AMD hands TSMC its entire 7nm portfolio: Vega 20, Zen 2 and Navi right around the corner

LemmingOverlrd

Posts: 86   +40

Mark Papermaster, CTO of AMD, has announced that the company has finally leveled out its GPU and CPU development and unified everything under a single manufacturing node, at 7nm, handing TSMC the keys to the kingdom.

In a blog post, the exec explains how the company's long term vision has finally paid off and AMD is now in a position to put its competition between a rock and a hard place with advanced 7nm designs in both GPU and CPU segments, thanks exclusively to the help of TSMC. This announcement sets the stage for the departure of GlobalFoundries, AMD's other, oldest, foundry partner.

The announcement had already been hinted at during several corporate communications, events and calls, however, it is the first time AMD comes forward and effectively drives this narrative. Papermaster said "Our work with TSMC on their 7nm node has gone very well and we have seen excellent results from early silicon."

Papermaster goes on to report that multiple 7nm products have taped out at TSMC. This includes the workstation-class GPU, Vega 20, and the next generation CPU architecture, Zen 2, all manufactured by TSMC. Going forward, AMD will focus its entire product portfolio on TSMC's 7nm process, and that includes this year's Vega 20, followed by an early 2019 launch of the Zen 2-based server CPU, Epyc, and later that year, the Navi GPU architecture, the long-awaited 7nm follow-up to Polaris which AMD is now beginning to talk up.

For more time than most AMD fans would care to remember, the company had relied on GlobalFoundries to cater to its semiconductor manufacturing needs. The results, however, have been a mixed bag. GlobalFoundries has tripped itself up more than once while playing catch-up with Intel, having twice nixxed its developing processes. It trashed its 14nm development in favor of licensing it off Samsung, and now it has announced it has ended its 7nm node. In the meantime, TSMC, the backup fab for AMD, not only delivered where GlobalFoundries failed, but managed to overtake Intel in manufacturing technology, and this has paid off.

While recognizing GlobalFoundries' role in producing 14nm and 12nm GPUs and CPUs, in particular for Zen and Zen+, which proved great successes in restoring AMD's reputation, it seems that the relationship between AMD and its oldest foundry partner will not go beyond those nodes. With TSMC, AMD now has the weapons to fend off its biggest competitor, Intel.

Permalink to story.

 
Last edited by a moderator:

Vulcanproject

Posts: 1,164   +1,924
TSMC's 7nm must be looking really good for AMD to entrust the whole lot to them.

This is also throwing a big dampener on the Nvidia RTX2000 series for me. Those products are clearly just one year deals at most on TSMC 12nm.

Nvidia delayed their launch so much I wouldn't be surprised to see the whole lineup replaced with 7nm parts by this time next year.

While Intel battle with their equivalent 10nm process I think Zen 2 will be a huge hit and take the overall consumer performance lead spring 2019.

Then with luck AMD can dish out some decent cheap graphics cards that move the price performance bar, which hasn't budged in two years.
 

CaptainTom

Posts: 414   +220
As long as TSMC has the capacity, this is FANTASTIC news.

I still suspect GF will be used eventually for chips like Navi 11 (small RX 560 replacement), mobile chips (GF is as efficient as TSMC, though not as powerful), and random things that don't need to be on the cutting edge.
 
  • Like
Reactions: sho12345

Sausagemeat

Posts: 409   +204
AMDs gamble to go all in for 7nm might be backfiring. Global foundries announcing that they are discontinuing 7nm forcing AMD to shift all orders to TSMC who now have huge pressure on them to meet AMDs demand and replace their largest foundry. Looking at the AMD share price and the fact that yet another AMD executive is leaving, Papermaster’s blog post looks to be more like damage control more than anything else.

So from what I can see, 2019 could be a very good year for AMD if TSMC can actually meet AMDs now enormous 7nm orders. However it if doesn’t, 2019 could be a very bad year indeed with no further 12nm or 14nm development in the company as far as I am aware.
 

CaptainTom

Posts: 414   +220
Please be worthy cheaper graphic card! I want a good 300 400 upgrade!
Let's hope we get more than GTX 1080 Ti equivalent with $300/$400 card.
The math is simple really:

1) Navi is rumored to be a ~250-300mm^2 chip

2) 7nm allows 2.8x the density of 14nm.

3) It also allows for 40% higher clocks OR 65% less energy usage.

Thus one can conclude that at a minimum Navi should be in-between the 1080 and 1080 Ti (RTX 2070), and at best it could actually compete with the 2080 Ti (not a joke). It should also do so while using at most 100-200w (RX 580 power usage).

We also know that Navi is the last version of GCN, and it is built for gaming first unlike the compute-focused Vega. So there could be extra dividends paid in gaming perf/watt.
 

CaptainTom

Posts: 414   +220
AMDs gamble to go all in for 7nm might be backfiring. Global foundries announcing that they are discontinuing 7nm forcing AMD to shift all orders to TSMC who now have huge pressure on them to meet AMDs demand and replace their largest foundry. Looking at the AMD share price and the fact that yet another AMD executive is leaving, Papermaster’s blog post looks to be more like damage control more than anything else.

So from what I can see, 2019 could be a very good year for AMD if TSMC can actually meet AMDs now enormous 7nm orders. However it if doesn’t, 2019 could be a very bad year indeed with no further 12nm or 14nm development in the company as far as I am aware.
People will pay top dollar for those 7nm EPYC and Vega 20 chips - so that's not an issue. The issue will be fending off Nvidia so they can make enough Navi and Ryzen 3.

But I have said this before - GF is ramping up a true 12nm (non-finfet!) for mobile and laptops. It's not crazy to think that AMD might use GF's 12nm-Plannar for their low cost Ryzen 3 laptops, APU's, and $100 Navi 11 cards.


https://www.globalfoundries.com/technology-solutions/cmos/fdx/12fdx

^15% higher clocks than existing 14nm, and half the power usage. Directly comparable to 10nmFF. That's perfect for an RX 660/650 and cheap Ryzen 3/5 APU's.
 

Duskfall

Posts: 6   +0
I hope it would be as good as the high end Nvidia ones. We need more competition in gpus so that prices go down. Hopefully AMD can achieve that
 
This switch fulling to 7nm TSMC was a multi-years plan again. Aman that we can part GF. A great day for AMD. Even AMD is to pay GF to part, I am happy we did.
 

Sausagemeat

Posts: 409   +204
The math is simple really:

1) Navi is rumored to be a ~250-300mm^2 chip

2) 7nm allows 2.8x the density of 14nm.

3) It also allows for 40% higher clocks OR 65% less energy usage.

Thus one can conclude that at a minimum Navi should be in-between the 1080 and 1080 Ti (RTX 2070), and at best it could actually compete with the 2080 Ti (not a joke). It should also do so while using at most 100-200w (RX 580 power usage).

We also know that Navi is the last version of GCN, and it is built for gaming first unlike the compute-focused Vega. So there could be extra dividends paid in gaming perf/watt.
I really don’t think it is that simple. As far as I am aware 7nm isn’t even ready yet, it’s just on course to be ready. And when it is I don’t think we will see 40% clock speed improvement. Clock speed clearly isn’t just down to the size of the process node as if it was then AMDs 12nm parts would be able to clock higher than Intels 14nm parts but the real story is that Intels 14nm parts can clock much faster than AMDs 12nm stuff and are more energy efficicient clock for clock too. My prediction is that AMDs next major GPU architecture will be a mid range competitor at best.

It will be interesting to see how things play out. Personally I can’t see the loss of global foundries from production of cutting edge silicon to be anything but a bad thing for its partners and the industry as a whole.
 
  • Like
Reactions: ctm17

CaptainTom

Posts: 414   +220
I really don’t think it is that simple. As far as I am aware 7nm isn’t even ready yet, it’s just on course to be ready. And when it is I don’t think we will see 40% clock speed improvement. Clock speed clearly isn’t just down to the size of the process node as if it was then AMDs 12nm parts would be able to clock higher than Intels 14nm parts but the real story is that Intels 14nm parts can clock much faster than AMDs 12nm stuff and are more energy efficicient clock for clock too. My prediction is that AMDs next major GPU architecture will be a mid range competitor at best.

It will be interesting to see how things play out. Personally I can’t see the loss of global foundries from production of cutting edge silicon to be anything but a bad thing for its partners and the industry as a whole.
  1. They have already showed off 7nm Vega, and it is set to launch by Q1 2019 (Rumors of Q4 happening too).
  2. You seem to misunderstand 7nm in general. TSMC's own slides show you can get either 40% higher clocks OR 65% less power usage per transistor. That means AMD will likely pick something in the MIDDLE - like a 10% faster clock and 50% less power usage. However they can also fit 2.8x as many transistors in the same space, and thus a 250mm^2 Navi card would have more SP's than Vega 64! Clock that a little faster and you have a 1080 Ti/2080 competitor using 150w. Anything less and Navi is a failure.
  3. I have no clue what you are talking about regarding your CPU statements. Ryzen is substantially more efficient than the chips Intel is currently making on their superior process (Intel's 14nm is ~30% better than GF's 14nmFF). The only thing Intel has is a raw performance crown in single threaded apps. That is literally it.
 

Sausagemeat

Posts: 409   +204
  1. They have already showed off 7nm Vega, and it is set to launch by Q1 2019 (Rumors of Q4 happening too).
  2. You seem to misunderstand 7nm in general. TSMC's own slides show you can get either 40% higher clocks OR 65% less power usage per transistor. That means AMD will likely pick something in the MIDDLE - like a 10% faster clock and 50% less power usage. However they can also fit 2.8x as many transistors in the same space, and thus a 250mm^2 Navi card would have more SP's than Vega 64! Clock that a little faster and you have a 1080 Ti/2080 competitor using 150w. Anything less and Navi is a failure.
  3. I have no clue what you are talking about regarding your CPU statements. Ryzen is substantially more efficient than the chips Intel is currently making on their superior process (Intel's 14nm is ~30% better than GF's 14nmFF). The only thing Intel has is a raw performance crown in single threaded apps. That is literally it.
I feel that it is you who is misunderstanding. TSMC state that up to a 40% clock increase is possible. They don’t garuantee it for any CPU. It depends on the architecture, clearly Ryzens clock speed isn’t being held back by their process node. Clock speed isn’t completely based on process node. You also need to look up the clock for clock comparisons out there. When both chips are clocked the same Intel chips use less power than the Ryzen parts but perform slightly faster, this is why they are more efficient, does this shed any light on your apparent confusion?

I know they have already shown off 7nm Vega however whether it will arrive on time is another thing. Most of AMDs recent launches have arrived quite late. Il believe it when I see it.

At the end of the day none of us really know what is going to happen. But I certainly think assuming a 40% clock bump will happen from just a die shrink is naive.
 

Lew Zealand

Posts: 1,491   +1,466
TechSpot Elite
Assumptions of future performance are no more than pipe dreams, like sports teams at the beginning of the season. Performance is almost always lower than assumed and cost is almost always higher for that performance. It seems that AMD fans engage in a bit more of this, which is of course understandable as AMD had been behind in CPU performance and is still behind in top end GPU performance.

I'm willing to give AMD the business if they deliver but all this talk about AMD will do this while Intel can't? Still wishful thinking. Product names next to Steve's blue bars or it didn't happen.
 
We also know that Navi is the last version of GCN, and it is built for gaming first unlike the compute-focused Vega. So there could be extra dividends paid in gaming perf/watt.
This is the first time I've heard anyone mention Navi being made for gaming first, mind giving me a source? :)

If you're talking about Sony, PS4 went through the same semi-custom process where some features/specs requested by Sony made it into GCN.
 
Last edited:

Bohefus

Posts: 11   +5
Please be worthy cheaper graphic card! I want a good 300 400 upgrade!
That's the market AMD is targeting so you should get your wish.
You do realize that they "Target" markets based off performance. The Vega cards were actually over priced based off the performance. If they beat the comparable Nvidia competition, they will be priced higher than the competition. AMD Fanboys think this company is acting in their best interest. Believe me, they're acting in the share holders best interest.
 

BigBoomBoom

Posts: 84   +68
  1. They have already showed off 7nm Vega, and it is set to launch by Q1 2019 (Rumors of Q4 happening too).
  2. You seem to misunderstand 7nm in general. TSMC's own slides show you can get either 40% higher clocks OR 65% less power usage per transistor. That means AMD will likely pick something in the MIDDLE - like a 10% faster clock and 50% less power usage. However they can also fit 2.8x as many transistors in the same space, and thus a 250mm^2 Navi card would have more SP's than Vega 64! Clock that a little faster and you have a 1080 Ti/2080 competitor using 150w. Anything less and Navi is a failure.
  3. I have no clue what you are talking about regarding your CPU statements. Ryzen is substantially more efficient than the chips Intel is currently making on their superior process (Intel's 14nm is ~30% better than GF's 14nmFF). The only thing Intel has is a raw performance crown in single threaded apps. That is literally it.
And you think leather jacket doesn't have any info on 7nm? I'm betting the fact that 7nm isn't as good as you think, or the fact that GF is now out off the race the price will surely surprises you, is the reason behind ginormous RTX cards pricing. Especially when AMD needs TSMC for both CPU and GPU, expect your wonderful Navi to be >$500, maybe $800 more like it.

Also the fact that NVIDIA raised the GPU pricing in general just means AMD will take advantage of it, but not the way you want. AMD GPU will also rise in price. An RTX 2080 Ti competitor in 2019 will be $800 MSRP, welcome to the screwed up future.
 

Sausagemeat

Posts: 409   +204
And you think leather jacket doesn't have any info on 7nm? I'm betting the fact that 7nm isn't as good as you think, or the fact that GF is now out off the race the price will surely surprises you, is the reason behind ginormous RTX cards pricing. Especially when AMD needs TSMC for both CPU and GPU, expect your wonderful Navi to be >$500, maybe $800 more like it.

Also the fact that NVIDIA raised the GPU pricing in general just means AMD will take advantage of it, but not the way you want. AMD GPU will also rise in price. An RTX 2080 Ti competitor in 2019 will be $800 MSRP, welcome to the screwed up future.
I’m suspecting this. GF pulled out of 7nm as they said it wasn’t profitable enough. That leaves TSMC as the only option AMD have who have gambled all their development into 7nm chips. TSMC can charge what they like. Either AMD will end up paying or the consumer will. I suspect the latter.
 
  • Like
Reactions: meric

TechCat

Posts: 30   +19
And you think leather jacket doesn't have any info on 7nm? I'm betting the fact that 7nm isn't as good as you think, or the fact that GF is now out off the race the price will surely surprises you, is the reason behind ginormous RTX cards pricing. Especially when AMD needs TSMC for both CPU and GPU, expect your wonderful Navi to be >$500, maybe $800 more like it.

Also the fact that NVIDIA raised the GPU pricing in general just means AMD will take advantage of it, but not the way you want. AMD GPU will also rise in price. An RTX 2080 Ti competitor in 2019 will be $800 MSRP, welcome to the screwed up future.
I’m suspecting this. GF pulled out of 7nm as they said it wasn’t profitable enough. That leaves TSMC as the only option AMD have who have gambled all their development into 7nm chips. TSMC can charge what they like. Either AMD will end up paying or the consumer will. I suspect the latter.
No GF said rather than invest 10B in 7nm they wanted to go for short term gains. They have a history of backing out of new tech. It has nothing to do with your inference of not profitable. Most of your other conclusions show how little you have been following this topic. Stated clearly you uninformed trolling is not appreciated here. Statements like 'no one knows what is coming' as an excuse to make unfounded unsupported assumptions dose not add to the topic. Big money investors who do their due diligence, don't agree with you either. AMD is on the right path and considering their available funds doing remarkably well. Intel is taking notice and is reacting as fast as a large lumbering company can with a rush to market an price slashing anything they can come up with. Nvidia is taking notice and is inserting expensive GDDR6 un-necessary Tensor and RT into older tech that is clocked to the max (FE is Dual Fan) to justify a high price needed to offset 1 Mil returned die, while all new tech is around the corner. AMD has consoles in the US and China and X86 64 CPUs in China, is gaining in server market CPUs and with new VM tech, in mobile APUs, embedded APUs, in workstation CPUs, in desktop CPUs and APUs, in pro Graphics with Radeon Pro Duo, Radeon Pro SSG and Radeon Pro Series for Apple iMacs.
 
  • Like
Reactions: CaptainTom

gamoniac

Posts: 366   +108
I hope GF can pull it together and work with AMD. TSMC is too close to China for comfort.
Xenophobia mentality is detrimental to technology advancement. The whole world is depending on each other. Other countries are heavily dependent on a myriad of US products and technologies (CPU, medical devices, automobile, etc.) I am not suggesting a naïve utopia. Everyone is sure to keep an eye out for self-protection, but our observation should be based on facts rather than sentiment alone.

Nonetheless, the US did threaten to ban exporting Qualcomm ARM chips to Chinese company ZTE a couple months ago, sending them into panic mode. So, who knows, maybe others are just as wary of us now.
 

Sausagemeat

Posts: 409   +204
No GF said rather than invest 10B in 7nm they wanted to go for short term gains. They have a history of backing out of new tech. It has nothing to do with your inference of not profitable. Most of your other conclusions show how little you have been following this topic. Stated clearly you uninformed trolling is not appreciated here. Statements like 'no one knows what is coming' as an excuse to make unfounded unsupported assumptions dose not add to the topic. Big money investors who do their due diligence, don't agree with you either. AMD is on the right path and considering their available funds doing remarkably well. Intel is taking notice and is reacting as fast as a large lumbering company can with a rush to market an price slashing anything they can come up with. Nvidia is taking notice and is inserting expensive GDDR6 un-necessary Tensor and RT into older tech that is clocked to the max (FE is Dual Fan) to justify a high price needed to offset 1 Mil returned die, while all new tech is around the corner. AMD has consoles in the US and China and X86 64 CPUs in China, is gaining in server market CPUs and with new VM tech, in mobile APUs, embedded APUs, in workstation CPUs, in desktop CPUs and APUs, in pro Graphics with Radeon Pro Duo, Radeon Pro SSG and Radeon Pro Series for Apple iMacs.
I clearly know more about his topic than you. You are of course incorrect in your statement. It is for economic reasons that Global Foundries is pulling out;

https://www.google.co.uk/amp/s/arstechnica.com/gadgets/2018/08/amd-to-use-tsmc-to-make-its-7nm-cpus-gpus-as-glofo-abandons-7nm-development/?amp=1

“GloFo's move is driven by economic considerations—the expected outlays for 7nm, 5nm, and beyond are expected to be enormous ($10 billion or more for a 7nm fab using extreme ultraviolet lithography), and GloFo would need to produce vast numbers of chips to amortize these costs. The company believes its money is better spent on refining its existing processes as such“

Or in other words, there isn’t much profit to be made. I’m clearly more informed than you are. And when it comes to the next gens chip performance none of us really do know what’s coming. If you claim you do you are lying. I think AMD are on the right path but if you think losing a foundry is anything but bad news for AMD or anyone in the industry then you really havent got a clue. They are gambling a lot on 7nm. If it comes off then it’s great news but it is a big gamble and their biggest fab pulling out is a bad sign. You seem quite eager to defend AMD and all their actions. However their steadily growing share price took a hit yesterday with this news, so perhaps you should refrain your claim that big investors know what they are doing. There is huge uncertainty at this point. TSMC have refined their 7nm for mobile parts but we haven’t seen a single X86 part out of them yet, they have quite a lot of pressure if they are now to exclusively deliver everything AMD requires. So much uncertainty that even Papermaster is appealing to the press. We even had another AMD exec leave although I wouldn’t read to much into that myself.