Qualcomm had extra hints than releases at Computex however they had been fairly huge hints. Nothing ought to be model new to SemiAccurate readers, simply updates and particulars.
The massive bang got here on the finish of Qualcomm CEO Christiano Amon’s keynote speech. The speech itself was primarily a pat on the again to companions who’re waving the Snapdragon PC flag, honest sufficient. The units proven off for the primary time final yr had been supplemented by one other batch of designs, mainly the yearly machine refreshes. However what was that huge bang we talked about?
Lastly it’s ‘formally’ introduced
In a shock to nobody, Christiano admitted that Qualcomm has datacenter ambitions for actual. This isn’t simply AI AISCs however the full CPU facet of issues as effectively. The datacenter plans at Qualcomm have had a rocky highway prior to now, extra on account of exterior occasions and finance than product benefit. That is the good distance of claiming that though there was a lot pomp, ceremony, and even units proven, nothing ever got here out. SemiAccurate noticed, used, and truly like the Centriq line however alas that too sunk underneath the waves.
This time round we expect the story may have a unique ending, a a lot happier one. Qualcomm was one of many companions that Nvidia introduced for his or her ‘opening’ of the NVLink protocol, however extra on that later. Couple this with the Nuvia derived server cores, mainly the highest of the three core tiers, then add an AI100 SoC and you’ve got one facet of the market lined. NVLink to Nvidia GPUs will get you to the opposite, and off Qualcomm goes. Particulars have been promised for ‘quickly’ so anticipate particulars within the not so distant future however there’s a darkish shadow looming over this too.
Again to happier themes, and no we don’t imply the Qualcomm fortieth anniversary, we imply Docker. As you would possibly guess when the COO of Docker makes an look at a Qualcomm keynote, Docker now helps Qualcomm/ARM CPUs natively. This can be a actually good factor for builders and would solely be surpassed if Qualcomm supported an actual OS. Don’t maintain your breath, the doorstop some declare to be a laptop computer by my ankle is testomony to that. (Authors notice: After over a yr bumping my ankle, it was moved to behind my chair towards the wall to maintain the cats from opening the knee wall door so it isn’t totally ineffective)
Lastly, accessible AI100 {hardware}
Final up we’ve the Dell Professional Max Plus laptop computer/cell workstation. Why is that this at a Qualcomm occasion? Properly for those who take the discrete GPU out and put in a Qualcomm AI100 module, you might have, err, a laptop computer with an AI100 card. Aside from being a simple entry level for AI100 {hardware} entry, it offers devs a identified start line to begin coding. What is understood about it? Have a look at the specs on the cardboard to the facet.
Yup, it’s certainly the primary discrete NPU in a laptop computer, the Intel choices weren’t known as NPUs however to be honest, they had been actually in a unique and far decrease class of merchandise. 450+ TOPS is nothing to be sneezed at, neither is the 80W it pulls however that’s nonetheless a strong perf/Watt ratio. The snarky bit is the listed CPU, or lack thereof, for the Professional Max itself. A fast perusal of the Dell website reveals there are two fashions, the 14 and 16, each with Intel CPUs, no AMD to be discovered there. Or Qualcomm, however it is a good factor. Why? All of the instruments devs use and are acquainted with will work so devs will dev reasonably than struggle compatibility with ARM complications. If you wish to develop for Qualcomm AI {hardware}, that is your device.
In the long run the Qualcomm Computex messaging was extra concerning the future than the quick releases. Snapdragon X Elite Gen 2 was promised for a September unveiling and a Q1/26 launch except somebody *HINT* does it a tad earlier. Servers had been ‘quickly’, and extra hints promised a reasonably hectic 2H/25 for the corporate. Even when Qualcomm’s launch schedule didn’t coincide with Computex, they did a reasonably strong job of retaining our consideration.S|A
