TSMC and SK Hynix reportedly join forces to build products for AI, including HBM4 memory.
TSMC and SK Hynix team up for HBM4 co-production: Report : Read more
TSMC and SK Hynix team up for HBM4 co-production: Report : Read more
And more often than not, enterprise-priced AI cards and GPUs.This kind of bandwidth is exactly what modern multi-core CPUs need as current two channel bandwidth is often the limiting factor or bottleneck to better performance. And of course, we all know GPUs will likely be the first to adopt this technology.
Intel gave us triple channel ram a decade ago, so why hasn't this become more of a norm? I would happily settle for 3 dimm slots over the current 4.And more often than not, enterprise-priced AI cards and GPUs.
CPUs/APUs ultimately need big L4 cache on package/die to remove the DRAM bottleneck.
I did not know about triple-channel Nehalem/Westmere. I believe those were considered HEDT parts of their time, replaced with quad-channel Extreme CPUs in subsequent generations.Intel gave us triple channel ram a decade ago, so why hasn't this become more of a norm? I would happily settle for 3 dimm slots over the current 4.