Hbm2 memory is still too expensive

HBM2 memory is a very technologically advanced product, with great potential but also a major drawback, as the memory itself is expensive and difficult to use due to its proximity to the attached products.

Samsung can't meet demand for HBM2 memory

HBM2 memory requires large interleavers to connect this memory to a GPU, FPGA, or other devices. These interleavers are complicated to use and not cheap, this situation many of the benefits offered by this new technology, and is one of the reasons that their use has been relegated to a very specific niche.

We recommend reading our post about AMD Radeon RX Vega 64 Review in Spanish (full analysis)

Added to all this is the fact that availability is very low, since HBM2 is used to manufacture professional-grade GPUs, FPGAs and other devices. Intel, AMD and Nvidia are the big buyers of this type of memory. A reduced availability inevitably translates into a very high sale price, one more inconvenience for your use.

During ISC 2018, Samsung reported that it could double its HBM2 memory production capacity and still not meet demand, so the price of this memory is not going to drop for now. This fact makes HBM2 an extremely high margin product for its manufacturers, Samsung and SK Hynix, which means they make a lot of profit by selling little memory, a situation they will really like.

Samsung's latest Aquabolt HBM2 memory modules are capable of offering 8GB of capacity and speeds of 307GBps with a single memory chip consuming 1.2V. This means that higher bandwidth than TB / s can be achieved using four of these chips, a figure that is unattainable even for GDDR6 memory.

Overclock3d font