TSMC reportedly won't make extra capacity for intel

[spartaman64]

Quote

Although there are reports of Intel placing orders at TSMC to fabricate their chips, a DigiTimes report today points to the partnership a temporary one. According to unnamed sources, TSMC doesn't consider Intel a long-term customer and is, therefore, unlikely to build additional fabrication capacity to meet the contracts.

Quote

The sources said that TSMC sees itself more as a "rescuer" than a long-term supplier. Intel typically manufactures its own silicon for its CPUs, rather than jumping to third-parties for help. Note that TSMC does already build some chipsets and FPGAs for Intel.

Nevertheless, TSMC's 7nm capacity is already quite booked, with the likes of AMD, Nvidia and Apple taking up much of the fabrication capacity. Some capacity will free up from the falling out of the TSMC-Huawei collaboration in September, so it wouldn't come as a surprise to see Intel jumping at the opportunity.

source: https://www.tomshardware.com/uk/news/report-intel-shortages-could-continue-until-2023-as-tsmc-wont-make-extra-capacity

 

Yeah this is about expected unless intel signs a contract with TSMC to use their process for many years they are not going to make more capacity eg getting more equipment, buildings, and workers for a temporary thing. If intel does decide to use tsmc for their processors there would probably be shortages.

[mariushm]

Yeah, there's no point in TSMC building more production lines / capacity (it would take 2-3 years and millions of dollars) for Intel to make some budget CPUs and chipsets at TSMC for a couple of years and then move back in house when they solve their production. They'd have a hard time recuperating their investment... and next year they're gonna be on 5nm anyway.

 

They freed some production capacity by not making chips for Huawei, but nvidia and amd probably already placed orders for the wafers available .... and amd / nvidia are not the only ones that need 7nm ... phone chips are huge market for example, fpgas, maybe camera sensors etc etc

[CTR640]

So in other words: Intel joined the party too late?

[rewird]

33 minutes ago, CTR640 said:

So in other words: Intel joined the party too late?

That's what happens when you spend too much time counting the beans

[Lord Vile]

1 hour ago, mariushm said:

Yeah, there's no point in TSMC building more production lines / capacity (it would take 2-3 years and millions of dollars) for Intel to make some budget CPUs and chipsets at TSMC for a couple of years and then move back in house when they solve their production. They'd have a hard time recuperating their investment... and next year they're gonna be on 5nm anyway.

 

They freed some production capacity by not making chips for Huawei, but nvidia and amd probably already placed orders for the wafers available .... and amd / nvidia are not the only ones that need 7nm ... phone chips are huge market for example, fpgas, maybe camera sensors etc etc

Would have thought apple would have taken that space for their new macs

[Haro]

It's like no one wants to work with Intel. 

Poor Intel..

[Haro]

6 minutes ago, Benji said:

don't even think it's that, why would a company decline money?

Not saying that that's the case. Just sounds like it. I know it most likely has nothing to do with this. In fact, my post was sarcastic really. 

Edited July 28, 2020 by TofuHaroto

[porina]

More clarification from Dr Wafer Eater:

 

The linked article is mostly behind a paywall though, which I don't have access to.

 

[porina]

1 minute ago, Benji said:

@porina If they're so advanced, why are they producing chipsets and so on at TSMC? Don't they have the capacity for it?

Not everything needs to be on cutting edge nodes.

If they're capacity limited still (I don't know if they are), moving stuff out will help their internal capacity.

Their capacity is below where they want it to be. We know about the 10nm problems. We now know they have a 7nm problem. This isn't likely to clear up any time soon.

[mariushm]

33 minutes ago, Benji said:

@porina If they're so advanced, why are they producing chipsets and so on at TSMC? Don't they have the capacity for it?

 

Chipsets don't need to be made at 7nm, or similarly small nodes like 12nm,14nm... they're made at 28/32/45nm ... some even 65nm

It's mostly a price issue ... let's say a factory can make let's 20.000 wafers (discs of silicon) per month, and you can cut only so many chips from a wafer... so it ends up costing you let's say $30k per wafer at 7nm  or it can cost you $10k to make a 28nm wafer.

Yeah, you get less chips because the structures are big (larger area per chip out of the wafer) but still gets cheaper... ex .. 10k / 2000 chips cut from wafer = 5$ per chip or something like that

The masks are also more expensive the lower the process node ... think of masks as the pattern or code you need to give the machines to make the silicon chip.

I'm not up to date with prices but think of it like costing a few hundred thousand dollars to prepare all data and code a 28nm/32nm process needs to starts churning out chips, and it can cost a couple of millions or even more to get a 7nm chip ready to be made... not to mention price of licenses for the software needed to design.

Converting existing chips to lower process nodes still costs tens to hundreds of thousands.

 

There's often little advantage to making the die small when you still need hundreds of pads on the chip for the "pins" going into motherboard.

They're already low power enough (tdp of around 5-8w for most chipsets) to only require passive cooling so a lower process would only save a bit of power

 

Modern processes like 7nm, 12nm, 14nm don't work well with high voltages like 3.3v, 5v etc and the chipset needs to work with 5v standby (the motherboard can have voltage regulators to lower the 5v to something lower like 3.3v or 2.5v but some parts of chipset still have to tolerate higher voltages) so it's also more convenient to use higher process nodes for these chips.

 

AM4 chipsets if I remember correctly were made at 55 nm ... and the io die that has the L3 cache that uses lots of space in the chip is made at 12nm or 14nm to save costs ... AMD can make lots of small chiplets (that each contains the groups of 4 cores/8 threads) on expensive 7nm and the cache memory and sata/usb controllers and memory/pci-e controllers can be made on cheaper 12nm or 14nm ... and AMD gets to use the best chipsets in EPYC and Threadripper processors that have higher profit margins

[mariushm]

4 minutes ago, Benji said:

Well, according to sites like Anandtech and other news outlets, Intels 300 chipset lineup apparently is being manufactured in 14nm, the only exception being the Z370 in 22nm and the early Z390 apparently also being manufactured in 22nm.

B365 is also made at 22nm...

Yeah, it's a bit different with Intel because they have their own fabs.

It's more convenient for Intel to turn off and upgrade some of their locations to newer processes ... so let's say they had 65nm and 45nm capabilities and 32nm in several locations, but in 2020 they may have very few locations that can do these or can do them at lower capacity or higher costs, and it makes sense to use those processes for chips with higher profit (fpga, military/auto grade chips etc)

They may have a lot of capacity on 22nm or maybe it just made more sense to Intel to prepare the chipset on 14nm and then remake it on 22nm  (less changes and faster turnaround to go to 22nm, versus 14->28/32)

 

[DildorTheDecent]

24 minutes ago, mariushm said:

AM4 chipsets if I remember correctly were made at 55 nm

The 300 and 400 series chipsets are 55nm. X570 is a repurposed Matisse I/O die on 14nm. The regular Matisse I/O is 12nm and the Rome I/O die is 14nm.

[spartaman64]

1 hour ago, porina said:

Not everything needs to be on cutting edge nodes.

If they're capacity limited still (I don't know if they are), moving stuff out will help their internal capacity.

Their capacity is below where they want it to be. We know about the 10nm problems. We now know they have a 7nm problem. This isn't likely to clear up any time soon.

saying your process is super advanced and more dense than everyone else's is all fine and good except they cant fucking make any LMAO

[porina]

13 minutes ago, spartaman64 said:

saying your process is super advanced and more dense than everyone else's is all fine and good except they cant fucking make any LMAO

Yes, it is a real problem. We're in interesting times as they try to transition into newer architectures even with their fabrication problems. We'll see if they can make up lost ground in the next two desktop generations. Not going to look for the exact quote, but they know it will take time to get back to parity, never mind try to push ahead again. That was before the recent 7nm news.

[Pickles von Brine]

5 hours ago, mariushm said:

Yeah, there's no point in TSMC building more production lines / capacity (it would take 2-3 years and Billions of dollars) for Intel to make some budget CPUs and chipsets at TSMC for a couple of years and then move back in house when they solve their production. They'd have a hard time recuperating their investment... and next year they're gonna be on 5nm anyway.

 

They freed some production capacity by not making chips for Huawei, but nvidia and amd probably already placed orders for the wafers available .... and amd / nvidia are not the only ones that need 7nm ... phone chips are huge market for example, fpgas, maybe camera sensors etc etc

Fixed

[CarlBar]

7 hours ago, spartaman64 said:

source: https://www.tomshardware.com/uk/news/report-intel-shortages-could-continue-until-2023-as-tsmc-wont-make-extra-capacity

 

Yeah this is about expected unless intel signs a contract with TSMC to use their process for many years they are not going to make more capacity eg getting more equipment, buildings, and workers for a temporary thing. If intel does decide to use tsmc for their processors there would probably be shortages.

 

No probably about it, TSMC doesn't begin to have anything close to the production capacity required.

[Uttamattamakin]

The key to this is TSMC picking up INTEL business just as it is losing Huawei.  Feels a little too perfect the timing on this.... 

[leadeater]

6 hours ago, Benji said:

Well, according to sites like Anandtech and other news outlets, Intels 300 chipset lineup apparently is being manufactured in 14nm, the only exception being the Z370 in 22nm and the early Z390 apparently also being manufactured in 22nm.

Intel usually uses a product and manufacturing sliding scale based on node progression targets which is where the problems comes from, chipsets are say n-1 from processors. While Intel is say developing 14nm fabrication technology they are making CPUs on 22nm and chipsets on 32nm, Intel has target dates for when new fabrication technologies are supposed to be ready and that is when products slide down. Manufacturing capacity is deployed based on these progression targets and older fabrication technology retired, Intel doesn't have customers and don't need to keep older technology around that suits the needs of different customers and product technology.

 

So what happens when targets are missed but expect the problems to be solved in a timely manor. Chipset production ends up getting moved down from 22nm to 14nm so chipset development targets like more HSIO/PCIe lanes, new USB revisions etc can be meet and prepare to retire 22nm/32nm by moving manufacturing off that. But the problems didn't get solved and the new manufacturing technology isn't ready yet but now products are sharing 14nm capacity, processors and chipsets, which isn't something normally done on a wide scale, there isn't enough capacity and now there is a problem. Short term solutions become big problems when they are no longer short term.

 

Chipsets that were being made on 14nm got moved back to 22nm. Intel stopped publishing manufacturing node details for chipsets after around Z270, or removed it.

 

For Intel the cycle is supposed to be develop, adopt, retire, replace. When develop doesn't finish steps retire and replace can't happen either, if those don't happen you can't update manufacturing facilities because the equipment that was supposed to go isn't gone.

[leadeater]

1 hour ago, CarlBar said:

No probably about it, TSMC doesn't begin to have anything close to the production capacity required.

Not just that but strategically it would be a very bad idea for Intel to adopt TSMC 7nm for a leading processor design, that does no good when 10nm or 7nm of their own is ready as it won't just be a simple matter of moving back to their own fabs. They are either committing to an entire product generation on TSMC or intentionally releasing something with a much shorter life-cycle with zero ongoing support.

 

Would you buy any of those CPUs? I know I wouldn't.

[CarlBar]

2 hours ago, leadeater said:

Not just that but strategically it would be a very bad idea for Intel to adopt TSMC 7nm for a leading processor design, that does no good when 10nm or 7nm of their own is ready as it won't just be a simple matter of moving back to their own fabs. They are either committing to an entire product generation on TSMC or intentionally releasing something with a much shorter life-cycle with zero ongoing support.

 

Would you buy any of those CPUs? I know I wouldn't.

 

I read that comment in terms of intel moving to TSMC permanently. But that could be my own bias talking as thats the only way i could see them doing it, though even then the scale issue mentioned would largely require intel to license the process for their own fabs rather than have TSMC do the work i their fabs.

[straight_stewie]

This could be a non-starter: It might be corporate espionage.

In order for Intel to make the most use of TSMCs 7nM offerings, TSMC has to give Intel quite a bit of information not only about their capabilities, but about how it works, and how Intel needs to interface with it.

This can give Intel very useful insight into how to fix their 7nM production problems. Possibly enough insight to fix the problems completely.

 

Intels plan might be to get TSMC to produce some small production run new generation 7nM processors for them, just so that they can steal TSMCs technology, rebuild it for themselves, and fix their company.

[leadeater]

31 minutes ago, CarlBar said:

I read that comment in terms of intel moving to TSMC permanently. But that could be my own bias talking as thats the only way i could see them doing it, though even then the scale issue mentioned would largely require intel to license the process for their own fabs rather than have TSMC do the work i their fabs.

Short or long term I don't think it actually changes much, longer term just won't be a thing because that basically means Intel exiting leading process technology all together. I think when people are saying long term they mean 2-5 years which actually isn't long term, that's single generation product life cycle. Process technology is so pivotal to Intel's architectures and designs I cannot see how it is actually possible to effectively use TSMC for processors without a giant list of potential and ongoing problems for both TSMC and Intel.

 

Node numbers are irrelevant, Intel's 14nm is market competitive right now which is just how good it has been and still is. There is no way Intel would have had such high performance products as well as low power laptop products as they have been over the multiple generations that have used 14nm using any other fab other than their own 14nm, and the same applies to 22nm. None of Intel's past success could have been done on anything other than their own, it's not a case of could they use someone else's it's the issue that what would have been achievable would be lesser. Sure Intel's 14nm is at the end of the road now but that is largely due to everyone else catching up.

 

This is why I say it is such a big problem for Intel to use TSMC for a leading processor design, there is a big assumption there that it would actually result in better processors than on Intel's 14nm. Even if it does that makes that product a dead end because there is no way Intel is going to give long term support to it without committing to using TSMC across multiple generations which I highly doubt will happen because that would call in to question if Intel should continue developing leading fabrication technology at all.

 

TSMC won't license 7nm to Intel and even if they did it's unlikely Intel could even do anything with it, is Intel supposed to just throw out what they have? Can the equipment they have be used? (probably). Where is Intel going to do their 7nm development? Are they?

[leadeater]

7 minutes ago, straight_stewie said:

7nM

<nitpick>The SI unit symbol for meter is a lower case m not an upper case</nitpick>

 

Dunno why but that capitol M annoys me more than it should lol

[leadeater]

1 minute ago, gabrielcarvfer said:

M stands for mega, so he's talking about a dimensionless milli in a weird way.

I'll have you know Intel has the best nanomegas of any company 

[straight_stewie]

41 minutes ago, leadeater said:

Dunno why but that capitol M annoys me more than it should lol

More than a thread that refuses to acknowledge a blatant industrial espionage attempt when it sees it?