Yup, you can easily build a web server on an RPi. Why would you want to? As a first project, it's good practice, and novices will learn a heap of stuff: downloading and compiling source, Linux command line tools, configuration etc. For the rest, well, for a headless web-app server, an RPi may be all you need. Or use it as a test server, or for hosting a blog. It's also forms the basis for some other types of projects (see below), where having a web page that can display data from the RPi's GPIO inputs is half the battle.Exclusive Taiwanese server maker Quanta is sick of people misrepresenting or guessing about the size and might of its server business, and so it is setting the record straight. And as it turns out, Quanta has an absolutely huge and absurdly fast-growing server business that should make all of the server incumbents quake with trepidation and sharpen their envy knives.The server arm of the $37bn Quanta Computer giant has been relatively quiet in recent years as it made its way into the original design and manufacturing, or ODM, server business and now has greatly expanded its direct sales.
The company opened up its US subsidiary last year, called Quanta QCT, which is located in Fremont, California, and is opening up an office in Seattle, Washington next month to do technical support, sales, and marketing to hyperscale cloud operators.As it becomes more known in the States, Quanta is starting to talk about its business a little more. Mike Yang, vice president and general manager of the cloud business unit at Quanta Computer, reached out to El Reg to set the record straight about its server biz."It is important to acknowledge that the market was changing, and is still changing, and that this business came to Quanta," explained Yang. "Quanta did not go looking for this."Rather, Yang said, companies running hyperscale data centers were not happy with general-purpose machines and wanted something that specifically fit their workloads and their data centers. That's why Facebook and Rackspace Hosting – two companies that Quanta can reference – and others, who are very secretive about their servers and their suppliers because the machines are themselves a competitive advantage, came to Quanta and got this custom server business rolling.And as it turns out, Quanta is relying less and less on its ODM business, where it makes machines for other IT suppliers, and more and more on its direct sales to the hyperscale data center customers, cutting out the middlemen.
"They know very well what they need, and we know how to design – that's what we do," affirmed Yang. "Most of the time, we exceed their design expectations. This is our value."That value has translated into a very large number of server shipments that are helping to prop up the overall market, and revenues that are growing five times as fast as shipments. Yup, you heard that right. In a server racket where revenues are flat to down and shipments are piddling along, Quanta is exploding.In 2008, using IDC figures, there were 8.1 million servers sold, with 7.7 million of them being based on x86 processors. In 2012, there were still 8.1 million servers sold worldwide, but 8 million of them were x86 boxes. So the x86 platform is crushing the market share like crazy.Yang didn't have figures at hand going all the way back to 2008 for Quanta server business, but said that the company shipped 1.2 million server nodes in 2012 to companies all over the world, and that this represented 19 per cent growth over 2011's figures.
If you do the math, that means Quanta shipped slightly more than 1 million machines in 2011. If you use IDC's shipment numbers, then about one of every seven machines that shipped last year on the entire planet was forged by Quanta.Granted, Quanta does not always get credit for that manufacturing, since the Googles, Facebooks, Amazons, Rackspaces, Yahoos, Baidus, and other hyperscale data center operators of the world do not talk about who actually designs and makes their machines.And Yang talking about shipments today with El Reg is meant to make it clear that the box counters of the world are not giving Quanta its proper share. Which was around 14.8 per cent, if you do the math and assume the IDC shipment numbers are correct across all vendors.AMD has released its latest A-Series laptop chips, codenamed "Richland" and replacing the "Trinity" chips released last spring.Is it a giant leap from Trinity to Richland? Well, no, but the new chips are clearly a step forward – if, in some cases, only incrementally.Before we dig into Richland's upgrades, a refresher course to those Reg readers who don't regularly follow AMD's consumer efforts. Rather than merely calling its processors, well, processors, a few years back it decided to call their CPU-GPU mashups "APUs", which stands for accelerated processing units. They've stuck with that terminology, so it appears that we're stuck with it, as well.
The new Richland A-Series APUs are best described as improved Trinity APUs. They're based on the same second-generation "Piledriver" compute cores that had their APU debut in Trinity, and their GPU remains based on the AMD "Cayman" VLIW4 (very long instruction word) architecture.AMD's Graphics Core Next, which we heard so much about at the company's 2011 Fusion Summit, doesn't make its appearance in Richland; it seems that we'll have to wait until AMD's next round – "Kaveri", "Kabini", and "Temash" – before GCN makes it into APUs.But despite Richland's close resemblance to Trinity, there are improvements to celebrate – improved power management, for one. According to AMD, Richland's power-management system is much more granular than that of Trinity, allowing for more-precise control over which parts of the chip are lit up – and, more importantly, which parts can be shut down, thus saving power.In most cases, the power savings over Trinity are nothing to get frightfully excited about, but AMD does claim that one workload – 720p video playback – can see power savings of as much as 25 per cent. Not too shabby, if true.
Any amount of power savings is to be welcomed, of course, and seeing as how both Richland's CPU and GPU cores can be cranked up and down in a base/turbo scheme, any power saved in other ways can keep them turboing that much longer without stepping over the total-power line, thus improving performance.And speaking of clock rates, Richland's are faster than those of Trinity, as well – but keeping with the rest of Richland's improvements, the speed-ups are modest but welcome. For example, the top-end Trinity A-10 had a base clock of 2.3GHz, and turbos up to 3.2GHz, while the new Richland A-10 has a base of 2.5GH and a turbo of 3.5GHz.The clocks of that Richland A-10's 384 Radeon GPU cores are a bit zippier than those same number of cores in the Trinity A-10: a base of 533MHz and a turbo of 720MHz in the new chip versus 497MHz and 686MHz in the older chip.In its announcement of the Richland line – excuse me, the "AMD Elite A-Series Accelerated Processing Units" – AMD touted a number of user-interface niceties for which the new APUs will provide support, including facial-recognition login, gestural control, video image-stabilization and color-enhancement, and more.