In 2024, Intel hopes to leapfrog its chipmaking competitors | Hacker News Reader

In 2024, Intel hopes to leapfrog its chipmaking competitors

spectrum.ieee.org

47 pointsvoxadam23 comments

I'll settle for them finishing the Ohio mega-fab and catching up, which would add a lot more cutting edge capacity beyond TSMC and capacity that is not in Taiwan.

"Company plans to beat competitors next year"

News at 6

Wow PowerVia is buying only 6% performance improvement? I'm not sure if that is when paired with a package designed optimized for the new thermal dissipation pathway? It has to buy more than that at the top of end of power scaling or even enable an increase in the power ceiling

Sigh. This again. Let's assume both SoC here has the same die size.

If Intel shipped 100,000 20A SoC in the first half of 2025. And 1M SoC in total in 2025.

TSMC shipped 50M of N2 SoC to Apple all in 2nd half 2025 ( Because of iPhone Pro )

In case anyone is wondering, back-side power ( or PowerVia ) is on roadmap for TSMC 14A. So technically Intel has technological leadership on 20A.

If we assume 20A and N2 are density equivalent. Which one is leading? Which one is better? ( So to speak ) And that is ignoring cost. And I said cost, not price of the node.

A lot of these information and debate are completely missing context. And then I am certain Intel fans will have their field day on HN, Reddit, Twitter, Anandtech, and possibly many other hardware forums which I no longer bother going to.

And the last bit.

>“Intel used to be the conservative one,” says Dan Hutcheson, vice chair of TechInsights. Previously, TSMC was more aggressive in its risk-taking, and the company more frequently missed the mark. Now, the situation has flipped,

That is just complete BS.

Edit: And just after I posted this, I read [1] TSMC announced their Chairman Mark Liu to retire next year. Another legend retiring. He was probably as important as Morris Chang in TSMC's success. Sigh.

[1] https://asia.nikkei.com/Business/Technology/TSMC-Chairman-Ma...

With the TSMC bidding war that happens every generation I could see Intel winning on cost.

For sure, competing on costs is GlobalFoundries' entire business model at this point, and it seems to be working well for them. I'm not sure Intel is able to make that transition though, they'd have no choice but to manufacture their CPUs elsewhere to stay competitive. They might as well sell off their foundry business at that point.

Winning on cost doesn't necessarily mean priced the lowest. Cost savings could be put toward larger die sizes and winning on performance.

> If Intel shipped 100,000 20A SoC in the first half of 2025. And 1M SoC in total in 2025.

Where did you get these numbers from because it seems like you just made them up.

> Intel fans will have their field day

It irks me that anyone is a "fan" of any such company. I have my favorite products from Intel and AMD over the years, but that's based on their merit as products, not on some allegiance.

It’s like sports team. You identify with them for some characteristic and root from them.

Assuming this actually ships on time, here's hoping Intel has the sense to offer this to foundry customers instead of keeping it exclusively for their own chips.

Does intel actually have success in this space outside of forcing government entities to use their foundries for what I assume is national security level projects?

Quick google on who their foundry customers are appears to be mostly defense sector.

I have heard that intel has not been the best at working with other customers because they have been so vertically integrated for the entirety of their business it was difficult for them to work with others.

Intel has stated that their foundry services will act independent of their chip designing division i.e. intel chips will have to compete with 3rd party chips for intel foundry time.

So yes, Intel (Intel foundry services) intends to be a direct competitor to TSMC.

https://www.anandtech.com/show/21191/intel-reiterates-we-are...

It's not clear from this that Intel would allow IFS to force their chip design division to "compete" on a level playing field.

They need to dogfood their foundry service from their chip design and keep the same interface for both in-house and external so the pain points are ground down.

Why would one assume this one will ship on time when the other projects from the company are either years late or cancelled?

Intel foundry announced sometime around March 2021. https://www.theverge.com/2021/3/23/22347250/intel-new-factor.... Anything actually come of that yet?

edit: found a claim for $300M in revenue from defence firms, https://www.investopedia.com/intel-shares-pop-after-foundry-.... Can totally see US chips for US defence keeping Intel sort-of alive indefinitely.

unless and until Intel scraps the hyperthreading architecture i cant see how they expect to leapfrog players like arm and amd. its been a sinking ship held together by PR for almost a decade now.

Is their chip design centered around hyperthreading? By this I mean they sell non-hyperthreaded processors...are they designed in a way that the non-ht processors still expect to be doing something every "tock" cycle so they just sit idle sometimes?

This doesn't make a lot of sense given hyperthreading has been a standard feature for AMD for almost as long...

Hyperthreading costs almost nothing in terms of chip area and increases the performance of things like JavaScript or Python based Webservers significantly. Sure your ultra optimized HPC workload or video game isn't going to benefit, but the performance cost for those is very small. The worst part of hyperthreading is handling CPU exceptions and keeping privilege levels separate between the two threads.

I'm old enough to remember the PowerPC to Intel transition at Apple, what a wonderful time that was.

Then you blew it because of arrogance, now you're high on hopium like the kids say.

Oh how the mighty have fallen.

> I'm old enough to remember the PowerPC to Intel transition at Apple

That was just yesterday …

Yeah that kids not old enough to have old-timer stories.

Ima dig out my 8080-based Interact, Atari 800, and see if I can find my dad's old Kim-1 board...

But I will not tell you any stories about punch cards.

Reply on news.ycombinator.com