Dr. Jeffrey Cheng, the chief software architect at AMD, was just named an AMD fellow. He talks with ChannelBuzz about AMD software’s role in the company’s recent turnaround.
Three years ago, AMD was in desperate financial straits. The company was bleeding money, and was getting hammered in its major markets, particularly at the high end. They had pulled out of the high-end server, PC processor and GPU markets as a result. Its rebound since under CEO Lisa Su has been remarkable, and well-chronicled. The company re-entered all of these markets with newly developed products. Last year they introduced their new EPYC server processor, their new PC processor, Ryzen, and Vega, their new GPU graphics card architecture. This spring they launched the next-generation versions of all three, which have been generally well received by the market.
“This has been an incredible watershed year for us, with these three leading edge technologies covering graphics, client and server,” said Dr. Jeffrey Cheng, the chief software architect at AMD. “We are entering the one-year anniversary of our server chip launch. Now, any data centre customer is talking EPYC. Server business typically takes a long time to ramp up – but many people already have EPYC deployed in their data centres.”
It has also been a very good year for Cheng personally. In May, AMD announced that he had been made a corporate fellow for his contributions to the company’s graphics technologies and software development. Cheng is the fifth such executive at AMD to receive this honour, but all the others were hardware-focused. He is the first AMD corporate fellow from the software side.
“After getting my PhD in computer science from the University of London in the U.K., I came to ATI in 1998,” Cheng said. ATI was a Markham ON-based graphics-focused semiconductor company that AMD acquired in 2006. “I had planned to stay at ATI for a while and move on, but found that it was a very interesting company, and 20 years later, it is now part of AMD, and I’m still here.” Recent AMD projects with which he was involved include the GPU architectural design for Microsoft Windows, where he was the lead software engagement person, and the Microsoft DX 12 API, critical since Microsoft is AMD’s biggest gaming partner and the bulk of games are written on Windows platforms. He also architected the Vega high-bandwidth cache controller, and oversaw GPU virtualization with SR-IOV technology, and GPU data centre projects.
Cheng said that while AMD is commonly thought of as a hardware company, the software is now essential to what they do.
“Software, and our software architecture strategy is critical for AMD,” he stated. “We have a very high investment in software development. The reason is that we design hardware to run software. People don’t think of what we do as being based on software, but the whole purpose of hardware development is to run software well. We make sure that the software that everybody uses is optimized for our hardware – both CPU and GPU.
“It’s also critical to enable our software for new applications that did not even exist a few years ago, like artificial intelligence, machine learning, and virtual reality,” Cheng added. “We are really building out our machine learning capabilities. Technologies like Blockchain need our technology to run well, and AMD technology is the preferred hardware for data mining.”
AMD’s graphics, which was once heavily tied to gaming, has moved beyond that – somewhat.
“We are no longer tied to gaming demands as we once were,” Cheng said. “AI has its own demand, and we are investing in GPUs to be more of a general purpose GPU. Gaming happens to be on the same level of workload, with the need to process a massive number of pixels. Still, while this has provided us with an opportunity to move beyond games, gaming is still critical. While a lot of people play games on the PC, and DirectX is important for us, the game consoles drive the graphics development agenda, and we are in all of these game consoles.”
The result of all this, Cheng said, is that AMD is aggressively growing, at a time when many large OEMs are laying people off.
“Our Markham office here is drastically expanding,” he said. “We have roughly 2000 people at this site, and because this is no longer a head office location, as it was with ATI, most of them are engineers. We have the opposite position of a brain drain issue here. We are looking for people constantly. We have 150 positions open today. If you have friends who are engineers, tell them to apply!”