Page 1 of 1
AMD's X3D - HBM on CPU - memory wall
Posted: Sun Mar 08, 2020 1:44 pm
by smatovic
Seems AMD plans to put HBM on CPUs, increases bandwidth 10 fold they say.
https://www.anandtech.com/show/15590/am ... 25d-and-3d
--
Srdja
Re: AMD's X3D - HBM on CPU - memory wall
Posted: Mon Mar 09, 2020 2:42 am
by Leo
Re: AMD's X3D - HBM on CPU - memory wall
Posted: Tue Jun 30, 2020 7:50 am
by smatovic
Followup...
Fujitsu put 32 GB HBM on their ARM based A64FX processor with 1 TB/s
https://en.wikipedia.org/wiki/Fujitsu_A64FX
--
Srdja
Re: AMD's X3D - HBM on CPU - memory wall
Posted: Tue Jun 30, 2020 9:25 am
by Dann Corbit
What wonderful news. And since CPUs and GPUs both can talk to the same greased-lightning fast memory, think of the stories we could tell.
Re: AMD's X3D - HBM on CPU - memory wall
Posted: Tue Nov 03, 2020 9:19 am
by smatovic
I like the new POWER10 design...
https://en.wikipedia.org/wiki/POWER10#I/O
10ns DDR4 RAM latency, external off-chip memory controller, up to 1 TB/s per socket...whoosh.
--
Srdja
Re: AMD's X3D - HBM on CPU - memory wall
Posted: Fri Jul 09, 2021 6:15 pm
by smatovic
"Intel to Launch Next-Gen Sapphire Rapids Xeon with High Bandwidth Memory"
https://www.anandtech.com/show/16795/in ... dth-memory
...vendors cross-connect already the memory between several GPUs, wonder if we will some kind of memory-centric design in future.
--
Srdja
Re: AMD's X3D - HBM on CPU - memory wall
Posted: Sun Jun 26, 2022 10:44 am
by smatovic
smatovic wrote: ↑Fri Jul 09, 2021 6:15 pm
[...]
...vendors cross-connect already the memory between several GPUs, wonder if we will some kind of memory-centric design in future.
[...]
We can look at the Apple M1 as an kind of memory centric design already, CPU and GPU use 4x128 bit DDR4 memory channels with unified adress space for RAM, dunno about the neural network engine.
--
Srdja
Re: AMD's X3D - HBM on CPU - memory wall
Posted: Wed Jun 29, 2022 8:33 pm
by dangi12012
Dann Corbit wrote: ↑Tue Jun 30, 2020 9:25 am
What wonderful news. And since CPUs and GPUs both can talk to the same greased-lightning fast memory, think of the stories we could tell.
This could be transformative for chessprogramming. Evaluating the board takes the most cpu time. If this can be offloaded to a gpu without a latency impact it will be a wonderful time for us programmers.