This site may earn chapter commissions from the links on this page. Terms of utilize.

Twenty-one-year AMD veteran and HSA (Heterogeneous Organization Architecture) president Phil Rogers has decamped from AMD to join Nvidia. Rogers will be taking over as Chief Software Architect of Nvidia's Compute Server division, at a fourth dimension when Team Green is rolling out features like NVLink and standing to button forward with its own plans to make CPU and GPU compute more capable. With HSA 1.0 now complete, yous could fence Rogers is taking a page from Jim Keller's book and moving on at present that he's finished his piece of work.

That's possible — but I honestly dubiety it. Jim Keller has a longstanding reputation as a problem solver, and a history of spending only a few years at whatsoever given company. Over the class of his career, he's worked for DEC, AMD, SiByte / Broadcom, PA Semi, Apple, and AMD (again). Rogers, in contrast, was a longtime AMD employee and the very public face of the entire HSA initiative. The problem is, AMD's ability to actually create an ecosystem around HSA capabilities is extremely express.

HSA: A great idea, merely a lousy near-term strategy

When AMD bought ATI back in 2006, it talked virtually a "Fusion" of product families that would create synergy between the two companies. At first, that meant better integrated graphics for desktop and mobile processors, merely AMD had something far more thou in mind. Integrating a GPU directly into the CPU was a tremendous technical achievement, just the real goal of Fusion (afterwards HSA), was to provide a programming model that allowed developers to write lawmaking that would run seamlessly on whatever processor block was all-time able to execute it.

AMD's original HSA roadmap

AMD's original HSA roadmap

Ane thing I want to emphasize is that HSA is a nifty idea. There'southward a reason why companies like Qualcomm, ARM, TI, and Samsung all signed on with the HSA Foundation, and why these companies continue to improve the heterogeneous compute capability of their own solutions. Every major semiconductor company on Earth has taken steps towards heterogeneous compute, from Qualcomm's Snapdragon 820 and its CPU – GPU – DSP triad, to Intel and the Xeon Phi. The thought of matching the right workload to the right processor is potent, and AMD wasn't incorrect when it recognized that many-core heterogeneous architectures would be critical to long-term performance improvements.

But technical excellence and early on design leadership don't always interpret into increased marketplace share. The problem with HSA is that AMD utterly lacked the resources to move the market, every bit a whole, towards adopting information technology. Kaveri'south OpenCL performance improved on Richland in a number of ways, equally we documented when that fleck launched, simply 18 months afterward Kaveri launched, there's well-nigh no HSA software in the market.

This isn't the first time AMD has led in technical evolution just depended on other companies to ultimately bulldoze adoption. Two of the company's previous technologies, HyperTransport and x86-64, took like paths. The deviation, notwithstanding, is that AMD was able to make substantial use of HT in its ain hardware — it drove the "glueless" architecture that fabricated Opteron servers so compelling compared with Xeon alternatives in the 4P infinite back in 2003 to 2005.

As for x86-64, non only did AMD win huge accolades for stealing a march on Intel, information technology ultimately forced Intel to adopt its own standard for the future of 64-chip on x86 CPUs. HSA wasn't intrinsically useful to AMD APUs without substantial software back up, and AMD merely lacked the funds and evolution resources to bulldoze wide adoption. AMD's APUs may be "adept enough" for vast swathes of the market, but that hasn't kept the company'south sales from collapsing. "Practiced enough" doesn't kindle the imagination or go developers excited to work on your platform. And while it'south truthful that GPU performance has been a highlight of AMD'due south APUs for many years, peachy graphics performance (relative to market segment) didn't require HSA features in the kickoff place.

It should be noted that these problems aren't at all unique to AMD. Later briefly flirting with consumer-level CUDA applications, like the media encoder Badaboom, Nvidia largely left the space. You lot tin can notwithstanding observe media encoders with GPU support, to exist certain, but there'due south been no push to bring GPU acceleration to coincidental content or to widely leverage OpenCL in every 24-hour interval applications. GPU dispatch remains the province of workstation-class software, for the nigh role.

So, why motility to Nvidia? Because the supercomputer / HSA space offers a much more fertile ground for the kinds of improvements that heterogeneous compute can offer. CUDA is already well-established, AMD doesn't seem to have any serious plans to assail the market in the near-term future, and Rogers likely wants to see the improvements he'southward designed bear fruit in actual shipping software. In the HPC space, edifice new software capabilities to take advantage of hardware is part of the job, and while Nvidia faces its ain headwinds in that market place thank you to Intel and Xeon Phi, it's still a much better overall position.

I don't retrieve AMD will drop HSA or heterogeneous compute going frontward, merely AMD's marketing materials don't actually focus on it these days. That'south probable wise. Fundamental CPU performance and ability efficiency are far more critical to the company's future than any improvements it can incorporate into heterogeneous compute. Nail Zen's debut and future APU launches, and the HSA question can exist dealt with at a later date.