Over the past few days, one may have noticed that STH has been featuring a number of different technologies that Intel showed at Intel Vision 2022 last week in Dallas, Texas. It has led to a number of pieces. One of the reasons we typically break these up is that we will reference them later, and it is easier to reference each individual part when we discuss Sapphire Rapids as an example. At the same time, we have some interviews that were recorded at Intel Vision 2022, so I wanted to have those as well. The result of all of this is that we basically have a huge portion of Intel’s 2H 2022 to 1H 2023 data center hardware in a lightning-round video with interviews.
Intel’s 2022-2023 Data Center Hardware First Look Lightning Video
Along with hardware spanning CPUs, GPUs, AI accelerators, IPU/DPUs, switch ASICs, and photonics, I also had the chance to sit down with some folks at the event and chat. I sat down at the event with Raja Koduri, recently promoted to EVP of Intel’s Accelerated Computing and Graphics Group, and Rick Stevens Associate Lab Director of Argonne National Lab who is working to get the 2+ Exaflop Aurora supercomputer online as one of the first publicly benchmarked Exascale systems. The result is that we have a video that shows the new technology (including one we have not published yet on STH but will be scheduled soon) and an interview about using the new technology all in one.
Here is the video of my trip to the Vision 2022 event:
As always, we suggest opening the video in its own tab, browser, or app for the best viewing experience. The embedded player is not great, but it is easy to use to help people get access to the video.
Final Words
If you are not yet subscribed to the STH newsletter, there is a bit of insight there. I am flying in May at a rate of one flight every other day, so we have slightly altered our article mix. At the same time, one of my big focus areas has been getting out and showing what we were not able to show during the peak pandemic era. About halfway through the Vision 2022 event, I realized that we had a large portion of Intel’s portfolio that will be hitting the market in the next year on the floor and so I started snapping photos and taking quick video shots. This is all unreleased gear, and Intel has products I know are coming but were not shown. Still, I wanted to put it all in one place for STH readers and our small but growing YouTube audience.
I’m scared for STH. Someday Intel – AMD – NVIDIA – Dell – HPE someone is going to wake up and they’ll realize that they could hire Patrick even at Lebron Supermax money and get a 10-100x+ ROI. I don’t know how STH stays alive since any of those companies can have this guy talking full portfolio. That’s so rare.
Silicon Photonics are so cool! Potentially a fully optical system bus where the only limitations are connection number/width and transceiver speeds. The “one connection to rule them all.” I can’t wait to see when this hits consumer markets and the photonics go RGB! LMAO!
It will be interesting to see when we finally transition from ‘photonics are required above certain speeds/distances but consume more power’ to ‘photonics are more power efficient above X throughput within a single chassis’
Is that going to be at 1 Tbps links throughout? Or maybe not until 10 Tbps? Or even higher? Or never?
For now, pcie 6.0 x16 gives almost 1 Tbps throughput. Curious to see if the power that will consume will be more or less comparable to a 1 Tbps silicon photonics transceiver…
Print, that would be excellent for STH since Santa Clara would gain its first natural communicator for a generation. Say what you will about Andy (and before even PC everyone did, as directly as dared) but nobody didn’t know what was*meant* to be happening. STH isn’t the MPR, even the new commercial light MPR, but it’s scratching the same kind of itch. This, Patrick, is the quantum leap your posterity demands you make.
Apologies Orin autocorrect plus mobile mangled your name.
We understand that you are super-busy, but it would be great if another editor threw in a few links to the product pages; some of the above is available now or coming soon:
https://habana.ai/training/gaudi2/
https://habana.ai/inference/