nVidia has just unveiled its complete Fermi line-up, albeit in a less-expected form: the company is launching three mobile GPUs: GF104M, GF106M and GF108M to replace the only mobile Fermi: GF100M, which as we all know – is not exactly a notebook part.

A very interesting bit about today’s launch was that for the first time since we can recall, nVidia launched mobile GPUs first, to be later followed by desktop parts [GF106 debuts on September 12, GF108 comes on October 13]. According to sources in the know, nVidia heavily optimized the power consumption on the GF106 and GF108 parts and planned to do a notebook launch first. Secondly, OEMs jumped on nVidia and pushed for the notebook parts, as this holiday season is increasingly likely to be plastered with "3D" stickers jumping from everywhere.

Discrete GPU Gaining Ground? Check
Both nVidia and AMD saw radical growth in adoption of discrete GPUs among OEMs – Windows 7 and consumer trends dictate the need for graphics power and there is no stepping back. The role of CPU is decreasing, with consumers requiring more and more visual content. As we all know, some of CPU-only apps did not scale at all, and you would get same experience with casual games such as Farmville regardless of you using a 1.33GHz dual-core notebook or a $1000, 3.33GHz sexa-core desktop processor.

Both AMD and nVidia agree the breakover point is coming for discrete GPUs
Both AMD and nVidia agree the breakover point is coming for discrete GPUs – according to nVidia, between Back-2-School and Holiday Season 2011

nVidia views Holiday Season 2011 as the infliction point, with the number of notebooks with discrete GPUs overtaking the number of ones with integrated graphics. These numbers are followed by nearly doubling the market share between 2007 and 2009, as well as steady growth during 2010. From my personal experience, I remember looking for notebooks with discrete graphics in Fry’s, Best Buy in the States and MediaMarkt and co., in Europe and back in 2008 – there weren’t a lot of notebooks with discrete graphics. However, during 2010, we saw that the tide has shifted and now retailers are pushing discrete GPUs like no tomorrow. In fact, during recent trip to retail store, we numbered 34 out of 44 notebooks running discrete GPU.

With this in mind, it is no surprise that nVidia is firstly launching two new GPUs in mobile, then followed by desktop.

The Lineup
From Left to Right: GF104M, GF106M and GF108M. Die sizes [not disclosed by nVidia]: 331mm2, 238mm2 and 126mm2
From Left to Right: GF104M, GF106M and GF108M. Die sizes [not disclosed by nVidia]: 331mm2, 238mm2 and 126mm2. Note that images are provided by nVidia and were not 1:1 in scale, as balls on GF108 are smaller

Today, nVidia is introducing seven new models based on three GPUs, two of which are seeing the light of day for the first time. We start off with a part that is already available in notebooks. Back in June, nVidia debuted GeForce GTX 480M, a cut-down GF100 chip which featured 352 cores and 256-bit memory interface – long story short, a desktop GTX 465 in a notebook. Today, nVidia is refreshing this part with 2GB of GDDR5 memory. The specs are still the same – 425MHz GPU clock, 850MHz core clock, and 600MHz QDR clock [76.8GB/s]. We wonder if nVidia plans to replace GF100 with a GF104 based part. In that case, the number of cores would have to change, as GF104 cannot feature 352 cores [but higher clocked 336 would be a neat option].

nVidia Mobile GTX Lineup at glance: GF100M, GF104M and GF106M GPUs

nVidia Mobile GTX Lineup at glance: GF100M, GF104M and GF106M GPUs

Do note that we disagree with nVidia’s deciphering of GDDR5 memory. In order to keep up with AMD’s specs [on paper], nVidia does not cite DRAM Cell clock as AMD does. GDDR5 memory operates in Quad Data Rate mode, not Double Data Rate mode. Thus, while nVidia may state GTX480M memory operates at 1.2GHz DDR, reality is that these 2GB of GDDR5 memory operate at 600MHz in QDR mode, resulting in 76.8GB/s. 32-bytes times 2400 is 76800. 32-bytes [256-bit] times 1200 [MHz] won’t give you 76.8GB/s even if you stand on your head.
Moving forward, GF104 already shipped in millions of units as the GeForce GTX 460, with the mobile version [GF104M] being nothing more than a power binned part. GF104M debuts as GeForce GTX 470M. GTX 470M features 288 cores working at 1.1GHz, while 768MB of GDDR5 memory checks at 625MHz QDR, resulting with 60GB/s of video memory bandwidth.

GF106M is a completely new GPU, which can be translated as "one half of GF104" – debuts as GeForce GTX 460M. This chip revolves around a number 192 – both CUDA cores and the width of memory controller. The GPU operates at 675MHz, while cores tick at double the clock [1.35GHz]. Even though the cores are different, this part features identical memory subsystem as the GTX 470M.

You can expect SLI notebooks based on GTX 480M, GTX 470M and GTX 460M, as these three are the only launched parts that support SLI.

nVidia Mainstream GT Mobile GeForce is headlined by GF106M and GF108M GPUs

nVidia Mainstream GT Mobile GeForce is headlined by GF106M and GF108M GPUs

GF106M is also available as GT 445M with either GDDR3 or GDDR5 memory. GT 445M clocks at 590MHz, meaning that its 144 cores clock in at 1.18GHz, with either 128-bit memory interface [800MHz GDDR3 memory, 25.6GB/s] or 192-bit interface. 192-bit interface uses GDDR5 memory clocked at 625MHz in QDR mode, resulting in bandwidth of 60GB/s [identical to GTX 460M and 470M]. On desktop, GF106 will debut on September 12th as GeForce GTS 445 / 450.

The final part makes the bulk of the announcement – GF108M is arriving today as GeForce GT 435M, GT 425M, 420M and 415M. As we already disclosed, GF108M is one half of GF106, i.e. one quarter of GF104. Two 48-core SM units make up for a single GPC cluster, with GT 415M using only one SM unit [48 core]. GT 420M, GT 425M and 435M all carry 96 available cores and pack 128-bit memory controller.

Optimus Technology
Since its introduction, Optimus Technology is a runaway success. Just like ATI jumped on nVidia with Eyefinity on desktop, nVidia jumped on AMD with Optimus Technology, resulting in a large number of OEM deals.

The idea of seamless switch between the integrated and discrete GPU, with controls in software instead of hardware remind us of the discussion between Bill Gates and Steve Jobs, at the time that Microsoft was working with Apple and secretly developing Windows operating system: Microsoft had issues understanding how to control the mice. Microsoft tried to use hardware and failed to do so, while Apple wrote a driver and had no issues. Until Microsoft learned of how Apple was controlling the mice, their UI efforts were stuck. The rest is, how they say – history. This legendary conversation appeared in "iCon: Second Greatest Act in History of Business", an unauthorized biography of Steve Jobs, as well as in movie flick Pirates of The Silicon Valley.

nVidia Optimus Technology at glance
nVidia Optimus Technology goal at glance – we can now confirm that nVidia achieved this goal

Getting back to business, over the past couple of months, the company introduced Verde notebook drivers, offering 1st party drivers for interested notebook owners, rather than relying on 3rd party refresh schedule, which a we all know, screwed both AMD and nVidia in the past, with companies releasing multi-GPU powered, $2000 notebooks with year old drivers.

This article is written on an Optimus-powered notebook, which in its nine months of service life, raised numerous eyebrows in executive meetings showing nine hours of battery life with Wi-Fi enabled, and yet being powerful enough to run World of Warcraft, Need for Speed: Shift and Left 4 Dead 2 in native display resolution [1366×768]. ASUS UL50vf was also the first notebook to raise discussion after Apple’s MacBooks started their entrance into executive meetings.

nVidia's Notebook Attack: GT 415 and GT 420 notebooks starting at only $749, going head to head against integrated
nVidia’s Notebook Attack: GT 415 and GT 420 notebooks starting at only $749, going head to head against integrated.

Best of them all, the notebook retailed for $849. With the new line-up of GeForce GT 400M boards, nVidia is reducing the price bracket to $749, going straight head to head against notebooks with integrated graphics which as we all know, have quite a ragged performance and suffer from the lack of essential DirectX capabilities, resulting in improperly rendered frames. As you can see in an image, Acer Aspire carries a larger screen as well, and a full-size keyboard.

3D Vision on the Go
Regardless of recent movies that brought sub-standard experience, the technology is here to stay. Playing games or reviewing old movies and TV shows in 3D [CyberLink’s PowerDVD software has a neat feature of "3D-izing" 2D movies] has its appeal and simply creates a better experience. Sony is even bringing 3D to five-year old PlayStation 3 console.

As a pioneer of 3D in computing world, nVidia has a neat position of already developed ecosystem and complete product line-up. Their glasses, while looking a bit geeky – offer greater comfort than some of the products we’ve tried. With GeForce 400M series, OEMs are bringing multiple 3D notebooks with integrated IrDA receiver, getting rid of the tiny pyramid that nVidia ships with its 3D Vision kit.

Two 3D Vision notebooks headline this announcement. Plenty more to come...
Two 3D Vision notebooks headline this announcement. Plenty more to come…

Two of the announced notebooks, GTX460M-powered ASUS G53Jw and GT 425M-powered Acer Aspire 5745DG are heading the initial 3D Vision Notebook line-up. We also learned of several new models that will debut for the Holiday season, such as notebooks from Dell, HP, Sony, Toshiba and others. Some of these designs will use 3D Vision, some will use passive technologies such as TriDef by DDD.

Conclusion
Arriving more than six months after AMD introduced their DirectX 11-capable Mobility Radeon parts, nVidia has a tough mountain to climb. We learned of several very interesting OEM deals for these parts. We had no performance data to rely on, and we did not receive any review units from the company or its OEM partners. Thus, at this point in time we cannot know how GeForce 400M Series performs against AMD’s Mobility Radeon HD 5000 Series. Then again, AMD is working in sixth gear to get mobile version of Southern Islands ready, and these parts should feature Optimus-like switching technology. However, Optimus is here and it works flawlessly, enabling OEMs to create thin notebooks with long-lasting battery that simply turn into gaming rigs on the go – something that was unimaginable just 10 months ago.

Thanks to very aggressive pricing against the integrated graphics there is really no reason to go with a notebook that does not carry discrete graphics, regardless of them being from AMD or nVidia.

There is only one thing we can conclude in the end of this preview article: Consumer wins. Big time.