You'd think with the Billions of $ Profits Intel makes every year, they would have more fabs under their belt.

cryohellinc:

I'll just leave this one here.

I haven't heard anything about high defect rates (poor yield) on TSMC 7nm. Shouldn't it be okay? Customers seem to be happy and it's running at 100% capacity, which limits how much AMD (and others) can sell their products. If the wafer cost is supposedly high (which I haven't heard anything about either), it's probably due to the demand. It certainly hasn't affected AMD's ability to beat Intel in price/performance.

Kaarme:

I haven't heard anything about high defect rates (poor yield) on TSMC 7nm. Shouldn't it be okay? Customers seem to be happy and it's running at 100% capacity, which limits how much AMD (and others) can sell their products. If the wafer cost is supposedly high (which I haven't heard anything about either), it's probably due to the demand. It certainly hasn't affected AMD's ability to beat Intel in price/performance.

The question there is the cut and profit per SKU. I personally believe Intel can easily pay a bit more for fab per SKU and still make a profit, since their CPUs have a cheaper node than AMD's, and up until recently they charged a lot more for essentially the same product they're selling now, so they already made a lot of profit. And buying additional fab capacity is more than logical if you can't keep up with demand... you already know you'll get your money back if you can feed the market.

cryohellinc:

I'll just leave this one here. [spoiler]https://i.imgur.com/TRFcB8T.jpg[/spoiler]

This virgin pleb is loving his 3950x

Brit90:

You'd think with the Billions of $ Profits Intel makes every year, they would have more fabs under their belt.

I read that intel was transitioning some fabs to new density processes, hence why reduction in their total fab capacity.

This would be a bit ironic.

Well thy not? Seems a good option for chipsets etc!

moo100times:

This virgin pleb is loving his 3950x

Joining the virgin club with my 3700x.

it's extremely curious why someone would post false and outdated propaganda regarding TSMC's 7nm production. this production line is booked fully. and even with the disappointments of iPhone X and XI, AMD and Qualcomm picked up the slack. and for the record, "chiplet"/SoC cpus are the future, including Intel. AMD is just faster at getting there. Furthermore, TSMC is about to unveil the 7nm+. and circuit density is no longer an Intel advantage.

I wonder why Intel didn't use GloFo earlier. They want business and they can make 14nm parts. I could see a strategic advantage of Intel going to TSMC (since that basically slows down their production of Intel's competitors) but that also means Intel isn't going to get as much product out the door either. Intel is becoming a less and less appealing option, so, right now they need to focus on keeping up with quantity, where they're the only option people can buy. As another thought... couldn't Intel just simply take some of their previous-generation CPUs all the way down to Skylake and just rebrand them as current-gen parts?

Brit90:

You'd think with the Billions of $ Profits Intel makes every year, they would have more fabs under their belt.

I don't think Intel was expecting demand of 14nm products to go up as much as they have, not to mention, Intel wasn't expecting to still be on 14nm by 2020.

tunejunky:

it's extremely curious why someone would post false and outdated propaganda regarding TSMC's 7nm production. this production line is booked fully. and even with the disappointments of iPhone X and XI, AMD and Qualcomm picked up the slack. and for the record, "chiplet"/SoC cpus are the future, including Intel. AMD is just faster at getting there. Furthermore, TSMC is about to unveil the 7nm+. and circuit density is no longer an Intel advantage.

Right, the comparison is apples to oranges. Intel and AMD have different architectures that run at different clock speeds but by looking at say, the 3900x and the 10900x AMD is comparable in performance and still keeping it cheaper. Intel was caught napping and AMD was good enough at keeping the Zen architecture with help from Jim Keller, a secret from Intel that for so long with intel's market dominance, they were slow to react. With the Brian Krzanich incident, the many vulnerabilities exploited in their processors, shortage of 14nm capacity (the current US tariffs do not help this either), Intel was desperate for change so much that they hired Jim Keller himself. It is troubling that only a few chip designers can help make an electronics technology company competitive again. I hope these two stay competitive for a long time and hope that Nuvia chip startup goes somewhere.

schmidtbag:

I wonder why Intel didn't use GloFo earlier. They want business and they can make 14nm parts. I could see a strategic advantage of Intel going to TSMC (since that basically slows down their production of Intel's competitors) but that also means Intel isn't going to get as much product out the door either. Intel is becoming a less and less appealing option, so, right now they need to focus on keeping up with quantity, where they're the only option people can buy.

I think they were hoping to get their 10nm issues resolved before resorting to a third-party. I don't think anybody at Intel had planned on staying on 14nm for this long and these chronic shortages are a testament to Intel's (stubborn) faith in their own capabilities. Also, I seriously doubt they would give TSMC more business (providing more business to their main rival in process technology would only exacerbate their problems).

D3M1G0D:

I think they were hoping to get their 10nm issues resolved before resorting to a third-party. I don't think anybody at Intel had planned on staying on 14nm for this long and these chronic shortages are a testament to Intel's (stubborn) faith in their own capabilities. Also, I seriously doubt they would give TSMC more business (providing more business to their main rival in process technology would only exacerbate their problems).

I thought I mentioned that but I guess I deleted it. Though, TSMC isn't exactly a competitor to Intel. TSMC is a tool to Intel's competitors. So - to give TSMC more business means less opportunities to their competitors. It is a smart move to overflow TSMC's contracts in a competition standpoint.

Are they just trying to make AMD CPUs more expensive? AMD might need to raise prices if they suffer from supply issues.

schmidtbag:

I thought I mentioned that but I guess I deleted it. Though, TSMC isn't exactly a competitor to Intel. TSMC is a tool to Intel's competitors. So - to give TSMC more business means less opportunities to their competitors. It is a smart move to overflow TSMC's contracts in a competition standpoint.

Intel's traditional strength has been their fabs, which TSMC is now rivaling (or exceeding). Feeding TSMC more money will only strengthen their rival in fab technology, leading to a situation where their own fabs won't be able to keep up while they need to continue buying out capacity at a rival fab. Furthermore, they would need to buy out most/all of the 7nm capacity when they have no products on that node (what are they going to do? make up a phantom product?). Overpaying for contracts for a node that they don't have products for, just to try to block out a competitor, seems like a sure path to failure.

D3M1G0D:

Intel's traditional strength has been their fabs, which TSMC is now rivaling (or exceeding). Feeding TSMC more money will only strengthen their rival in fab technology, leading to a situation where their own fabs won't be able to keep up while they need to continue buying out capacity at a rival fab. Furthermore, they would need to buy out most/all of the 7nm capacity when they have no products on that node (what are they going to do? make up a phantom product?). Overpaying for contracts for a node that they don't have products for, just to try to block out a competitor, seems like a sure path to failure.

I'm not understanding your thought process here: Intel's traditional strength was marketing and creating desirable platforms. Customers don't give a crap about what factory a product was produced in. Nobody buys an Intel chip because of their fabs. Intel owning their own fabs is very much a strategic advantage (as it is for Samsung too), but it's not the reason for their strength. Intel's control over their own facilities allows for optimizations for their architectures, and maximizes profits (as if Intel really needs that to be any higher...) since they don't have to pay the fees of a 3rd party. But if Intel's fabs were so crucial to their strength, they wouldn't be able to go to TSMC or GloFo. Meanwhile, Intel's fabs today seem to be their greatest weakness - they are several years behind. Intel has plenty of money to burn on 3rd party facilities. As of today, the only thing that Intel cares about is making sure they don't lose sales to their competition. If a product of theirs is unavailable but a competitor's isn't, that's a loss for Intel because that means they're losing marketshare. Spending extra to make sure they are still an available choice is worth the money in the long run. I didn't say Intel was using TSMC's 7nm node. TSMC has finite resources regardless of the node being used, whether that be the raw materials or possibly even employees. Intel isn't a small customer.

schmidtbag:

I'm not understanding your thought process here: Intel's traditional strength was marketing and creating desirable platforms. Customers don't give a crap about what factory a product was produced in. Nobody buys an Intel chip because of their fabs. Intel owning their own fabs is very much a strategic advantage (as it is for Samsung too), but it's not the reason for their strength. Intel's control over their own facilities allows for optimizations for their architectures, and maximizes profits (as if Intel really needs that to be any higher...) since they don't have to pay the fees of a 3rd party. But if Intel's fabs were so crucial to their strength, they wouldn't be able to go to TSMC or GloFo. Meanwhile, Intel's fabs today seem to be their greatest weakness - they are several years behind.

As I said, Intel's traditional strength was its fabs - they had the most advanced fabrication plants in the world and were able to create fast/powerful chips as a result. Also, fabs are not a cost-saving measure for maximizing profit - if anything, they are extremely expensive to run and operate (AMD had to spin off their fabs because they could no longer afford them). Intel poured billions into their fabs because they needed them to maintain a competitive edge against their rivals. One of the reasons why AMD was able to make a strong comeback was because of TSMC. It is in Intel's interest to see TSMC brought down, not lifted up. Trying to kill off AMD by buying up capacity at TSMC would do more harm than good - it will tank their profits while practically guaranteeing a competitive edge for TSMC going forward. The best thing they can do in this situation is use GlobalFoundries, which directly competes with TSMC - a stronger GF can keep TSMC in check and siphon off some of their customers (it's better to have two mediocre foundries splitting the profits and growing slowly compared to one strong foundry taking in most of the profits and growing rapidly).

schmidtbag:

Intel has plenty of money to burn on 3rd party facilities. As of today, the only thing that Intel cares about is making sure they don't lose sales to their competition. If a product of theirs is unavailable but a competitor's isn't, that's a loss for Intel because that means they're losing marketshare. Spending extra to make sure they are still an available choice is worth the money in the long run.

As I mentioned elsewhere, Intel's focus going forward is attaining a larger slice of the overall silicon market (they are no longer focused on dominating the CPU market). This necessarily means they will be competing with other companies who produce silicon so keeping companies like TSMC at bay will be extremely important - if their fabs can't keep up then they have no hope of expanding.

D3M1G0D:

As I said, Intel's traditional strength was its fabs - they had the most advanced fabrication plants in the world and were able to create fast/powerful chips as a result. Also, fabs are not a cost-saving measure for maximizing profit - if anything, they are extremely expensive to run and operate (AMD had to spin off their fabs because they could no longer afford them). Intel poured billions into their fabs because they needed them to maintain a competitive edge against their rivals.

As you yourself have acknowledged in your previous post: it's cheaper to own your own than to go through a 3rd party. So, it is a cost-saving measure. Yes, Intel did pour billions into their fabs to maintain a competitive edge, but how are you not seeing my point here? Intel needed to make more chips than 3rd party facilities could keep up with. That increased demand would yield and increased fee on their end. It was worth it for Intel to make their own fabs because they could pump out more chips than their competition, because like I said: getting more products available to be sold and increasing marketshare is an easy way to remain successful and profitable, especially when you're not paying a 3rd party factory.

One of the reasons why AMD was able to make a strong comeback was because of TSMC. It is in Intel's interest to see TSMC brought down, not lifted up. Trying to kill off AMD by buying up capacity at TSMC would do more harm than good - it will tank their profits while practically guaranteeing a competitive edge for TSMC going forward. The best thing they can do in this situation is use GlobalFoundries, which directly competes with TSMC - a stronger GF can keep TSMC in check and siphon off some of their customers (it's better to have two mediocre foundries splitting the profits and growing slowly compared to one strong foundry taking in most of the profits and growing rapidly).

I see your point, but Intel has no ability to slow down TSMC. Think of Intel's options: A. Only stick to their own fabs despite they can't keep up with demand. B. Fund competing 3rd party fabs to increase Intel's product yield, and, reduce the available resources of their competition. Option B grants 2 wins and 1 loss, and they'll be losing against TSMC anyway even if they choose option A. Remember - this isn't a hypothetical situation. Intel has already contracted TSMC to pick up some of the load. They wouldn't do that if it didn't mean a long-term strategic advantage.

if their fabs can't keep up then they have no hope of expanding.

Intel has the money to open up a new facility if they really wanted to, but as you said: they're trying to get out of 14nm and onto 10nm, so they're hoping production isn't going to be suffering much longer until that transition can be completed. Yes, Intel is expanding beyond their classic product range, but they're not dumb enough to deliberately doom their expansion. Seeing as they don't appear to be building a new facility (right?), that implies they theoretically do have the capacity to keep up with their load, they just need to temporarily lean on 3rd party facilities.

Lol. Intel is making CPUs on AMDs (former) fabs. Not sure if it's ironic, but it's definitely something.

Having your own fab can be extremely expensive if you fail at getting a new, smaller node to produce results and if you can't keep the whole facility running at a decent level of capacity. The same as a hotel needing enough rooms occupied all the time to be profitable. Apparently GlobalFoundries was too afraid of trying to invest in 7nm to risk it. If they had failed, losing potentially bilions, it might have been the end for GF. Intel is so dirty rich that the partial failure with the 10nm node hasn't even really affected their bottom line. TSMC is also quite rich traditionally and went forward. Furthermore, everything worked perfectly for them.

schmidtbag:

As you yourself have acknowledged in your previous post: it's cheaper to own your own than to go through a 3rd party. So, it is a cost-saving measure.

I'm not sure where I indicated this - I said they were trying to resolve their 10nm issues before resorting to a third party.

schmidtbag:

Yes, Intel did pour billions into their fabs to maintain a competitive edge, but how are you not seeing my point here? Intel needed to make more chips than 3rd party facilities could keep up with. That increased demand would yield and increased fee on their end. It was worth it for Intel to make their own fabs because they could pump out more chips than their competition, because like I said: getting more products available to be sold and increasing marketshare is an easy way to remain successful and profitable, especially when you're not paying a 3rd party factory.

Intel's supply issues are a relatively recent phenomenon, and was brought about mostly due to their failure to transition to 10nm in time. Supply issues were not a concern when they originally built their fabs.

schmidtbag:

I see your point, but Intel has no ability to slow down TSMC. Think of Intel's options: A. Only stick to their own fabs despite they can't keep up with demand. B. Fund competing 3rd party fabs to increase Intel's product yield, and, reduce the available resources of their competition. Option B grants 2 wins and 1 loss, and they'll be losing against TSMC anyway even if they choose option A.

Improving their fabs can help resolve their supply problems - smaller node can allow them to produce more chips with more performance per watt. This is what they've been trying to do for the past five years or so - trying to transition to 10nm. They are only resorting to third-parties because they are unable to do this (assuming this rumor is true, of course).

schmidtbag:

Intel has the money to open up a new facility if they really wanted to, but as you said: they're trying to get out of 14nm and onto 10nm, so they're hoping production isn't going to be suffering much longer until that transition can be completed. Yes, Intel is expanding beyond their classic product range, but they're not dumb enough to deliberately doom their expansion. Seeing as they don't appear to be building a new facility (right?), that implies they theoretically do have the capacity to keep up with their load, they just need to temporarily lean on 3rd party facilities.

Swan indicated that they will increase capacity by 25% in 2020 (in addition to the 25% in the past year), although this doesn't necessarily mean they've built new facilities. I think one of the problems with this is that they are now creating bigger chips than they were building before - up to 8-core for desktop and 18-core for HEDT, a significant jump from the 4-core and 10-core chips from before Zen. I think this has added to their supply problems (so these increases are only bringing them back to where they were before).