Intel’s new boss wants to teach the chipmaker new tricks
economist.com
economist.com
>That is unlikely to harm Intel directly, since such accelerator chips are complements to its server chips, not replacements.
This is just straight down the line wrong. In the data centre there is just 'compute' it doesn't matter if it's a GPU or CPU doing the compute. What does matter is that if you're doing all your compute on a GPU or ASIC then the CPU you're buying can be a lightweight chip just to do a little control & synchronization. Intel doesn't make huge money on those chips. It's the massive 512core 100GHz monsters where it's making the most cash. So these custom solutions absolutely present a threat to Intel's server revenue.
As well as that Tesla are going for ASIC in their car. Automotive was a key growth area for Intel which is one of the reason s it bought Altera and MobilEye. If they miss automotive (which they will given most big movers are on Nvidia platforms with ASICs coming after, and Intel is failing in their fab business).
>he hopes to continue Mr Krzanich’s strategy of expanding the firm’s reach.
Intel cannot execute on this until Intel's culture changes. The problem they have is they're trying to cost cut in legacy businesses whilst investing for growth in new businesses. Firstly this pisses off the legacy businesses like you wouldn't believe. Secondly, legacy business spends it's entire time dressing up in make-up and a new dress and screaming "Look at me! I'm a growth area!". And finally, the actual growth areas are hamstrung by inappropriate cost-saving, 1-size fits all management and spending 99% of their time trying to persuade the rest of the business they really are a growth area.
Intel is just not set up to win new markets.
Being aware of history is important. Very few things in business are new once you distill things down to their essence.
https://en.m.wikipedia.org/wiki/High-Tech_Employee_Antitrust...
Sounds like the "Hooli Box 3 - Signature Edition" https://youtu.be/xbBhH_spfBo?t=123
"Intel dominates the market for chips that power desktop PCs." It has a bit over 85% market share, in a two-player market. Furthermore, because they didn't say CPU like they should have, they're technically inaccurate. Intel network PHYs don't "dominant" consumer boards, last I checked.
"demand for server chips is growing, propelled by the profusion of internet-connected gizmos, from smartphones to cars." Really? IoT is the catalyst?
"missed the arrival of the smartphone" Unfair and reductive. Intel didn't "miss" the smartphone: it saw it coming, didn't have a product that fit, looked at the margins, and didn't think it was worth investing. That probably wasn't a bad move in hindsight.
"GPUs, specialised chips designed for video-game graphics" That tortures the definition of a different microarchitecture.
"And its manufacturing technology, which used to advance with such metronomic regularity that Intel called its business plan “tick-tock”"
This was it for me. Intel literally moved to the tick-tock cadency because Moore's Law was slowing down. Before, it was just an unending stream of process shrinks. Tick-tock accepted that there was now time for an entire microarchitecture update between process shrinks.
"The firm’s latest generation of products, built on its “ten nanometre” manufacturing process" Also, since when do we air quote measurements?
Ever since the process name stopped being indicative of the minimum metal pitch, or any aspect of the process itself.
If you are old enough, you have been jerked around by Apple enough to know better.
Expecting $500$+ dev packages in the next 10 years from Apple.