[mardifoufs]

I agree that the article is very lacking in substance and as you said, a hit piece. Yet there really seem to be a problem with Samsung's chips, especially their SoCs. They have underperformed for years now, and have always been inferior even in the high end (except maybe for the galaxy s6). That's despite repeated performance promises and a complete control over the entire production process (from design to fab to assembly). So what gives? I'd be very interested to hear your thoughts on that especially since it's hard to find any legitimate information beyond rumors.

[jason-phillips]

I was a software engineer and did not work in LSI, so I cannot comment on that in a way which would be most informative, but I will try from a different perspective. An accurate answer would be incredibly complex, obviously.

Samsung, from my perspective, mostly focuses on this year's production and developing next year's model. Yes, there is work done planning for new fabs and global supply chain systems, but 98% of cycles are spent on this year's production and designing next year's model.

Samsung accounts for a huge percentage of South Korea's GDP, about 20%. Therefore, Samsung needs to favor stability over taking risks associated with living at the bleeding-edge of innovation. Samsung always did play the safe game when I worked there (and for good reason when you're babysitting a $20 billion factory with $100 million of product that could be scrapped at any time).

Again, I wasn't in LSI but this is my perception overall having worked there, traveled to Korea and learned Korean culture.

[dylan522p]

Good thing the article even says Samsung is a well oiled machine in many areas outside of the ones with issues discussed in this article. Almost like large companies can have varrying success and culture.

[jason-phillips]

Absolutely. It's a very complicated subject.

[glowingly]

In many ways this is still true of Intel today. Like Samsung, their chips draw more power (even with "Intel 7" supposedly drawing parity to TSMC 7N) and are often slower than their competition's chips.

Maybe it's an IDM mindset issue? It seems the market has moved on to pure play foundries. For example, Nvidia's first fab partner was STMicro, an IDM, and Nvidia noted STMicro was unable to focus on being a fab partner. After that, Nvidia moved onto TSMC, also working with IBM, UMC, and Samsung at later points in time.