A hygiene factor is an attribute that can cause dissatisfaction if missing but does not necessarily drive decisions if increased beyond what is required.
I don’t honestly think the race for more absolute performance will end any time soon…there is a long way left to go. We are taught that more (and more and more and more) processor performance is the holy grail, right? That argument can certainly be well made in a historical context. However, I do believe the way we think about performance will dramatically change in the next few years given more significant and more difficult challenges involving power. For example:
- Thermal constraints drive server design, deployment, and total costs
- Consumers dislike fans in their living room (or any living space)
- TVs are getting thinner and thinner (thermals again)
- When I travel, I have my “airport power plug” radar switched on
- The electronics in cars have a noticeable effect on fuel efficiency
- Higher temperatures hit package costs, reliability, end-user satisfaction, attractiveness and profitability
- Bigger batteries and thermals constrain industrial design (bigger, heavier, costlier, uglier)
- Geometry scaling appears to be providing less and less benefit in the power efficiency domain
- The uniprocessor has hit the power wall; hello multiprocessing
- We have a serious problem with climate change
Many folks I run across want the maximum performance possible but inside a particular power profile, and not just for battery powered devices. The fixation on power profiles is a new ingredient in the mix and is becoming more and more important. So I see power becoming a front-and-center game changer for 2010.
Happy holidays everyone. BTW, you can find me on LinkedIn here.
Hi – I’ve invested a little money in your very fine company, so I read your blog with an orientation perhaps a little bit different from the average of the 20,000 (did I read right?) who’ve come before me…. I recognize that you were in this post raising some VERY high-level (and intriguing) questions, but I wonder if you could connect the dots, as they say, between your excellent points about “power” and your employer and its products…. That is, does ARM have – pretty much literally – ANYthing to do with cars? … And, more generally, since I share your “looking for places to plug in” and would add – Maybe, a rear-end warmer is a nice thing to have in the winter, but one’s laptop shouldn’t necessarily warm one’s thighs quite as much as it does – what’s the “nexus” between power and ARM?
By: pfusco on December 22, 2009
at 10:14 PM
I think the problem is that most people are not well versed enough to even understand what raw performance means. Clock speed is the most widely used metric, but is absurd since it tells you nothing about efficiency of the processor, the amount and type of caches (L2, L2, L3, etc), the speed ratio between processor and bus, and everything else that impacts real performance. You can have a 1GHz clock and barely run an app any faster than if it had a 400MHz clock (or even slower).
Multi-core seems to confuse people even more, since it may well speed up the overall system at a much slower clock. On the other hand, a single threaded hard-core app will not run as fast on a multi-core system (assuming slower clock).
It is likely that as process geometry shrinks, we are better off with higher efficiency (slower clock, but as much or more work being done) because routing density improves but thermals and power do not. Put another way, routing density improves better than linearly with smaller geometry, but power needs increase far worse than linearly with clock speed increases, which leads to higher thermal issues (from LDOs and power buses, from leaky transistors needed to run at those high speeds, from too many transistors switching at high speed in close proximity, etc).
By: Paul K on December 23, 2009
at 2:52 AM
Chaps,
Thanks for the comments. Yes, Automotive is a key application for ARM though unfortunately a lot of the specifics remain confidential at the moment. Have a look at http://www.arm.com/markets/embedded_solutions/keyapp/2269.html for starters. Oddly enough, one of the competing architectures is termed “POWER”.
The nexus between “power” and ARM is the latter’s low-power leadership. ARM has been connected to a very small battery for decades so it is in the DNA. I don’t have the space or inclination for a closed-form proof here (and there are always corner cases) but that would certainly make a great follow-up post. Take a look at http://jkontherun.com/2009/12/17/ill-bet-an-arm-and-a-leg-that-x86-wont-power-googles-chrome-os-netbook/ and http://www.energymicro.com/technology/arm-cortex-m3
Paul: thanks for the insight. You are right that MHz is nearly meaningless. I’d like to see a dual Cortex-A9 browsing next to an Atom for example….
Regards, Eric.
By: eschorn on December 23, 2009
at 6:01 AM
Dear Eric,
I like this sentence you wrote: “I’d like to see a dual Cortex-A9 browsing next to an Atom for example…”
I have to admit that I also liked very much the Youtube video titled “Cortex-A9 browsing example” available at http://www.youtube.com/watch?v=W4W6lVQl3QA
It gives tangible details to people unaware of ARM true performance/efficiency… or to downsize the pedestal!
Kindly, Alban
By: Alban on January 5, 2010
at 8:49 PM
I have reminiscence from the history.
From the book “The Pentium Chronicles..”, Robert P. Colwell
“The Pentium Pro, launched by Intel Wednesday, is now the world’s fastest. But apart from sheer speed, it offers nothing that isn’t already available.
- A TV anchor introducing P6
“Apart from sheer speed…?” What did that TV anchor want his microprocessor to do, tap dance while juggling spinning plates and singing “Georgia on My Mind”? A computer’s processing speed is its raison d’etre. With few exceptions, it is the additional computational “horsepower” of each new generation that allows new usage models and system features. The TV anchor in question was betraying his near-total lack of understanding of what computers are and how they do what they do.
But he was also inadvertently revealing how buyers view our products. It’s easy to get so wrapped up in the technology that you lose sight of how buyers will eventually judge your efforts. Few buyers will be motivated by your slick design. Most care only that the product adds value to their lives. It is extremely difficult to make a chip design run faster, so difficult that you become fixated on the task and forget that speed must translate into tangible benefits to the buyers or the chips will not sell.”
By: Anton on December 23, 2009
at 1:00 PM
Thanks, Eric – and the other poster (I’ve worked with PC’s since they were born, and I remember once being floored by someone asking me about L1 & L2 caches – still don’t know. And just was looking for a new digital camera where 10.2 or some crazy number of pixels was all that graced the camera body save the brand name. I guess there are places (like light bulbs) where 60 or 100 sufficed for a number of years. Obviously – well, Eric, you’ve helped make it obvious for me – chips are less amenable to that kind of “pegging.”
Also, thanks for those links. One of them features someone who thinks and writes at approximately your level. His comments on power are both straightforward and, well, ARM-centric:
Until recent years, computing was all about processor power and clock cycle speeds. We wanted faster processing, richer third-party client apps and the ability to multitask til the cows came home. x86 computing is great for that. But it comes at a cost in terms of battery life. If you can only multitask for two hours before your battery burns out, what good is it in this mobile day and age? And so a shift towards efficient computing took place. The devices might be slower in this case, but they can run for six or eight hours. That’s the main reason I embraced netbooks when they arrived in late 2007 — they can handle the tasks I do today for nearly eight hours on a single charge.
…other interesting things in his post, too, but I think it’s safe to say – I’m a late adopter but awfully high in terms of enthusiasm – that what he said about netbooks goes at least double for smartphones.
In the States, a battery company popularized the 2 words “STILL GOING” in connection with THEIR product. The alternative reminds me of electric cars when they had a range of 20 miles – DEFINITELY NOT the vehicle you want to drive if you have a 30 mile commute.
By: pfusco on December 29, 2009
at 1:30 PM
[...] Eric Shorn (VP ARM Processor Marketing) Blog. [...]
By: ARM Products and Platforms Primer and Resource List for Mobile Internet Devices in 2010. | UMPCPortal - Ultra Mobile Personal Computing on January 3, 2010
at 4:57 PM