TSMC to Kickstart 5 nm Volume Production in April, Production Capacity Already Fully Booked

[Raevenlord]

TSMC will be doing good on their previous expectations for a H2 2020 ramp-up for high volume production (HVM) on their 5 nm manufacturing process. The new 5 nm fabrication process is an Extreme Ultraviolet lithography (EUV) one, with up to 14 layers being etchable onto the silicon wafers, as opposed to five and six, respectively, for TSMC's N7+ and N6 processes.

Volume production will start with Apple's A14 SoC, meant to be driving next-generation iPhones that should hit shelves by September this year (should the COVID-19 pandemic let it be so). Apple is using two thirds of TSMC's capacity for 5 nm as is with this SoC; it's currently unclear which client (or clients) are getting the leftover one third capacity. TSMC announced back in December that they were seeing yields upwards of 80% in 5 nm EUV fabrication, so now it's "just" a matter of monetizing the process until their 3 nm iteration comes online, expectedly, in 2022.



View at TechPowerUp Main Site

 

[RH92]

So it's safe to assume that now that Apple is moving to 5nm that will let more room for GPU manufacturers on 7nm EUV right ?

 

[Chomiq]

So it's safe to assume that now that Apple is moving to 5nm that will let more room for GPU manufacturers on 7nm EUV right ?

Unless they'll keep their cheaper models on old node.

 

[seronx]

So it's safe to assume that now that Apple is moving to 5nm that will let more room for GPU manufacturers on 7nm EUV right ?

NVIDIA is joining Apple on N5.

Sr Mask Design Engineer => 5nm FinFet TSMC and Sec8 Samsung
Ampere is on Samsung's 8LPU and Hopper is on TSMC's 5FF.

 

[R0H1T]

So it's safe to assume that now that Apple is moving to 5nm that will let more room for GPU manufacturers on 7nm EUV right ?

It's safe to say that Apple, among many other manufacturers, will see a lot of demand erosion. This may lead to more room for AMD to grow at 7nm or even 5nm, but then again the notebook/server/DIY/HPC supply chains will also be majorly affected.

 

[TheLostSwede]

NVIDIA is joining Apple on N5.

Sr Mask Design Engineer => 5nm FinFet TSMC and Sec8 Samsung
Ampere is on Samsung's 8LPU and Hopper is on TSMC's 5FF.

Or not...

Rumor: Next gen Nvidia RTX chip will be based on Samsung 10nm node

All new cards will be RTX. SLi will be available to GA102 ONLY. https://nl.hardware.info/nieuws/71288/gerucht-alle-ampere-kaarten-krijgen-rtx-en-worden-gemaakt-op-samsungs-10nm-node

www.techpowerup.com

 

[TechLurker]

I wonder if AMD also already booked. If I'm not misremembering, Zen 4 onwards were to be 5nm (along with PCIe 5.0 capability), with Zen 4 estimated for a Q3 or Q4 release in 2021, if everything continues to go well.

 

[TheGuruStud]

I wonder if AMD also already booked. If I'm not misremembering, Zen 4 onwards were to be 5nm (along with PCIe 5.0 capability), with Zen 4 estimated for a Q3 or Q4 release in 2021, if everything continues to go well.

AMD definitely is. They doubled the order for 2H this year of 7nm and they'll need even more once moving to 5nm. You can't wait around like Nvidia and end up with nothing lol.

Huang, eat a big bag of phallic shaped objects XD

 

[krusha03]

So it's safe to assume that now that Apple is moving to 5nm that will let more room for GPU manufacturers on 7nm EUV right ?

From what I'm reading the difference between their 5nm and 7nm node is using EUV for more layers. Meaning using the same machine for longer time. So I dont expect 7nm node being positively affected by this.

 

[pjl321]

I wonder if AMD also already booked. If I'm not misremembering, Zen 4 onwards were to be 5nm (along with PCIe 5.0 capability), with Zen 4 estimated for a Q3 or Q4 release in 2021, if everything continues to go well.

It will be very very late 2021 and with everything going on in the world right now then that might slip to 2022 because Zen 3 might slip this year - imo.

NVIDIA is joining Apple on N5.

Sr Mask Design Engineer => 5nm FinFet TSMC and Sec8 Samsung
Ampere is on Samsung's 8LPU and Hopper is on TSMC's 5FF.

Man I would love that, I still can't see myself buying an Nvidia card simply because they never seem to offer the best performance per dollar which is the only thing i care about but the performance jump they could achieve by leaping from 16nm to 5nm would be utterly insane!

From what I'm reading the difference between their 5nm and 7nm node is using EUV for more layers. Meaning using the same machine for longer time. So I dont expect 7nm node being positively affected by this.

I guess that depends how/where the machines are set up, 5nm might be a totally different fab with totally separate machines?? I personally think that is the case or they couldn't ramp up production of 5nm without ramping down N7+ production which they are not doing, if anything they are ramping N7+ up too.

 

[ARF]

AMD has got a commanding lead with Zen 2, so Zen 3 is not important right now, at all.

AMD needs to pull forward Big Navi lineup and adjust its graphics cards portfolio accordingly so they will be at least somewhat competitive again.

So, what's up with Navi 21 and Navi 23 ?

 

[medi01]

AMD needs to pull forward Big Navi lineup and adjust its graphics cards portfolio accordingly so they will be at least somewhat competitive again.

That is the bit I hate most about the market: a company needs to roll out useless (1% of the market, give me a break) piece of silicon, wasting lot's of resources, to create a halo for lower end products to sell better.

 

[ARF]

That is the bit I hate most about the market: a company needs to roll out useless (1% of the market, give me a break) piece of silicon, wasting lot's of resources, to create a halo for lower end products to sell better.

Well, but there are many forces which influence this process.
Enthusiasts buy Super Navi and begin to recommend all other Radeons to the masses;
We have still quite a lot of performance on the table which we desperately need for our 4K, VR, high-Hz gaming, ray-tracing included.
Free advertisement for the whole lineup - ultimate performance leadership is all that matters.

 

[pjl321]

AMD has got a commanding lead with Zen 2, so Zen 3 is not important right now, at all.

AMD needs to pull forward Big Navi lineup and adjust its graphics cards portfolio accordingly so they will be at least somewhat competitive again.

So, what's up with Navi 21 and Navi 23 ?

The issue for AMD is as the underdog (which they still are as size of company, sales numbers, turn over, etc) they can't just match Intel or beat them by a small amount they have to absolutely destroy them and do it for several years to win the hearts and minds of lay person that doesn't follow the tech industry who still chooses Intel because "that's who they've heard of".

Intel's Comet Lake will be out April/May time and with it's 5.3GHz and 10 cores it will mean Intel win a few benchmarks and the issues comes from if you give Intel an inch they will 'claim' a mile, if they have the best gaming CPU and the fastest at SuperPi and a few other obscure benchmarks then their marketing department will go to town on using only those benchmarks and the clueless buyers (which is probably 80%+ of the market) will fall for the marketing and buy Intel.
Zen 3 has such a large IPC gain that AMD has the ability to cut Intel off at the knees, they can finish them until Rocket/Alter Lake in 2H 2021 (still 14nm can you believe) so why would AMD not want to break the camel's back with Zen 3 as early as possible and really put some distance between them and Intel?


90% of people who were going to buy Zen 2 will have bought it by July and the rest will be waiting for Zen 3 whether it's released in July or December.
There is no downside for AMD to genuinely pull ahead of Intel rather than just trading blows with them and offering more cores or lower prices.

The reason I think AMD could release Zen 3 for the server market even earlier in maybe March/April is because that is where the big money is and I know a lot of server customers like to wait for 3 generations of a new architecture before buying so that any major issues / security flaws are fixed or at the very least known. So if Zen 3 is ready then why not? The design was finalised back in July last year so should be ready. Also AMD can split demand for Epyc over two nodes, 7nm and 7nm+ which will help manage capacity.

With Zen 3 AMD can really cement their superiority as the best CPU manufacturer for pretty much any usage case. They need to build this mindset before Intel regroups with Willow or Golden Cove.

Finally, my last reason it make sense for AMD to release Zen 3 sooner rather than later is that they have promised mobo manufacturers that Zen 4 will be released at the end of 2021,. Zen 4, as i'm sure you know, needs a whole new socket which mobo manufacturers love and they will demand AMD sticks to this 2021 time scale so with that in mind it just makes sense that if either Zen 2 or Zen 3 is going to have a shorter than normal life that it's Zen 2 and not Zen 3.

 

[ARF]

5.3 GHz on Comet Lake is a single-core turbo boost frequency kept for nanoseconds ?

AMD only needs prices correction. Ryzen 9 3950X for $399 and Ryzen 9 3900X for $299 and Intel's entire lineup is DOA no matter what they do.

 

[pjl321]

5.3 GHz on Comet Lake is a single-core turbo boost frequency kept for nanoseconds ?

AMD only needs prices correction. Ryzen 9 3950X for $399 and Ryzen 9 3900X for $299 and Intel's entire lineup is DOA no matter what they do.

There are a lot of people out there that just want the fastest and will pay, with in reason, anything for it. The 9900K and the Comet Lake CPUs that are replacing this are the fastest for gaming, Adobe software and quite a few other things. That is why AMD should and need to push Zen 3 out as early as possible to complete kill Intel's offerings and take very single performance title going so there is no question what anyone should buy no matter what your usage case is.

Sadly, even if AMD did this Intel would still sell reasonable numbers 'because it's Intel', there are quite a lot of people out there that would buy Intel because they have heard of Intel, they've not heard of AMD.

Also, Alter Lake and Rocket Lake are in coming around the end of this year and 1H of next year so AMD really can't sit still.

 

[R0H1T]

Sadly, even if AMD did this Intel would still sell reasonable numbers 'because it's Intel', there are quite a lot of people out there that would buy Intel because they have heard of Intel, they've not heard of AMD.

I have a feeling with the impending global recession, premium isn't something what most people - even enthusiasts - will look kindly at & that applies to Intel as well as AMD!

 

[rainzor]

Or not...

Rumor: Next gen Nvidia RTX chip will be based on Samsung 10nm node

All new cards will be RTX. SLi will be available to GA102 ONLY. https://nl.hardware.info/nieuws/71288/gerucht-alle-ampere-kaarten-krijgen-rtx-en-worden-gemaakt-op-samsungs-10nm-node

www.techpowerup.com

Or not what? 8LPU is an extension of their 10nm process

 

[ARF]

That is the bit I hate most about the market: a company needs to roll out useless (1% of the market, give me a break) piece of silicon, wasting lot's of resources, to create a halo for lower end products to sell better.

Not 1% but more likely towards 10%.

“With the Radeon 5000-series we are essentially covering 90-something-percent of the total PC gamers today,” says Chandrasekhar.

Like Ryzen, AMD’s Big Navi is “going to similarly disrupt 4K” gaming

Like Ryzen, AMD’s Big Navi is “going to similarly disrupt 4K” gaming

"All of us need a thriving Radeon GPU ecosystem. So, are we going after 4K, and going to similarly disrupt 4K? Absolutely."

www.pcgamesn.com

 

[medi01]

We have still quite a lot of performance on the table which we desperately need for our 4K, VR, high-Hz gaming, ray-tracing included.

The "high Hz" gaming, guys, isn't it about time to figure it's not about card performance, but about game developers taking better looks over higher clocks?

Not 1% but more likely towards 10%.

Both 2080 and 2080Ti are under 1%.

 

[ARF]

The "high Hz" gaming, guys, isn't it about time to figure it's not about card performance, but about game developers taking better looks over higher clocks?

Better looks also need more processing power, unless graphics developers introduce a technology which doesn't require it.

Both 2080 and 2080Ti are under 1%.

Those are uber super expensive, and low quality.

Nvidia’s flagship RTX 2080 Ti graphics cards are failing.

Nvidia RTX 2080 Ti Failures Update: Founders Edition Errors Most Common? | Digital Trends

Owners of Nvidia's RTX 2080 Ti have been complaining about crashes, blue screens, artifacts, and repeated RMA issues for weeks now, but investigations aren't turning up anything too definitive with what's going on. It does appear, however, that the cards causing the most problems are Founders...

www.digitaltrends.com