NVIDIA RTX A6000 GPU for RIVA inference

AI & Data Science Deep Learning (Training & Inference) Riva
#1

I noticed that the NVIDIA RTX A6000 is not listed as a Preffered Deployment Platform (see here ), this is surprising as the A6000 GPU memory exceeds comfortably the minimal 16GB requirement. What should I expect if I use the A6000 card for RIVA? Lower performance, or maybe some compatibility issues?

1 Like
#2

Has anyone benchmark A6000 against A100 for RIVA?

#3

My apologies that this slipped through the cracks. This post should be in the RIVA category I believe.

I hope you don’t mind if I move it over.

#4

Hi @mk.faraz and @balintcenturio

Thanks for your interest in Riva

RTX A6000 will run Riva comfortably, regarding the benchmarks, I will have to check with the internal team and let you folks know the details,

Thanks

#5

Any update on the benchmark for RTX A6000?

We are able to process 60 streams on Tesla T4, 180 streams on A100 with 800ms real time chunks before getting observable performance delay in transcriptions.

#6

Hi @mk.faraz

Apologies for the delay,
I hear from the team that currently we don’t have benchmark for RTX A6000, but it should work fine with Riva

Thanks