Discussion Intel's past, present and future

[fastandfurious6]

guys GPT5 got released and the timeline is going bonkers!

 

[marees]

guys GPT5 got released and the timeline is going bonkers!

Curious. How does this link to Intel's future 🤔

 

[fastandfurious6]

intel board just consulted gpt5 and they were told intel needs to end and remade into a completely new company. they already got convinced lol

 

[whoshere]

 

[DZero]

No, I'm pretty sure the doom is overstated. ARM couldn't even stop Qualcomm and Nuvia. There's no way AMD with its far more nebulous claim of Intel using its IP could stop Intel and any buyer.

I'll take it as the worst scenario. Best one is that x86 survives and status quo is god.
Realistic one is that x86 loses marketshare due Intel mishaps and AMD would reduce those loses.

View attachment 128496

Imagine if the USFS survives but Intel as a company fails hard? It would be hilarious to see.

IBM Z is their mainframes, with an ISA that's been extended and extended and is still able to run code dating back to the 1960s. PowerPC is IBM POWER, used in their AIX servers. That traces its development back to the early 80s (the first true RISC)

ISAs that for the most part have a single company behind them are the ones that are dying / have died out. That ones that survive are ones with wide support or a purpose for existing. i.e. ARM, x86 for obvious reasons. RISC-V because it is the "free" alternative to ARM for the really cheap stuff. Loongson because it is a China developed ISA without any western control. IBM Z because mainframes are gonna outlive all of us. POWER won't last forever, but so long as customers keep buying enough AIX servers for IBM to stay in that business they'll keep selling them.

Funny story China wants RISC-V to be their ARM, so expect a own branch called "RISC-V C" or something like that

 

[marees]

Imagine if the USFS survives but Intel as a company fails hard? It would be hilarious to see.

Won't happen, if we go by evidence of Lunar Lake

The software guys are not magicians. They are heavily constrained by fab hardware

An Intel that exclusively uses TSMC hardware would have designs that go toe-to-toe with AMD

USFS otoh will be a white elephant

 

[dangerman1337]

View attachment 128496

Big question what can USFS offer that Samsung and TSMC can't outside the (sadly plausible) scenario of China invading Taiwan? Like if Nvidia won't even chance RTX 70 (SF2X is probably the node for RTX 70) on Intel 14A-E/P with some Intel 18A-P silicon for a Chiplet based GPU for late 2028/2029 then who will use USFS? Qualcomm seem iffy, Broadcom ditto.

Won't happen, if we go by evidence of Lunar Lake

The software guys are not magicians. They are heavily constrained by fab hardware

An Intel that exclusively uses TSMC hardware would have designs that go toe-to-toe with AMD

USFS otoh will be a white elephant

I mean the design side is profitable effectively for Intel but USFS has the big dual problem of their older nodes having little to no value for external foundries with the cheapness of Samsung's nodes like 8nm and people who will want to risk big orders on 18A derivatives, 14 derivatives and beyond.

But that said Pat did split IFS & Product side division wide and probably will help the product side not use foundries their biatch in trying to make steppings to fix problems like what happened with Sapphire Rapids.

 

[Win2012R2]

An Intel that exclusively uses TSMC hardware would have designs that go toe-to-toe with AMD

No they won't magically get those designs, they'll just be able to use same node (at best), look how their GPU on TSMC worked out

 

[511]

No they won't magically get those designs, they'll just be able to use same node (at best), look how their GPU on TSMC worked out

this is their second dGPU IP they will address the PPA issue due time it takes time to change mindset. Xe3 has already made PPA Improvements

 

[Win2012R2]

this is their second dGPU IP

3rd - there was one before Alchemist, small volume for devs "to prepare"

There won't be scope for dGPUs at Intel anyway, my view.

 

[511]

3rd - there was one before Alchemist, small volume for devs "to prepare"

oh yeah i forgot the initial dGPU on 10nm

There won't be scope for dGPUs at Intel anyway, my view.

Xe3P dGPUs likes to have a word with you they are alive as of now

 

[Win2012R2]

Xe3P dGPUs likes to have a word with you they are alive as of now

We'll see if or when they are out for sales, and most importantly - how many.

IMHO success of RDNA4 will give extra incentive to can it, if half the stuff about RDNA5 is true then it's dead end for Intel, not like they got many choices anyway.

 

[fastandfurious6]

I really want to read intel senior architects' troubles, the stuff that only becomes public after many many years, like what happened with royal core? who were shafting them all the time? and all that stuff

no doubt terrible decisions kept being made over and over again

 

[511]

I really want to read intel senior architects' troubles, the stuff that only becomes public after many many years, like what happened with royal core? who were shafting them all the time? and all that stuff

royal was bad architecture i am surprised it even got green lit

 

[fastandfurious6]

what do you know about it are you a covert deep-intel senior architect??

 

[igor_kavinski]

what do you know about it are you a covert deep-intel senior architect??

He works there as a security guard

 

[Geddagod]

royal was bad architecture i am surprised it even got green lit

such a bad architecture that the people behind it left Intel, formed their own company, raised >20 million dollars, and got Jim Keller on the board, despite Jim Keller also being the CEO of a different high performance CPU RISC-V company? (Tenstorrent).
I also want to add- this didn't happen when ocean cove got canned. This didn't happen when RYC 1.0 didn't get productized. So obviously the folks at AADG didn't just get up and leave every time things didn't go their way...

 

[511]

such a bad architecture that the people behind it left Intel, formed their own company, raised >20 million dollars, and got Jim Keller on the board, despite Jim Keller also being the CEO of a different high performance CPU RISC-V company? (Tenstorrent).

both are RISC-V

I also want to add- this didn't happen when ocean cove got canned. This didn't happen when RYC 1.0 didn't get productized. So obviously the folks at AADG didn't just get up and leave every time things didn't go their way...

have you checked their bio most of them spent 35 years at Intel maybe they were done with Intel and it was their project they wanted to continue even if it got canned they wanted to complete it

 

[511]

He works there as a security guard

Intel Ultra Security

 

[igor_kavinski]

have you checked their bio

Debbie worked on Haswell. You know, the architecture that Broadwell had trouble beating. My office work PC is still i7-4770. Until two years ago, the COO had i7-4790 (now on Alder Lake). She had legit chops to do a kickass CPU. But Pat took away her funding. And now Pat probably wishes he was a doormat outside her office.

 

[511]

Debbie worked on Haswell. You know, the architecture that Broadwell had trouble beating. My office work PC is still i7-4770. Until two years ago, the COO had i7-4790 (now on Alder Lake). She had legit chops to do a kickass CPU. But Pat took away her funding. And now Pat probably wishes he was a doormat outside her office.

Pat gave her funding to Chadmont you know that right so between Haswell and Chadmont which is better compared to their respective predecessor.
I also had a Haswell I3-4005u laptop btw

 

[igor_kavinski]

Pat gave her funding to Chadmont

That makes no sense. Royal Cove was not an E-core.

 

[Geddagod]

both are RISC-V

Exactly, so why would Jim Keller waste his resources joining a company that is trying to do the exact same thing the company he is already the CEO of is trying to do?
Unless the company showed promise, and Keller wanted to get on the ground floor, either to buy them out later, or just to profit if they succeed later...

have you checked their bio most of them spent 35 years at Intel maybe they were done with Intel and it was their project they wanted to continue even if it got canned they wanted to complete it

The founders there prob spent a good bit of their own money, and their job security, to work on an architecture that was so bad that you struggled to understand why it got greenlit?

 

[igor_kavinski]

She is talking trash about x86: https://www.aheadcomputing.com/post/a-seismic-shift-in-the-computing-ecosystem-brings-opportunity

Goes to show how much she is hurt by what Pat did to her. She's gonna take on x86 with a vengeance.

 

[DZero]

She is talking trash about x86: https://www.aheadcomputing.com/post/a-seismic-shift-in-the-computing-ecosystem-brings-opportunity

Goes to show how much she is hurt by what Pat did to her. She's gonna take on x86 with a vengeance.

The issue? x86 is the USA and mostly the west, if Intel falls, x86 gets hurt, Microsoft gets a critical hit and nVIDIA would jump full ARM.