Intel has some new codenames in the works for their high performance computing lines. The new AI GPU is code named Rialto Bridge and the next generation flexible architecure past 2023 will be Falcon Shores. The Falcon Shores lineup continues the chiplet idea, with x86 and Xe cores mated together to create a very configurable solution somewhat akin to the Apple M1 with CPU and GPU running together with high-speed memory. Speaking of GPUs, Intel also showed of the Ponte Vecchio. This and more on this week’s Rundown.
NVIDIA announced this week that a new supercomputer being built by Los Alamos National Labs is going to be powered by the latest chips honoring Adm. Grace Hopper. The new chips, dubbed Grace and Grace Hopper, are part of the NVIDIA HGX platform that is designed for high performance applications and provide twice the memory bandwidth of traditional units. The computer at Los Alamos, named Venado, will be capable of up to 10 exaflops of AI performance once completed. Stephen, what’s so cool about these new chips in your eyes?
Microsoft announced this week that their Support Diagnostic Tool has a little hole in it. That hole was discovered on April 12 and has been actively exploited ever since. Microsoft announced a patch that will plug the hole, going back on their previous statement that the issue wasn’t a security bug after all. The bug allows an attacker to send a specially crafted Word document to you that would execute remote code and avoid detection via tools like Windows Defender. Tom, what kind of damage are we looking at here?
It’s new CPU day, again! Ampere announced a new server CPU as well this week, the AmpereOne. The ARM-based system is looking to keep the momentum going that has seen it powering AWS and Azure-based ARM instances. The value of AmpereOne comes from the fact that the CPU can pack an incredible number of cores per unit. Instead of relying on tricks like hyperthreading to create virtual performance, AmpereOne will use regular old cores to do the trick. The goal is to provide a unit with lots of headroom for things like cloud-based mobile gaming apps. Stephen, are you excited for AmpereOne?
Let’s Encrypt announced a new initiative to free up some IPv4 addresses this week. Presenting at an Asian Internet confernece the company announced plans to reclaim several sections of wasted space in the legacy IP address range. Specfically, the company announced they would like to reclaim address spaces starting with 240, 225-232, and 127. Listeners with eagle ears may recognize those addresses as Class D multicast, Class E experimental, and loopback addresses. Tom, would this be a problem?
US based Oak Ridge National Laboratory has a new computer that tops the charts for speed. The Frontier supercomputer is the first exascale to be publicly benchmarked. This new milestone beats out the previous record holder, Japanese Fugaku, by a wide margin. Frontier is based on HPE’s Cray EX platform, running liquid-cooled third gen AMD Epyc processors. Stephen, what are we gaining from these speed races?
You don’t have to look far to see the current economic climate has some issues. Whether it’s inflation or geopolitical conflict or the shaky attitude of the stock market the signs are everywhere that potential trouble is on the horizon. Want another indicator? Sequoia, the large venture captial firm, had a Zoom call last week telling customers and potential investors that startups need to look at trimming costs and reducing spending. They also warned that startups should be looking to raise cash now instead of hoping for more funding later. Incubator Y Combinator sent a letter to their portfolio companies urging them to get to something they call “Default Alive” status, which is a fancy term for “profitable with your current funding”, also known as running a successful business. Stephen, are the VCs sounding the alarm? Or screaming about the sky falling?
Intel has some new codenames in the works for their high performance computing lines. The new AI GPU is code named Rialto Bridge and the next generation flexible architecure past 2023 will be Falcon Shores. The Falcon Shores lineup continues the chiplet idea, with x86 and Xe cores mated together to create a very configurable solution somewhat akin to the Apple M1 with CPU and GPU running together with high-speed memory. Speaking of GPUs, Intel also showed of the Ponte Vecchio. Stephen, what should we make of Intel’s roadmap here?
37:04 | The Weeks 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.