Samsung ends Intel's 2-decade-plus reign in microchips | Hacker News Reader

Samsung ends Intel's 2-decade-plus reign in microchips

hosted.ap.org

75 pointsrgbrenner45 comments

Let's be fair. Intel did more to end Intel's reign than any other company.

This. Strategic misteps, in mobile the most glaring one, plus I'm not sure how, but it seems that the main foundries (TSMC, Global Foundries, Samsung) have all almost caught-up on Intel. I also think this is what have permitted AMD to produce an outstanding new processor.

> plus I'm not sure how, but it seems that the main foundries (TSMC, Global Foundries, Samsung) have all almost caught-up on Intel

This is in part because of Physics getting in the way. Moores law is dying, because the Physics of shrinking transistors further is getting harder and harder

Also, many people don't realize that Intel doesn't make their manufacturing equipment. They work closely with the companies that do, and provide them research and have exclusive deals, but tsmc and Samsung probably buy the same brand fab equipment. Factory construction and such matters a lot too though, Intel closed a plant in new Mexico recently, and the yields were likely lower due to a difference in elevation

Intel acquisitions:

McAfee - Why?

Havok - Why?

Infineon Technologies - .............

Altera - Make Sense but not proving anything ( yet )

Mobileye - Make Sense but they needed to prove they can complete under Intel ( They were doing fine before being acquired )

SSD Contoller - For a specific range of uses ( QD1 )in DC, otherwise no longer competitive.

Baseband Modem, 7 years since acquiring Infineon, their modem is still fabbed on TSMC!

They could have worked with Apple on Fabbing their SoC, they had the past 3 - 4 years to decide and make the move. Now Apple is pretty much all the way with TSMC till 2019.

Where is Knight Corner / Landing / Hill? May be they are on the market somewhere, but is Nvidia out numbering them 20:1? Or even more?

May be they will milk the x86 for Desktop and Server for as long as they could. But Zen is coming, they have no growth area, what's next?

McAfee at least seemed to me like Intel executives didn't know what to do with the money they had, but didn't want to return it to the shareholders.

Instead it was burnt on a company in dying industry selling bad products.

Well, they could have turned it into a half decent product. Instead it lets it limp along while Symantec continues to improve.

P.S - Sorry lots of spelling mistake in the original post.

Just like Google, its hard to come up with something new profitable when you are fat and lazy on your gold lying goose.

> Where is Knight Corner / Landing / Hill? May be they are on the market somewhere, but is Nvidia out numbering them 20:1? Or even more?

Take a look at the Top500 list of computers [0]. Of the top 10, 3 have Xeon Phi coprocessors (aka Knights Corner/Landing, Knights Hill hasn't been released, nor has any 10nm Intel processor), 2 have Nvidia coprocessors.

[0] https://www.top500.org/

Intel worked very hard to earn this. For the last decade, they basically did nothing, it was so bad to the extent that you can dig out a 10-years old Q6600 then get yourself a decent nvidia video card and still hit 60fps on average on latest triple-A game titles.

https://www.techspot.com/article/1313-intel-q6600-ten-years-...

Intel would do well to bring Linus Sebastian on board as a paid advisor to help them strategize for how to embrace what's left of their market.

Please don't! They end up putting RGB LEDS on the CPU die!

No amount of marketing veneer is going to fix this.

He is already being paid by Intel (among other brands, pushing products is how you make money on YT nowadays).

Hmm, the article states the ten years old Q6600 won't cut it, but

"Soldiering on with its undying attitude, the 2500K consistently out-edged the i3-6100 and in titles such as Overwatch and Total War: Warhammer, the Sandy Bridge quad-core stayed several steps ahead. Considering how heavily the 2500K can be overclocked, those who invested in this processor six years [ago] are probably still grinning today."

Yep! I still have my i7 2600K running at 4+ GHz and Ryzen is the only thing that has made me even consider upgrading.

You know it's bad when the Dow Jones Industrial Average is outperforming desktop Intel CPUs in annual gains.

Hehe, I just upgraded my 2400 to a used 2700K. I have Samsung NVMe storage (though not for the boot drive), 16G of RAM and an nvidia 1070 in there. It does everything I can throw at it quite marvellously.

I have no reason to change the base system any time soon, or move to a newer intel or amd architecture.

(I really wanted a used e3-1290 v2 but those are still quite pricey, for an old chip)

> (I really wanted a used e3-1290 v2 but those are still quite pricey, for an old chip)

The price for used Intel chips is ridiculous. I bought a Dell R210 II and was looking to upgrade from the E3-1220v1 to an E3 v2 chip, but the cheapest ones available are still selling for $90+ USD (this is a CPU released 5 years ago).

The retail price of the E3-1220v2 new, in 2012, was $189. Heck, the E3-1220Lv2 was $189 retail and sells today for $150!

5 years later and it hasn't even lost half its value. Insane what stagnation in the CPU market has done to resale prices...

Yeah, the 1290v2s were still over £200, even the v1 was around that mark and the 3770k was still around £150... I managed to bag the 2700k for 100, and get 40 for my 2400.

Crazy for such old chips, but then they are still good!

i7-920 in my home machine here. Maxed out at 24GB of RAM for when I run a handful of VMs. But I'm not planning on upgrading anytime soon! Handles Overwatch and pubg just fine too.

I have the same i7-920 (overclocked at 3GHz) with 12Gb RAM. I have it since 2008 and I'm not planning to replace it anytime soon.

i7-860 (2.8ghz) and it's been "put to stud" as my web/SQL server, gaming server (Minecraft/ARK/whatever), and hosts my bot, from a closet in my apartment. The only reason I upgraded was to get something stock 3+ GHz, with a newer motherboard for a 1080.

Mine's OC'd to 3.66ghz, but I also added a SATA3 and USB 3.1 expansion card which has helped to keep it current.

I think this is as far as I can take it though. It'll never be able to boot off an NVMe drive, which I'll definitely grab for my next machine. My XPS 15 has one and there's a subtle but noticeable difference in snappy-ness.

Had the same, but recently upgraded to a HP workstation, found a decent deal on eBay.

https://hardwareisboring.blogspot.com/2016/06/hp-z440.html

I'm still running on a Core 2 Quad and while I don't game much, I can still handle my programming and browsing workload just fine. The best upgrade you can get is solid state storage, I/O is still the biggest bottleneck.

Massive grin here. Bought mine used for 100€ and it's been fantastic.

As 3D programming enthusiast and former gamer, I never understood the appeal of the lame 3D support on Intel chips, specially their OpenGL bugs like lying about supported hardware features.

The only positive thing that they have going for them is that at some point they decided to start supporting open source drivers for Linux.

Most people just want their desktop's drop shadows and window dragging effects to work fast enough. There's plenty of appeal in saving battery and money from a faster GPU that you aren't going to use.

Probably has nothing to do with the slow AMD Cpus in both largest consoles.

That's an oddly specific reference to read here, since I'm typing this on a Q6700. I have definitely noticed the same thing as this machine is still perfectly usable and faster than quite a few modern laptops.

Of course, this years processors are faster, but I can't imagine another point in time when ten years of progress would produce so little effective change.

That's a ridiculous statement. Look at the benchmarks:

https://www.cpubenchmark.net/compare.php?cmp[]=1038&cmp[]=30...

That's a nearly 8X difference in performance.

This is a nonsensical comparison. It lumps together three different kinds of ICs: DRAM, NAND flash, and CPUs. Samsung is in no danger of beating Intel in the CPU market any time soon, and Intel long since yielded the DRAM and NAND flash markets. The semiconductor is brutally cyclical. Right now memory is up. Samsung and Intel are both well aware that the wheel will continue to turn, and that the reward for complacency is annihilation. That's why they're both ruthlessly unsentimental about exiting markets in which they can't profit.

In Q1 2017, Samsung had 44.8% market share in DRAM, and 35.4% market share in NAND Flash. Two adds up to more than 10 billion dollars in a quarter.

http://www.dramexchange.com/WeeklyResearch/Post/2/4675.html http://www.dramexchange.com/WeeklyResearch/Post/2/4686.html

I watched this video yesterday, and it seems to explain a lot about how Intel has been less focused on building better technology in the last 10-15 years:

https://www.youtube.com/watch?v=osSMJRyxG0k

I'm not sure how balanced the viewpoint in the video is, but the author comes up with a lot of references.

If Intel wants to get us excited about their chips these days, they're going to have to start releasing more firmware and microcode so that we can start trusting their products again.

I'd rather run South Korean tech than anything the United States produces these days. It's sad that politics, and the threat of government spying is hurting their tech industry.

Here's some fortuitous timing: https://news.ycombinator.com/item?id=14872418

(Don't know all the implications yet, but it's probably relevant to my comment.)

What makes you think the US doesn't introduce spyware in South Korean tech as well?

Do you think most people care about this? If it did we would have a completely open architecture by now with lots of products already on shelves. But we don't, because nobody cares.

I care about this, my colleagues care about this.

GP's point is that the vast majority of people do not.

Heck, the vast majority of tech-inclined people I know don't. And the uninitiated? I'd imagine the only people who care about this outside of IT are the sort who prep armored and spiked vehicles full of canned food.

I'm not trying to deride, just saying, most people do not care about this.

I'm surprised the big clouds (which have to be some of the biggest, most profitable customers of x86 chips) don't push for this.

Do you honestly think that Google and Amazon don't have full control of the firmware that runs on their boards and chips, including microcode? Both Google and Amazon are literally buying customized CPUs from Intel for their clouds.

What are Samsung’s gross margins if their operating margins are about 50%?!

That's just semiconductors, which are about half their profit but only about a quarter of their revenue. Their overall gross margin is 46.9% and their op margin is about 25%.

In all honesty, Samsung owes its success to Intel here.

Reply on news.ycombinator.com