TSMC talks about the challenges in supersizing chip wafers
For those of you who have never seen how a computer chip is made, they start out their lives as round silicon “wafers.” The chips are etched into the wafer, with all of the square(ish) chips laid out in rows, and at the end of the process the wafers are cut into individual chips. As a result, the larger the wafer is, the more chips can be made with one pass through the manufacturing process, while also generating proportionately less wasted space per chip (the wasted space is the parts of the round edges of the wafer that are left blank because a square chip won't fit). As with other electronics, being able to make more things with less waste brings down the cost – so obviously bigger wafers are a good thing.
Unfortunately, it’s no small challenge to move up from the current 300mm wafers to 450mm wafers (for the metrically-challenged, that’s 18 inches). TSMC is one of the companies trying to tackle the problem head on, having just secured permission from the Taiwan government to build a new plant to produce the new biggie wafers at a projected cost of $8-10 billion dollars.
Intel is also making an aggressive push into 450mm wafer production, indicating that it will spend a similar amount (quoted to be $8 billion) to expand two of its U.S. manufacturing facilities to accommodate upsized chip production. TSMC Chairman Morris Chang told reporters that he thinks Samsung is also working hard to get ready for 450mm wafers.
How soon are we likely to see chipsets in our phones and tablets that utilize 450mm wafers? Maybe not for a while; at TSMC’s annual meeting Chang indicated that “the technology is not ready yet”, and he refused to be pinned down to a specific timeline as to when full scale manufacturing might start on the larger wafers. Chang did indicate that he expected technical difficulties that stem from working with 450mm wafers to last for the next half decade.
Unless Samsung, Intel, or someone else gets there first, don’t look for an immediate drop in the price of your favorite mobile devices (at least due to chip prices). But the long term outlook is rosy, and we should see ever more advanced chips that actually cost less than current chips do. At least by sometime in 2018.
1. remixfa (Posts: 14252; Member since: 19 Dec 2008)
considering how small a portion of the device is the cost of the chip, any savings materialized will be considered extra profit for the manufacturer of either the chip or the phone (or both) and we will not likely see a dramatic reduction. It might be able to make the low end handsets just a little more cheap (or profitable) but thats about it. Consumers wont "reap" any benefits of it.
2. schecter7 (Posts: 99; Member since: 20 Apr 2012)
Not really! 12 inches to 18 inches will result in over 2X number of the same sized chips on a wafer. Price per chip will be scaled down by a factor of ~2 - which is huge. It was basically the same reason behind the push for 12 inch (from 6- 8 inch wafer) few years back.
5. remixfa (Posts: 14252; Member since: 19 Dec 2008)
true. but your talking a chip is selling at large scale for about 40 bux or less. I think the S4 chip for example, is in the $20-30 range if i remember right. Its not going to crop the cost to manufacturer in half, nor is it going to cut the cost to the phone company in half.
Your basically just speeding up production with the method, which reduces labor and some manufacturing costs like upkeep per chip, but not the cost of supplies. Besides, those savings wont be immediate anyways as the cost of the new factory has to be factored in as well.
If it saves the manufacturer $10 per chip, they will pass the savings as probably $5 per chip to the phone maker, which raises the manufacturer's profit while doing the same with the phone maker. The end user is not very likely to see any change in price.
9. schecter7 (Posts: 99; Member since: 20 Apr 2012)
You're right. But if it happens to all the chips used to make a phone - maybe we can see $30/40 savings per phone ?
10. schecter7 (Posts: 99; Member since: 20 Apr 2012)
Oh one more thing about bigger wafer - better yield since you have a bigger core area. You know how crappy the yield is around edge area of the wafers.
3. ph00ny (Posts: 622; Member since: 26 May 2011)
Despite what others have said about samsung being the "small" fish in the pond, they obvious have morris chang's attention
4. ngo2dd (Posts: 896; Member since: 08 Jul 2011)
Beside the Samsung, apple and that one chinese company. Name who else use Samsung base chips? They are small compare to Intel and TSMC.
6. remixfa (Posts: 14252; Member since: 19 Dec 2008)
When your talking ARM space, Samsung is the world's largest chip maker by a large margin. There was an article about that a few weeks ago. When your talking "all" chips, TMSC, some Chinese company, and Intel rule the roost.
"Besides Apple and Samsung".... Umm.. between the 2 of them, they make up about 60% of the entire cell market. Think about that. :)
8. schecter7 (Posts: 99; Member since: 20 Apr 2012)
You need to back up your claims, dude. Samsung has a long way to go to rule the ARM space. By the way, don't keep assuming all samsung phones use their own chipsets. In the US all their galaxy phones are powered by Qualcomm (fabbed by TSMC) chipsets - like it or not - that's the fact.
11. ph00ny (Posts: 622; Member since: 26 May 2011)
i looked it up since your last response. There were few articles showing that they're the biggest multi core AP manufacturer
"This Strategy Analytics research shows that Samsung led the dual-core smartphone applications processor market in 2011, with 60% volume share, followed by Qualcomm, Texas Instruments and Nvidia"
13. schecter7 (Posts: 99; Member since: 20 Apr 2012)
LOL - I noticed a new word in your response- 'multicore'. So single core AP were on the 80% of the 2011 smartphones. And 72% of those have integrated baseband. Does that ring any bell?
15. ph00ny (Posts: 622; Member since: 26 May 2011)
I'm pretty sure i said multicore last time
Is single core even relevant outside of WP7?
16. schecter7 (Posts: 99; Member since: 20 Apr 2012)
It is actually very relevant since we are talking about MARKET SHARE in ARM space. 80% of the 2011 smartphones were single core.
17. schecter7 (Posts: 99; Member since: 20 Apr 2012)
And Samsung will never give you the breakdown of Snapdragon (TSMC), OMAP (mostly UMC), and Exynos based Samsung smartphones. You'd like to think all dual core samsung phones were using samsung chipsets. I'd think the opposite. To tell you the truth Exynos has 0% market share in the US. Sales in the rest of world won't make that up in my opinion.
12. ph00ny (Posts: 622; Member since: 26 May 2011)
ARM goes well beyond mobile devices
7. schecter7 (Posts: 99; Member since: 20 Apr 2012)
LOL - Meizu or something. It's kind of funny why the fans get so upset when in reality Samsung has 5% of foundry market share (most of it from Apple). Morris Chang's attention makes them happy though - LOL.
14. ph00ny (Posts: 622; Member since: 26 May 2011)
Intel is actually a small fish in the world of mobile processor business but they're a giant in the world of processors in general
TSMC is a big fish in both since their customers include both Qualcomm and Nvidia as well as AMD