The eternal war for the next generation of computing saw two recent announcements by both Advanced Micro Devices
(NYSE:AMD) and Intel
(NASDAQ:INTC), which make is clear that both firms' aggressive road maps for their next big-core processors, Kaveri and Broadwell, will be delayed. What investors need to know, however, is that the delay for Intel's Broadwell is looking to be close to a year, while AMD's Kaveri will ship in mass a quarter delayed past its most recent ship date of Q4.
Not Going to the Well Too Often
Last month a leaked Intel road map had Haswell getting refreshed in Q2 2014 and no mention of Broadwell in 2014 at all. This has since been confirmed. Broadwell is the 14nm die shrink of Haswell which will also bring some improvements to the GPU side of things, according to an older report by Semiaccurate
Frankly, this is disastrous for Intel, which already has to cope with the lackluster response to Haswell in the first place. While the chips are very, very good at conserving power when they are not being used heavily, the increased relative size of the onboard GPU has performance per watt under load dropping substantially
. An excerpt from the linked Anandtech review follows:
The new S7 delivers over 2x the battery life of the old model. Normalizing for battery capacity, the improvement due to Haswell is 57.5%. These results track perfectly with what we saw in PCMark 8. Workloads with greater idle time will show the biggest improvement in battery life thanks to Haswell ULT.
More importantly, Broadwell is the chip that would be able to truly drive growth for Intel in the "ultramobile" market as defined by Gartner. This group consists of Google
(NASDAQ:GOOG) Chromebooks, slate and hybrid laptops running Microsoft
(NASDAQ:MSFT) Windows 8, and thin and light designs. It is also the segment that is projected to have the highest growth in 2014.
So, no Broadwell means continuing to saddle these designs with Haswells which, while they are a significant improvement over Ivy Bridge are, again, not great under heavy load. Moreover, while Intel has made great strides with its GPU architecture, it still trails AMD’s Richland APUs, built on an ancient 32nm process technology with effectively 3-year-old GPUs (VLIW4 cores), at all but the most expensive options, which put it out of the reach of the average consumer.
So, why is this good for AMD?
Kaveri and the refresh of AMD’s small core Kabini/Temash APUs, known as Beema and Mullins, will ship in early 2014 on 28nm processes sporting both Graphics Core Next 2.0 cores, AMD’s latest as well as hUMA – Heterogeneous Unified Memory Architecture. Put simply, hUMA radically simplifies chip design by allowing both GPU and CPU to use the same on-die memory and resources. Because of this, I believe hUMA is the next disruptive technology in chip design for a number of reasons.
Simplified design means more compute power per area as well as per watt. No longer is there an entire second bank of memory that has to be powered up and down to use the GPU. Moreover, smaller die sizes overall mean higher margins per wafer, even at the larger transistor size. We’re going to get a real taste of how powerful hUMA is in 2013 with two next-generation consoles, Microsoft’s Xbox One and Sony’s
(NYSE:SNE) Playstation 4.
We’ve also gotten a taste of how good GCN 1.1 cores are in a mobile product with the current Jaguar-based A4 and A6 chips. Refreshed with GCN 2.0 cores and hUMA will put Beema/Mullins in a different class of performance for the ultramobile segment next year over Intel’s Bay Trail.
Intel pushing back Broadwell until 2015 means that what we see from Intel for the mid-range consumer part is what we’re going to see for nearly two years. Whereas AMD will have to push back Kaveri a quarter, its initial launch was due to be just ahead of Broadwell; now it has nearly a year to compete against something that, on a price-per-performance basis, is already better than either of Intel’s current offerings.
On the eve of AMD’s next earnings report, we’ll know if the firm is on track to meet its objectives of being operationally profitable by Q3. If so, look for a close this week above $4.43 per share to confirm the next leg of the stock’s resurgence. Pre-orders of both consoles are record-breaking. The key will be the company’s guidance for design wins in upcoming ultramobiles, which at this point, are few and uninspiring.
(See also: Will Intel Corporation Lead the Market Lower?
No positions in stocks mentioned.