Well, he's not technically wrong. You need the actual production metrics to say if he's full of shit though.
this post was submitted on 07 Dec 2024
95 points (97.0% liked)
Technology
59978 readers
3663 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 2 years ago
MODERATORS
Tldr it should be yield rate per area
It's not the right measure for progress. It's the right measure for viability. If yield is terrible, the end product may cost too much to market.
If yield rates aren't a good metric, what does he think is then? It's certainly not layoff numbers, or C-suite compensation.
If after all that investment you're only able to produce TEN PERCENT of the product successfully, that's a failure, by definition. Even if they quintuple the yields, that's still incredibly poor
Gelsinger was hired as a known long time engineer, rather than as a business expert. I would trust his numbers from an engineering perspective, even though I was laid off under his rule
All depends on the maturity of the process. 10% for a new design on a bleeding edge process is possibly viable. You'll then tweak the design and process to get the yield up.
Only exception would be if they can produce those wafers at 1/10th of the previous cost, but I highly doubt that's the case.
If it's 1/10 of the cost of purchasing them from TSMC, it's viable
Yield over die area should be the metric.
If you have a chip that is 50% of the wafer area, a single fault will lead to a yield of 50%. Now compare it with a chip that is 1% of the wafer area, the same single fault gets a yield of 99%.
So comparing the yields of two processes without factoring in the die area is not a fair game.
Gelsinger replied to a post by a prominent analyst, Patrick Moorhead, where it was initially claimed that Intel's 18A wasn't tested on a PDK 1.0 but rather an older design kit, which is why the yield rate figures are reported so low.