I love the EETimes, it helps keep me current on one of the loves of my life; electronics.

And I read and sometimes ask questions. Here is my latest inquiry.

The article was about how the super-computing Cray Corp. was creating new deep learning milestones.

My response/questions were:

Accelerators – as I understand them – are essentially hardware constructs of software code. The underlying reason being is that anything you can do in software can be built as hardware to run faster. So accelerator(s) speeds up certain functions on say an SoC by utilizing ASIC type hardware rather than processing those same functions as software.

That said, the way I read the last bit of the article is that this specialty Microsoft software is able to run on accelerators of computer systems other than just the Cray’s.

Am I misunderstanding what an accelerator is or is the Microsoft code delivered as hard IP in GDSII or OASIS data format and is then integrated into the customer ASIC before the accelerator is taped out?

Or, is the Microsoft code delivered as RTL and the customer synthesizes the IP into gates or physical constructs before integrating it into the accelerator?

Or, is the accelerator in this context just a highly specialized MPU that runs this Microsoft code?

Maybe it’s just me, but the accelerator part of the article threw me a bit and left me with questions I can’t answer.”

PS – I truly love electronic technology.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s