Discussion Intel current and future Lakes & Rapids thread

[DrMrLordX]

I wonder if Kuo is getting something mixed up? Lunch vs volume?

I hope so. Launch vs. volume for Ice Lake-SP had a lengthy gap so we'll see.

 

[nicalandia]

I hope so. Launch vs. volume for Ice Lake-SP had a lengthy gap so we'll see.

Hey, at least the CPU Coolers are already available

Intel "Sapphire Rapids" CPU coolers are already on Newegg, while the shipment of new Xeon series is reportedly delayed to Q2 2023 - VideoCardz.com

CPU coolers for Intel Sapphire Rapids coolers listed on Newegg Intel Sapphire Rapids coolers may already be on Newegg, but shipments of new Xeon CPUs is reportedly delayed to Q2 2023. The CPU coolers for LGA4677 socket used by next-gen Xeon Scalable processors can already be found on Newegg. The...

videocardz.com

 

[moinmoin]

What I meant was that past is often looked as more positive it really was, just like how 14nm was said to be awesome and same with Skylake, when some of us didn't think Skylake was super impressive on the day of launch nevermind 5 years after release date. For me Skylake was "meh" and 14nm overall was disappointing. I didn't think 22nm was impressive either - after all they used all the performance gains for Atom anyway.

I think both Intel's big advantage and detriment was the isolation. All its foundry advances stood alone simply due to the fact even if somebody wanted nothing compared, since nothing comparable existed. During IFS 1.0 Intel was sitting on its high horse exactly because its foundry was the best and nothing compared, so external companies naturally were expected to adapt to Intel's internal quirks. That's why a newly competitive AMD using TSMC's competitive node was such a groundbreaking event. Barely anybody would have cared about TSMC's node if AMD's chips using it wouldn't directly show how much more efficient TSMC is with x86 is than Intel's so far. And that's why a major target of Intel is to reach performance-per-watt parity with TSMC by 2024.

Maybe Apple themselves had no idea of the volume. They must have given Intel some number that Otellini didn't consider worth their while. Also, being as proud as Intel is, he probably thought no one else could supply the volume Apple wanted.

Either it's too little or it's too much, can't be both at once.

They are having issues validating their whole stack, they have issued multiple Pre-production steppings to try to fix them, but they are still not production ready. Intel first try at Chiplets are harder than expected. If they iron these issues they will have better results for Emerald Rapids.

How I wish there to be a post-mortem for these kinds of stuff. There must be a lot of stories behind this many very likely unvoluntary steppings.

Hey, at least the CPU Coolers are already available

Intel "Sapphire Rapids" CPU coolers are already on Newegg, while the shipment of new Xeon series is reportedly delayed to Q2 2023 - VideoCardz.com

CPU coolers for Intel Sapphire Rapids coolers listed on Newegg Intel Sapphire Rapids coolers may already be on Newegg, but shipments of new Xeon CPUs is reportedly delayed to Q2 2023. The CPU coolers for LGA4677 socket used by next-gen Xeon Scalable processors can already be found on Newegg. The...

videocardz.com

View attachment 63198

It's like with Aurora: The supercomputer is done, everything is prepared, just need to plug in the SPR CPUs to start up.

Which reminds me, my impression so far is that SPR is doing worse than PV which as a package seems vastly more complex. Is PV actually faring better, or is it just SPR getting the news?

 

[Exist50]

Which reminds me, my impression so far is that SPR is doing worse than PV which as a package seems vastly more complex. Is PV actually faring better, or is it just SPR getting the news?

I've said it before, but any attempt to blame the current state of Sapphire Rapids on merely process or manufacturing issues is severely misguided.

 

[nicalandia]

It's like with Aurora: The supercomputer is done, everything is prepared, just need to plug in the SPR CPUs to start up.

Which reminds me, my impression so far is that SPR is doing worse than PV which as a package seems vastly more complex. Is PV actually faring better, or is it just SPR getting the news?

Word on the street is that many of the issues and delays is that they have not been able to "Hard" Lock on the Software Defined Silicon (SDSi) mechanism and it's getting bypassed rather quickly.

 

[moinmoin]

Word on the street is that many of the issues and delays is that they have not been able to "Hard" Lock on the Software Defined Silicon (SDSi) mechanism and it's getting bypassed rather quickly.

Oh, it's about that turd? Then I'm a big fan of that "feature" backfiring this gloriously.

 

[igor_kavinski]

Word on the street is that many of the issues and delays is that they have not been able to "Hard" Lock on the Software Defined Silicon (SDSi) mechanism and it's getting bypassed rather quickly.

If true, super dumb of them to keep trying to make it work. Just ship and worry about that feature in the next gen.

 

[Exist50]

If true, super dumb of them to keep trying to make it work. Just ship and worry about that feature in the next gen.

Yeah, there's no way that's what's stopping them from shipping, much less a year's delay.

 

[nicalandia]

Yeah, there's no way that's what's stopping them from shipping, much less a year's delay.

One of Many Issues... Sapphire Rapids can't catch a break

 

[nicalandia]

https://twitter.com/x/status/1537561772294492160

This should apply to Sapphire Rapids HBM right?

MCDRAM as High-Bandwidth Memory (HBM) in Knights Landing Processors: Developer’s Guide

MCDRAM as High-Bandwidth Memory (HBM) in Knights Landing Processors: Developer’s Guide

This publication is part of a developer guide focusing on the new features in 2nd generation Intel® Xeon Phi™ processors code-named Knights Landing (KNL). In this document we discuss the…

colfaxresearch.com

 

[jpiniero]

I've said it before, but any attempt to blame the current state of Sapphire Rapids on merely process or manufacturing issues is severely misguided.

It sure seems rather convenient.

Thought about it and I think the play is to never ramp Sapphire and cancel Emerald and pull in Granite Rapids if possible. Lets face it, anyone buying a 10 nm Xeon is only doing so because they can't (or won't) buy AMD. Make them buy Icelake.

This would mean they would have to have the foresight to dual source the CPU chiplet at TSMC though, which is total lol.

 

[moinmoin]

the play is to never ramp Sapphire and cancel Emerald and pull in Granite Rapids if possible.

There's no way that's going to fly, is there?

 

[Exist50]

Thought about it and I think the play is to never ramp Sapphire and cancel Emerald and pull in Granite Rapids if possible

No possible way that can/will happen. They just delayed it a year. Can't just turn that around and pull it back in again. If anything, continued issues with SPR mean GNR is more likely to be even further delayed. SPR will basically have to keep Intel's server efforts alive for the next two years.

 

[Karnak]

Saphire Rapids delayed again??

https://twitter.com/x/status/1537457755690532864

What would NV think if this is actually true? Switching back to Intel and then... welp.

 

[jpiniero]

What would NV think if this is actually true? Switching back to Intel and then... welp.

HGX is low enough volume that maybe nVidia would be able to get what they need? Off Label if need be.

 

[moinmoin]

HGX is low enough volume that maybe nVidia would be able to get what they need? Off Label if need be.

Lower volume than Aurora needs?

 

[RTX2080]

If SPR is proven to be delayed to 23Q2, these "leakers" are wrong, again

https://twitter.com/x/status/1530101010865463296

not surprise if these "release date" are all wrong

 

[DrMrLordX]

Hey, at least the CPU Coolers are already available

Is anything else going into Eagle Stream? Or just Sapphire Rapids/Emerald Rapids?

Which reminds me, my impression so far is that SPR is doing worse than PV which as a package seems vastly more complex. Is PV actually faring better, or is it just SPR getting the news?

Ponte Vecchio makes extensive use of chiplets sourced from TSMC. Coincidence?

Thought about it and I think the play is to never ramp Sapphire and cancel Emerald and pull in Granite Rapids if possible.

That would mean no new enterprise products until 2024. Assuming Granite Rapids doesn't suffer delays of its own.

Make them buy Icelake.

Intel can't ramp volume on IceLake-SP. Or, at least, they haven't yet. They're still shipping, uh, 6x as much Cascade Lake as their are IceLake-SP I think?

not surprise if these "release date" are all wrong

We already know that Raphael won't launch in August.

 

[jpiniero]

Lower volume than Aurora needs?

Aurora, as far as I can tell, is being installed.

 

[nicalandia]

That would mean no new enterprise products until 2024. Assuming Granite Rapids doesn't suffer delays of its own.

Granite Rapids has already suffered it's first delay

Intel delays 2023 server chip, says it needs to boost spending to catch competitors

Intel hopes to catch up to TSMC and Samsung in terms of the latest generation of chip manufacturing technology.

www.cnbc.com

 

[LightningZ71]

Does anyone think that Intel may have a refresh of Ice Lake-SP on Intel 7 / 10esf as a stopgap for Saphire Rapids/Emerald Rapids? Just managing higher clock speeds and better power/thermals would be a notable improvement for that product, which is something that 10ESF was supposed to help with. I know these are long lead time products, so it would have to already be well under way to be market relevant.

 

[nicalandia]

Does anyone think that Intel may have a refresh of Ice Lake-SP on Intel 7 / 10esf as a stopgap for Saphire Rapids/Emerald Rapids? Just managing higher clock speeds and better power/thermals would be a notable improvement for that product

Perhaps, but Ice lake and Cascade Lake are having a Terrible times at this stage, they are getting clobbered by Rome and Milan

"The 2022 Cloud Report evaluates the performance of Amazon Web Services (AWS), Microsoft Azure (Azure), and
Google Cloud Platform (GCP) for common OLTP workloads"

Thank You

www.cockroachlabs.com

 

[nicalandia]

Intel First 8 Core processor built on Intel 4 at just 2 mm^2

https://twitter.com/x/status/1536700762432098305

 

[moinmoin]

Aurora, as far as I can tell, is being installed.

Exascale Watch: Aurora Installation Underway, Now Open for Reservations

Installation has begun on the Aurora supercomputer, Rick Stevens (associate director of Argonne National Laboratory) revealed today during the Intel Vision event keynote taking place in Dallas, Texas, and online. Joining Intel exec Raja Koduri on stage, Stevens confirmed that the Aurora build is...

www.hpcwire.com

All told, Aurora will implement 60,000+ Ponte Vecchio GPUs and 20,000+ Sapphire Rapids CPUs, said Stevens. “We’ve already installed the Intel DAOS storage system, management nodes and cooling infrastructure, and we’ve put in place a validation system with Sapphire Rapids and PVC processors in place right now for applications testing. The whole team can’t wait until we get the system up and running.”

That reads to me that they already installed "the Intel DAOS storage system, management nodes and cooling infrastructure" and are now waiting for SPR and PVC to finally arrive.

 

[Ajay]

Granite Rapids has already suffered it's first delay

Intel delays 2023 server chip, says it needs to boost spending to catch competitors

Intel hopes to catch up to TSMC and Samsung in terms of the latest generation of chip manufacturing technology.

www.cnbc.com

Geez, this was announced back in February! Well, whatever else is going on, Intel is at least tacitly revealing that they know they are behind. What’s the saying, the beginning of wisdom is to admit your faults?