Yes, this was always the plan and incredibly aggressive. It is usually 18 months from tapeout (what this article is talking about) to mass production. That would be the very end of next year. But Intel wants to not just start mass production, but have products on shelves by the end of next year.
Believe me? No. Pat? Maybe. But my comment was based on what they've achieved so far as noted in this article.
As for 10nm being "on schedule" the schedule kept getting pushed back. So, not really applicable to 18A and 20A. In case of doubt, here's a citation for 18A from the year 2021: https://www.anandtech.com/show/16823/intel-acceler...
If Intel is eager enough to hide defects across two generations of CPUs on a mature, high-volume node (Intel 7), what would Intel be willing to hide on their most critical node (Intel 18A) in the company's history?
Intel 7's hidden via oxidation is especially damaging to Intel Foundry.
I can't imagine a firm choosing Intel 18A and two years later, Intel quietly reveals to them, "Oh, right, some of your dies are badly damaged, but we didn't tell you until it was exposed by media. You better tell your customers ASAP, because they won't like that news. Guess you should've went with TSMC like we do for ~30% of our portfolio.
And we might've stolen a few of your DTCO & uArch optimizations. Don't be mad. Honestly, you choose to send us your: why shouldn't our design team be allowed to take a good look?"
>Intel 7's hidden via oxidation I'm sure all fabs have internal hiccups and problems to solve all the time, and they don't make press releases for each and every one of them. For all the reporting on this via oxidation issue, I haven't seen anything to suggest that it really raises above normal operating troubles.
Fabrication defects are normal. The problem is a systematic failure that affects all CPUs.
Uh, it is certainly a serious problem: Intel was forced to recall all inventory of the affected CPUs.
“Intel was able to confirm full removal of impacted processors in our supply chain by early 2024. However, on-shelf inventory may have persisted into early 2024 as a result.“
Intel didn’t do that for free. It was a mega-failure. These wafers should’ve never left the fabs.
There are LOTS of rumors that NVIDIA will be the first customer, and I totally believe. They are incredibly supplied constrained at TSMC. If they can have a successful tapeout and ramp-up of a GPU family at Intel - say their consumer "Blackwell" graphics cards - they could then use all their allocated capacity at TSMC for the high-profit AI accelerators.
I think Apple would be wise to try Intel, too, maybe for their data center chips where the volumes and stakes are lower than for the consumer facing products. Building up a credible alternative to TSMC seems like a very good idea for Apple (and everyone, really) given the geopolitical risks
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
14 Comments
Back to Article
James5mith - Tuesday, August 6, 2024 - link
"first 18A chips released in the latter half of near year."Next year?
NextGen_Gamer - Tuesday, August 6, 2024 - link
Yes, this was always the plan and incredibly aggressive. It is usually 18 months from tapeout (what this article is talking about) to mass production. That would be the very end of next year. But Intel wants to not just start mass production, but have products on shelves by the end of next year.ballsystemlord - Tuesday, August 6, 2024 - link
Believe it or not, 18A and 20A seem to be on schedule. No huge delays like 10nm complete with broken iGPUs.shabby - Wednesday, August 7, 2024 - link
Believe who? You? Pat? Gimme a break, 10nm was on schedule too for 5 years...ballsystemlord - Thursday, August 8, 2024 - link
Believe me? No. Pat? Maybe. But my comment was based on what they've achieved so far as noted in this article.As for 10nm being "on schedule" the schedule kept getting pushed back. So, not really applicable to 18A and 20A. In case of doubt, here's a citation for 18A from the year 2021:
https://www.anandtech.com/show/16823/intel-acceler...
ikjadoon - Tuesday, August 6, 2024 - link
If Intel is eager enough to hide defects across two generations of CPUs on a mature, high-volume node (Intel 7), what would Intel be willing to hide on their most critical node (Intel 18A) in the company's history?Intel 7's hidden via oxidation is especially damaging to Intel Foundry.
I can't imagine a firm choosing Intel 18A and two years later, Intel quietly reveals to them, "Oh, right, some of your dies are badly damaged, but we didn't tell you until it was exposed by media. You better tell your customers ASAP, because they won't like that news. Guess you should've went with TSMC like we do for ~30% of our portfolio.
And we might've stolen a few of your DTCO & uArch optimizations. Don't be mad. Honestly, you choose to send us your: why shouldn't our design team be allowed to take a good look?"
Dolda2000 - Tuesday, August 6, 2024 - link
>Intel 7's hidden via oxidationI'm sure all fabs have internal hiccups and problems to solve all the time, and they don't make press releases for each and every one of them. For all the reporting on this via oxidation issue, I haven't seen anything to suggest that it really raises above normal operating troubles.
ikjadoon - Friday, August 9, 2024 - link
Fabrication defects are normal. The problem is a systematic failure that affects all CPUs.Uh, it is certainly a serious problem: Intel was forced to recall all inventory of the affected CPUs.
“Intel was able to confirm full removal of impacted processors in our supply chain by early 2024. However, on-shelf inventory may have persisted into early 2024 as a result.“
Intel didn’t do that for free. It was a mega-failure. These wafers should’ve never left the fabs.
NextGen_Gamer - Tuesday, August 6, 2024 - link
There are LOTS of rumors that NVIDIA will be the first customer, and I totally believe. They are incredibly supplied constrained at TSMC. If they can have a successful tapeout and ramp-up of a GPU family at Intel - say their consumer "Blackwell" graphics cards - they could then use all their allocated capacity at TSMC for the high-profit AI accelerators.Blastdoor - Tuesday, August 6, 2024 - link
That makes sense.I think Apple would be wise to try Intel, too, maybe for their data center chips where the volumes and stakes are lower than for the consumer facing products. Building up a credible alternative to TSMC seems like a very good idea for Apple (and everyone, really) given the geopolitical risks
drwho9437 - Wednesday, August 7, 2024 - link
It is much more likely to be Microsoft, Amazon or Google for custom AI training chips.Terry_Craig - Tuesday, August 6, 2024 - link
Everything is balanced. One moment Intel is behind, the next TSMC is ahead. /sDrivebyguy - Tuesday, August 6, 2024 - link
@Terry Craig LOL!!!!!SydneyBlue120d - Wednesday, August 7, 2024 - link
I'm very curious to see if we'll ever see Qualcomm chips coming from Intel foundry....