Intel Data-Centric Innovation Summit

[maddie]

I am familiar w/ EMIB and it looks promising.
They claim it is cheaper than the industry standard 2.5D interposer methodology and they claim that it affords them far more flexibility with respect to connecting different process nodes on the same chip. Lots of claims about cost reductions/etc. ...............................................................................................

You're a very generous person.

For advanced topologies and connecting to non-adjacent chiplets, it seems to be a horrible solution.

In & out, in & out, in & out. In other words.

 

[ub4ty]

You're a very generous person.

For advanced topologies and connecting to non-adjacent chiplets, it seems to be a horrible solution.

In & out, in & out, in & out. In other words.

Thought I'd at least try to throw Intel bone here and there beyond my general roasting.
Was starting to attract some unfavorable heat.

https://semiengineering.com/using-silicon-bridges-in-packages/
As always, the broader industry tends to be right and have a more solidified process at cheaper costs.
Meanwhile, Intel had a pipe dream of cornering ever aspect of computing down to very detailed process level implementations that aren't its speciality.

The trend is clear with their mistaken proprietary approaches... A Meme w/ low volume sales and insane price premium.

The EMIB-based silicon itself is not expensive, although the cost of the entire process is difficult to quantify, according to officials from Intel. Plus, EMIB is a proprietary technology that is only available at Intel. - Article

Then, in R&D, Imec is developing its own silicon bridge technology with a twist—it’s not only an alternative to 2.5D, but it also enables a high-density, fan-out package.
Imec’s technology is similar to EMIB. The bridge, which has routing layers, connects one die to another. It also undergoes a flip-chip process. The technology enables 8 GBps per channel.
- Article

Competition from all angles. The coming years of computer hardware is going to be a Golden Age.

 

[inf64]

Next 3 years are going to be brutal for intel. I feel kinda sad for them, but then I remember how much MONEY they have been sitting on all this time, milking their customers in the process. I welcome this change, it is good for the market.

 

[witeken]

Next 3 years are going to be brutal for intel. I feel kinda sad for them, but then I remember how much MONEY they have been sitting on all this time, milking their customers in the process. I welcome this change, it is good for the market.

Please explain how a delay is good for the market??

If 10nm had been delivered as planned, Intel would've released a mainstream 8-core on a 4x denser process than Ryzen, one year before Ryzen's launch.

Intel had been utterly prepared for AMD's comeback, but their manufacturing team completely failed to deliver and let their CPU teams down. Tech is space is getting more boring with the slowing of Moore's Law, so I don't see how it's "good for the market".

 

[LTC8K6]

The delay brought back the competition, I think.

What happens to AMD and competition if "Intel would've released a mainstream 8-core on a 4x denser process than Ryzen, one year before Ryzen's launch."

Competition is good, overall.

Maybe not good for Intel, but good for buyers.

 

[maddie]

Please explain how a delay is good for the market??

If 10nm had been delivered as planned, Intel would've released a mainstream 8-core on a 4x denser process than Ryzen, one year before Ryzen's launch.

Intel had been utterly prepared for AMD's comeback, but their manufacturing team completely failed to deliver and let their CPU teams down. Tech is space is getting more boring with the slowing of Moore's Law, so I don't see how it's "good for the market".

I understand the need to have and promote a champion, but do you at all see the contradiction in that statement.

Maybe the statement should have read as below.

"Intel had fully planned for AMD's comeback, but their manufacturing team completely failed to deliver and let their CPU teams down."

 

[witeken]

The delay brought back the competition, I think.

What happens to AMD and competition if "Intel would've released a mainstream 8-core on a 4x denser process than Ryzen, one year before Ryzen's launch."

Competition is good, overall.

Maybe not good for Intel, but good for buyers.

AMD wouldn't have gone bankrupt if 10nm hadn't been delayed, so still don't see your point. Even Tesla which barely has had any profitable quarter ever is still in business. For this exact reason (Intel doesn't even want a monopoly - so competition is also good for Intel to alleviate monopoly concerns), even if Intel had this two process node advantage with Ryzen vs. the supposed 2016 Cannon Lake-S, we know Intel loves its segmentation and gross margin. So I'm not saying we don't need AMD for competition, AMD would've played the value card it always has, and AMD fans would've argued that 7nm would close the gap.

 

[witeken]

I understand the need to have and promote a champion, but do you at all see the contradiction in that statement.

Maybe the statement should have read as below.

"Intel had fully planned for AMD's comeback, but their manufacturing team completely failed to deliver and let their CPU teams down."

Well, if you look before 14nm, Intel had a very good track record. You don't expect to suddenly have a manufacturing catastrophe that causes a three year delay, don't you? You can try to plan ahead for any issues, but something as what happened with 10nm is certainly not something anyone could have foreseen, by definition. 10nm wasn't just a three or even six month delay like 14nm was. Even with a one or two year delay 10nm would've been fine, but the current three-four years are stretching it for sure.

For Intel its process lead has always been a differentiator, so if the manufacturing guys stop delivering on Intel's process excellence, then the CPU guys are left scratching their head. You can't plan around a loss of manufacturing lead as far as I'm aware.

 

[maddie]

Well, if you look before 14nm, Intel had a very good track record. You don't expect to suddenly have a manufacturing catastrophe that causes a three year delay, don't you? You can try to plan ahead for any issues, but something as what happened with 10nm is certainly not something anyone could have foreseen, by definition. 10nm wasn't just a three or even six month delay like 14nm was. Even with a one or two year delay 10nm would've been fine, but the current three-four years are stretching it for sure.

For Intel its process lead has always been a differentiator, so if the manufacturing guys stop delivering on Intel's process excellence, then the CPU guys are left scratching their head. You can't plan around a loss of manufacturing lead as far as I'm aware.

To be clear.

Someone cannot be "utterly prepared" and also say " but I didn't get this critical part completed".

A definitive oxymoronic statement.

edit: The above statement is being criticized not you.

 

[LTC8K6]

IIRC, here's no evidence that Intel planned to release 6 or 8 core chips, prior to Ryzen's appearance in the middle of the road on a foggy morning, as Intel drove placidly along it's 4 core road...

 

[jpiniero]

IIRC, here's no evidence that Intel planned to release 6 or 8 core chips, prior to Ryzen's appearance in the middle of the road on a foggy morning, as Intel drove placidly along it's 4 core road...

There was a rumor from a guy on LinkedIn suggesting a "4,6,8 core Cannonlake SoC". Given the density improvement of 10nm, I don't think that is much of a stretch to think that was true. A theoretical 8+2 (40 EU) Cannonlake might only be 100 mm2.

 

[witeken]

IIRC, here's no evidence that Intel planned to release 6 or 8 core chips, prior to Ryzen's appearance in the middle of the road on a foggy morning, as Intel drove placidly along it's 4 core road...

October 1, 2015: https://www.itworld.com/article/298...generation-could-be-going-to-eight-cores.html

 

[LTC8K6]

As I said, there's no evidence...

 

[PeterScott]

As I said, there's no evidence...

An Intel employee writing: "Intel Cannonlake SoC integrates 4/6/8 cores and Converged Coherent Fabric." is evidence. It just isn't proof.

 

[LTC8K6]

An Intel employee writing: "Intel Cannonlake SoC integrates 4/6/8 cores and Converged Coherent Fabric." is evidence. It just isn't proof.

It's evidence of several things, but not that Intel was planning to release a 6 or 8 core desktop chip in mid 2016...