating or supporting that actually went industry-wide. When you're the little guy—as Apple was, and still is in computer OS marketshare, with under 10 percent—having a hand in larger industry standards is important. It keeps your platform and programming goals from getting steamrolled by, say, the de facto "standards" enforced by the bigger guy who grips 90 percent of the market.

If you succeed in creating a standard, you're making everybody else do things the way you want them done. If you're doubting how important standards are, look no further than the old Sony throwing a new one at the wall every week hoping it'll stick. Or Microsoft getting basically everybody but iTunes to use its PlaysForSure DRM a couple years ago. Or its alternative codecs and formats for basically every genuine industry standard out there. To be sure, there is money to be made in standards, but only if the standard is adopted—and royalties can be collected.

Web Standards: The Big Headache
The web has always been a sore spot in the standards debate. The web is a "universal OS," or whatever the cloud-crazy pundits call it, but what shapes your experience is your browser and in part, how compliant it is with the tools web developers use to build their products. Internet Exploder shit all over standards for years, and web programmers still want IE6 to die in a fiery eternal abyss.

Enter WebKit, an open source browser engine developed by Apple based off of the KHTML engine. It's so standards-compliant it tied with Opera's Presto engine to be the first to pass the Acid3 test. What's most striking about WebKit isn't the fact it powers Safari and Google Chrome on the desktop, but basically every full-fledged smartphone browser: iPhone, Android, Palm Pre, Symbian and (probably) BlackBerry. So WebKit hasn't just driven web standards through its strict adherence to them, but it has essentially defined, for now, the way the "real internet" is viewed on mobile devices. All of the crazy cool web programming you see now made is made possible by standards-compliant browsers.

True, OpenCL and WebKit are open source—Apple's been clever about the way it uses open source, look no further than the guts of OS X—but Apple is hardly devoted to the whole "free and open" thing, even when it comes to web standards.

All the AV Codecs You Can Eat
The recent debate over video in the next web standards, known collectively as HTML5, shows that: Mozilla supports the open-source Ogg Theora video codec, but Apple says it's too crappy to become the web's default video standard—freeing everyone from the tyranny of Adobe's Flash. Apple says Ogg's quality and hardware acceleration support don't match up to the Apple-supported MPEG-4 standardized H.264 codec, which is tied up by license issues that keep it from being freely distributed and open. (Google is playing it up the middle for the moment: While it has doubts about the performance of Ogg Theora, Chrome has built-in support for it and H.264.)

Apple has actually always been a booster of MPEG's H.264 codec, which is the default video format supported by the iPhone—part of the reason YouTube re-encoded all of its videos, actually—and gets hardware acceleration in QuickTime X with Snow Leopard. H.264 is basically becoming the video codec (it's in Blu-ray, people use it for streaming, etc.).

Why would Apple care? It means Microsoft's WMV didn't become the leading standard.

A sorta similar story with AAC, another MPEG standard. It's actually the successor to MP3, with better compression quality—and no royalties—but Apple had the largest role in making it mainstream by making it their preferred audio format for the iPod and iTunes Store. (It saw some limited use in portables a little earlier, but it didn't become basically mandatory for audio players to support it until after the iPod.) Another bonus, besides AAC's superiority to MP3: Microsoft's WMA, though popular for a while, never took over.

FireWire I Mean iLINK I Mean IEEE 1394
Speaking of the early days of the iPod, we can't leave out FireWire, aka IEEE 1394. Like OpenCL, Apple did a lot of the initial development work (Sony, IBM and others did a lot of work on it as well), presented it to a larger standards body—the Institute of Electrical and Electronics Engineers—and it became the basis for a standard. They tried to charge a royalty for it at first, but that didn't work out. It's a successful standard in a lot of ways—I mean, it is still on a lot of stuff like hard drives and camcorders still—but USB has turned out to be more universal, despite being technically inferior. (At least until USB 3.0 comes out, hooray!)

Update: Oops, forgot Mini DisplayPort, Apple's shrunken take on DisplayPort—a royalty-free video interface standard from VESA that's also notably supported by Dell—which'll be part of the official DisplayPort 1.2 spec. Apple licenses it for no fee, unless you sue Apple for patent infringement, which is a liiiiittle dicey. (On the other hand, we don't see it going too far as industry standard, which is why we forgot about it.)

That's just a relatively quick overview of some of the standards Apple's had a hand in one way or another, but it should give you an idea about how important standards are, and how a company with a relatively small marketshare (at least, in certain markets) can use them wield a lot of influence over a much broader domain.

Shaping standards isn't always for royalty checks or dominance—Apple's position doesn't allow them to be particularly greedy when it comes to determining how you watch stuff or browse the internet broadly. They've actually made things better, at least so far. But, one glance at the iPhone app approval process should give anybody who thinks they're the most gracious tech company second thoughts about that.

Still something you wanna know? Send questions about standards, things that are open other than your mom's legs or Sony Ultra Memory Stick XC Duo Quadro Micro Pro II to tips@gizmodo.com, with "Giz Explains" in the subject line.