Intel Confirms On-Package HBM Memory Support for Sapphire Rapids

Intel
(Image credit: Intel)

In a document released a couple of days before the New Year, Intel confirmed that its upcoming Xeon Scalable processors based on the Sapphire Rapids microarchitecture would support on-package HBM memory. A memory subsystem featuring HBM will significantly increase the bandwidth available to the CPU compared to a subsystem that uses conventional types of memory, like DDR4 or DDR5, meaning that future CPUs could resemble today's GPUs with a healthy amount of HBM memory riding on the same package. 

HBM Support Confirmed

Anton Shilov
Contributing Writer

Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.

  • JayNor
    0220H - HBM command / address parity error.
    0221H - HBM data parity error.

    That seems to be the only HBM info in the manual. It indicates that the memory controller has support for HBM. It would be a big jump to conclude this means there is HBM on-package.
    Reply
  • InvalidError
    JayNor said:
    That seems to be the only HBM info in the manual. It indicates that the memory controller has support for HBM. It would be a big jump to conclude this means there is HBM on-package.
    Well, given the ball density of HBM modules and the requirement that traces between endpoints be too short for transmission line effects to be a major concern, any HBM has to be on-substrate with whatever chip hosts the memory controller by definition.
    Reply
  • Chung Leong
    JayNor said:
    0220H - HBM command / address parity error.
    0221H - HBM data parity error.

    That seems to be the only HBM info in the manual. It indicates that the memory controller has support for HBM. It would be a big jump to conclude this means there is HBM on-package.

    Does it indicate even that much? I mean, the CPU could encounter an HBM error when it accesses HBM memory on an accelerator over CXL.
    Reply
  • ezst036
    Level 8 cache.
    Reply
  • JayNor
    Chung Leong said:
    Does it indicate even that much? I mean, the CPU could encounter an HBM error when it accesses HBM memory on an accelerator over CXL.

    Yes, that is likely the use. The Xe-HPC Ponte Vecchio GPUs will have stacks of HBM2e and the sketchy plans presented so far are for the Sapphire Rapids chips to access those GPUs via CXL .
    Reply
  • Russell_232
    Thank you for the update regarding the intel chip set.
    Now I can update my hardware with this chip set so that my pc can run faster.




    FaceTime
    Reply