The artificial intelligence community is buzzing about the just-released ML Commons MLPerf Training v2.0 results, which were announced today. NVIDIA, Intel Habana, Google, and Graphcore all submitted at least partial results, and everyone is crowing about their performance. NVIDIA is the only one to cover the field, and has a strong showing despite still relying on their A100 GPUs. Intel’s Habana Gaudi 2 also shows great performance per watt in the tests they submitted, with the Graphcore Bow IPU and Google TPUv4 trailing somewhat. How do we make sense of all this ML training hardware? Head to GestaltIT.com for shownotes.
AMD is the latest company to have a large cache of data stolen by attackers. This week the new group RandomHouse popped up on the scene to claim they were holding 450 Gb of data stolen from the chip maker back in January. RandomHouse claims they didn’t write the malware or invade the systems and that they are simply a mediator that is asking for payment or they’ll release the data. Industry analysts like Catalin Cimpanu suggest that perhaps RandomHouse simply purchased the data from a failed ransomware scheme and is trying to profit off their investment. Tom, what do you think about this situation?
HPE has announced a raft of storage services in GreenLake, including block storage, backup and recovery, and disaster recovery. We got a sneak preview of the latter, which is based on Zerto technology, at Cloud Field Day last week. What’s the story?
Read More: HPE GreenLake spreads its storage wings
Watch The Tech Field Day Presentations Here
Former Security Field Day presenter Tempered Networks is now a part of Johnson Controls. Tempered, which utilizes the Host Identity Protocol to create private networks for communications in a novel implementation of zero trust network architecture (ZTNA), is slated to become a part of Johnson Controls OpenBlue platform. OpenBlue is one of the leading building control system platforms to manage lighting, HVAC, and other IoT devices. Financial terms of the deal were not disclosed. Tom, why would Johnson want Tempered?
8:23 | HPE Arms ProLiant
HPE is joining the ARMs race, introducing a ProLiant server with Ampere’s Altra ARM CPUs. The new ProLiant RL Gen11 will offer both the Altra, with up to 80 cores, and Altra Max, which tops out at 128 cores, in a single socket. This is the first time a tier-1 vendor has offered Ampere’s processors, and one of the first ARM servers directed at mainstream enterprise customers. Is it time for ARM?
As previously reported on the Rundown, IBM has been under fire in an age discrimnation lawsuit. The case is based on email exchanges between former CEO Ginny Rometty and former senior vice president for HR DIane Gherson that dealt with laying off older workers and replace them with younger millenial employees. The trial was expected to start next month but IBM has decided to avoid that by settling with the defendant. This is the second time that IBM has settled a case for age discrimination in the last two years. Tom, is IBM trying to hide the fact they’re changing their workforce?
14:05 | VMware Adds vSphere+ and vSAN+
Stephen, I hope you’re ready to add some commentary to this story. VMware has released the public version of Project Arctic, which has been renamed vSphere+. The sum total of the project is that companies want to run VMware on-prem but manage it in the cloud. In a move that looks shockingly like Amazon Outpost, vSphere+ allows you to add vSphere instances as a service and manage them from one central location. Sounds great! Well, there are caveats. The first is that vSphere+ and vSAN+ can’t manage instances running in the cloud, which negates any possibility of running it as a mangement platform for hybrid cloud workloads. The second? VMware did not announce pricing, which will become available at the end of fiscal Q2, or August. Stephen, does vSphere+ add up to you?
Read More: Project Arctic: Launch of vSphere+ and vSAN+
The artificial intelligence community is buzzing about the just-released ML Commons MLPerf Training v2.0 results, which were announced today. NVIDIA, Intel Habana, Google, and Graphcore all submitted at least partial results, and everyone is crowing about their performance. NVIDIA is the only one to cover the field, and has a strong showing despite still relying on their A100 GPUs. Intel’s Habana Gaudi 2 also shows great performance per watt in the tests they submitted, with the Graphcore Bow IPU and Google TPUv4 trailing somewhat. How do we make sense of all this ML training hardware?
30:08 | The Week Ahead
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.