Intel’s "Processor Supply Shortfall" Could Mean i9-9900K Being Delayed to 2019

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
PCGamesN has a hunch that a paper launch is likely for Intel’s upcoming 9th-generation processors. Leaning on a DigiTimes article hinting at tight supplies, the author points out that Intel’s 14nm lines are probably inundated due to the delay of its 10nm process. With so many products still on 14nm silicon, the author warns that an i9-9000K may not even hit shelves until 2019.

...with the release of the rest of the 300-series chipsets on 14nm, and the upcoming Z390 motherboards also on the same node, there is increasingly more pressure being put on the production of 14nm silicon. Intel has already halted production of one entire chipset range because of it. And that’s going to have some serious consequences for the PC market in the second half of this year.
 
Fortunately, I'm close to spent out on computer parts for 2018, not that I was really that interested in an i9-9900K anyway.
 
AMD needs to take advantage of this now and get their full lineup of Ryzen and TR chips on store shelves ASAP.

If you had to pick the Ryzen 7 2800x to be a refined 12nm 8 core processor with speeds of 4.0Ghz base and 4.65Ghz OverDrive Boost (4.5Ghz nominal boost) or a dual-die (think Q6600) with each die with 6 active cores, and a 3.6Ghz/4.2Ghz 12-Core/24-Thread chip. Both retailing for $500.

Can two die's fit on the AM4 substrate to begin with?
 
Availability will do nothing for them if they don't also have spins and new SKUs with increased clockspeeds at the very least, and they need better IPC ASAP.

Honestly, it feels like they are just about their with IPC. I enjoy the 8 extra threads on my 2700x over the minuscule IPC increase I get on my 6700k.

I'd like the see a mid-cycle refresh with refined 12nm dies for Ryzen.

Like a Ryzen 2750x, 8-Cores with 3.9Ghz/4.5Ghz and OC potential to 4.7Ghz...one can dream....
 
Is this a surprise? My bet is that Intel paper launched the 9000 series specifically to buy time.

Given the benchmarks of the 9900K, it's definitely powerful enough to give mid-range TR chips a run for their money.

"Better and cheaper" than a 2950X, too bad you can't buy it.

It would be hilarious if Intel then backtracks out of soldering their CPU when it is actually finally launched.
 
Honestly, it feels like they are just about their with IPC. I enjoy the 8 extra threads on my 2700x over the minuscule IPC increase I get on my 6700k.

But your 2700x has a clock ceiling that your 6700k does/did not share. Single-core performance, which is what we're getting at when we talking about IPC and clockspeed, is a product of the two.

Now, granted, whether this makes a damn bit of a difference depends on use case. And that's subjective.
 
But your 2700x has a clock ceiling that your 6700k does/did not share. Single-core performance, which is what we're getting at when we talking about IPC and clockspeed, is a product of the two.

Now, granted, whether this makes a damn bit of a difference depends on use case. And that's subjective.

Yeah but lets be honest, aside from the niche high fps sector, there isn't really much of a benefit in real world gaming (and sometimes zen takes the cake anyway) between the two. Once you get to realistic resolutions for most [H]folk, it's all GPU. The RAM speed + timings boost and patches have nearly levelled the playing field already and they don't even have clock equality with Intel yet.
Rumours had it that the GF 7nm was going to be faster (hence the 5Ghz rumours) because it had cobalt etc and different design but they pulled out, so obviously something wasn't working right maybe like Intel dicking around with Cobalt and 50% yields..
So now we may get lower expected speeds with 7nm but higher densities of various parts, e.g. topping out 4.6-4.8 with most cooling setups vs 5Ghz.
 
But your 2700x has a clock ceiling that your 6700k does/did not share. Single-core performance, which is what we're getting at when we talking about IPC and clockspeed, is a product of the two.

Now, granted, whether this makes a damn bit of a difference depends on use case. And that's subjective.
IPC speed could be close or even with all the fixes and microcode updates to Intel's cpu's. I think reviewers really need to post what all has been updated, unless they are paid by Intel.

2019 is a ways away, how can you come out with a chip and have it delayed that much?
 
My tinfoil hat is buzzing at me.

If the i7-9700k launches on time and the i9-9900k gets delayed to an unknown time next year I'd take a hard look at the security vulnerabilities and ask yourself if Intel is just trying to avoid releasing an expensive chip with HT that might end up needing to be disabled soon.

It's one thing to ignore all the CPUs that are out before they know about it. It's another entirely to launch and sell $500 CPUs knowing about it.

Who knows. But I'm getting more conspiratorial about it with every new press release.
 
closest so far because the actual story sounds fishy

My tinfoil hat is buzzing at me.

If the i7-9700k launches on time and the i9-9900k gets delayed to an unknown time next year I'd take a hard look at the security vulnerabilities and ask yourself if Intel is just trying to avoid releasing an expensive chip with HT that might end up needing to be disabled soon.

It's one thing to ignore all the CPUs that are out before they know about it. It's another entirely to launch and sell $500 CPUs knowing about it.

Who knows. But I'm getting more conspiratorial about it with every new press release.
 
Well boys, just get your money ready and be waiting to order on release day. I will be down at Microcenter early to get mines. If you do that then nothing else matters.
 
I'd be disappointed if Intel's 9000 series didn't have higher average overclock headroom.

A jewel 8000k series chip can touch 5.4-5.6ghz in the 1.4-1.5 vcore range.

Now we just need more than 5% of retail chips to hit those numbers..
 
Last edited:
Honestly, it feels like they are just about their with IPC. I enjoy the 8 extra threads on my 2700x over the minuscule IPC increase I get on my 6700k.

I'd like the see a mid-cycle refresh with refined 12nm dies for Ryzen.

Like a Ryzen 2750x, 8-Cores with 3.9Ghz/4.5Ghz and OC potential to 4.7Ghz...one can dream....
I sold my 6700k for my 2700x and I feel exactly the same.

I used to defend ipc and efficiency. But really it's about multi cores now.
 
"And that’s going to have some serious consequences for the PC market in the second half of this year." -From article.

I doubt it. For the 9900k that will take some pretty hefty wallets and I doubt the spending stops with the CPU's. On most threads I rarely read of people with Intel upgrades unless they committed to blue and had something really ancient and under-performing. Usually the upgrades are AMD. I do see a number of 1st time builds with 'em and of course the prebuilts but even then that high end of a chip is going to be rare in those markets too.

I still have a ways to go. My 2600k is just barely starting to show its age(I currently see 40-60% in gaming) and my 4930k barely even wakes up for games(20-30%). I will say the 2700x is looking really attractive. The 9900k is more for bragging rights and professional work, not a gaming chip by any means and last I checked gaming was the majority of the pc market, besides the low end web browsing segments.
 
Yeah but lets be honest, aside from the niche high fps sector, there isn't really much of a benefit in real world gaming (and sometimes zen takes the cake anyway) between the two. Once you get to realistic resolutions for most [H]folk, it's all GPU. The RAM speed + timings boost and patches have nearly levelled the playing field already and they don't even have clock equality with Intel yet.
Rumours had it that the GF 7nm was going to be faster (hence the 5Ghz rumours) because it had cobalt etc and different design but they pulled out, so obviously something wasn't working right maybe like Intel dicking around with Cobalt and 50% yields..
So now we may get lower expected speeds with 7nm but higher densities of various parts, e.g. topping out 4.6-4.8 with most cooling setups vs 5Ghz.

Okay I found some numbers...
https://www.semiwiki.com/forum/content/7439-tsmc-technologies-mobile-hpc.html

N7 node has been making good progress in providing excellent PPA values with >3x density improvement, >35% speed gain and >65% power reduction over its 16nm predecessor.
N7 HPC track provides 13% speed over N7 mobile (7.5T vs 6T), while it has passed the yield and qual tests (SRAM, FEOL, MEOL, BEOL) and MP-ready D0. Part of the contributing factor is TSMC successful leveraged learning from N10 D0 and it is targeted for Fab15.
So it's 35% faster + 13% faster and 65% less power than TSMC 16nm. So depending on how AMD has got their architecture working, could be anything, we don't have previous numbers to go on. Arm got their stuff running up to 2.6GHz.
Assuming TSMC is 10% behind the process used for Zen, then this would put it well over 5Ghz. Reality makes me think we're looking at somewhere under that but who knows.

feature sizes tsmc 16nm.jpg
 
E-tailers love news like this, allows them to break out the old "supply and demand" rhetoric and ratchet the price to whatever they like.
 
But your 2700x has a clock ceiling that your 6700k does/did not share. Single-core performance, which is what we're getting at when we talking about IPC and clockspeed, is a product of the two.

Now, granted, whether this makes a damn bit of a difference depends on use case. And that's subjective.


don't forget those sweet sweet REAL security holes and possible performance tromping patches.
 
IPC speed could be close or even with all the fixes and microcode updates to Intel's cpu's. I think reviewers really need to post what all has been updated, unless they are paid by Intel.

2019 is a ways away, how can you come out with a chip and have it delayed that much?

Remember AMD has Spectre so it needs a lot of these fixes as well (even though it is even less likely to be compromised).

What it does not need is the Meltdown fixes. As a I am highly doubtful that this will make up enough ground to make the per core performance even.
 
AMD, you need to get Zen2 out the gate, ASAP. This is your chance.

AMD is on a yearly cadence, more or less. Zen+ was almost exactly a year after launch. Zen 2 will likely launch on mainstream in April 2019. TR Zen 2 a few months after. Epyc Zen 2 is supposedly sampling now, though.
 
As to the original topic, I expected this to be a paper launch. Intel did the same thing with the 8700k. October launch... but for the most part, availability didn't catch up until December/January. I figure we will see more of the same.
 
If you had to pick the Ryzen 7 2800x to be a refined 12nm 8 core processor with speeds of 4.0Ghz base and 4.65Ghz OverDrive Boost (4.5Ghz nominal boost) or a dual-die (think Q6600) with each die with 6 active cores, and a 3.6Ghz/4.2Ghz 12-Core/24-Thread chip. Both retailing for $500.

If base clocks increased only by 100MHz for 14nm-->12nm, then they cannot increase by 300MHz for 12nm-->12nm. And 50MHz higher or 100MHz higher 2800X is not going to change things when i9 are for sale.

Can two die's fit on the AM4 substrate to begin with?

No.
 
AMD is on a yearly cadence, more or less. Zen+ was almost exactly a year after launch. Zen 2 will likely launch on mainstream in April 2019. TR Zen 2 a few months after. Epyc Zen 2 is supposedly sampling now, though.

Zen EPYC comes in Q2* next year. Zen2 Desktop is expected for Q3 or Q4 next year, and Zen2 APUs are expected for 2020.


* Or Q3 depending of volume.
 
If base clocks increased only by 100MHz for 14nm-->12nm, then they cannot increase by 300MHz for 12nm-->12nm. And 50MHz higher or 100MHz higher 2800X is not going to change things when i9 are for sale.

Sort-of agree. Though the TR 2950 hitting 4.4 boost... I wonder if some well-binned chips sold as 2800Xs could do 4.5 boost, if AMD pushed for that. And maybe a +100MHz base. Not saying they will, or they should. Only that it might be possible.

It still wouldn't do much against the 9900k. That will dominate Zen+. But it could keep the 9700k from being much of a threat, especially since the lack of hyperthreading will probably create some curious inconsistencies with its performance vs the 8700k. I'm very interested to see how that's going to play out.

More than likely, AMD will just cut prices on the 2700X a bit, instead. Though I *highly* suspect the October launch for the 9900k and 9700k will be a paper launch, and we won't see widespread availability until December/January. It was that way with the 8700k.
 
PCGamesN has a hunch that a paper launch is likely for Intel’s upcoming 9th-generation processors. Leaning on a DigiTimes article hinting at tight supplies, the author points out that Intel’s 14nm lines are probably inundated due to the delay of its 10nm process. With so many products still on 14nm silicon, the author warns that an i9-9000K may not even hit shelves until 2019..

Just digitimes being digitimes. They make up rumors, call it analysis, then sell it... I have seen their stuff for ages, and they are wrong way more often than right.

Normal digi would warn of shortages because new process, but 14nm at Intel is very old hat so they have to get creative. So now it's "too many products on 14nm". :rolleyes: Is this the 4th or 5th year on 14nm? I would imagine they have their 14nm production/volume issues well covered by now.

Of course there is going to be some shortage at launch because there is always an initial rush on top end CPUs. Especially if they show some promise of stepping up performance significantly, which an 8 Core, soldered, high clock CPU does.
 
Just digitimes being digitimes. They make up rumors, call it analysis, then sell it... I have seen their stuff for ages, and they are wrong way more often than right.

Normal digi would warn of shortages because new process, but 14nm at Intel is very old hat so they have to get creative. So now it's "too many products on 14nm". :rolleyes: Is this the 4th or 5th year on 14nm? I would imagine they have their 14nm production/volume issues well covered by now.

Of course there is going to be some shortage at launch because there is always an initial rush on top end CPUs. Especially if they show some promise of stepping up performance significantly, which an 8 Core, soldered, high clock CPU does.

The supply problems on the 14nm line are real. Intel cannot satisfy the huge demand.

HPE just published an internal report to train its vendors to deal with the supply problem when customers ask for Xeon servers. The report even suggest vendors to recommend EPYC-based servers for those customers are in a tigh schedule and cannot wait for a Xeon server.

This supply problem is also the reason why we are seeing an increase of sales of AMD processors in stores as Amazon, Mindfactory,...
 
The supply problems on the 14nm line are real. Intel cannot satisfy the huge demand.

HPE just published an internal report to train its vendors to deal with the supply problem when customers ask for Xeon servers. The report even suggest vendors to recommend EPYC-based servers for those customers are in a tigh schedule and cannot wait for a Xeon server.

This supply problem is also the reason why we are seeing an increase of sales of AMD processors in stores as Amazon, Mindfactory,...
Yes, the huge demand. That's why.
 
PCGamesN has a hunch that a paper launch is likely for Intel’s upcoming 9th-generation processors. Leaning on a DigiTimes article hinting at tight supplies, the author points out that Intel’s 14nm lines are probably inundated due to the delay of its 10nm process. With so many products still on 14nm silicon, the author warns that an i9-9000K may not even hit shelves until 2019.

...with the release of the rest of the 300-series chipsets on 14nm, and the upcoming Z390 motherboards also on the same node, there is increasingly more pressure being put on the production of 14nm silicon. Intel has already halted production of one entire chipset range because of it. And that’s going to have some serious consequences for the PC market in the second half of this year.

It feels like yesterday when they paper launched the Coffee Lake R processors last October ahead of the Spectre/Meltdown disclosures! Never change.
 
This supply problem is also the reason why we are seeing an increase of sales of AMD processors in stores as Amazon, Mindfactory,...
Ehh, probably not. Maybe mindfactory, but a lot of Amazon sales are consumers, and they buy what they want when they can afford it. I don't see them going for a whole other platform just because they might have to wait a while for their beloved intel processor. If they are getting more sales, it's because their friends are saying "well, it's pretty close, depends on your workload", and consumers are recognizing the value, "well, if it's close and it costs this much less, why not?"
 
The real indicator of a supply shortage is if the big cloud providers are also seeing supply constraints. The Amazon/Microsoft/Facebook/Google's of the world develop their own server hardware for their enormous data center foot prints. The server OEMs like Dell, HPE, Lenovo etc. serve the smaller business crowd who haven't migrated to the cloud. End result is that who used to get preferential treatment are now second tier for supply.

Another variable is Cascade Lake which Intel still is officially holding to a 2018 introduction. Unofficially this product appears to have been delayed a few months to put some Spectre/Meltdown fixes into hardware. The big cloud providers mentioned above are likely getting Cascade Lake Xeons this year but probably no one else if I were a betting man. The transition to Cascade Lake also likely put kink in Intel's manufacturing. It takes months for a water to go into a factor and come out as a ready chip. Intel likely didn't stop Sky Lake-SP Xeon production but what is coming out of their factories I would fathom be going again to the big cloud providers. Supply should pick up once Intel moves all of their Xeon production lines over to Cascade Lake.
 
Ehh, probably not. Maybe mindfactory, but a lot of Amazon sales are consumers, and they buy what they want when they can afford it. I don't see them going for a whole other platform just because they might have to wait a while for their beloved intel processor. If they are getting more sales, it's because their friends are saying "well, it's pretty close, depends on your workload", and consumers are recognizing the value, "well, if it's close and it costs this much less, why not?"

You cannot purchase what is not available on the store. You don't need to go for another platform, just wait for the chip you want. Whereas you wait, other customers continue purchasing AMD, so the balance of Intel:AMD sales change.

The same happens in the server space. HPE is recommending customers to wait for availability of Xeon or order an EPYC-based server.
 
You cannot purchase what is not available on the store. You don't need to go for another platform, just wait for the chip you want. Whereas you wait, other customers continue purchasing AMD, so the balance of Intel:AMD sales change.

The same happens in the server space. HPE is recommending customers to wait for availability of Xeon or order an EPYC-based server.
What I mean is, there are likely other intel processors unaffected by this which perform similarly to the ones the consumer was initially looking for. Besides simply backordering, the consumer might decide to buy an older or lower-spec part that is in stock rather than looking at another platform (AMD). Intel might not be making as much per sale, but they are likely still selling just as many parts, and I doubt it's a large contributor to increased AMD sales.
 
Back
Top