I will definitely take your word for it over a multi-billion dollar multi-national corporation like Intel
Thank you, I've been right so far for the last 5 years and Intel has been wrong [why? Because Intel is directed by non technical people and is all marketing; I was an engineer at Intel].
I already left the company last year, but I worked briefly in that "father of all GPUs" back in 2017. Last I knew before leaving back on December, it was going worse than expected. It has some "redeeming qualities", but nothing that NVIDIA can't quickly push out the market if the market is truly interested in the "advantages" Intel can provide.
Personal record:
* Q1'15 meeting, me: "we need to go past quad-cores to drive sales, we need bigger numbers in the specs so that people can see a clear difference, even if they don't take advantage of it. Something like what Qualcomm is doing." (before Skylake's launch, back when people was saying the PC era was over and sales declining every quarter). Intel's official response: "our power and performance teams report that the multi-core efficiency peak is at four cores and anything beyond that is diminishing returns".
* Somewhere back in 2015, Intel: "SoFIA will be the program that will allow us to improve our time-to-market and steadily increase our market share in phones and tablets." Me: "well, there goes our quality validation out the window."
Some time later...
* Me after the Broxton internal presentation, also somewhere in 2015: "looks good, but it would be great if we were to launch it soon; its launch date is way too late for being competitive."
April 2016...
* Intel at Denverton internal presentation, 2016: "we still believe in Goldmont [after having cancelled Broxton - Goldmont with graphics IP] and we want to push it to ADAS, be players in that new market that will boom in the upcoming years... four octa-core DNV CPUs with redundancy will be working together to process all incoming information and make autonomous driving (AD) decisions... we have signed an agreement with Daimler to incorporate our solution in their 2021 lineup..." Me: "haha, you wish. This is not competitive with what NVIDIA has right now, even less with what NVIDIA will offer in 2020/2021"
Fast forward...
* Intel at Q2'18: "we just did a demonstration of our Neon City reference platform to BMW (you see a picture of a... "file cabinet"-like liquid-cooled dual-socket platform... monstruos) and they are showing great interest..." General engineers' reaction: "that definitely doesn't fit in a car, without compromising storage; and definitely looks like an energy vampire for any gas or electric car to operate and cool".
BMW a year later.
* Intel acquiring Altera, Nervana, Habana Labs... Me: "oh great, another promise of unlocking the future with more integrated solutions". Suffice to say, Nervana only launched one generation of products under Intel and was replaced with Habana Labs. I can't remember the name of this platform, but it was a Xeon Scalable solution with integrated FPGA, trying to be sold for months to NASDAQ, Microsoft, and Disney without success and finally cancelled. Ironically, Intel internally uses Xilinx products instead of Altera's, because it's cheaper!
* Many more cancelled products and wasted efforts that were "dead on pitch meeting" not worth covering, but the jury is out there for this one: Intel giving away its $1B in R&D for 5G modem to Apple, because apparently owning the whole chain from design to manufacturing of the modem chips wasn't profitable [commentary: if being in control of IP, engineering, logistics, and manufacturing in a not-yet crowded market doesn't allow you optimize and cut costs to make CHIP MANUFACTURING PROFITABLE, you as a CHIP MANUFACTURING company have serious problems]. The bet here is to be players in "edge computing" (processing power at basestations), something Huawei [I know what you are going to say about it, but there are countries other than the US that don't care about such allegations] or Qualcomm can easily beat.
