Question Intel's future after Pat Gelsinger

[511]

Leave Swan alone! He had to be CEO coz they had no one else. He didn't take the job claiming he would turn Intel around.

Ok but he got handed the job he should have said so he can't accept it trying to make awful decisions the only decision he got right was outsource to TSMC.

Market demands more than stability these days.

Like look at my 245KF. I beat a 14900K's y-cruncher score with it. Took out the Patriot sticks and put in Teamgroup sticks and now I can't get D2D to go over 30 (it was 34 with Patriot). Like what the heck? What a crazy brittle platform sensitive to the smallest of things.

D2D shouldn't be affected with the DIMM maybe they share the same Rail for voltage cause the D2D to downclock?

 

[Io Magnesso]

There is no indication that anything was wrong with Alder Lake. It's probably the most stable Intel CPU in years. Everything bad started happening under Pat. "Somehow" Lunar Lake happened. I would love to know how.

Well, Alder Lake has no problems.
The version of Intel 7 process used in Alder Lake had no problems...
First of all, considering the design period of the CPU, let's assume that Alder has a design and verification period of 3 to 4 years as an assumption.
(Correction, Alder seems to be 30 months
Well, it's not wrong to say it's just over 3 years...
In addition, the process design period If you include , it may take 4 years, so it may not be wrong to guess that it is 3 to 4 years.)
If you think so, Alder Lake's design continued before Bobswan became the official CEO.
Raptor Lake was designed from Intel's official design while sharing each component and IP with Alder.
It seems that I was able to tape out in 2 years.
Considering it as two years, I think the plan itself was made during the period when Bobswan was appointed CEO.
There was a production lot with oxidation problems in part of the version of Intel 7 processes used to manufacture Raptor Lake
At least it seems that there is an oxidation problem in some lots, so I don't think there will be any problems with the ones that come out after that.
And Raptor Lake has a problem that has been discovered since the 2023 release Raptor Refresh comes out…
It was a Vmin shift problem
(At the same time as July 2024, when the VMIN SIFT problem was confirmed, the aforementioned oxidation problem was also announced.) Since it was announced at the same time, some people misunderstood that the VMIN SIFT problem = the oxidation problem.
At least these two issues seem to be separate things
Who the hell was Lunar Lake planned/designed when he was CEO? When it is said
Since 2020, at least Intel has reviewed the development system.
Instead of creating an IP to create a single product It is like designing each IP and design points in parallel and at the same time.
Simultaneously with chiplets, and shorten the period accordingly
Maybe Meteor and Arrow are also benefiting
Lunar Lake was originally scheduled to be released after Arrow Lake was released.
Released before Arrow Lake came out with a change in schedule
In fact, if you look at the key points of LUNAR LAKE, there is a place that has been reworked from METEOR/ARROW.
It's not perfect yet but Lunar seems to have improved cache coherence
Personally, I think LUNAR is the one that's progressed under Pat CEO.
Also, if I had to say it, Xeon6 might be straddled by Bob Swan and PAT...?
Sorry for the late reply and the delay in replying

 

[Io Magnesso]

Only those who have actually worked on the design will know...
From LUNAR LAKE, it seems that the interconnect has definitely been improved.
Xeon6 also looks like it's got a lot of work around the interconnect compared to Emerald and Sapphire.
No, not only interconnects but also cash management and bus topology…？
There was also a slide that was written as Scalable Fabric Gen 2 on the Lunar Lake slide.
Perhaps the completed form will be Nova or Panther at the earliest.

 

[igor_kavinski]

Perhaps the completed form will be Nova or Panther at the earliest.

I really hope Intel is able to flood the market with Panther Lake. We need better, affordable mobile CPUs!

 

[511]

FWIW Panther and Nova are pin to pin compatible

 

[Io Magnesso]

FWIW Panther and Nova are pin to pin compatible

Certainly, NOVA-HX and Panther had the same BGA socket, right? I've heard that story
Well, not all of them, but it's easy to share the design. It may be the advantage of being able to divert the motherboard.

 

[511]

Certainly, NOVA-HX and Panther had the same BGA socket, right? I've heard that story
Well, not all of them, but it's easy to share the design. It may be the advantage of being able to divert the motherboard.

I meam Panther Lake/Nova Lake H/P SKUs it's like ARL and MTL.
NVL-HX is different also NVL Halo

 

[reaperrr3]

Alder Lake was created under him then Pat destroyed that architecture with Raptor Lake's fab troubles.

BK was definitely one of their, if not the worst CEO(s).
https://www.tomshardware.com/tech-i...ial-media-comments-after-blistering-criticism

He's responsible for failing to get their 10nm mess under control in a more timely fashion, for failing to maintain at least architectural lead (5 years for the Skylake -> Ice Lake transition means the uArch team sat twiddling their thumbs for about 2-3 years, probably because it was tied too hard to process tech), hired Raja and let him spend too much money on dreamy dGPU efforts that never paid off, wasted money on stock buybacks and stupid acquisitions, failed to replace Murthy with someone more competent in time...

Pat just came in too late and was a bit too naive about some things.
Not saying he was all that good, but saying Pat was worse than BK just because BK's failures* didn't take full effect until later is shortsighted at best.

(*along with the BoD's failure to directly replace him with a competent tech CEO instead of poor inundated Bobbie, who only cost them even more time and money with those stock buybacks)

The individual CPU architectures aren't the CEO's responsibility anyway. At most he greenlights or cancels projects. The only difference between Alder Lake and Raptor was that Raptor was cranked up beyond healthy voltages (and put too many stops on the ringbus), but that was more overconfidence on the engineering side, not the CEO's fault.
And the only reason they even needed Raptor was because the failures of BK, Swan and Murthy prevented them from having at least a decent 7nm EUV process along with an updated architecture by that point.

"Intel 3" should've been ready and better than 10nm++ by the time Pat took over, along with an architecture better than Alder Lake, if his predecessors and their division leaders had done a better job.

 

[igor_kavinski]

Pat just came in too late and was a bit too naive about some things.

That's not what was expected from a seasoned CEO so yes that makes him worse than BK.

 

[511]

That's not what was expected from a seasoned CEO so yes that makes him worse than BK.

Time to take CEO Leadership at Intel you have my vote 😛

 

[igor_kavinski]

Time to take CEO Leadership at Intel you have my vote 😛

Hey, if that happens, regardless of the outcome, I will make more money than I could ever dream of.

And I may also steal their Tejas test platform for personal use

 

[Schmide]

My issues with the two.

Brian Krzanich - Started the software segmentation by keeping the moat around AVX512. The adoption never happened until AMD mainstreamed it. Now it's in a freaking PS3 emulator. Go figure.

Pat Gelisinger - Took segmentation to 11 with SDSi (software defined silicon). The walled garden was now an a la carte dim sum dish where you end up paying more for less. Intel makes good software, and sometimes at least for a while they made the best, but more often than not better directed implementations created by third parties ended up distilling the best implementation. If your library is the only code that can use the hardware, sooner or later you will be walled off in from your own garden.