Nvidia has released a few technical details of its upcoming ‘Kal-El’ Tegra processor, including a secret it’s done well to keep under its hat thus far: it’s a five-core, not four-core, chip.
Tegra, Nvidia’s ARM-based mobile processor design, is a real winner for the company. Nvidia boss Jen-Hsun Huang has stated that he aims to make Nvidia’s Tegra the company’s main focus, moving away from the discrete graphics that has been the company’s bread and butter in the past.
A combination of novel technologies and a 12-month development cycle – extremely rapid in a market where the key players work on an 18-to-24-month cycle – helped Nvidia score some high profile milestones. Nvidia’s Tegra 2 was the first dual-core chip to hit the market, and can be found in a wealth of tablets and a smattering of smartphones.
All eyes are now on the company’s next release, codenamed Kal-El – following a trend for superhero alter-egos – which the company announced back in February as a quad-core successor to the current dual-core Tegra 2 design.
According to a white paper released by the company today, it wasn’t telling the whole truth when it outed the Kal-El project, however: rather than four processing cores, the new SoC design features five.
In a paper entitled “Variable SMP – A Multi-Core CPU Architecture for Low Power and High Performance,” Nvidia engineers explain that Kal-El will be the first chip to hit the market with vSMP technology. “Not previously disclosed publicly, vSMP includes a fifth CPU core – the ‘Companion’ core – built using a special low power silicon process that executes tasks at low frequency for active standby mode, music playback, and even video playback,” the paper explains.
If that sounds familiar, then you’ve been paying attention: back in September of last year ARM licensee Marvell announced the Armada 628, an ARM v7 MP based system-on-chip design that paired two 1.5GHz high-performance cores with a slower 624MHz processing core for low-power tasks and standby use.
“Marvell’s groundbreaking tri-core architecture is a unique solution to a long-time problem – how to achieve enterprise performance without breaking the limited power budget of smartphones, tablets and other mobile consumer devices,” company co-founder Weili Dai explained at the time – a sentiment which carries forward in Nvidia’s whitepaper, almost exactly twelve months later.
“In addition to four main Cortex A9 high-performance CPU cores,” Nvidia’s paper explains, “Kal-El has a fifth low power, low leakage CortexA9 CPU core called the ‘Companion’ CPU core that is optimized to minimise active standby state power consumption, and handle less demanding processing tasks.”
Where Nvidia’s design differs from traditional antisymmetric multiprocessing implementations like Marvell’s Armada 628 is in the vSMP technology itself. “Project Kal-El is the world’s first mobile SoC device to implement a patented Variable Symmetric Multiprocessing (vSMP) technology that not only minimises active standby state power consumption, but also delivers on-demand maximum quad core performance,” the company claims in its paper.
All five processing cores, Nvidia claims, are identical in their implementation of ARM’s Cortex-A9 design. Where they differ is in their production process: the four main cores are built on a fast process technology, which allows them to run at very fast switching speeds using a normal voltage draw; the companion core, in contrast, is built on a low-power process which minimises leakage while idle and at lower clock speeds, but requires vastly more power to ramp up to the same performance as the fast process chips.
The result: four processing cores which offer impressive performance, and which can be switched off while the device is idling in order to conserve power. While the main cores will scale up to at least 1.5GHz – and likely beyond in future revisions – the companion core hits its peak at just 500MHz, waking up a main core and disabling itself should the required performance exceed that level.
Nvidia claims that the companion core isn’t just for standby mode in smartphones, however: according to the whitepaper, it will be up to the task of running simple apps like Twitter, music playback, and even video playback when used in conjunction with hardware graphics acceleration. As a result, a Kal-El-equipped device should hit battery levels far beyond those of devices based on today’s dual-core chips, despite the extra hardware.
Because the vSMP technology is OS-transparent, Nvidia claims that Android ‘Honeycomb’ 3.x can already take advantage of the extra power a Kal-El chip would bring. “The operating system
assumes that all available CPU cores are of equal performance capability and schedules tasks to available cores based on this assumption,” the whitepaper explains. “Therefore, in order to make the management of the Companion core and main cores totally transparent to the operating system, Kal-El implements both hardware-based and low level software-based management of the Companion core and the main quad CPU cores.”
Nvidia’s whitepaper is filled with references to patented technologies, including hardware and software CPU management logic and the vSMP concept itself. While that doesn’t mean that rival SoC designers – like Qualcomm, Texas Instruments, and a Marvell which is no doubt feeling a sense of deja vu at Nvidia’s anouncement – can’t come up with a similar yet distinct design of their own, it does suggest that Nvidia’s Tegra platform is going to be getting some serious interest over the next twelve months.
The white paper also includes some other facts surrounding the upcoming chip: the five processing cores will be joined by a GPU, an HD video decoder, an HD video encoder, a display driver, a security engine, HDMI output, memory IO, ISP, and audio – but not, in initial versions at least, by an integrated modem for smartphone use.
According to Nvidia’s internal testing, a production-ready 40Nm Kal-El chip – manufactured, in case you were wondering, by TSMC – hits a score of 5589 in the Coremark benchmark when running at 480MHz, drawing just 579mW. That contrasts nicely with the likes of Texas Instruments’ OMAP4 or Qualcomm’s QC8660, which require almost three times the power – and a clock speed of around 1GHz – to hit the same performance levels.
For the speed freaks, Nvidia claims that a Kal-El chip running at 1GHz – likely to be the top speed of the initial production run, with Nvidia producing higher-performance parts throughout 2012 – will hit a Coremark score of 11,667 while drawing just 1,261mW, offering more than twice the performance of its competitors for less power.
That could change rapidly: Nvidia is using last-generation chips to form the comparison, and with Qualcomm planning a quad-core design of its own a true comparison can’t yet be made.
Sadly, Nvidia’s whitepaper doesn’t include a release date – but with sample chips already available for testing, our money’s on seeing the first Kal-El-equipped tablets hitting shop shelves before the year is out.
Nvidia’s full whitepaper, plus a second publication entitled “The Benefits of Quad Core CPUs in Mobile Devices” – which reads remarkably like last year’s “The Benefits of Dual Core CPUs in Mobile Devices,” if you ask us – can be found on the company’s website.Leave a comment on this article