Eridu Cuts To The AI Networking Chase With High Radix Switch System
…We presume the current comparison above pits 51.2 Tb/sec Ethernet switch ASICs, not the 102.4 Tb/sec devices that have been revealed and that will start appearing in products…
Search
New Google Networks Tuned Up For GenAI Inference And Training
…The company said in a deep dive blog that the Virgo network has 400 Gb/sec per accelerator on the Sunfish TPU 8t, which is four times the 100 Gb/sec provided…
Arm Comes Full Circle With Homegrown, AI-Tuned Server CPU
…It is etched in one of the N3 processes from Taiwan Semiconductor Manufacturing Co, which have transistors in the range of 3 nanometers and which also have practical reticle limits of a…
Another Step In Building The HPC Ecosystem For Arm
Compute Another Step In Building The HPC Ecosystem For Arm Many of us are impatient for Arm processors to take off in the datacenter in general and in HPC in particular. And…
Contemplating Meta’s Homegrown MTIA Compute Engine Roadmap
…With the MTIA 300, Meta Platforms shifted from the INT8 processing to FP8 processing, which means data conversion to integer from floating point formats is no longer necessary. The tensor units on…
Cisco Wins Over AI Customers With Merchant Silicon And Optics
…and added that Cisco has shipped over 750,000 400 Gb/sec units and over 40,000 800 Gb/sec units, which he said exceeds the shipment levels of the next biggest…
Scaling The Storage Pyramid With Micron’s Datacenter Memory And Flash
…So that’s the overall thought process: if you look at these four vectors for memory, and you add acquisition cost and TCO, the choices will evolve very naturally, and you will…
Sometimes, Air Is The Only Way For AI Systems To Keep Their Cool
…As you can see, on whatever these tests were, the MI350P was able to deliver 90 percent of the 4 TB/sec of peak bandwidth. As for compute, on the 16-bit…
Ayar Labs Gets $500 Million To Ramp Photonics Into 2028 AI Systems
…The chip above has more than 8 Tb/sec of bandwidth coming off of each of those TeraPHY optical engines, with less than 25 nanoseconds of latency across the optical I/O…
SambaNova Pits Its Engineering Against Nvidia For Agentic AI
…Which means the cost of inference has to scale down many orders of magnitude as the number of tokens processed as input and generated as output goes up by many orders of…