Intel Has Fixed its 7 nm Node, But Outsourcing is Still Going to Happen

[RandallFlagg]

That's why i said "we can't make 7/10nm", i know about the density and all that.
But it's strange they can't make anything close to what Samsung and TSMC makes, even if they lowered the density.
This could be just bull...it, biggest semiconductor company in the world stuck at the same node for 6 years, no progress in 6 years!

Intel 10nm is equivalent to TSMC 7nm and is much better than Samsung 8nm. Intel has been making 10nm for over a year now and and at this point I would bet that there are a lot more x86 Intel 10nm in people's possession than there are x86 TSMC 7nm (AMD). So no, they haven't been on the same node for 6 years.

Intel makes volume that at least right now, TSMC and Samsung can't match. If you're not aware, about half of TSMCs nodes are larger than 12nm while Samsung's 8nm is only 60% the density of Intel 10nm / TSMC 7nm.

Intel 7nm is expected to be not just equivalent to TSMC 5nm, but about 20% more dense. There are no AMD 5nm chips out and reportedly Apple has 80% of TSMCs 5nm capacity for 2021 booked up. That leaves 20% for *everyone else*.

AMD will have to compete for that 20% of 5nm capacity in 2021 with Broadcom, Qualcomm, and potentially Intel - among others I'm sure.

This effectively means you won't see an x86 5nm until 2022. It's actually quite likely that, outside of maybe a low volume paper launch, the first one to actually supply the x86 market with meaningful volume of 200MT/mm "Intel 7nm / TSMC 5nm" chips will be - Intel.

Edit: I should throw in, the real threat to Intel right now is probably Apple. I said x86 for a reason. Apple is clearly well supplied with 5nm M1 and A14. AMD is basically going to be getting scraps of what's left from TSMCs capacity on 5nm for at least another year, so regardless of their performance they most likely just won't have much product to sell. Based on Gelsinger's comments, he seems quite aware that Apple is the real threat.

 

[xSneak]

I hope intel keeps its fabs making advanced nodes. It's dangerous to have the world's cutting edge nodes located in taiwan and south korea. The tsmc factory in arizona is only going to be 5nm in 2024.

 

[Vya Domus]

It's "fixed" when there are volume products using it. Until then this is just hogwash.

AMD is basically going to be getting scraps of what's left from TSMCs capacity on 5nm for at least another year

"Scraps" that will stomp whatever Intel comes up with, no matter how you spin it Intel is in deep shit. They have simply fallen back too much on process technology, they should have booked 7/5nm wafers long ago.

 

[RandallFlagg]

It's "fixed" when there are volume products using it. Until then this is just hogwash.

So when do you think we will see Zen 3 in "volume products"?

 

[Vya Domus]

So when do you think we will see Zen 3 in "volume products"?

Excuse me ? There are already Zen 3 CPUs out there.

 

[Luminescent]

Intel 10nm is equivalent to TSMC 7nm and is much better than Samsung 8nm. Intel has been making 10nm for over a year now and and at this point I would bet that there are a lot more x86 Intel 10nm in people's possession than there are x86 TSMC 7nm (AMD). So no, they haven't been on the same node for 6 years.

Intel makes volume that at least right now, TSMC and Samsung can't match. If you're not aware, about half of TSMCs nodes are larger than 12nm while Samsung's 8nm is only 60% the density of Intel 10nm / TSMC 7nm.

Intel 7nm is expected to be not just equivalent to TSMC 5nm, but about 20% more dense. There are no AMD 5nm chips out and reportedly Apple has 80% of TSMCs 5nm capacity for 2021 booked up. That leaves 20% for *everyone else*.

AMD will have to compete for that 20% of 5nm capacity in 2021 with Broadcom, Qualcomm, and potentially Intel - among others I'm sure.

This effectively means you won't see an x86 5nm until 2022. It's actually quite likely that, outside of maybe a low volume paper launch, the first one to actually supply the x86 market with meaningful volume of 200MT/mm "Intel 7nm / TSMC 5nm" chips will be - Intel.

Are you kidding me, TSMC is making 7nm silicon for everything for quite a long time, phones, gpu's, cpu's, what more volume do you want.
Sure, if you compare PC cpu's only then Amd might be at a disadvantage, from what you say AMD beat Intel with whatever scraps TSMC can provide after they satisfy Apple, Nvidia, Qualcomm and so on.
I don't care about Intel vs Amd, powerful cpu's get more irrelevant by the day, i was gonna upgrade to a 3950x or better but latest Adobe suite got so well optimized for GPU that CPU sits almost IDLE.
CPU for productivity in video editing is dead and photo editing is moving fast to replace the cpu, other software that i don't use is also heavy on the gpu like blender and the likes.

 

[DeathtoGnomes]

So when do you think we will see Zen 3 in "volume products"?

being a fanboi of any product team usually means they shut out the entire world and only pay attention to what they fawn over.

Excuse me ? There are already Zen 3 CPUs out there.

Not everyone can or will pay attention to AMD.

 

[TheLostSwede]

Intel 10nm is equivalent to TSMC 7nm and is much better than Samsung 8nm. Intel has been making 10nm for over a year now and and at this point I would bet that there are a lot more x86 Intel 10nm in people's possession than there are x86 TSMC 7nm (AMD). So no, they haven't been on the same node for 6 years.

Intel makes volume that at least right now, TSMC and Samsung can't match. If you're not aware, about half of TSMCs nodes are larger than 12nm while Samsung's 8nm is only 60% the density of Intel 10nm / TSMC 7nm.

Intel 7nm is expected to be not just equivalent to TSMC 5nm, but about 20% more dense. There are no AMD 5nm chips out and reportedly Apple has 80% of TSMCs 5nm capacity for 2021 booked up. That leaves 20% for *everyone else*.

AMD will have to compete for that 20% of 5nm capacity in 2021 with Broadcom, Qualcomm, and potentially Intel - among others I'm sure.

This effectively means you won't see an x86 5nm until 2022. It's actually quite likely that, outside of maybe a low volume paper launch, the first one to actually supply the x86 market with meaningful volume of 200MT/mm "Intel 7nm / TSMC 5nm" chips will be - Intel.

Edit: I should throw in, the real threat to Intel right now is probably Apple. I said x86 for a reason. Apple is clearly well supplied with 5nm M1 and A14. AMD is basically going to be getting scraps of what's left from TSMCs capacity on 5nm for at least another year, so regardless of their performance they most likely just won't have much product to sell. Based on Gelsinger's comments, he seems quite aware that Apple is the real threat.

Isn't that making assumptions? So far none of Intel's 10nm products have really shown that their process is superior to the competition in terms of foundry quality.
Most of their 10nm parts have been... meh at best? Their first 10nm parts weren't power efficient, nor did they perform well.
I guess they're going to have to prove that they can produce either power efficient or high performance parts this year, as otherwise all the claims of Intel having a good 10nm processes that's competitive is just going to be hot air.

Samsung is equally getting crap for their process being poor, yet there doesn't seem to be any overwhelming proof of such, beyond possibly the shortage of Nvidia GPUs. However, the actual GPUs doesn't seem to underperform, although they seem to be running a tad hot.

I don't see what TSMC can't match, they're already making 5nm part for Apple in large(?) volumes, a mere three years ahead of Intel's 7nm schedule. Will Intel deliver 7nm on time, just as they did with 10nm?

There's way too much speculation as to what is what, but yes, you are correct with regards to competition for 5nm capacity from TSMC, it's going to be in tight supply for this year and most likely next year. If this is going to be a problem for AMD or not, I guess we'll have to wait and see. There's possibly a chance for them to use the 6nm node as a step for now.

Do you have any inside info on what AMD is planning next, as all the roadmaps I have seen suggests that AMD is only planning for 5nm in 2022. I guess at least that makes your statement correct, we won't see 5nm x86 chips until 2022.

I hope intel keeps its fabs making advanced nodes. It's dangerous to have the world's cutting edge nodes located in taiwan and south korea. The tsmc factory in arizona is only going to be 5nm in 2024.

What danger is in having the fabs in Taiwan or South Korea? Are we talking earth quakes, typhoons or invasion by China or North Korea? Beyond that I can't think of any reasons why it would be dangerous, unless you're worried about the US falling behind in chip manufacturing tech.

 

[1d10t]

Good news, so after Skiplake this March and Elderlake after, we got a proper 7nm Nonelake processor from Intel.

 

[RandallFlagg]

Isn't that making assumptions? So far none of Intel's 10nm products have really shown that their process is superior to the competition in terms of foundry quality.

And I never said that it was, I said it was equivalent to TSMC 7nm not better.

If you are talking about 7nm, that's based on Intels statements that 7nm density is between 200MT/mm2 and 240MT/mm2, which is higher density than TSMCs stated ~175MT/mm2.

Also, Intel uses a flip flop and a logic gate to measure transistor density - most of the others do not state their method of measuring, so they could use an SRAM circuit which is smaller. This is why you have Samsung "8nm" at a 58 MT/mm2 density while "Intel 10nm" and "TSMC 7nm" are both around 95MT/mm2. "Nanometers" are meaningless.

Most of their 10nm parts have been... meh at best? Their first 10nm parts weren't power efficient, nor did they perform well.
I guess they're going to have to prove that they can produce either power efficient or high performance parts this year, as otherwise all the claims of Intel having a good 10nm processes that's competitive is just going to be hot air.

Intel Tiger Lake (10nm) destroys similar Zen 2 laptop parts on single core, and its Xe is far better. Its also near desktop Zen 3 performance in single core. I suppose you think Zen 3 is meh as well.

Samsung is equally getting crap for their process being poor, yet there doesn't seem to be any overwhelming proof of such, beyond possibly the shortage of Nvidia GPUs. However, the actual GPUs doesn't seem to underperform, although they seem to be running a tad hot.

I agree with that, and I actually don't think that the node is as important as many think it is. Node size has a lot more to do with production efficiency. We've had extremely powerful large CPUs in the midrange market on GloFlo 14nm for years now, they are just expensive to make.

I don't see what TSMC can't match, they're already making 5nm part for Apple in large(?) volumes, a mere three years ahead of Intel's 7nm schedule. Will Intel deliver 7nm on time, just as they did with 10nm?

Seems like everyone failed to comprehend the ***x86*** qualifier of my post.

To spoon feed it - AMD is no threat because AMD has limited access to nodes beyond 7nm, and Intel already has a match for 7nm.

There's way too much speculation as to what is what, but yes, you are correct with regards to competition for 5nm capacity from TSMC, it's going to be in tight supply for this year and most likely next year. If this is going to be a problem for AMD or not, I guess we'll have to wait and see. There's possibly a chance for them to use the 6nm node as a step for now.

I think the current market already indicates it's a problem for AMD. Zen 3 and RDNA2 are not volume shipping components. If they were, then indeed Intel would be in trouble. Instead of Zen 3 and RDNA 2, AMDs limited contract capacity with TSMC seems to be tied up with Xbox / PS5.

I expect AMDs Q4 earnings to reveal a big jump in "embedded" aka Xbox/PS5 market, with much more muted and possibly negative growth in laptop/desktop chips as well as GPUs. This all has to do with production allocation, not benchmarks.

Do you have any inside info on what AMD is planning next, as all the roadmaps I have seen suggests that AMD is only planning for 5nm in 2022. I guess at least that makes your statement correct, we won't see 5nm x86 chips until 2022.

There have been a number of articles and constant speculation on the forums about AMD doing Zen 4 launch in late 2021. Not inside info just being aware of what the media is saying -

AMD’s 5nm Zen 4 CPUs could arrive in 2021, really turning the heat up on Intel

AMD could pile on the pressure if this rumor is correct

www.techradar.com

AMD Zen 4 CPUs Rumored On Track For 2021 Release As Early TSMC 5nm Yields Impress

TSMC reportedly has three companies lined up with first production run: Apple, HiSilicon, and AMD.

hothardware.com

AMD Zen 4 could arrive in 2021 as TSMC 5nm shows promising yields

AMD has been on a tear recently and with Zen 3 getting claimed huge performance uplift and with TSMC having a solid 5nm yeild, Ryzen is not stopping.

www.tweaktown.com

 

[Redwoodz]

And I never said that it was, I said it was equivalent to TSMC 7nm not better.

If you are talking about 7nm, that's based on Intels statements that 7nm density is between 200MT/mm2 and 240MT/mm2, which is higher density than TSMCs stated ~175MT/mm2.

Also, Intel uses a flip flop and a logic gate to measure transistor density - most of the others do not state their method of measuring, so they could use an SRAM circuit which is smaller. This is why you have Samsung "8nm" at a 58 MT/mm2 density while "Intel 10nm" and "TSMC 7nm" are both around 95MT/mm2. "Nanometers" are meaningless.


Intel Tiger Lake (10nm) destroys similar Zen 2 laptop parts on single core, and its Xe is far better. Its also near desktop Zen 3 performance in single core. I suppose you think Zen 3 is meh as well.



I agree with that, and I actually don't think that the node is as important as many think it is. Node size has a lot more to do with production efficiency. We've had extremely powerful large CPUs in the midrange market on GloFlo 14nm for years now, they are just expensive to make.



Seems like everyone failed to comprehend the ***x86*** qualifier of my post.

To spoon feed it - AMD is no threat because AMD has limited access to nodes beyond 7nm, and Intel already has a match for 7nm.



I think the current market already indicates it's a problem for AMD. Zen 3 and RDNA2 are not volume shipping components. If they were, then indeed Intel would be in trouble. Instead of Zen 3 and RDNA 2, AMDs limited contract capacity with TSMC seems to be tied up with Xbox / PS5.

I expect AMDs Q4 earnings to reveal a big jump in "embedded" aka Xbox/PS5 market, with much more muted and possibly negative growth in laptop/desktop chips as well as GPUs. This all has to do with production allocation, not benchmarks.


There have been a number of articles and constant speculation on the forums about AMD doing Zen 4 launch in late 2021. Not inside info just being aware of what the media is saying -

AMD’s 5nm Zen 4 CPUs could arrive in 2021, really turning the heat up on Intel

AMD could pile on the pressure if this rumor is correct

www.techradar.com

AMD Zen 4 CPUs Rumored On Track For 2021 Release As Early TSMC 5nm Yields Impress

TSMC reportedly has three companies lined up with first production run: Apple, HiSilicon, and AMD.

hothardware.com

AMD Zen 4 could arrive in 2021 as TSMC 5nm shows promising yields

AMD has been on a tear recently and with Zen 3 getting claimed huge performance uplift and with TSMC having a solid 5nm yeild, Ryzen is not stopping.

www.tweaktown.com

Intel 10nm is NOT equal to TSMC 7nm. It's fine for low power parts but is NOT suitable for large high clock chips. The yields are terrible. It also took Intel like 4 yrs to make that 10nm process even viable for notebooks.

 

[RandallFlagg]

Intel 10nm is NOT equal to TSMC 7nm. It's fine for low power parts but is NOT suitable for large high clock chips. The yields are terrible. It also took Intel like 4 yrs to make that 10nm process even viable for notebooks.

Intel targeted laptops because, for 2020, they are 75% of the client space. They are putting their best parts in the highest volume space and have been effectively supplying it, and are now moving on to servers and higher power laptops on 10nm. AMD put their best parts in a small fraction (DIY and 3rd tier SIs) of the lowest volume space (desktop) and still couldn't effectively supply it.

 

[laszlo]

Corrected version:

"Intel CEO, Pat Gelsinger, has talked about the state of the 10 nm node, giving shareholders reassurance and a will to remain in such a position. He has made an argument that he has personally reviewed the progress of the "health and recovery of the 10 nm program"

 

[dicktracy]

RIP AMD.

 

[Luminescent]

RIP AMD.

Far from it, for a while people thought that AMD could never equal or beat Intel with chiplet design at single core, now they equaled them or beat them in some applications and we are talking inefficient chiplet design with high latency compared to Intel's monolithic.
So AMD at some point in the future if need arises just needs to make a monolithic CPU and Intel is fuc...ed, for a while of course.

 

[1d10t]

Corrected version:

"Intel CEO, Pat Gelsinger, has talked about the state of the 10 nm node, giving shareholders reassurance and a will to remain in such a position. He has made an argument that he has personally reviewed the progress of the "health and recovery of the 10 nm program"

Oh come on, at least give him some credit. Personally I think Pat would bring major improvement for Intel, in near future all of their CPU get VT-x enable by default.

 

[TumbleGeorge]

RIP AMD.

That is fact right now for desktop businesses of all high-tech companies. Prices increased, shortages growing, delays are extended. The big and beautiful adventure with desktops is not too far from over. And it was so satisfying, so diverse.

 

[cst1992]

7nm in 2023... looks like I'll get an AMD processor for my next build.

 

[r9]

This is satisfactory as licking ice cream through glass window.

 

[DeathtoGnomes]

RIP AMD.

RIP INTEL.

 

[Aquinus]

Next up: 7nm Evaporated Lake

 

[pjl321]

7nm delayed 1 year? Try 6 years!!!
And even that is only if Intel sticks to 2023 this time which I won't be holding my breath for.

Official Intel slide:

 

[dinmaster]

so they put enough pluses on 14++++++ they just decided to call it 7nm.. while amd is going to 5nm.. two steps back calling it one step back. That pr spin on a cow turd...

 

[Dave65]

FIXED... Ok!

 

[micropage7]

why i feel Intel try to move the news from new processor and its power to something like this, oh maybe AMD is a little weak in this form