MLPERF Inference results for the GOYA processor,
 November 2019 publication

MLPerf defines different product status and test methodologies of reported results to help clarify conclusions that might be made from its reporting. Here are some of the distinctions it reports.

Product status

  • Available – available nor for purchase/deployment
  • Preview – on a path to availability. not yet there
  • Research – to test, learn and iterate

Test specifications adherence

  • Closed – tested to match the specifications, enabling comparison with other closed results
  • Open – tested to a set of parameters defined by the vendor to present product results in most    favorable light
  • Number of accelerators contained in the tested solution
  • Type of host processor employed

Other important factors , such as power, were not included in the measurements.

Goya performance shown here was reported in the available and closed categories

Goya

MLPerf Inference 0.5 results

BenchmarkDatasetDivisionStatusSingle stream (msec)Multi stream (SPQ)Server (QPS)Offline (Throughput)
Resnet - 50 V1.5ImageNetClosedAvailable0.2470070013090*14451
SSD - LargeCOCO - 1200X1200ClosedAvailable3.8518296.7326.3
SPQ = Samples per query @99% tail latency
QPS = Queries per second @99% tail latency
(*) ResNet-50 v1.5 Server scenario result not verified by MLPERF

For more details, see the MLPERF industry – wide results and whitepaper