TDA4VH-Q1: GFLOPS/GOPS

cao qi

Part Number: TDA4VH-Q1
Other Parts Discussed in Thread: TDA4VH

请问下DSP C7X 的算力320GFLOPS和1024GOPS是并列关系么，可以同时存在的么？分别是什么精度下的算力（FP16/INT8?）

Is the computing power of DSP C7X 320GFLOPS and 1024GOPS juxtaposed, and can they exist at the same time? What is the computing power at what accuracy (FP16/INT8?)

over 2 years ago

0 Pratik Kedar over 2 years ago

TI__Mastermind 24041 points

Hi,

Could you please elaborate the question in details (Few terms from posted question are not clear to us) ?

What is use case you are looking for here ?

Regards,

Pratik

0 cao qi over 2 years ago in reply to Pratik Kedar

Intellectual 786 points

datasheet first page

0 cao qi over 2 years ago in reply to Pratik Kedar

Intellectual 786 points

datasheet first page

0 Pratik Kedar over 2 years ago in reply to cao qi

TI__Mastermind 24041 points

Firstly, Half Precision Float (FP16) is not supported. We do support the Full Precision Float (FP32).

Explanation on 320 GFLOPS, for TDA4VH there are 4 C7x Cores and each can do 80 GFLOPS, so combined GFLOPS would be (80 * 4) = 320 GFLOPS

The GOPS benchmark is correctly pointed in your above shared image.

We don't have C7x Core only benchmark on fixed point 8 bit presently.

Pratik

Processors

Processors forum

TDA4VH-Q1: GFLOPS/GOPS