The ARM vs x86 Wars Have Begun: In-Depth Power Analysis of Atom, Krait & Cortex A15
by Anand Lal Shimpi on January 4, 2013 7:32 AM EST- Posted in
- Tablets
- Intel
- Samsung
- Arm
- Cortex A15
- Smartphones
- Mobile
- SoCs
Modifying a Krait Platform: More Complicated
Modifying the Dell XPS 10 is a little more difficult than Acer's W510 and Surface RT. In both of those products there was only a single inductor in the path from the battery to the CPU block of the SoC. The XPS 10 uses a dual-core Qualcomm solution however. Ever since Qualcomm started doing multi-core designs it has opted to use independent frequency and voltage planes for each core. While all of the A9s in Tegra 3 and both of the Atom cores used in the Z2760 run at the same frequency/voltage, each Krait core in the APQ8060A can run at its own voltage and frequency. As a result, there are two power delivery circuits that are needed to feed the CPU cores. I've highlighted the two inductors Intel lifted in orange:
Each inductor was lifted and wired with a 20 mΩ resistor in series. The voltage drop across the 20 mΩ resistor was measured and used to calculate CPU core power consumption in real time. Unless otherwise stated, the graphs here represent the total power drawn by both CPU cores.
Unfortunately, that's not all that's necessary to accurately measure Qualcomm CPU power. If you remember back to our original Krait architecture article you'll know that Qualcomm puts its L2 cache on a separate voltage and frequency plane. While the CPU cores in this case can run at up to 1.5GHz, the L2 cache tops out at 1.3GHz. I remembered this little fact late in the testing process, and we haven't yet found the power delivery circuit responsible for Krait's L2 cache. As a result, the CPU specific numbers for Qualcomm exclude any power consumed by the L2 cache. The total platform power numbers do include it however as they are measured at the battery.
The larger inductor in yellow feeds the GPU and it's instrumented using another 20 mΩ resistor.
Visualizing Krait's Multiple Power/Frequency Domains
Qualcomm remains adament about its asynchronous clocking with multiple voltage planes. The graph below shows power draw broken down by each core while running SunSpider:
SunSpider is a great benchmark to showcase exactly why Qualcomm has each core running on its own power/frequency plane. For a mixed workload like this, the second core isn't totally idle/power gated but it isn't exactly super active either. If both cores were tied to the same voltage/frequency, the second core would have higher leakage current than in this case. The counter argument would be that if you ran the second core at its max frequency as well it would be able to complete its task quicker and go to sleep, drawing little to no power. The second approach would require a very fast microcontroller to switch between v/f modes and it's unclear which of the two would offer better power savings. It's just nice to be able to visualize exactly why Qualcomm does what it does here.
On the other end of the spectrum however is a benchmark like Kraken, where both cores are fairly active and the workload is balanced across both cores:
Here there's no real benefit to having two independent voltage/frequency planes, both cores would be served fine by running at the same voltage and frequency. Qualcomm would argue that the Kraken case is rare (single threaded performance still dominates most user experience), and the power savings in situations like SunSpider are what make asynchronous clocking worth it. This is a much bigger philosophical debate that would require far more than a couple of graphs to support and it's not one that I want to get into here. I suspect that given its current power management architecture, Qualcomm likely picked the best solution possible for delivering the best possible power consumption. It's more effort to manage multiple power/frequency domains, effort that I doubt Qualcomm would put in without seeing some benefit over the alternative. That being said, what works best for a Qualcomm SoC isn't necessarily what's best for a different architecture.
Facebook
Twitter
RSS
140 Comments
View All Comments
karasaj - Friday, January 4, 2013 - link
Yeah, keep in mind that Haswell will be like Windows 8 Pro, i.e. a more traditional laptop experience anyways, so it won't necessarily be strictly competing with the iPad.mrdude - Friday, January 4, 2013 - link
Win8 and its devices aren't selling, so I'm not sure how Intel plans on suddenly making that OS and the products that run on it any better. Like I said, Win8 tablet sales have been really really poor.And as far as consumers go, it is directly competing with the iPad and Android. People are more inclined to buy a new shiny tablet than a laptop and they could care less for x86 compatibility. For your average user, the iOS app store has thousands of more applications than your x86 desktop if only for the fact that it makes it incredibly easy to search through, install and play with. x86 compatibility means very little to most folks, and you can argue that it's actually a detriment due to the x86 legacy's inherent safety issues (non-Metro) and higher prices.
The "It's also a PC" shtick only works if people want PCs. Judging by the sales figures and the longer upgrade cycles, Win8's sales figures, it's clear they're obviously aren't that interested.
Krysto - Friday, January 4, 2013 - link
"One reason the Pro version of the device will be more expensive is that it uses a PC-style chip from Intel Corp. INTC -0.87% (INTC), part of a family of chips that sells for between $177 and $225. The Nvidia Corp. NVDA +3.48% (NVDA) chip typically used in the Surface RT model costs about $28, according to an estimate by research firm UBM TechInsights."http://blogs.wsj.com/digits/2013/01/04/windows-8-f...
iPhone 5's BOM is $188. iPad 3's BOM was $160. I wish Intel good luck if they think they are going to have competitive devices on the market where the chip alone costs as much as all the components in iPads and iPhones.
Oh, but that's just for IVB chips. Did I mention Haswell will be 40% more expensive - so Haswell chips will be more like $250-$300. Yeah...good luck with that Intel.
jeffkro - Saturday, January 5, 2013 - link
"I'm more than willing to pay for the performance premium."I'm not, I'm perfectly happy with the speed of my galaxy nexus which is completely outclassed by the Krait S4. So why would I want to pay a huge premium for an intel powered phone? Just give me dual core krait performance in a all day battery life phone and I'll be thrilled. For me getting more battery life takes precedence over all out speed.
djgandy - Friday, January 4, 2013 - link
True, pricing is another issue, but Intel has room to cut prices a lot. Intel makes the conscious decision to make upwards $30 on a chip rather than $3. In turn they don't have to ship as many chips.I'm pretty sure the cost of manufacturing an Atom SOC is around $5-6, so Intel has plenty of room to make money in the $20 if they so choose. There is no technical reason a Tegra 3 is cheaper than a Haswell when it comes to manufacturing. It's all about market segment.
Intel idling fabs a bit is probably due to the fact they went massive with 22nm. Intel also makes in profit a quarter 75% of what Nvidias entire yearly revenue is.
If Intel wants to play ball, I am sure they will.
mrdude - Friday, January 4, 2013 - link
But can they afford to?Those billion dollar fabs require billions of dollars to run, and if they want to maintain - or even stretch - their fab lead over their competitors, they need to make MORE money going forward. This means that if they were to compete at ARM level they can't afford to do so over a longer period of time else they'd find themselves short of cash to funnel back into the fabs and losing that distinct advantage. I remember seeing something that stood out in the latest Qualcomm earnings report, it read (paraphrasing): our fabless strategy is actually an advantage.
For Intel's fab advantage to remain an advantage they must make more and more money going forward. As soon as the sales figures look gloomy then it all goes downhill quickly, as the fabs go from being a distinct advantage to a potentially expensive disadvantage.
Intel also has investors to answer to, and this is more complicated than the microarchitecture involved. The investors expect >60% gross margins, and Intel dropped margins a bit below 60% but also let fabs idle so they wouldn't have to drop them even further. The mere fact they're latting fabs are idle means that they're not meeting sales estimates. That's not good. This is rather obvious and seen by the drop in chip sales over the same period last year. In order to keep that dip from looking worse (dropping margins even further), Intel just let the fabs idle. Smart short term strategy as investors don't look at that stuff, but it shows that things aren't 'all gravy.'
If Intel were to drop prices during a resurgent and strong PC market then I'd completely agree with you. They'd even be able to lose money in the SoC mobile arena and still show good numbers at their earnings call. That's just not the case, with PC sales slowing quite a bit and tablet sales picking up substantially, you'd have to question whether they can "weather the storm" by leaning on their dominant x86 PC/server position long enough to make up for the lower-than-usual prices in the mobile SoC space. If Intel went at ARM head on with competitive prices a couple of years ago there would be no question that Intel would remain competitive, but with weak PC sales that are expected to look even weaker compared to mobile this year?
If Intel is to compete with ARM on price, it's, oddly enough, going to be determined by how well their high profit products sell in the near future.
Ananke - Friday, January 4, 2013 - link
They can't compete on pricing with x86 vs ARM. I'm in the business, I know. Intel has absolutely competitive process facilities. If they were strictly making chips, nobody can beat them. If they license the design and make chips - nobody can beat them either. However, coupling the own design R&D expenses with own production, and their cost is higher. Products might be better, but cost is higher.On the marketing side, only price matters today. You may think performance is important, but in reality it defines 1% of the decision of 1% of the market....
Hence, the trend towards ARM designs. That trend was not accidental, it is structural, and I see no reason it will turn around.
It is the reason why AMD is performing so poorly, just Intel is much larger and owns its fabs, it takes longer to become obvious they will have revenue problems.
GillyBillyDilly - Saturday, January 5, 2013 - link
Exactly. Performance is one thing, Price another. ARM is not AMD. There is no way intel can compete with ARM price-wise. And the less competetative they are, the less they will sell and the less they sell, the less R&D and the less R&D, well, this goes on and on. I am glad I don't own any Intel shares.felixyang - Friday, January 4, 2013 - link
There is no doubt A15 is more power consuming. A A7 core can save power sometimes, but when you have a CPU intensive workload like sunspider, the A7 core's effect is limited.GillyBillyDilly - Saturday, January 5, 2013 - link
I hope you do realise that this website is a business and not a charity organisation, and no money = no business.Of course Anand is being paid for their reviews ( not only Intel related ) but as long as they don*t maipulate the data (which I don*t think they do ), their reviews are worth reading. You read it, get a general picture and make YOUR own conclusions.