While many top chefs might hold on to their most precious family recipes, Francois Piednoel is touring the Asia giving away what he calls his “recipe” for building an Atom processor-based tablet.
“When companies in Asia heard about the recipe, they wanted it,” said Piednoel, a veteran performance analyst at Intel who has traditionally worked on the extreme edge of high-performance PC’s. “But if I give you a recipe of French cooking and you go into your kitchen alone, you are not going to make what my mom used to cook,” he says with his native French accent, then chuckles.
“A lot of people building tablets were using the wrong recipe,” said Piednoel “They were using things that were designed for today’s high-end PCs and they tried to run them on Atom. It’s not going to work.”
But are tablet makers really interested in hearing from a guy who loves over-clocking and once conceived and developed a screaming dual-socket, dual-CPU desktop machine with multiple graphics options called Skulltrail?
Apparently so. Piednoel steps in as a kind of pied piper, drawing new tablet makers towards his mastery of math and computer code, which he claims can speed up performance of Atom-powered tablets so they respond smoothly at 60-frame-per-second.
Aside from Apple’s line of iPads, innovation in tablets has largely disappointed many industry analysts, including IDC researcher Mario Morales, who believes a shakeout in the tablet space is brewing.
“At Computex last year (2010), most companies made a big splash by announcing so many types of different tablets that never saw the light of day,” said Morales.
Despite similar criticism rising from this year’s Computex, tablet makers continue to bring out new and improved designs to take on Apple’s iPad 2 and leading Android tablets like Motorola’s Xoom and Samsung’s Galaxy Tab.
Making a Recipe with Legacy Ingredients
“I’m going across Japan and Taiwan and sitting with programmers, explaining to them the entire philosophy and they see very quickly that my very simple research is making a big difference,” said Piednoel.
“The goal is to get every one of those companies that have OK software today to have very good software by the end of this year. Then you don’t get one tablet that’s good, but you get many good tablets.”
Looking at Piednoel’s past, it’s like he traded Formula One cars for go-carts. For more than a decade, he has pushed the limits of technology. He and a small team dreamed up then brought to life Skulltrail, a dual CPU socket desktop platform for enthusiasts. It brought together eight processor cores and support for multi-graphics setups, something that continues to inspire innovation among high-end computing aficionados.
While low-power and just enough performance are in vogue in today’s tablet market, some might question whether Intel should transform a high-end desktop guy into a tablet cook. Many believe mobile, thin, light, and long battery life – are antithetical to the desktop computing legacy.
“If we did something in the legacy before, we had a good reason,” said Piednoel. “It always brought us some performance. We have some advantages that we don’t talk about very much like the ability of using SSSE3 instructions to get another level of performance. There are a lot of mechanisms, which we can take advantage of, that are built into legacy.”
“When going across Asia this is what surprised most of the people. That the x86 architecture legacy is an awesome tool. If you use it right you can do things that nobody else can.”
He says it’s all about finding the edge, but emphasizes that tablet makers must first know the scientific limits and then get creative about working within and around unmovable constraints.
According Piednoel, his tablet recipe takes into account known memory bottleneck issues, and includes programming tricks that can increase performance from 30 to 60 frames per second.
“Pay attention to smoothness,” he says, referring to a key ingredient. “This is the only thing you cannot sacrifice.”
Piednoel’s recipe first saw the light of day in early June during Computex. Gizmodo said that what Piednoel is doing “is about putting user experience above all else, but doing it in a way that is based on the fundamental math of what performance measures are actually making a difference.”
“You start measuring bandwidth and what you can do with the GPU and CPU subsystems,” said Piednoel. “Then you figure out the discrepancies between what people experience every day and what’s actually possible. When you get a discrepancy, you can analyze it to a point where you say, if I change how it’s doing this now, I can get it to go that much faster.”
“All of those tricks in my recipe are nothing new,” Piednoel said. “It’s just about putting them together and making sure people understand them and why they are there. It’s about putting the tablet user’s needs first, not the operating system.”
In his lab, Piednoel demonstrate the responsiveness and media sharing ability of an older Atom-processor based system (Pineview) against a recently released Oaktrail-powered tablet. Both have about the same performance but Oaktrail has up to 10 hours of battery life, more than double the older tablet.
He tests these tablets with a program that his performance lab team built to test high-end PCs. Rather than folders, the program displays an expandable calendar filled with dated photos and videos that can play simultaneously by tapping on the thumbnail.
The program was built to show the limits of performance on high-end PC’s; by seeing how many videos can be played at once. Today it helps him find the edge of tablet performance, revealing how much a tablet can handle before video playback stutters or the touch-screen experience lags.
Ironically, it is just this kind of legacy PC test however, that is helping Piednoel refine his recipe for tablets today.
“Atom has the processing power of high-end computers from about 8 to 10 years ago, but it can’t handle what today’s Sandy Bridge Core processors can handle,” he said. “Still, if we treat the Atom like a gaming machine of the past, there are lots of ways we can find the edge and then fix things to bring out a smooth user experience.”
“Basically what I’m doing here is giving out two years of R&D to anyone who wants to do it, which is OK,” he says. “It enables competition by helping other innovate more quickly. I know that my recipe will get into the hands of the other side, but it’s OK.”
This content was originally published on the Intel Free Press website.