AMD Zen - Key Dates and Information

[Abwx]

Noctua, Be Quiet and Enermax will provide adaptators for free for whom has bought one of their cooling appartus, this is according to Hardware.fr, but the most interesting is that there s some info between the lines :

Ces kits, qui concernent toutes les solutions de la marque (exception faite du Liqtech120X), seront disponibles au lancement de Ryzen, ou au plus tard le 1er mars.

http://www.hardware.fr/news/14940/am4-enermax-lepa-genereux-leur-tour.html

"Those kits....(.....)......, will be available at Ryzen launch , or 1st of march at the latest."

So we know now that it will launch before march...

 

[zinfamous]

I always wonder what people are thinking when they use the term "ripped off." The fact that the retail price may be 10x the cost of goods sold is irrelevant. Especially when considering products that require magnitudes more in sunk costs, such as R&D and manufacturing equipment.

yep. calculating the cost of a die as "$34" because that's the cost of a single die popped off of a wafer is just, well, incredibly naive. I guess AMD was sitting around for the last 5 years with their thumbs up their butts waiting for Zen to just come into being through inertia.

 

[zinfamous]

Noctua, Be Quiet and Enermax will provide adaptators for free for whom has bought one of their cooling appartus, this is according to Hardware.fr, but the most interesting is that there s some info between the lines :



http://www.hardware.fr/news/14940/am4-enermax-lepa-genereux-leur-tour.html

"Those kits....(.....)......, will be available at Ryzen launch , or 1st of march at the latest."

So we know now that it will launch before march...

eh, that actually says that they may release the coolers by March 1st at the latest, in the event that Ryzen launches after March 1st...

 

[jackstar7]

349$ for 8 cores Zen is more than a sufficient ripoff. It would be like paying 700$ or more for a Polaris 10.
A 6-8 cores die will cost ...maybe 25-40$, including the box contents.

Make your own then.

 

[imported_jjj]

I always wonder what people are thinking when they use the term "ripped off." The fact that the retail price may be 10x the cost of goods sold is irrelevant. Especially when considering products that require magnitudes more in sunk costs, such as R&D and manufacturing equipment.

You might be overestimating your ability to understand the conversation.
A semi company model would be somewhat like this - 50% gross margins, 30-35% operating expenses, then tax and profits - to keep it simple.
AMD at 99 $ to 349$ for 4 to 8 cores would have way higher margins than 50%.
Manufacturing is outsourced to GloFo and packaging to others. GloFo fabs the wafers at X$, sells them to AMD at twice that and that's it.
Costs are what AMD pays for the damn thing. the wafers, test and packaging, the box contents, the IP, transport. Then,depending on at what point they count revenue, there are the distributors and retailers that could be included as costs - I wasn't including the channel in the 25-40$ costs. You deduct all that from revenue and you got gross margins..In AMD's case the operating expenses (R&D,sales,marketing,administrative) are about 350 million per quarter,non-GAAP.

In the end, the decision on pricing is simple. How do you make the most money out of it. They would do that by targeting the established price points and offering 2x the cores Intel offers.They go higher ,they get much lower volumes. They go lower, they don't really gain volumes just lower revenues and margins so no point in pricing an octa core at 99$, even if they could easily do that.

If AMD prices Ryzen at 99-349$ and they get 60% margins ,lets say they ship 20million units this year(could be twice that really) and generate 4 billions in revenue.Gross profit (so deducting the 40% costs) would be 2.4 billion.
If they price them 50% higher and get 66% margins, they ship 10 million units and generate 3 billions revenue.Gross profit would be 1.98 billion.
If they price them 2x higher,so at 199 to 700$, they ship 5 million units,if that, at 70% margins and revenue would be 2 billion. Gross profit would be 1.4 billion.
And that's without counting the billions in marketing value as well as GPU and laptop sales that the better pricing would generate.

Sure you can go into further details but that's beyond the point. Like, If they include mask costs in the costs section, volume reduces the cost per unit.Or how volumes help with components makers and retailers investing more in the platform. And many more advantages.

 

[imported_jjj]

Make your own then.

How does that help AMD maximize profits and fully exploit the opportunity?

I get it that some that are new to this PC thing might be confused by AMD removing the GPU from an APU and replacing it with extra cores, you''ve been brainwashed by Intel's marketing but you got to be rational about what is best for AMD.

 

[Abwx]

eh, that actually says that they may release the coolers by March 1st at the latest, in the event that Ryzen launches after March 1st...

They are talking of providing adaptators for coolers that were already bought for previous AMD , or intel, plateforms, they will be available once Ryzen launch and if they are not at this date (that is, when Ryzen launch..) they will be available the 1st of march..

Or do you think that AMD will eventually postpone their date because said firms are late in their plans about a piece of metal that is worth 0.1$...??..

You might be overestimating your ability to understand the conversation.
A semi company model would be somewhat like this - 50% gross margins, 30-35% operating expenses, then tax and profits - to keep it simple.

Manufacturing is outsourced to GloFo and packaging to others. GloFo fabs the wafers at X$, sells them to AMD at twice that and that's it.

We can understand that if manufacturing is outsourced then they dont need to make a 50% GM, 35% is plenty enough to make a lot of $$$, same for Intel, if they were to spun off their factories the CPU division would no more make 60% GM, because those 60% include the factories necessary and own GM...

 

[krumme]

You might be overestimating your ability to understand the conversation.
A semi company model would be somewhat like this - 50% gross margins, 30-35% operating expenses, then tax and profits - to keep it simple.
AMD at 99 $ to 349$ for 4 to 8 cores would have way higher margins than 50%.
Manufacturing is outsourced to GloFo and packaging to others. GloFo fabs the wafers at X$, sells them to AMD at twice that and that's it.
Costs are what AMD pays for the damn thing. the wafers, test and packaging, the box contents, the IP, transport. Then,depending on at what point they count revenue, there are the distributors and retailers that could be included as costs - I wasn't including the channel in the 25-40$ costs. You deduct all that from revenue and you got gross margins..In AMD's case the operating expenses (R&D,sales,marketing,administrative) are about 350 million per quarter,non-GAAP.

In the end, the decision on pricing is simple. How do you make the most money out of it. They would do that by targeting the established price points and offering 2x the cores Intel offers.They go higher ,they get much lower volumes. They go lower, they don't really gain volumes just lower revenues and margins so no point in pricing an octa core at 99$, even if they could easily do that.

If AMD prices Ryzen at 99-349$ and they get 60% margins ,lets say they ship 20million units this year(could be twice that really) and generate 4 billions in revenue.Gross profit (so deducting the 40% costs) would be 2.4 billion.
If they price them 50% higher and get 66% margins, they ship 10 million units and generate 3 billions revenue.Gross profit would be 1.98 billion.
If they price them 2x higher,so at 199 to 700$, they ship 5 million units,if that, at 70% margins and revenue would be 2 billion. Gross profit would be 1.4 billion.
And that's without counting the billions in marketing value as well as GPU and laptop sales that the better pricing would generate.

Sure you can go into further details but that's beyond the point. Like, If they include mask costs in the costs section, volume reduces the cost per unit.Or how volumes help with components makers and retailers investing more in the platform. And many more advantages.

I am in line with this. They might have a 500 shiny part but the 8c beef is at 299 to 399. They will try to develop a premium brand here beeing slightly more expensive than equivalent named core line but keeping that fab utilization at near 100% if possible also because of the wsa. They need to move that gf capacity and not risk a slow uptake. Remember risk is money and this is also about risk management.
Attractive prices where the mainstream buyers is and offer them a unique product moves that capacity. No risk here. The product needs to sell itself by merrits name and price. Amd is in no position to do it differently.
If they end up in a situation where their prices tank after 6 months its very bad for brand development. Prices should stay more or less fixed and its fine if there is a bit of shortage as that signals premium. Think apple. Thats what you want to do.
If they price their zen line to high out the gate and selling slows down i agree it will be a bigger management disaster than bd. This is an opportunity to both maximize total profit and build brand. Looking squarely at margins is a huge error.

 

[imported_jjj]

yep. calculating the cost of a die as "$34" because that's the cost of a single die popped off of a wafer is just, well, incredibly naive. I guess AMD was sitting around for the last 5 years with their thumbs up their butts waiting for Zen to just come into being through inertia.

The R&D is not included in costs, that's not how the math works.
A custom core takes 3-4 years and 300-400 million. However, that core and updates will be used by AMD in all product lines for 4 cycles.Ofc there is lots of other IP and costs for each die.
Do note that if you would factor in R&D costs , the cost per unit decreases with volume so you want to maximize units. Over the 4 years Zen is used ,they could ship over 1 billion cores (cores not SoCs) if they do great in DYI, gain 20% share in remaining PC and just 10% share in server..Or they can ship half that, if they mess up pricing now.

 

[zinfamous]

They are talking of providing adaptators for coolers that were already bought for previous AMD , or intel, plateforms, they will be available once Ryzen launch and if they are not at this date (that is, when Ryzen launch..) they will be available the 1st of march..

Or do you think that AMD will eventually postpone their date because said firms are late in their plans about a piece of metal that is worth 0.1$...??..

uh, lol--no. I understand that this is about releasing a piece of metal for coolers that already exist. But that statement: "To be available at the time of Ryzen release, or March 1 at the latest," could just as easily be interpreted as the metal plates will be out by March 1st at the latest, regardless of Ryzen release (before or after March first).

The metal plates are the subject of that sentence, not the Ryzen release. (eh, unless it wasn't properly translated into English sentence structure to reflect something more clear from the French writers. )

 

[zinfamous]

The R&D is not included in costs, that's not how the math works.
A custom core takes 3-4 years and 300-400 million. However, that core and updates will be used by AMD in all product lines for 4 cycles.Ofc there is lots of other IP and costs for each die.
Do note that if you would factor in R&D costs , the cost per unit decreases with volume so you want to maximize units. Over the 4 years Zen is used ,they could ship over 1 billion cores (cores not SoCs) if they do great in DYI, gain 20% share in remaining PC and just 10% share in server..Or they can ship half that, if they mess up pricing now.

that's not how math works?

OK. But you should know that this isn't about math...

 

[krumme]

that's not how math works?

OK. But you should know that this isn't about math...

Its about marginal cost and marginal profit.
The r&d cost is more or less sunk cost.

 

[imported_jjj]

I am in line with this. They might have a 500 shiny part but the 8c beef is at 299 to 399. They will try to develop a premium brand here beeing slightly more expensive than equivalent named core line but keeping that fab utilization at near 100% if possible also because of the wsa. They need to move that gf capacity and not risk a slow uptake. Remember risk is money and this is also about risk management.
Attractive prices where the mainstream buyers is and offer them a unique product moves that capacity. No risk here. The product needs to sell itself by merrits name and price. Amd is in no position to do it differently.
If they end up in a situation where their prices tank after 6 months its very bad for brand development. Prices should stay more or less fixed and its fine if there is a bit of shortage as that signals premium. Think apple. Thats what you want to do.
If they price their zen line to high out the gate and selling slows down i agree it will be a bigger management disaster than bd. This is an opportunity to both maximize total profit and build brand. Looking squarely at margins is a huge error.

Capacity could be a problem but hopefully won't be as they seem to be using Samsung too. if Ryzen is 200mm2 (quad core native die at, lets be generous, just 30% less area) and yields are not terrible (why would they be ,Polaris is way better priced than Ryzen), they don't need huge capacity. 10 million units per quarter just for Ryzen would be maybe a bit tough but it would be good even if they get half that. GloFo might have 30k wafer starts per month capacity and Samsung was aiming at 40-45k but likely they reduced it after losing Apple and ramping 10nm. AMD might need some 5.5k wafer starts per month for 5 million units. Ofc the die could end up bigger or even quite a bit smaller,we'll see about that.
Intel doesn't server folks that buy discrete GPUs and they haven't been doing so for some years. AMD can force everybody to upgrade and they need to offer sufficient value at every price band to achieve that AND rebuild the brand before Intel reacts. And it's not like they need to be aggressive and kill margins to do so. The smart play is,blowing Intel out of the water, they really won't get the opportunity ever again.

 

[dahorns]

You might be overestimating your ability to understand the conversation.
A semi company model would be somewhat like this - 50% gross margins, 30-35% operating expenses, then tax and profits - to keep it simple.
AMD at 99 $ to 349$ for 4 to 8 cores would have way higher margins than 50%.
Manufacturing is outsourced to GloFo and packaging to others. GloFo fabs the wafers at X$, sells them to AMD at twice that and that's it.
Costs are what AMD pays for the damn thing. the wafers, test and packaging, the box contents, the IP, transport. Then,depending on at what point they count revenue, there are the distributors and retailers that could be included as costs - I wasn't including the channel in the 25-40$ costs. You deduct all that from revenue and you got gross margins..In AMD's case the operating expenses (R&D,sales,marketing,administrative) are about 350 million per quarter,non-GAAP.

In the end, the decision on pricing is simple. How do you make the most money out of it. They would do that by targeting the established price points and offering 2x the cores Intel offers.They go higher ,they get much lower volumes. They go lower, they don't really gain volumes just lower revenues and margins so no point in pricing an octa core at 99$, even if they could easily do that.

If AMD prices Ryzen at 99-349$ and they get 60% margins ,lets say they ship 20million units this year(could be twice that really) and generate 4 billions in revenue.Gross profit (so deducting the 40% costs) would be 2.4 billion.
If they price them 50% higher and get 66% margins, they ship 10 million units and generate 3 billions revenue.Gross profit would be 1.98 billion.
If they price them 2x higher,so at 199 to 700$, they ship 5 million units,if that, at 70% margins and revenue would be 2 billion. Gross profit would be 1.4 billion.
And that's without counting the billions in marketing value as well as GPU and laptop sales that the better pricing would generate.

Sure you can go into further details but that's beyond the point. Like, If they include mask costs in the costs section, volume reduces the cost per unit.Or how volumes help with components makers and retailers investing more in the platform. And many more advantages.

The personal attack wasn't necessary or warranted. And I'm quite comfortable with my understanding of the basics of economics and accounting, thanks.

At any rate, I don't disagree that there is a Goldilocks zone for AMD. I probably disagree on where it is. And I definitely disagree with calling established prices on similar products that have supported robust demand as a "rip off." In fact, I think the term "rip off" is overused in general and should probably be reserved for cases of misrepresentation by sellers and manufacturers. And it certainly doesn't need to be directly related to the cost of goods as that isn't particularly relevant to what a good is worth (e.g., somethings can be really expensive to produce, but not really worth much or vice versa.).

 

[Abwx]

The metal plates are the subject of that sentence, not the Ryzen release. (eh, unless it wasn't properly translated into English sentence structure to reflect something more clear from the French writers. )

The plates release should be made in function of Ryzen release, i think that it s clear enough, that s what is stated, that they should be ready for when the chip launch, but it s possible that they wont be ready at this date, in wich case they will be available starting from 1st march.

As for an english translation i dont need one since i m perfectly fluent in french for some reasons..

 

[CentroX]

Are we complaining at cost for an AMD product? How about that 1100 dollar 8c 6900?

 

[jackstar7]

How does that help AMD maximize profits and fully exploit the opportunity?

I get it that some that are new to this PC thing might be confused by AMD removing the GPU from an APU and replacing it with extra cores, you''ve been brainwashed by Intel's marketing but you got to be rational about what is best for AMD.

What? You're saying it costs so little, sounds like it should be easy for you to make your own and save all that $. Heck, make more and pass the savings on to customers!

 

[krumme]

Capacity could be a problem but hopefully won't be as they seem to be using Samsung too. if Ryzen is 200mm2 (quad core native die at, lets be generous, just 30% less area) and yields are not terrible (why would they be ,Polaris is way better priced than Ryzen), they don't need huge capacity. 10 million units per quarter just for Ryzen would be maybe a bit tough but it would be good even if they get half that. GloFo might have 30k wafer starts per month capacity and Samsung was aiming at 40-45k but likely they reduced it after losing Apple and ramping 10nm. AMD might need some 5.5k wafer starts per month for 5 million units. Ofc the die could end up bigger or even quite a bit smaller,we'll see about that.
Intel doesn't server folks that buy discrete GPUs and they haven't been doing so for some years. AMD can force everybody to upgrade and they need to offer sufficient value at every price band to achieve that AND rebuild the brand before Intel reacts. And it's not like they need to be aggressive and kill margins to do so. The smart play is,blowing Intel out of the water, they really won't get the opportunity ever again.

I think both its very difficult to make people upgrade and take new sales in a mature if not declining desktop market where gaming is a important part of the mid/highend where core line resides but zen isnt going to show the biggest impact for gaming. Then going against an established brand with product that perhaps doesnt have the computational power but never the less is efficient and well known.
I think capacity will be am issue but thats because serverline must be a blast if they can get these clocks with that bwe like ipc for desktop. In a 180w tdp 32c server core that is surely going to make a huge impact. Man. Thats profit. And that can kind of take away the fun for desktop as that will take all the best 8c bins.

 

[imported_jjj]

The personal attack wasn't necessary or warranted. And I'm quite comfortable with my understanding of the basics of economics and accounting, thanks.

At any rate, I don't disagree that there is a Goldilocks zone for AMD. I probably disagree on where it is. And I definitely disagree with calling established prices on similar products that have supported robust demand as a "rip off." In fact, I think the term "rip off" is overused in general and should probably be reserved for cases of misrepresentation by sellers and manufacturers. And it certainly doesn't need to be directly related to the cost of goods as that isn't particularly relevant to what a good is worth (e.g., somethings can be really expensive to produce, but not really worth it or vice versa.).

It wasn't an attack , your comment made you seem very uninformed and my tone was in line with the comment i was replying to. "I always wonder what people are thinking when they use the term "ripped off." The fact that the retail price may be 10x the cost of goods sold is irrelevant. Especially when considering products that require magnitudes more in sunk costs, such as R&D and manufacturing equipment."

When the cost of goods is many times lower than what consumers pay, it is a ripoff. You have a bunch of entities that earn exaggerate amounts of money without merit.
Our system is supposed to be based on competition. And competition is supposed to keep prices sane and keep prices close to the cost of good.
Corrupted politicians, weak regulators, tacit non aggression pacts between market players,marketing , bloated channel and so on, have enabled the current situation where many products are priced very high.
It's not a functional economy that has enabled this, it's not supposed to be like that.

Semi companies do have high margins at 50% in general and that allows for them to invest some 20% of revenue in R&D. A car maker might have 10% margins, a retailer a bit more. A PC maker or phone maker might invest 3% of revenue in R&D. We, the consumers, do actually pay for the R&D.

Ryzen at 349$ vs Polaris 10. Polaris 10 is 232mm2 on 14nm, so likely has a larger die, you got 8GB of GDDR5 and that's quite a lot of $, the actual board and the AiB folks that need to make a buck. So it would offer quite a lot more value than a Ryzen at 349$, thus it's not at all unjustified to say that 349$ is a ripoff. It's a price enabled by the lack of competition in the marketplace and marketing.

Sure when it comes to low volume chips on an advanced process ,the costs devloping that chip are very high so ,the chipmaker would need very high ASPs and margins. That's a recent problem, the ecosystem has failed to sufficiently innovate on the cost side but that's another topic.

This is very off topic but hopefully it will be tolerated. The "ripoffs" are one of our biggest problems. It's one of the major factors leading to income inequality and leading to folks being fed up with the system .It enables corporations to make undeserved profits while reducing the consumer's purchasing power.

EDIT - cost of goods has a slightly different definition than what some might assume

 

[imported_jjj]

What? You're saying it costs so little, sounds like it should be easy for you to make your own and save all that $. Heck, make more and pass the savings on to customers!

https://en.wikipedia.org/wiki/Economies_of_scale

 

[imported_jjj]

I think both its very difficult to make people upgrade and take new sales in a mature if not declining desktop market where gaming is a important part of the mid/highend where core line resides but zen isnt going to show the biggest impact for gaming. Then going against an established brand with product that perhaps doesnt have the computational power but never the less is efficient and well known.
I think capacity will be am issue but thats because serverline must be a blast if they can get these clocks with that bwe like ipc for desktop. In a 180w tdp 32c server core that is surely going to make a huge impact. Man. Thats profit. And that can kind of take away the fun for desktop as that will take all the best 8c bins.

The PC market has declined but in this segment (no need for anintegrated GPU) , it was harmed by Intel's strategy./Mobo sales have gone from 80 million units per year just a few years ago, to likely 50 or less last year. Some will argue that it's all about demand but is it? Or is it also about folks not having what to upgrade to? China and other emerging markets were a big opportunity for growth but folks are spending money on everything else (GPUs, DRAM, SSDs) and not on new CPUs and mobos as Intel hasn't been offering large gains in single core perf or more cores.
I have quite a few friends that build their own PCs and every single one of them is waiting for Zen, no matter if they are running on Nehalem, Sandy or Skylake. If Zen delivers, they'll upgrade,if not ,they won't upgrade for a while.
I would also argue that this segment is a lot more rational and informed than the average consumer. People tend to ask for advice, read reviews and it would be easy for AMD to gain share if they have a great product. In all other segments it would be a lot harder, that is true.

The focus on just gaming is weird.The CPU is not supposed to make much of a difference to begin with and the PC is more than gaming. More cores would lead to devs focusing more on that ,that would be a long term plus. AMD could have high single core turbo with their XFR
Ofc we must not focus on the top SKU and remember all price points. Cheapest quad Skylake was 183$ a few days ago , the last time i've checked. If AMD has quads from 99 to 199$ it can realistically serve quite a lot of folks. The ones that would buy bellow 100$, would likely make an effort and get the 99$ quad. The ones that have a 120-150$ dual core, would likely upgrade to quad or even hexa. It's not just about the 349$ price band- it might be rather popular here but the market is much bigger than that.

The server market this year would likely be up to 25 million units for CPUs. Last year the average selling price was about 600$. How much share can AMD realistically get and what's the average core count? If they get 10% share and 16 cores, that's the equivalent of 5 million 8 cores dies in 2+ quarters. Ofc it would be great for them to get more share but 10% is plenty optimistic in 2+quarters (even 5% would be a good start) and we don't even know yet how competitive they are in this market.
APUs could be quite a bit of volume if they launch early and they get an Apple win but If they arrive in Sept, unlikely they ship more than a few million this year as gaining share in laptop will be hard.

 

[raghu78]

AMD will probably be looking at a lot of factors before determining the price of Summit Ridge CPUs. Firstly what is the overall sales volume of CPUs in each price segment from USD 100 - USD 1000 in 100 USD increments, the competitive positioning against Intel Skylake / Kabylake and Broadwell-E CPUs, anticipated pricing reaction if any from Intel and most importantly what they think is a realistic market share percentage attainable in those price segments. People have to realize that Intel will always have the brand advantage even when AMD have comparable performance. If AMD is providing slightly slower performance than perf/$ must be much better than Intel to move significant market share towards them.

imo AMD needs both market share and massive nett profit dollars to pay off huge debt and fund R&D for future CPU and GPU architectures. imo AMD will probably price their products at USD 200 - USD 600. I doubt they need to go any lower. AMD has Zen based APUs coming towards the end of 2017 / early 2018 which will serve the lower price segments. I think AMD will have atleast 2 dies - a 100 sq mm APU with 2C/4T and a Vega iGPU with probably 384-512 sp and a 200 sq mm APU with 4C/8T APU and a Vega iGPU with 768-1024 sp . The 100 sq mm 2C/4T die can easily go after the USD 60 - 120 segment. The 200 sq mm 4C/8T APU can go after the USD 150 - USD 250 segment.

 

[Shivansps]

When Athlon was released in August 1999 it was faster than Intel Pentium III at lower prices. The Athlon FX was release several years after the original Athlon was introduced and at a time when Athlon CPUs were widely accepted as the better product vs the Intel Piii and P4.
So i believe the same will happen with RYZEN, it will be cheaper than Intel Kabylake.

Athlon XP was mainstream.

 

[R0H1T]

It's not about Intel's HEDT line it's about shipping tens of millions of units not 5 times less. It's about killing Kaby Lake (when it comes to customers that don't need an integrated GPU), not about an irrelevant platform like HEDT. It's about creating marketing value worth many billions. It's about not giving Intel an easy way to strike back and obliterate AMD. It's about boosting GPU and APU sales in the short and long run,It's about taking advantage of a once in a lifetime opportunity.
And doing all that at very high margins.
You are already getting severely ripped off even at 349$.

You think Intel can't throw $ like the last time, when they were lagging AMD & most OEM's had to budge or risk getting wiped out? They did it a decade ago, they did it with Atom, trying to cram it in mobile/tablets & they'll likely do it again if push comes to shove. Don't you think AMD should make all the $ while they can, or just give away cheap octacores hoping it'd wipe Kabylake out with a single (master) stroke?

 

[tential]

just give away cheap octacores hoping it'd wipe Kabylake out with a single (master) stroke?

Explain what this would do exactly?