Both Intel (INTC) and Advanced Micro Devices (AMD) have focused on performance per watt with their most recent products. The biggest reason most users have at this point to upgrade from previous generations is not performance increases but extended battery life. And the battery life gains are enormous. I am typing this article on a new Apple (AAPL) MacBook Air. With heavy internet browsing, I may have to charge it once per day. On a side note, if you have the cash to burn and have been on the fence about a new laptop, and battery life is extremely important to you, now is not a bad time to start shopping.
I feel the metrics by which consumers are making their purchases are shifting away from performance and price to the combination of power consumption (here I'll make the argument that consumers don't even know this is important to them, but yet this allows the smaller form factors they desire), display, battery life, price, functionality, and performance; users desire a better overall experience, but the price they are willing to pay is dropping.
Consumers have been shifting away from PCs to more portable devices, and now I feel we're seeing things come full circle, in which PCs are taking the shape of these mobile devices. Even if the devices do not exactly resemble tablets, functionality and products, such as touchscreen laptops with Microsoft (MSFT) Windows 8 being designed from the ground up for touch, are becoming more common.
As the consumer has shifted from the traditional PC to mobile, the price point they're paying is dropping. For comparison, the IDC calculates tablet ASP this year reaching $381, compared to PCs at $635.
As in the rest of my technical articles, I will summarize at the end of each section (except the first section, look for the *POINT*s)
Image taken from isuppli.com
Although the graphic above is a little dated, it is meant to provide a rough estimate of how much it actually costs to build a laptop. The BOM above I believe is for a Samsung Chromebook 5, which at launch in 2011 retailed for around $430. If isuppli's numbers are correct, between Samsung and the retail store that sold it, there was about $90 left for profit, and this is not including costs such as shipping.
As the price we're willing to pay for computing devices drops, consumers are making sacrifices to get the features they want. Take solid state drives for example. A 128 gb solid state drive offers 25% of the storage of a 500 GB HDD at a higher price. The tradeoff we're making is performance for space and price. You could buy a device with 500 GB of solid state storage if you're willing to spend an extra $400 or so, but this is not something the majority of users opt for. If OEMs are smart about the performance and features, they can deliver a good experience at a good price.
Putting Performance and Price into Context
Narrowing a price for Jaguar was fairly difficult. There is a Chinese website that shows the price coming to $16 USD, which seems too cheap. An AMD E350 (previous generation of Jaguar) *with* motherboard is $75, and z2760 Atom is $41. So for a ballpark price, I am using $35 as the price OEMs pay for an A4-5000, and am assuming equal pricing between A4-5000 and Atom.
The A4-5000 has 4 CPU cores, 2 GCN (graphics) cores, a single channel memory controller, and the I/O function integrated on die.
For comparison, Intel's z2760 is $41 per the ARK website, an i3-3217u is $225 ($124), and an i3-4100u is $287 ($158). For practical purposes, assume OEMs pay 55% of pricing for buying in bulk (numbers in parentheses). So $35 should buy you an A4-5000 or Clover Trail Atom, or you can spend what appears to be 3-4 times as much for the experience of the Core i3.
The websites I'm using for performance benchmarks are techreport.com and Anandtech.com. I am very much paraphrasing here, because several articles on Seeking Alpha have focused solely on GPU/CPU performance between AMD and Intel. I am keeping the performance argument in ball park figures, but focusing more on power consumption, since that is a major push from both camps in which I feel there is a skewed perception.
As far as CPU performance is concerned, a ballpark estimate is that the A4-5000 should deliver roughly 80% of the performance of an IVB i3 or 60-70% of the IVB i5. Regarding GPU performance, AMD's Kabini should deliver better real world GPU performance over HD4000 in a single channel memory configuration, but worse performance than HD4000 using dual channel memory. Kabini hands down wins against z2760, but this not a fair comparison to make given that z2760 operates at much lower powers, although costs appear similar. Again, these are rough estimates based on looking through third party independent reviews.
*POINT* Clover Trail and Kabini appear to be similarly priced in the ballpark of $35 for OEMs. OEMs are likely to pay much more for Intel's Core line. The Core CPUs have more CPU computational power than Kabini, while Kabini is more competitive in GPU performance. GPU-wise, in real world applications Kabini performs better than HD4000 in a single channel memory configuration, but worse in a dual channel memory configuration.
And Now Power Consumption
To compare power consumption and battery life, I am using Anandtech.com. The following 4 images are from their review of the A4-5000.
The units of measurement are extremely important - Minutes/Watt*Hr. It accounts for power consumption but normalizes for battery size.
The following numbers are taken from Anand's review of an i7-4500u Acer.
The Acer ultrabook comes equipped with a 46 Whr battery, meaning the Acer scores roughly 10.9 Min/Whr. Taken on the whole, these numbers represent at best a 20% battery life improvement when looking at Haswell vs Kabini while surfing the web. This lead diminishes under load to roughly 10% (again, these are ballpark numbers). Unfortunately it looks like different tests were run between AMD and Intel. There were also differences in testing platforms (hdd, display size, keyboard lighting, etc), so it's hard to make an apples to apples comparison. To test Kabini, Anandtech used 'light', 'medium', and 'heavy' tests, but used PCMark 08 battery tests for Haswell. The 20% advantage to Haswell appears to be the most direct comparison in internet usage, and the numbers makes sense given Haswell's advanved power saving features. Given that power savings is about 20% while only web surfing but diminishes as work loads intensify, call it a *15% advantage for Haswell*.
INTC bulls like to point out is that 22nm finFETs allow for reduced power consumption. The Kabini tested was built on a 28nm process. Interestingly, although Anand wasn't very specific, the Dell XPS 12 compared against Kabini I believe was built on a 22nm process, yet the Kabini bested it in every instance of normalized battery life, despite Kabini using bigger transistors. Note the improvements between Haswell and Ivy Bridge appear to be mostly not related to transistor characteristics. Keep in mind though, Intel is delivering more processing power at lower power consumption.
*POINT* - Despite being built using a 28nm planar process, Kabini has better normalized battery life compared to an Ivy Bridge Core CPU, and delivers 80-90% of a Haswell's i7 normalized battery life. For context, Acer's i7 Haswell Ultrabook lasted 8 hours and 24 minutes in an internet test. If the Kabini reference design had the same 46 Whr battery, it would last approximately 8.8 Min/Whr*46Whr = *6 hrs 45 minutes*. The tradeoff consumers make by buying a Kabini is saving $100 give or take, while getting roughly 70% of the CPU performance in normal workloads and about 85% of the normalized battery life. The one area where Intel's process lead gives them the largest advantage is in smartphones. Right now the general consensus seems to be that AMD is not targeting this sector anytime soon.
Prices and performance figures above are rough estimates. I have shown that AMD offers around 70% of the CPU performance, slightly less GPU performance, and 85% of Haswell's normalized battery life at a steep discount to OEMs over Haswell. And for those that care about die size (important for financials in the semiconductor industry), AMD achieves this performance in about 60% of the die area (107 mm^2 vs 181 mm^2 - this is why Intel receives better margins). I have used both a Kabini and Haswell laptop, and based on my experience, AMD's problem in the consumer market isn't performance but is being relegated to bargain basement design wins. The A6-5200 was pleasant after swapping the HDD for an SSD.
Recall the BOM from above regarding the Samsung Chromebook. The battery the Chromebook used was almost 8300 mah for $50. The battery used in the Acer S7 I used to compare Kabini against is 6280 mah, which is 25% smaller. With the savings by choosing Kabini, OEMs could use a bigger battery, faster storage, and a better display, giving the end user a better overall experience. Given an i3 or i5 ULV with a mechanical HDD vs a Kabini with an SSD, most users would perceive the Kabini laptop as faster.
Later this year, we will see Samsung introduce the Ativ 9 lite, which is the kind of design I'm excited about. It will offer consumers many of the features desirable in an Ultrabook with an AMD CPU purportedly at the helm. Although no official prices have been released yet, it will be interesting to see the price differences between various SKUs. Right now on Amazon (AMZN), you can purchase an Acer Aspire V5-122P with a quad core Temash for $460. Acer could swap the HDD for a 128 GB SSD and throw in a bigger battery (current battery is tiny at 30 Whr), and have a product that could compete with the $1149 Sony Vaio Pro 11, but retail for $525-550. The A6-1450 could be swapped for an A4-5000, offering much better CPU performance and come in around $600. The Sony (SNE) would have more CPU power and a better screen, but it would cost 2x as much.
Brazos was very popular for AMD, shipping 50M units. AMD has been virtually non-existent in upper tier laptops, but if more OEMs follow Samsung's lead and offer AMD processors with upgraded features like IPS panels, larger batteries and solid state storage, I feel AMD stands a chance to steal back some market share. AMD has a low power processor that started off as something more designed for netbooks that can compete with Intel's Core line of CPUs. And the retail price of the hypothetical Acer I described above at $525 comes in between the ASPs of tablets and PCs (although closer to PC ASP), while giving users tablet-like features (good battery life, small form factor, IPS panel, solid state storage), but access to legacy x86 software and a keyboard.