Over the past few years, major chip manufacturers seemed to focus more and more on providing the best bang-for-the-buck-per-power-consumption ratio. I wouldn’t go as far as to call it a defining trend, but it is definitely a shift from the strict performance-based orientation chip makers had a decade ago. While yesterday, we’ve talked about the most power efficient CPU architecture out there, the ARM Cortex M0+, today I’m here to introduce you to the most power efficient GPU architecture out there, the newly uncovered Nvidia Kepler. With Kepler, Nvidia hopes to conquer all three major GPU markets: desktop, notebook, and smartphone/tablets.
It’s not a novelty that desktops PC components are the most power-hungry ones out there. After all, desktops don’t run on a battery, and are, thus, freely able to suck as many watts from the power grid as it’s needed to feed the high-performance chips inside.
One step down the mobility lane, you’ll find that, while notebook CPU’s are somewhat comparable to their desktop counterparts, mobile graphic cards are heavily outpaced by top-class desktop video cards. Take another step down that same lane, and you’ll find that smartphone GPUs are no match for PC video cards (desktop or notebook), as are smartphone CPUs when compared to the Intel Core i7 architecture, for instance. This entire hierarchy is defined by two things: chip size and power consumption.
Leaders in the GPU race, Nvidia used to design and manufacture different GPUs for different devices: from the performance-based, power-hungry desktop chips, to the GeForce ULP inside their Tegra 3 SoC. All this multi-orientation hassle is about to end though, as Nvidia has recently announced the new scalable architecture called “Kepler”, one they tout as being the most power-efficient GPU design up to date. According to Nvidia officials, the Kepler architecture is, in fact, so power-efficient that we should expect it to reach smartphones in the not-so-distant future.
The interesting part is that, although the Kepler GPU showcases an impressive performance-per-watt ratio, it does so without any major performance compromise, up to the point where the first desktop-variant of the Kepler chip – the GeForce GTX 680 – is considered the best performing GPU in the world, period. The first notebook Kepler variant is called the GeForce GT 640M (you’ll find it inside the upcoming Acer Aspire Timeline Ultra M3-581TG) and promises to enable notebook gamers to play even the most demanding video game titles, at decent frame rates.
If you’re wondering when the Kepler architecture will reach flagship superphones, however, I’d advise against holding your breath, unless you can live without oxygen for at least a couple of years. If you’re interesting in finding out the exact hardware specs, check out this PCMag article.
On a side note, it is apparent that Nvidia officials are so pleased with the development of the Kepler architecture that their CEO, Jensen H Huang, has issued a thank you email to Nvidia employees, congratulating them for all the hard work they’ve put into the Kepler project.
To conclude, the Kepler architecture is major warning sign for other smartphone GPU manufacturers, especially Qualcomm, the makers of the Adreno-series of GPUs. If Qualcomm doesn’t want to end up struggling as an underdog (much like AMD-ATI are doing now in the desktop GPU market), it’s time for them to step up their game, and do it fast!
Like this post? Share it!
“To conclude, the Kepler architecture is major warning sign for other smartphone GPU manufacturers, especially Qualcomm, the makers of the Adreno-series of GPUs. If Qualcomm doesn’t want to end up struggling as an underdog (much like AMD-ATI are doing now in the desktop GPU market), it’s time for them to step up their game, and do it fast!”
I laughed at that conclusion, no offence, but this statement is wrong in so many ways.
AMD is having quite the lead over Nvidia every generation, the prove is how kepler came 4 months after AMDs new products, where AMD is struggling at the moment is in the high end CPU arena, AMD doesn’t have anything that match Intels top end cpus
As for Qualcomm appart really A5 and A5X which leaves every single SoC on the graphic area in terms of GPU (with SGX543MP2 and MP4) it has pretty much the same performance as it competitors and the fact that the adreno225 out performs Tegra 3 is enough to prove it, (I’m not gonna talk about the Adreno3xx which is available but not been benchmarked yet)