As a long-time technology enthusiast and as somebody who has been active in the markets for years, it never ceases to amaze me just how easily a few buzzwords and some hype from management can, time and again, lead to wild expectations from new investors. The name that I'm talking about, of course, is AMD (AMD).
Now, don't get me wrong - many folks have made their millions by trading AMD (either long or short), but from the perspective of a fundamental investor (i.e. myself and many of my readers), it's not hard to see that AMD is...well, it's worth taking a look at these two charts to show you just how many investors have been "suckered in" over the years believing the AMD hype:
Anyway, what AMD's management has been touting at various conferences is the upcoming "Kaveri" chip, which features a next-generation CPU core code-named "Steamroller" and a GPU block based on the "Graphics Core Next" architecture found in AMD's high end discrete GPUs, the game console APUs and even the low-cost Kabini/Temash chips.
Many investors in AMD have hailed Kaveri as the company's savior in the PC space. Further, the buzzword "HSA" has been thrown around quite a lot nowadays, particularly in connection with a number of much bigger (and/or more important) players in computing such as Imagination Technologies, Qualcomm and ARM. However, the performance numbers for Kaveri seem to be in courtesy of a very nice find on Baidu. So, how does this chip stack up?
CPU Performance: Not A Bad Uplift
Here are some of the results for the A10-7800 (@ 3.5GHz) against the prior generation "Richland" chip (A10-6800K @ 3.5GHz):
It looks like the new "Steamroller" cores are anywhere from 10-33% faster than their predecessors. This actually is pretty good for a brand-new micro-architecture, although given the absolutely huge gap between "Piledriver" and Intel's (INTC) "Sandy Bridge"/"Ivy Bridge"/"Haswell" CPUs, this helps to narrow the gap, but AMD will still be significantly behind Intel's CPU performance.
That's fine. We already knew that. AMD's big push is graphics, right? Let's see how those numbers turned out.
Graphics: Hey, Wait A Minute, So AMD Isn't Magic?
Here are some results comparing:
- AMD's A10-6800K APU
- AMD's A10-7800 ("Kaveri")
- AMD A10-7800 + Radeon HD7750
So, uh, this doesn't really look like a big increase. In fact, given that the Intel Iris Pro graphics is already right about where the A10-6800K is (except with much more CPU performance and lower power consumption), things are going to look exceptionally bad for AMD's high end PC and notebook APUs once Broadwell (with an entirely new GPU architecture) on 14 nanometer hits.
The Fundamental Problem
The problem here is that AMD will likely be selling this silicon, which is probably going to weigh in well in excess of 200mm^2, for about $150-$180. This is obviously gross margin positive, but it will probably spend most of its life competing against higher performing, lower power, and much smaller solutions from Intel on 22 nanometer during the first half of the year and then on 14 nanometer during the second half of the year.
While these results are preliminary and not official, it's clear that the move from 32nm -> 28nm didn't give AMD's engineers a whole lot of extra real-estate to work with from a die size perspective (while keeping things remotely economical) and the performance of the process itself is probably nothing to write home about relative to the 32nm SOI process that "Richland" was built on.
AMD's competitive position with Richland on the desktop was bad (hence the low ASPs), and while Kaveri will improve that position against Haswell somewhat, all of that progress is likely to be wiped away by the time Broadwell arrives. It's tough to see AMD taking much - if any - share in the $150-$200 price-band with these parts that it didn't already have with Richland, but on the bright side, it could potentially defend its share within certain niches if it is willing to keep prices aggressive enough - at least until the second half of 2014, then all bets are off.