TSMC 5 nm+ Node Manufacturing Goes High-Volume in Q4, AMD one of the Major Customers

[AleksandarK]

TSMC is working hard to bring the best silicon out there, with the company supplying many of the companies like NVIDIA, AMD, Huawei, and Apple - all customers who demand the latest and greatest when it comes to the silicon technology. According to sources close to DigiTimes, TSMC is expected to kick-off volume production of its next-generation 5 nm+ manufacturing node, which is an enhancement of the 5 nm node, as soon as Q4 of this year hits.

Update May 29th: The DigiTimes report indicates that TSMC is preparing the 5 nm+ node for AMD Ryzen 4000 "Vermeer" series of CPUs. Originally planned for using the 7 nm+ node, the CPUs are supposedly ported to a smaller node, providing better transistor performance and lower power consumption. The Ryzen 4000 series of desktop processors were planned for launching later this year, however, being that the new information provided by DigiTimes suggests 5 nm+ node could be used, we can expect to see Zen 3 based processors sometime in early 2021.



Despite the current world pandemic happening, TSMC is relentless and seems like it can not be stopped. Sources at DigiTimes have confirmed that the company will do High-Volume Manufacturing of the chips in Q4, meaning that we can expect first chips on the 5 nm+ node somewhere in the beginning of 2021. This is very good news as it reassures us that TSMC is untouched by the happenings in the world and that they will manage to deliver just fine.

View at TechPowerUp Main Site

 

[yeeeeman]

Screw it, just go volume production at end of the year also for 3nm, 2nm, 1nm, 0.5nm, 0.0001nm. Everything now.

 

[ShurikN]

Rumor from digitimes also states Ryzen 4000 (Zen 3) on the same 5nm+ process.
XT refresh makes more sense now, as these new chips will launch probably very late in the year.

 

[ratirt]

Rumor from digitimes also states Ryzen 4000 (Zen 3) on the same 5nm+ process.
XT refresh makes more sense now, as these new chips will launch probably very late in the year.

Huh. That explains a lot. Wonder how this 5nm will do against 7nm+ Not sure how much of a difference we can expect. 14nm vs 7nm more noticeable change I think.

 

[ppn]

No 5nm for Zen, 6000 series 6nm EUV in 2022.

 

[ARF]

If by any chance AMD indeed takes N5 capacity, they have to focus on the graphics cards.

Intel is no threat right now..

 

[ppn]

Volume production means for ARM chips, 5nm is at least 1 year away.
Besides the design rules of N7 are only compatible with N7P and later on N6.
So they will have to redesign everything from the ground up for it to work with N5.
N7 risk production 2017, NAvi 2019, see - 2 years.
N6 risk 2020, CPUs, GPUs in 2022 18% +density, yeah, not much better, and limited to 420mm2, N7P until 2022.
N5 risk 2019, earliest 2021.

 

[Tomorrow]

No 5nm for Zen, 6000 series 6nm EUV in 2022.

You're assuming at this point they are carrying the same design foward. From what i've seen that's not the case. Each design is made for a respective node. Even if Zen3 4000 series will not be on 5nm the Zen4 5000 will be because that's a late 2021 or early 2022 release. This was already confirmed earlier on official slide. So it's about as official as it gets at this point.

Volume production means for ARM chips, 5nm is at least 1 year away.

That's usually how it goes. But rumors suggest 5nm is already having better yields than 7nm had at the same point in it's lifecycle. That would explain AMD's move.

Besides the design rules of N7 are only compatible with N7P and later on N6.

Zen3 is a redesign anyway not a full carryover like XT refresh seems to be.

So they will have to redesign everything from the ground up for it to work with N5.

Well not ground up but it would give them half a year or more. And who knows how long they have planned this move already. Reports of good 5nm yields surfaced earlier this year already and they still needed some time to cook up the XT refresh as a stopgap. So Zen3's move to 5nm may have been in the works for longer than we know of.

 

[OGoc]

Huh. That explains a lot. Wonder how this 5nm will do against 7nm+ Not sure how much of a difference we can expect. 14nm vs 7nm more noticeable change I think.

14nm to 7nm was a quadrupling of density.

5nm is roughly double the density of 7nm because the number is squared. 5nm is a 25nm2 transistor and 7nm is a 49nm2 transistor... well, 5nm "feature", not the whole transistor but the doubling ratio still applies.

 

[medi01]

Intel is no threat right now..

Intel's income is two times higher than NV's revenue.

"No threat", south eastern part of my body.

 

[ARF]

Intel's income is two times higher than NV's revenue.

Success and competitiveness are not being measured by revenue.

AMD Ryzen 9 3950X PassMark Score: 39,400
AMD Ryzen 9 3900X PassMark Score: 32,800
Intel Core i9-10900K PassMark Score: 23,400

AMD Ryzen Threadripper 3990X PassMark Score: 81,000
Intel Core i9-10980XE PassMark Score: 35,000


Nvidia RTX 2080 Ti ~50% faster than AMD Radeon VII

 

[Ashtr1x]

All of that super advanced technology that man has ever known in the era of computers and semiconductor engineering, made for Apple A series phones which are used for - IG, FB, Whatsapp, Selfies, iMessage, AirPods, Apple Music. That iOS doesn't even have a Filesystem exposed to user, but it gets Anandtech approval for the SPEC bragging rights lol. What a world we live in. Oh also that Chinese Huawei which was a literal zero during the Mate 7 somehow now beat Exynos Samsung SK.

On the other side, Intel is still struggling with whatever plan they have with 10nm, and a few years ago, they were leading the bleeding edge leadership, the best example of monopoly and lack of competition. Thankfully AMD is utilizing that technology to innovate now , x86 improvements to shut those SPEC ARM talk. AMD's Zen4 will probably be Q4 2021 on 5nm+ I guess, Zen 3 will be on N7P (EUV N7), RDNA2 will be on the same N7P I guess along with Nvidia Ampere, which is also sourced to Samsung 8nm EUV node.

 

[djisas]

I'll take a ZEN3 4700X on 5nm tyvm...
Wishful thinking...

 

[TheLostSwede]

N5 risk 2019, earliest 2021.

You need to learn how to read...

Our 5nm technology entered risk production in March 2019 and target for volume production in 2020.

5nm Technology - Taiwan Semiconductor Manufacturing Company Limited

TSMC’s 5nm (N5) Fin Field-Effect Transistor (FinFET) technology successfully entered volume production in the second quarter of 2020 and experienced a strong ramp in the second half of 2020.

www.tsmc.com

 

[AnarchoPrimitiv]

If by any chance AMD indeed takes N5 capacity, they have to focus on the graphics cards.

Intel is no threat right now..

That is a completely incorrect assessment of the markets. The x86 market, comprised of enterprise, consumer, and mobile, has an exponentially larger Total Addressable Market (T.A.M.) than the dGPU market, in fact if I remember correctly I believe enterprise x86 is larger by itself than the dGPU market. In other words, the x86 market has a much larger pot of gold than the GPU market. Also, you're acting as though Intel is Vanquished, when it's only a matter of time before they inevitably answer back, after all, they have many times the financial resources than AMD. This is why it's important for AMD to be unrelenting and to not let go of Intel's belt buckel (it's an analogy the Vietcong made for how to neutralize overwhelming American Air Power in the Vietnam War).

If this 5nm rumor is true, it's a smart move by AMD, as what thry need to achieve is "top of mind awareness" in the consumer and the perception of being better, as if this is achieved, even if Intel answers back, it won't be enough to break AMD's hold on consumers. Just consider the mid to late 2000s, when AMD offered not only obviously better performing cards, but at better prices..... And what happened? Nvidia STILL sold more cards despite having the weaker product all due to consumer perception... And that's what AMD needs to achieve, they need 10 years of dominance, nor just a few years, so that even if Intel offers a better product eventually, it won't matter that much. Intel would have to offer success after success for several years to gain confidence back, which is a lot more difficult that a one time performance jump. Just look at Ryzen, it took the third release (Zen2) for the general consumer to finally relent.

See if AMD focuses on x86, then who knows, in five years they may have 50% of the OEM market and then they could take a page out of Intel and Nvidia's playbook and do something like offer OEMs a deep discount on AMD CPUs ONLY IF they bundle them with one of our discrete GPUs. Then they can leverage their strengths in x86 to put pressure on Nvidia from two sides, on the OEM front, and in competing on the DIY PC Videocard front.

 

[medi01]

Success and competitiveness are not being measured by revenue.

It ultimately is.

Nvidia RTX 2080 Ti ~50% faster than AMD Radeon VII

Radeon VII is 330mm2.
2080Ti is nearly 3 times bigger.

Randomly picking products is one of the most nonsensical ways to compare performance.

 

[Hossein Almet]

Can't wait for Vermeer if indeed it is fabricated on 5 nm node. Very excited indeed.

 

[TheinsanegamerN]

It ultimately is.


Radeon VII is 330mm2.
2080Ti is nearly 3 times bigger.

Randomly picking products is one of the most nonsensical ways to compare performance.

2080ti also has a decent chunk of silicon dedicated to RT cores, which VII doesnt have.

The 1080ti is only 267mm2 and is also 5-10% faster then radeon VII.

. Also, you're acting as though Intel is Vanquished, when it's only a matter of time before they inevitably answer back, after all, they have many times the financial resources than AMD.

This part, right here, is incredibly relevant when it comes to AMD. The athlon 64 era was great for AMD, but AMD made the mistake of getting distracted by ATi and was goofing around with phenom and L3 caches.

The core 2 caught AMD totally off guard, and the legitimate core 2 killer didnt arrive arguably until 2009 with the phenom II line, as the phenom I line just couldnt clock high enough to challenge intel.

The worst thing AMD could do right now is make this mistake again.

 

[ARF]

Radeon VII is 330mm2.
2080Ti is nearly 3 times bigger.

Randomly picking products is one of the most nonsensical ways to compare performance.

Pick products that are not random or stop trolling.

Radeon VII is a full process shrink to N7, while RTX 2080 Ti when equalised/normalised for the process is no more than 377 mm2.

If this 5nm rumor is true, it's a smart move by AMD

No, it is not..

 

[Dante Uchiha]

I think they should move only RDNA2 to 5nm+, zen 3 should use 7nm+ EUV and hit the market as soon as possible, AMD can not give the intel chance to get up and counter attack with any thing lake produced in 10+++(...).

 

[Elysium]

I find this deeply difficult to believe. Su has never explicitly ruled out 5nm for Zen 3 but it should be obvious to everyone that deviating from the roadmap a mere six months before Zen 3 launch - in what is still a very uncertain financial climate - would be a potentially devastating self-inflicted wound.

 

[R0H1T]

Success and competitiveness are not being measured by revenue.

AMD Ryzen 9 3950X PassMark Score: 39,400
AMD Ryzen 9 3900X PassMark Score: 32,800
Intel Core i9-10900K PassMark Score: 23,400

AMD Ryzen Threadripper 3990X PassMark Score: 81,000
Intel Core i9-10980XE PassMark Score: 35,000


Nvidia RTX 2080 Ti ~50% faster than AMD Radeon VII

You'd be naïve to think Intel will never have an answer to AMD, if anything Intel is a bigger threat to Nvidia (+AMD of course) than the two GPU competitors are to each other!

mere six months before Zen 3 launch - in what is still a very uncertain financial climate - would be a potentially devastating self-inflicted wound.

Why, it's actually better for AMD especially if RKL or TGL prove to be better than most are expecting? In fact at a time when their notebook sales are looking to take off, it makes perfect sense to get an unassailable lead over Intel in desktops!

Of course it's a rumor atm so take it with a dead sea full of salt

 

[Elysium]

Why, it's actually better for AMD especially if RKL or TGL prove to be better than most are expecting? In fact at a time when their notebook sales are looking to take off, it makes perfect sense to get an unassailable lead over Intel in desktops!

Of course it's a rumor atm so take it with a dead sea full of salt

Intel deviated from their roadmap - not even on purpose but by plain old mistake - and look where that got them. There's more risk than reward here for AMD, they know they are safely penned in for 2021 at the very least so it'd be better to stick to the broad plan than try to take advantage of what is a temporary vacancy in 5nm investments.

Yeah, I for one am taking it with more salt than the Pacific Ocean.

 

[ARF]

You'd be naïve to think Intel will never have an answer to AMD, if anything Intel is a bigger threat to Nvidia (+AMD of course) than the two GPU competitors are to each other!

The point is not that Intel could have an answer in 2 or 5 years. The point is that AMD is not competitive in the graphics market.

Intel could release something ground breaking but more likely and I do expect that they will not.


Wait, what do you mean by this? Do you hint that uncle and niece could arrange it, so that AMD effectively doesn't compete, because of the family conflict of interest ?

 

[Space Lynx]

HOLY CRAP 5NM+ IN LATE 2020?????

this is 2 years ahead of schedule wtf... wow im def buying a 4800x or 4900x now... and a msi x570 tomahawk

I think they should move only RDNA2 to 5nm+, zen 3 should use 7nm+ EUV and hit the market as soon as possible, AMD can not give the intel chance to get up and counter attack with any thing lake produced in 10+++(...).

i agree with this 100% they could really come out swinging a big punch to nvidia with 5nm+ big navi