MLCommons released the latest wave of machine learning training and inferencing benchmarks, and Nvidia’s next-generation H100 really aced the results, delivering up to 4.5 times the performance of the leading Nvidia A100. But there are some other gems in the results, including a surprisingly strong showing for the Biren Technology BR104 and good enough performance from Intel’s Sapphire Rapids Xeon to eliminate the market for low-performing inferencing hardware. What do the MLPerf results tell us? This and more on the Gestalt IT Rundown. Head to GestaltIT.com for show notes.
0:42 | Meta Moves PyTorch to Start AI Fire
PyTorch is one of the most popular deep learning frameworks in the industry today. It powers Tesla along with thousands of other projects looking to leverage AI and machine learning. On Monday, Meta’s AI group announced that the project had been transferred to a new PyTorch Foundation which is under the auspices of the Linux Foundation. Housing the project under a new Linux Foundation umbrella allows it to become more neutral, including a governance board made up NVIDIA, Google, Amazon, and more. Stephen, this is a bold move by Meta. What’s the play here?
Read More: Meta spins off PyTorch Foundation to make AI framework vendor-neutral
3:05 | Cyberattack Earns Iran Albianian Ban
Threat research group Mandiant, which is now a part of Google, reported this week that the Albanian government was disrupted by a targeted malware attack. The attack has been attributed to groups backed by Iran. The attack was in response to calls from the Russian government to support ongoing operations and may have targeted the former Soviet republic specifically for this reason. The attacks happened back in July but we’re starting to get new data about them today. Tom, is this going to be the new normal for cyberattacks?
5:46 | VMware Charged with Misleading Investors
It turns out that SEC fraud is just a matter of dates. VMware will be settling up with the government to the tune of $8 million dollars for misleading investors about their financial performance. Per the investigation, VMware leveraged their order backlog process to push licensing revenue into the next quarter by delaying the delivery of said licenses to customers. This allowed VMware to hide their slowing growth. The SEC said they found evidence of this going back to 2019. Stephen, could this cause issues for the Broadcom acquisition?
Read More: SEC charges VMware with misleading investors by obscuring financial performance
9:42 | HP Pays Up Over Printer Security
HP Inc has settled a complaint about printer security. The issue arose in the EU back in 2016 when the printer giant decided to push an update to the HP OfficeJet printer series. Those printers had an embedded security chip and the update enabled secure communications between the chip and the printer cartridges. That means that the printer would only work with Genuine HP products and not 3rd party equipment. HP will pay 1.35 million Euro into a fund to compensate consumers affected by the update but not claiming any fault. Tom, what odes this mean for HP printer security?
Read More: HP pays $1.3m to settle dispute over printer security chip
13:20 | Micron Is Building the Biggest RAM Fab
Micron this week broke ground on a massive new memory factory in Boise, Idaho. Although DRAM is entering a down cycle, it’s a long term market and what goes down must come up, at least as far as semiconductor manufacturing is concerned. Micron’s US move matches SK Hynix’ expansion in South Korea, ensuring that these two countries will lead in production of next-generation memory chips. What does this say about the future of the IT industry?
Read More: Micron Breaks Ground on Its $15 Billion EUV DRAM Fab in the U.S.
17:02 | Biden Looks to Curb Chinese Chips
The White House is looking to restrict chip exports to China in the near future. Reports say that President Biden is going to release new regulations on semiconductors that are used in artificial intelligence and chipmaking applications that are bound for China. The policy was outlined in a series of letters earlier this year that said that equipment used to produce chips for these applications would need a Commerce Department license. The letters allow for rapid action to be taken to put the restrictions in place but only apply to companies that receive them, which means this is not a blanket federal policy yet. Stephen, are we about to enter a new trade war over chipmaking?
Read More: Exclusive: Biden to hit China with broader curbs on U.S. chip and tool exports
26:25 | Nvidia Aces MLPerf with H100
MLCommons released the latest wave of machine learning training and inferencing benchmarks, and Nvidia’s next-generation H100 really aced the results, delivering up to 4.5 times the performance of the leading Nvidia A100. But there are some other gems in the results, including a surprisingly strong showing for the Biren Technology BR104 and good enough performance from Intel’s Sapphire Rapids Xeon to eliminate the market for low-performing inferencing hardware. What do the MLPerf results tell us?
Read More: Nvidia H100 ‘Hopper’ Benchmark Results Published
32:20 | The Week Ahead
Cloud Field Day Extra with NetApp – September 20, 2022
Cloud Field Day 15 – September 21-23, 2022
The Gestalt IT Rundown is a live weekly look at the IT news of the week. It broadcasts live on Facebook every Wednesday at 12:30pm ET. To watch along, “Like” our Facebook page. Be sure to subscribe to Gestalt IT on YouTube for even more weekly video content.
