Are there any cons of chip stacking? What's the reason that GPU has adopted the technology but CPU didn't?
ant
@nba16235 I found some great articles related to your question.
Why are CPUs still using DDR3/DDR4, while GPUs are on GDDR5? - https://www.reddit.com/r/hardware/comments/3bosvn/gpus_use_high_bandwidth_gddr5_and_hbm_why_are/
tl;dr - GPUs are designed for massive parallel processing and will soon potentially hit a barrier where memory bandwidth becomes the limiting factor for most applications (not sure if that has already happened). They can afford to trade some latency for this, as long as there is a good pipeline going. CPUs aren't quite there yet and latency matters a lot more.
Are there any cons of chip stacking? What's the reason that GPU has adopted the technology but CPU didn't?
@nba16235 I found some great articles related to your question.
Why are CPUs still using DDR3/DDR4, while GPUs are on GDDR5? - https://www.reddit.com/r/hardware/comments/3bosvn/gpus_use_high_bandwidth_gddr5_and_hbm_why_are/
Why HBM? - http://www.anandtech.com/show/9266/amd-hbm-deep-dive
tl;dr - GPUs are designed for massive parallel processing and will soon potentially hit a barrier where memory bandwidth becomes the limiting factor for most applications (not sure if that has already happened). They can afford to trade some latency for this, as long as there is a good pipeline going. CPUs aren't quite there yet and latency matters a lot more.