Training 3.0 benchmark results show performance gains of up to 1.54x compared to six months ago and 33-49x improvement over the first round, driving innovation and energy efficiency in the industry. Intel’s Habana Gaudi2 ML training engine competes with Nvidia’s offerings, boasting better performance than A100 and lower pricing than H100. Nvidia, on the other hand, unveils their NeMo model with half a trillion parameters and expands the MLPerf Training suite to include GPT-3 and a new Recommendation engine. Their collaboration with CoreWeave showcases the superior performance of the H100, providing a 3.6x speed increase for GPT-3 compared to Intel Xeon and Gaudi2. Nvidia is also developing foundation models for their DGX cloud, collaborating with major players in the industry, and Intel is widely rumored to be developing its own Gaudi2-as-a-Service offering. Then there’s the Tiny 1.1 inferencing benchmark, which saw over 150 results and performance improvements up to 1000x.
0:48 – What Red Hat is doing with CentOS
IBM Red Hat has placed the source code for Red Hat Enterprise Linux (RHEL) behind a paywall, limiting access to subscribers and leaving CentOS Stream as the only public repository for the source code. While IBM Red Hat justifies the move as a means to protect their developers’ work, it has generated controversy and concerns among those outside of the company. This certainly has the community up in arms!
Read More: Furthering the evolution of CentOS Stream
Read More: IBM Red Hat Puts RHEL Source Behind Paywall
3:36 – Moving Windows to the cloud for consumers
A Microsoft internal presentation from June of last year that was recently made public due to an ongoing FTC hearing, shows Microsoft’s desire to move “Windows 11 increasignly to the cloud”. The goals here seem to be to deliver improved AI services and provide streamlined device roaming experience for customers across hardware platforms. Is virtual desktop infrastructure the way of the future for the consumer?
Read More: Microsoft wants to move Windows fully to the cloud
6:31 – IBM acquires Apptio
IBM is acquiring Apptio, a company specializing in tracking and managing data in hybrid cloud environments, for $4.6 billion in cash. The move aims to strengthen IBM’s hybrid cloud services and provide customers with tools to optimize their IT investments. Apptio’s platform will be integrated with IBM’s IT automation software and AI platform to offer comprehensive technology investment management solutions. Is this a major acquisition?
Read More: IBM acquires Apptio from Vista for $4.6B in cash to double down on hybrid cloud services
8:51 – Cisco set to acquire SamKnows
Cisco has announced that they intend a company called SamKnows which produces broadband-network monitoring solutions. SamKnows has products geared toward both service providers and consumers that show data around network performance and attempt to help resolve issues quickly. Cisco plans to integrate the SamKnows technology into its earlier acquistion solution of Thousand Eyes. What seems to be the overall play here?
Read More: Cisco to buy network-monitoring firm SamKnows for better last-mile visibility
12:04 – Databricks Acquires MosaicML
Data Lakehouse developer Databricks is acquiring generative AI startup MosaicML for $1.3 billion to empower its customers in building and deploying AI models using their own data. MosaicML specializes in running AI models on minimal systems and training them with proprietary data, addressing the resource-intensive requirements and potential inaccuracies of large language models. Databricks aims to democratize AI by offering more accessible and cost-effective solutions, expanding its capabilities beyond massive GPU resources. How will this acquisition impact the worlds of AI and Data?
Read More: Databricks buys AI darling MosaicML for $1.3B
15:37 – Cato Networks introduces AI tracker for malware command and control
Cato Networks is launching a product to leverage its deep learning algorithms to find and block malicious command and control domains. The differentiator here is that Cato is striving to do this in a way that is faster than traditional methods of blocking these domains based on domain reputation. What makes Cato’s approach interesting?
Read More: Cato Networks launches AI-powered tracker for malware command and control
18:39 – MLPerf 3 Upsets the AI Apple Cart
Training 3.0 benchmark results show performance gains of up to 1.54x compared to six months ago and 33-49x improvement over the first round, driving innovation and energy efficiency in the industry. Intel’s Habana Gaudi2 ML training engine competes with Nvidia’s offerings, boasting better performance than A100 and lower pricing than H100. Nvidia, on the other hand, unveils their NeMo model with half a trillion parameters and expands the MLPerf Training suite to include GPT-3 and a new Recommendation engine. Their collaboration with CoreWeave showcases the superior performance of the H100, providing a 3.6x speed increase for GPT-3 compared to Intel Xeon and Gaudi2. Nvidia is also developing foundation models for their DGX cloud, collaborating with major players in the industry, and Intel is widely rumored to be developing its own Gaudi2-as-a-Service offering. Then there’s the Tiny 1.1 inferencing benchmark, which saw over 150 results and performance improvements up to 1000x.
Read More: MLPerf Results Show Rapid AI Performance Gains
Read More: New MLCommons Results Highlight Impressive Competitive AI Gains for Intel
32:10 – The Weeks Ahead
Security Field Day 9 – June 28-29, 2023
Networking Field Day 32 – July 26-27
The Gestalt IT Rundown is a live weekly look at the IT news of the week. Be sure to subscribe to Gestalt IT on YouTube for even more weekly video content.
