This new chip package will be part of Intel’s 8th-generation lineup, and will combine a Core H-Series processor with a custom AMD Radeon graphics solution in addition to stacked HMB2. All of the silicon will be housed on one package/substrate, and will use Intel’s EMIB (embedded multi-die interconnect bridge) technology to share a power and communication framework. Anytime there are multiple chips or dies on one package, there is the complex issue of managing power, bandwidth, and communication between them. That is an issue AMD has already solved with their Infinity Fabric, and it’s unlikely that sharing proprietary technology is within the purview of this arrangement. Essentially, EMIB will allow heterogeneous pieces of silicon to communicate back and forth. That said, there are many questions that remain unanswered regarding Intel’s EMIB insofar as how it will perform; however, we do know that the technology allows Intel to use silicon interconnects--like an interposer--without actually using an interposer. This theoretically offers the density and performance of an interposer while forgoing the price, and offering the smaller footprint of traditional multi-chip designs.
All of this is to say that OEMs will be afforded the opportunity to design smaller, thinner laptops aimed at gaming and content creation, that don’t require a high-powered, discrete mobile GPU. To be clear, the GPU is still discrete; integrated, yet discrete. As it stands, Intel claims that using EMIB will reduce the system motherboard by 2.9 square inches, and halves the power usage of a traditional design. The Intel CPU will still keep its IGPU intact for encoding and video playback while shedding the more laborious video loads off to the embedded AMD Radeon chip. Allegedly, functionality is in place for intelligently switching between the two, and an article over at Anandtech briefly touches on this point.
Despite Intel and AMD being rivals and competitors, this partnership stands be be beneficial to both parties; Intel gets to reinvent and respin the H-Series line in a very appealing way, while AMD finally gets a chance at a market segment--that is, the mobile GPU segment--previously dominated by Nvidia. For an integrated solution, AMD is undoubtedly the partner of choice. According to PC Gamer, AMD is treating the Radeon GPU as a semi-custom design, similar to the designs they supply for the Xbox One/One X and Playstation 4. That would certainly make sense: AMD does operate a semi-custom foundry business, after all.
This is big news, and a lot of questions remain unanswered. We still know very little about the multi-chip design, and what to expect of EMIB. We know Intel will be running x86 cores, but in what configurations? We don’t know much of anything regarding the Radeon chip--Polaris or Vega cores, how many CUs, etc. The presence of HBM2 indicates a Vega possibility. We don’t know the intricacies of Intel and AMD’s partnership, and how it might affect each company’s product stack. We will undoubtedly have follow-up coverage, as questions are answered in due time. So, stay tuned!
- Eric Hamilton