Discussion Intel current and future Lakes & Rapids thread

[Markfw]

GNR P cove version 8Ch version launched

Intel Announces Xeon 6500P + Xeon 6700P Processors - Phoronix

www.phoronix.com

No independent benchmarks.

 

[511]

No independent benchmarks.

Serve the home has one and I would say It's quite underwhelming the nodes can't do anything if your design is that bad.

I thought 9754 was 96C/192T but that is not the case it is 128C/256T these CPUs seems okayish not great vs Zen 5

Intel Xeon 6700P and 6500P Granite Rapids-SP for the Masses Initial Benchmarks and First Look

We take a look at the new Intel Xeon 6700P and 6500P processors codenamed "Granite Rapids-SP" to see what they offer

www.servethehome.com

 

[Saylick]

Big OOF if true. What's PTL's compute tile size again? Trying to understand what the equivalent D0 would be.

https://twitter.com/x/status/1894054674384523347

 

[511]

Big OOF if true. What's PTL's compute tile size again? Trying to understand what the equivalent D0 would be.

https://twitter.com/x/status/1894054674384523347

The 18A tile size was 114mm2 and this makes the rumor as false for PTL Die if we go by August D0 of 0.4.
How can he survey the industry if the customers are yet to tape out on 18A the only company with working 18A sample is Intel themselves

 

[Saylick]

The 18A tile size was 114mm2 and this makes the rumor as false for PTL Die if we go by August D0 of 0.4.
How can he survey the industry if the customers are yet to tape out on 18A the only company with working 18A sample is Intel themselves

Thanks. Yeah, that's a pretty small die so the yields should be much higher w/ D0 = 0.4.

 

[Darkmont]

This looks to be the platform for Coral Rapids

 

[Io Magnesso]

This looks to be the platform for Coral Rapids
View attachment 125720

I don't know when it will be, but future Rapids will be called Coral Rapids. I knew it for the first time
Did you miss it? I'm not conscious

 

[igor_kavinski]

Coral Rapids from Tech Yes City only. No one else seems to have mentioned it.

 

[Darkmont]

Coral Rapids from Tech Yes City only. No one else seems to have mentioned it.

MLID did in March 2024

 

[511]

This looks to be the platform for Coral Rapids
View attachment 125720

 

[DavidC1]

By the way, regarding yields and execution, the last time Intel got a non defeatured/delayed process was 32nm process.

That's right, 32nm was the last generation. Not 22nm. 22nm was delayed 6+ months. I read that there were tremendous yield issues but they pretty much got it fixed at the last minute. 14nm was delayed over a year, but nobody knows that, since 10nm was delayed 3+ years.

14nm likely got delayed for the same reason 10nm did, it was too ambitious. You see, until 22nm, in density wise compared to TSMC, they weren't true 22nm. Intel's 22nm was only about 30% more density compared to TSMC's 28nm. Meaning calling it 24nm would have been more accurate. On 14nm though, they decided to close the gap, only for the sake of mobile, since desktops didn't benefit from it anyway(other than GPU which sucked anyway).

If 22nm was a tremendously difficult step, what do you think happens to execution when you up the difficulty on a shrink of the process when every step gets harder and harder? And on 10nm, they decided to triple down, rather than learning from 22nm/14nm.

When BK got fired, he moved to a much smaller, lesser known company. He was rated similarly terrible by employees on comparably as when he was at Intel.

 

[511]

By the way, regarding yields and execution, the last time Intel got a non defeatured/delayed process was 32nm process.

That's right, 32nm was the last generation. Not 22nm. 22nm was delayed 6+ months. I read that there were tremendous yield issues but they pretty much got it fixed at the last minute. 14nm was delayed over a year, but nobody knows that, since 10nm was delayed 3+ years.

14nm likely got delayed for the same reason 10nm did, it was too ambitious. You see, until 22nm, in density wise compared to TSMC, they weren't true 22nm. Intel's 22nm was only about 30% more density compared to TSMC's 28nm. Meaning calling it 24nm would have been more accurate. On 14nm though, they decided to close the gap, only for the sake of mobile, since desktops didn't benefit from it anyway(other than GPU which sucked anyway).

If 22nm was a tremendously difficult step, what do you think happens to execution when you up the difficulty on a shrink of the process when every step gets harder and harder? And on 10nm, they decided to triple down, rather than learning from 22nm/14nm.

When BK got fired, he moved to a much smaller, lesser known company. He was rated similarly terrible by employees on comparably as when he was at Intel.

Well the problem was not technical but also fab had less funding as well during Kranzich/Otleni/Swan and the process were glued with tape during the OG 10nm they made several changes to OG 10nm to make it yield well.
The only reason he didn't get fired Sooner is he was licking shareholder's feet if he had been fired sooner and than Pat would have come they wouldn't be in this bad of a position.

 

[igor_kavinski]

The only reason he didn't get fired Sooner is he was licking shareholder's feet if he had been fired sooner and than Pat would have come they wouldn't be in this bad of a position.

NO

NO NO NO

Jim Keller was the only CEO that could've taken Intel to space and beyond!

 

[DavidC1]

Well the problem was not technical but also fab had less funding as well during Kranzich/Otleni/Swan and the process were glued with tape during the OG 10nm they made several changes to OG 10nm to make it yield well.
The only reason he didn't get fired Sooner is he was licking shareholder's feet if he had been fired sooner and than Pat would have come they wouldn't be in this bad of a position.

Originally, expected Pat's successor was Sean Maloney. Then he got a stroke and that went out the window, having BK succeed.

 

[Io Magnesso]

Originally, expected Pat's successor was Sean Maloney. Then he got a stroke and that went out the window, having BK succeed.

It was because of the stroke...
I'm still young, but I have to be careful about stroke...

 

[511]

Originally, expected Pat's successor was Sean Maloney. Then he got a stroke and that went out the window, having BK succeed.

You mean BK right cause Melony retired in 2012

 

[DavidC1]

You mean BK right cause Melony retired in 2012

I made a mistake. I meant Otellini's successor was Maloney. Also how would he be a successor to BK when he got in at 2013/2014?