This is how badly Intel screwed itself over

I

igor_kavinski

Quasi-regular
Joined
Aug 15, 2025
Messages
304
Reaction score
174
1764005372274.png

All Intel had to do was LISTEN to the consultants it had hired (according to Jim Keller) and change its fab tooling and PDKs to mimic the industry standard so companies failing to get allocation at TSMC could place their orders with Intel with minimum investment of time and effort. BUT NO.

OH NO.

"THE ENTIRE WORLD DEPENDS ON INTEL CPUs! We don't really want business from the rest of the world. We want to keep making our own CPUs and flood the world markets with them!".

Yeah, how's that going for you, INTEL?
 
Yeah its mind blowing that there is this level of industry demand and Intel can't get folks to fab at their plants. Like customers are literally willing to lose money on unrealized sales than even bother with Intel foundry services.

And the US took a 10% ownership stake in this ****pile of a company.
 
Intel might not be the hero we want... but if they bent their excess FAB capacity to making DRAM modules...they might drive the price back down!
 
Except I think retooling the existing fabs for RAM production is no small task unfortunately. It's one reason why Samsung's chip yields are abysmal yet they are doing fine with RAM production. I've been told by people more informed than I that Andy Grove (?) famously saved Intel by getting them out of the memory business because of the price volatility and how sharp the price drops were once production exceeded demand. Pat did the same and killed Optane.

Ah, I found an article: https://anthonysmoak.com/2016/03/27/andy-grove-and-intels-move-from-memory-to-microprocessors/

And about Pat's origins: https://www.thelayoff.com/t/1vPygjc5

By the way, one guy at overclock.net was cussing Pat so bad because of how bad Arrow Lake is. So powerful yet gimped so pathetically that it can never achieve its full potential without major respin. He made me chuckle.
 
I jumped the gun.

PTL has 6MB less cache yet performing well. So with more cache, it could do pretty well in games.

1764088123262.png
 
1764094084107.png

Panther Lake vs. Arrow Lake frequencies in GB6.5

If the performance holds at higher frequencies, then 5.1 GHz Panther Lake is almost as good as 5.8 GHz Arrow Lake!
 
igor_kavinski said:
Kind of a miracle that Intel managed to stay barely competitive with work conditions as bad as that ex-employee is claiming.
Click to expand...
Really just companies with two different histories and wildly different market positions for most of their existence.

Imagine if, when the Athlon came out, Intel came out with Netburst / P4, and instead of going 'oh crap, that's not what the market needs!', they stuck with it. And then AMD, now with more business, is actually able to remain performance competitive and avoids the absolute footgun that was the 'Dozer series (mirroring Intel's Netburst years earlier, the irony).

Now AMD gets to the point that they're (along with Apple) working with TSMC to push the latest nodes, and they have volume shipping to enterprises - and Intel has shrunk to 10% of the market.

How might the (average) employees feel differently?


I think Intel getting their junk stomped in by AMD being able to build literally two dies (their current Zen5 eight-core CCD and now 12-core Zen5c CCD), and just swap I/O dies around between desktop, HEDT, and server, is starting to sink in. Intel has to build so many different unique dies, or in the case of Arrow Lake, resorted to an expensive stacking solution that still falls short in many usecases.

And AMDs work with TSMC to stack 64MB of cache on the die and have it be cohesive with the 32MB already there is just :chef's kiss:. They can provide products with oodles of L3 cache using the same exact dies that they use for everything (except APUs)!
 
You must log in or register to reply here.
Become a Patron!
Back
Top