To get it out of the way, in my opinion Intel (INTC) is ahead of the competition regarding their fabrication process. Every article I have read about Intel speaks volumes as to how far. But that's only one piece of the puzzle. ARM Holdings (ARMH) is an IP vendor that licenses designs to vendors, and so their revenue is closely tied with foundries such as Taiwan Semiconductor Manufacturing Company (TSM). Intel's unique position essentially puts it in direct competition with a wide array of companies.
At the very basic level you have the transistor. Those transistors are clustered together to form CPUs or SoCs (system on a chip). You cannot evaluate a chip from the transistor level and draw the conclusion that it is the best processor on the market. It simply does not make sense. The architecture of the chip has much to do with its success, along with the price the consumer is going to pay.
These SoCs execute software - that is their main purpose. We measure the utility of SoCs by real world performance and benchmarks -- CPU scores, GPU scores, video playback, battery life, performance per watt, and performance per dollar just to name a few. These indicators are how we can tell how efficient the CPU is at doing its job, and relate it back to price.
Reading Between the Lines
In my debut article on Seeking Alpha, I gave my thesis as to why consoles could help Advanced Micro Devices (AMD) survive. These console wins could do much more than help AMD survive. Console wins ensure programmers are writing code for HSA-based hardware, provided the next gen consoles are indeed HSA devices. AMD has either failed miserably or definitively succeeded when they have innovated how data is handled. HSA appears to be their next idea of how SoCs should handle data.
Taken from Techspot.com:
HSA stands for Heterogeneous System Architecture, which aims to increase the efficiency of SoCs by unifying the memory space of the GPU and the CPU and directing calculations to the best suited processing unit for a given task.
One of the barriers to features like HSA catching on is ensuring software is actually coded for HSA hardware. If extemetech.com is correct and both Microsoft's (MSFT) Xbox One and Sony's (SNE) PS4 are both HSA compliant, then programmers are already writing code with this shared memory architecture, and they're actually enjoying it. Not to mention a fellow contributor pointed out to me that the Xbox One is running a Windows kernel, and is reported to be an HSA device. I do not believe it's out of the question to think Microsoft could release a version of Windows optimized for HSA. Microsoft is pushing into the tablet and smartphone sectors. These devices use either x86 or ARM based SoCs at their core. Microsoft would benefit from the optimizations as much as chip vendors. Having programs optimized for HSA hardware could be big for Intel's competition. Having an operating system optimized for HSA could be huge.
A Look Back
For those of us that don't remember, luckily there is Tom's Hardware Guide. On THG, look at historical benchmarks as we made the shift from 32 bit to 64 bit processors. Intel also chose to handle data a little differently than AMD. AMD chose to do more work per clock cycle, whereas Intel focused on netburst to do smaller steps quicker. There is a good summary on geek.com. The arguments to who was the performance king between Intel and AMD were not so one sided. In 2005, there weren't as many players in the space Intel had to worry about.
Fast forward to AMD's Bulldozer. AMD bet incorrectly by sacrificing single-threaded performance for multiple-threaded performance and suffered. AMD's excuse was that the operating system did not know how to handle AMD's radical new design, making the new design inefficient. Supposedly Windows 8 was supposed to help, but benchmarks on Tom's Hardware Guide showed very little improvement (about 1%). The Bulldozer architecture is where I feel Intel really started gaining an insurmountable lead over AMD.
Learning from Mistakes
I bring up the x64/netburst and Bulldozer examples to show how the handling of data can make huge differences in CPU performance. In the first example, AMD had the advantage over Intel. In the Bulldozer example, not so much.
Looking at benchmarks of high end Ivy Bridge CPUs vs. Sandy Bridge, you can see a marginal performance increase; sometimes even the flagship i7-3770k Ivy Bridge is outperformed by an i5-2550k Sandy Bridge. Leaked benchmarks of an i7-4770k Haswell compared to an i7-3770k show again only a relatively small performance increase.
Compare these to benchmarks of generational changes of AMD processor lines, and you can see that AMD has larger generational changes compared to Intel, but still significantly lag Intel in terms of performance.
How Could HSA Change the CPU Landscape?
There are a few scenarios that could be important to investors.
1. HSA could fizzle out, and nothing comes of it. I feel this is highly unlikely since both next generation consoles should be HSA devices. In this scenario both Intel and AMD are likely to see revenue go up since neither have much of a current presence in the mobile space.
2. Individual programs could benefit from HSA on a case by case basis. This could give AMD or ARM based SoCs an advantage in these specific programs. I feel this is the most likely scenario, and will help diminish Intel's lead over AMD. Intel and AMD are still both likely to see revenues go up however, because they're both seeing revenue growth in new segments. But it could make AMD more attractive to consumers than the first scenario.
3. The biggest win for AMD and ARM would be if operating systems were optimized for HSA hardware. Intel currently has a very small market share of mobile devices. Seeking Alpha contributor Mr. Tom Luongo directed me to a post at semiaccurate hinting that there is one major unnamed player in the HSA foundation. I would never rely on rumors alone, but it's also foolish to ignore them. I use them as a tell-tale of things to look for. It was a well spread rumor that Xbox was using an AMD APU before Wall Street took notice.
If Microsoft optimizes Windows for HSA hardware, this would be the biggest win for AMD and ARM. Many people feel AMD's only position in the semiconductor industry is to prevent Intel from stagnating. You can see from my examples above of the x64 architecture and Bulldozer that how a CPU handles data can make much more of a difference than the silicon the CPU is made on. After all, CPUs and SoCs execute software. And I do not believe an HSA compliant Windows is out of the question, since it could aid Microsoft in their press into the mobile space. Again, to be clear, so far this is only a rumor, and it's not as wide spread as the rumor of Xbox One featuring an AMD APU was.
ARM is also a member of the HSA foundation. Intel is very close to eating ARM's lunch with Bay Trail. Intel's Clover Trail Atom looked very impressive when compared against Nvidia's Tegra 3. Intel's Bay Trail may be an even more compelling alternative to an ARM powered processor. If HSA takes off, an HSA optimized Android OS could neutralize some of Intel's process advantage by making ARM processors run more efficiently than Intel.
At the end of the day, Intel and AMD are going to have a hard time losing entering into the rapidly growing mobile segment. Neither of these companies have seen significant revenue in the tablet or smartphone sectors, and both of these sectors have managed to grow YoY amidst declining PC sales. AMD's new low powered Jaguar devices have received fairly good reviews so far. Do not expect to play AAA titles such as Bioshock Infinite on them, but they are more than adequate to play less demanding titles, watch movies, or to run MS Office. We should see Jaguar powered devices by autumn, and Bay Trail powered devices by Q4.
Since the mobile industry is growing at such a rapid pace, it's possible that ARM, AMD, and Intel could all see revenues grow. Initiating a short position in ARM Holdings before a sign of slowing growth may not be the best play.
AMD and ARM Holdings both stand to gain big depending on how much HSA takes off. HSA optimized operating systems and programs could be a huge success for Intel's competition. Although there have been rumors of Microsoft getting rid of Windows RT, according to winbeta.org, Microsoft disagrees.
To completely dismiss the possibility of HSA succeeding on a grand scale is as unwise as investing in rumors. The SoC is at the heart of the growing mobile device sector. Intel and Nvidia (NVDA) have both abstained from joining. HSA is a way for Intel's collective competition to buy time to allow other foundries to catch up to Intel. Intel released their trigate technology last year, and no other foundry seems close to commercial production of finFET based devices. Companies like TSMC *need* the time to catch up. Any announcement of an operating system being HSA optimized could be huge for the SoC it's designed to run on. And at the end of the day, programmers will be exposed to coding for HSA hardware on the next generation consoles if extremetech.com's report is correct.
Additional disclosure: I may initiate a short in ARMH at any time.