0
Tag: google
Google YouTube VCU for Warehouse-scale Video Acceleration
The Google YouTube VCU is designed for warehouse-scale video acceleration behind many of the company's popular services
Google TPUv3 Discussed at Hot Chips 32
At Hot Chips 32, Google TPUv3 took center stage in two talks. The company announced the TPUv3 in 2018 but is still guarding its hardware secrets
MLPerf 0.7 Training Results When Winners Only Apply
We discuss the MLPerf 0.7 results and how one company has become "the fastest" in a race of one when they are are also a race organizer
Google Cloud Confidential Computing Enabled by AMD EPYC SEV
Google Cloud Confidential Computing is available enabled by AMD EPYC 7002 SEV features. The new VMs encrypt data in memory to increase security
Google Cloud NVIDIA A100 Instance Launched
First to announce A100 support, the Google Cloud NVIDIA A100 instance offering will couple up to 16x Ampere GPUs in a single HGX A100-based cloud instance with NVSwitch
The 6 Layers of Google GCP Data Center Security
We get a glimpse into the 6 layers of Google GCP data center security to see how cloud production environments are secured
New MLPerf v0.6 Results Dominated by NVIDIA and Google
In the MLPerf v0.6 results, we still see NVIDIA Tesla V100 and Google TPU dominance along with new results from bigger clusters and Intel Cascade Lake generation CPU offerings
Intel CXL Compute Express Link Interconnect Announced
Intel CXL announced with a small set of partners. The Compute Express Link Interconnect release seems to signal Intel is getting closer to Gen-Z adoption
Google Enables Low-Cost Fast TPU v2 Pod Training in GCP
This week, Google announced a low-cost and fast deep learning model training in GCP. With the results of MLPerf v0.5 in hand and the Google TPU v2 Pod hardware in its cloud, Google showed off the fast and cost-effective training solution
NVIDIA MLPerf v0.5 Results Show DGX-2H Power
NVIDIA DGX-2H chassis chained together in a 32x system set along with clusters of DGX-1 and DGX-2 systems set six world record benchmarks on the initial set of MLPerf v0.5 results launched today