The first profile of Linglan Zhang, System Architect Manager (PMTS) of AMD, revealed that the development of first GPU with HBM has been completed.
Developed the world’s first 300W 2.5D discrete GPU SOC using stacked die High Bandwidth Memory and silicon interposer – Linglan Zhang
We’ve been speculating about Fiji silicon purportedly using High-bandwidth-memory (HBM) for almost a year. As it turns out, it could be true after all. The most disappointing part of the quote is definitely the power consumption, as 300W are more than we expected. The fact that AMD was forced to manufacture GPU with such a high TDP could only mean that it is still manufactured in 28nm process. NVIDIA’s Maxwell GM200 will likely feature lower power consumption (I will be surprised if it’s higher than 250W).
In case you are wondering what’s the difference between 2.5D and 3D stacking, here’s a good quote from ChipDesign:
In 2.5D stacking, two dies are flipped over and placed on top of an interposer. All of the wiring is on the interposer, making the approach less costly than 3D stacking but requiring more area. Heat dissipation is not much of a concern, since cooling mechanisms can be placed on top of the two dies. This approach is also lower cost and more flexible than 3D stacking because incorrect connections can be reworked.
3D vs 2.5D HBM stacking
AMD Radeon R9 380X
The second profile of Ilana Shternshain, Full Chip Timing Leader, MTS Asic/Layout design engineer, tells us about Tahiti successor codenamed R9 380X, also known as the largest in “King of the hill” line of products. King of the hill is likely the codename of Radeon 300 series. It is possibly the largest, since it does not use HBM like R9 390X GPU.
Backend engineer and team leader at Intel and AMD, responsible for taping out state of the art products like Intel Pentium Processor with MMX technology and AMD R9 290X and 380X GPUs.
AMD R9 380X GPUs (largest in “King of the hill” line of products)