Fanning the flames that fired me
|
|
|
Nvidia shows off its quad-core Kal-el Tegra chip
CHIP DESIGNER Nvidia has shown off the performance of its quad-core Kal-el Tegra chip by releasing a demo video.
Nvidia's next generation Tegra Kal-el chip was shown off back in February at Mobile World Congress, where the question was not how fast the chip was but how it would affect battery life. Now Nvidia has released a demo video called Glowball that shows off the dynamic lighting capabilities of Kal-el.
There's little doubt that Nvidia's quad-core Kal-el Tegra chip impresses with its 12-core GPU, but potential smartphone and tablet designers will want to know what the battery life will be like. Nvidia's senior product manager for notebook products, Matt Wuebbling didn't give an answer on his blog, but he did say the chip is "battery-friendly".
When Nvidia talked to The INQUIRER back in February about Kal-el, it said the job of the extra cores is to spread the load, most of the time. So instead of having a single core clocked at, say, 1GHz, a device could run four cores at 600MHz but at a significantly lower voltage. However this notion of underclocking and undervolting chips to increase battery life has not been seen in the latest crop of dual-core smartphones such as the Samsung Galaxy S II or the LG Optimus 2X.
The problem faced by chip designers is that processor technology is significantly outpacing battery development, which means that it's up to firms such as Intel, Nvidia and Qualcomm to come up with ways of increasing performance without increasing power draw.
If they could only get some help from the battery manufacturers, we could start to see smartphones that don't need to be charged every day after moderate use. µ
Tags: Nvidia
Share this:
Share
While we'd all love to see better batteries, the current crop of the best types already has energy similar to high explosives -- there's just no other place on the periodic table to go for more -- chemistry is like that.
Same reason battery autos aren't going to match IC performance -- the old oil burners don't have to carry around 15/16 of the total fuel-oxidser-- air is free.
But mostly it's bad software. How is it that a 4 mhz Z-80 in a Kaypro can keep up with how fast I type in a text editor, but a 1 ghz PC has issues?
Yes, it does take more to draw little pictures of characters (fonts) vs just feed ascii to a character generator, and so on and so forth, but as a decades long software engineer, that's not the whole story. As PC's got faster, devs with the current state of the art gear didn't optimize if it ran OK on the dev machine...and here we are.
Not that the CPU guys couldn't make use of a few more tricks to turn off things that aren't needed, get rid of some stages of prediction that aren't power-effective, and so on. But the IP/patent system has it that this guy owns one trick, the other guy owns another, and therefore is stifling innovation there like everywhere else -- you can't make a CPU without violating tons of obvious patents, and the big boys lock out all upstarts with their patent portfolios.