New Zen microarchitecture details

[krumme]

I think they maybe able to produce more than 50M.
https://youtu.be/zyYiDfws9tY

Also I read somewhere, can't remember where now, that AMD has reached out to Samsung to produce more.

They certainly seem confident. I just hope they don't pull a Bulldozer .

My point is people have to calculate this to the end to get a view of the consequences.
Amd is not getting to be a 40B company next year or two when zen ramps up.

Lets say they used 50% capacity as in this video and got 75 8c dies a month.

Did this man actually know it was 50% more than a low end bobcat that sold like crazy? When pc market was far bigger.

Its an insane number of dies even if server take more than half of it.
You cant sell 2M 8c desktop cpu a month with the name AMD printed on it in the price range 400 to 800 usd. AMD will sell like what 200.000 at most?

The cost is sunk for r&d process and fab and you have a business running. What matters now with a variable cost of aprox 30 usd a die is basically getting most revenue. End of story. The wsa just put emphasis on it.

The revenue is where the core line is.
Its not a matter of performance but segments. Few people grasp that.

If you want to move several millions each month you have to adress those segments. No matter what you have in hand. If you have a cpu that can compete then you can adress it. If not well you cant. But you cant change the segments. Especially when you are amd.

I dont know if amd will hurt their brand and lower profit by lowering prices q3 and q4. It seems to be their prior strategy but if they ramp to 50% capacity and desktop takes half and it translates into aprox 3M cpu a month this is imo the price the market can take competing with Intel:

Sr7 300 to 500
Sr5 200 to 300
Sr3 100 to 200

Sr3 is 4c comming later. Sr7 8c 16t. Sr5 is what suits. But they will go for selling a lot in the 300 to 400 range as thats what they have. And its expensive enthusiast range.
They might have a special binned black fx skull edition with funky cooler for 700 usd but its borderline interesting.
They might find it more profitable selling to server market as its limited what consumer market can take. If its highly efficient us consumer might get less

 

[Dresdenboy]

Did AMD fix that bug? That would doom a Ryzen server farm - so the uop cache would need to be turned off indefinitely.
8/8 etc. would be a good fix for consumers.

Due to their apparent single-die strategy they don't have consumer and server dies, they just have Zeppelin dies with different steppings, usable for either market (likely going through different test+validation steps). The only drawback for servers could be a delayed availability of usable steppings for validation.

 

[AtenRa]

AMD should aggressively target the $200/230 Core i5 to $300/350 Core i7 Kabylake market. Take a nice % of this market and they can make tons of money.

Core i5 4460 = $195
Core i7 6700K = $350
Core i7 5820K = $425

Those 3 are what AMD should aggressively attack, they are 1st, 4th and 6th in revenue.

1. 4c 8T RYZEN at $199 to attack the $200/230 market
2. 6C/12T or 8T/8C RYZEN at $289 to attack the $300/350 market
3. 8C 16T RYZEN at $399/450 for "we have a High-End Desktop CPU like Intel people"

1+2+3 . With an ASP of $230 (AMD CPU prices , not retail prices) with just 1 Million CPU sales volume per month , they will make 2,8Billions in revenue in one year (just from the Desktop RYZEN CPUs alone).

With just 35% margins, this will translate to 1Billion of profit.

First half of 2016

https://blog.neweggbusiness.com/components/10-best-selling-cpus-2016-first-half/

Intel Core i7-6700K dominated CPU sales for Newegg B2B customers in the first half of 2016. It outsold the next most popular desktop processor, a Haswell Intel Core i5, more than four times over in sales volume.

Best Selling CPUs 2016—First Half
Sorted by Sales Volume (# sold)

1. Intel Core i7-6700K

2. Intel Core i5-4460

3. Intel Core i7-4790K

4. AMD FX-6300

5. Intel Pentium G3258

6. Intel Core i5-6500

7. Intel Core i7 5820K

8. AMD FX-8350

9. Intel Core i7-4790

10. AMD FX-4350


Best Selling CPUs of 2016—First Half
Sorted by Total Revenue ($ sold)

1. Intel Core i7-6700K

2. Intel Core i7-4790K

3. Intel Core i7-5960X

4. Intel Core i7 5820K

5. Intel Core i7-5930K

6. Intel Core i5-4460

7. Intel Core i5-6500

8. Intel Core i7-4790

9. Intel Core i7-6700

10. Intel Xeon E5-2620

 

[ShintaiDK]

You cant sell 2M 8c desktop cpu a month with the name AMD printed on it in the price range 400 to 800 usd. AMD will sell like what 200.000 at most?

You cant as Intel either. A reality check for people, look at steam. There isn't even half a million 8 core CPUs after several years with 8 core CPU offerings. FX8xxx, FX9xxx, X series HEDT etc.

Even if you include 6 cores, then you are at 2½-3 million combined 6 and 8 cores installed over the last 5-6 years or so.

Its a very tiny market to begin with that wouldn't exist without servers.

Sr7 300 to 500
Sr5 200 to 300
Sr3 100 to 200

I sure dont hope AMD will end up anywhere near the left numbers. The production cost of these Zeppelin dies are twice or more of those they replace.

What AMD really needs is a 4C/8T and lower Zen APU. Zeppelin chips cant be used in laptops, business desktops and most non gaming desktops.

 

[AtenRa]

Noctua offers free upgrade-kits for AMD Ryzen platform
http://www.guru3d.com/news-story/noctua-offers-free-upgrade-kits-for-amd-ryzen-platform.html

Cooler Master ready for Kaby Lake and Ryzen
http://www.guru3d.com/news-story/cooler-master-ready-for-kaby-lake-and-ryzen.html

 

[Tuna-Fish]

Did AMD fix that bug? That would doom a Ryzen server farm - so the uop cache would need to be turned off indefinitely.
8/8 etc. would be a good fix for consumers.

Yes. It was fixed with a new stepping. The speculative is in case the old stepping was mass-produced before the bug was found. I don't know if it was.

 

[Atari2600]

I sure dont hope AMD will end up anywhere near the left numbers. The production cost of these Zeppelin dies are twice or more of those they replace.

Are they?

http://semiengineering.com/finfet-rollout-slower-than-expected/

If that’s not enough, there is also a sizable jump in manufacturing costs. In a typical 11-metal level process, there are 52 mask steps at 28nm. With an 80% fab utilization rate at 28nm, the loaded manufacturing cost is about $3,500 per 300mm wafer, according to Gartner.

At 1.3 days per lithography layer, the cycle time for a 28nm chip is about 68 days. “Add one week minimum for package testing,” Wang said. “So, the total is two-and-half months from wafer start to chip delivery.”

At 16nm/14nm, there are 66 mask steps. With an 80% fab utilization rate at 16nm/14nm, the loaded cost is about $4,800 per 300mm wafer, according to Gartner. “It takes three months from wafer start to chip delivery,” he added.

For a 8C Ryzen die to cost twice that of a BD die, it would have to be ~1.5x bigger in die area than Excavator. Is that the case?

 

[ShintaiDK]

Are they?

http://semiengineering.com/finfet-rollout-slower-than-expected/

For a 8C Ryzen die to cost twice that of a BD die, it would have to be ~1.5x bigger in die area than Excavator. Is that the case?

Yes they are. IC design is 2x or more (~3x in your article). Per transistor cost is higher.

 

[Agent-47]

For a 8C Ryzen die to cost twice that of a BD die, it would have to be ~1.5x bigger in die area than Excavator. Is that the case?

Depends if you compare a 2M Excavator to a 4C zen part. if that is the case, it can easily be larger.
there is L3 in zen which can increase the die by 60-80% easily, although increased L3 size will probably be countered by a lack of iGPU.

another thing to consider is that 28nm has been around for "ages" and most definitely enjoying peak yields. 14nm. Compared to 28 nm, 14 nm is still in its infancy and can be argued that it will not have as high a yield. Time will however negate this factor.

 

[ShintaiDK]

EUV will perhaps fix part of the issue, time in itself wont.

 

[leoneazzurro]

The cost may be probably higher, yes, but that article is dated 2015. It's a lot in terms of manufacturing prowess. All new processes cost much more at the ramp-up, while when it's widely adopted it costs less.

 

[AtenRa]

Are they?

http://semiengineering.com/finfet-rollout-slower-than-expected/



For a 8C Ryzen die to cost twice that of a BD die, it would have to be ~1.5x bigger in die area than Excavator. Is that the case?

ZEN dies at 200mm2 TODAY could be close to 1.8x more expensive than 300mm2 Bulldozer dies at 28nm BUT,

If cheapest ZEN die price is $199, then its 2.5x times higher than cheapest Bulldozer FX4350 at $80.

4C 8T ZEN at $199 will outsell Bulldozer 2M 4T FX4350 by more than 2x
Add higher profit per die and you still make more money than selling Bulldozer dies at $80

 

[ShintaiDK]

The cost may be probably higher, yes, but that article is dated 2015. It's a lot in terms of manufacturing prowess. All new processes cost much more at the ramp-up, while when it's widely adopted it costs less.

It looks more like something like this:

ZEN dies at 200mm2 TODAY could be close to 1.8x more expensive than 300mm2 Bulldozer dies at 28nm BUT,

If cheapest ZEN die price is $199, then its 2.5x times higher than cheapest Bulldozer FX4350 at $80.

4C 8T ZEN at $199 will outsell Bulldozer 2M 4T FX4350 by more than 2x
Add higher profit per die and you still make more money than selling Bulldozer dies at $80

Ye what could possible go wrong with that prediction.

 

[AtenRa]

It looks more like something like this:

Ye what could possible go wrong with that prediction.

Yes AMD will have hard time outselling the FX4350 TODAY with 4C 8T ZEN

edit: Not to mention those tables above are from 2015.

 

[ShintaiDK]

Yes AMD will have hard time outselling the FX4350 TODAY with 4C 8T ZEN

edit: Not to mention those tables above are from 2015.

So now you add today

If the 4C/8T Ryzen had an IGP it would be different. But lets be honest, a lot less desktops and a lot less graphics cards are sold now. And Ryzen better outsell FX4350 today. Even a 64$ Pentium wipes the FX line and it comes with an IGP too.

Yes and the world is quite different today. Feel free to link something else if you think its wrong.

 

[AtenRa]

Yes and the world is quite different today. If the 4C/8T Ryzen had an IGP it would be different. But lets be honest, less desktops and less graphics cards are sold now.

Feel free to link something else.

FX4350 DOESNT HAVE AN IGPU

Did you actually see the link i gave above ???

https://blog.neweggbusiness.com/components/10-best-selling-cpus-2016-first-half/

First haff of 2016 and highest volume CPU at Newegg was the Core i7 6700K, a $350 CPU.
Desktop RYZEN will aim the DIY retail market, those that build their own Systems like you and me and almost everyone in this forum. This market highest volume is from $199 to $450, meaning a 4C 8T $199 RYZEN CPU will easily outsell a 2M 4T FX4350 today.

 

[ShintaiDK]

FX4350 DOESNT HAVE AN IGPU

Did you actually see the link i gave above ???

https://blog.neweggbusiness.com/components/10-best-selling-cpus-2016-first-half/

First haff of 2016 and highest volume CPU at Newegg was the Core i7 6700K, a $350 CPU.
Desktop RYZEN will aim the DIY retail market, those that build their own Systems like you and me and almost everyone in this forum. This market highest volume is from $199 to $450, meaning a 4C 8T $199 RYZEN CPU will easily outsell a 2M 4T FX4350 today.

You have no idea how small the DIY market is do you?

 

[AtenRa]

You have no idea how small the DIY market is do you?

The DIY is the same market the FX4350 is sold, and what i said is that 4C 8T will outsell the FX4350. I dont really care how large the DIY market is, but we do know Intel sold 4x more $350
CPUs than $195 CPUs in DIY market..

So again we are not talking about the entire CPU market but only for the DIY and in this market AMD will easily outsell the FX4350 with RYZEN.

 

[DrMrLordX]

Eh no that's squarely where they are aiming with their top Ryzen product as its been stated many times by them. Don't worry yourself so much over AMD's marketing plan, the products should sell themselves.

Nobody from AMD has said that they are aiming squarely at the HEDT market; in fact, they would be insane to do so since the volumes are so low. I am certain that AMD will be quite content to take a piece out of the much-larger 4790k/6700k/7700k pie.

They only want to prove that their 8c/16t CPU can perform as well as an Intel CPU with the same core count (in this case, the 6900k).

How much are the GTX 1070 and 1080 selling for right now? And you think there ain't enough people willing to spend $400-$600 for a product that will likely last for 2-3x times longer in their system?

GPUs != CPUs. Different markets, different dynamics.

AMD can have a halo SKU for $1K+. Heck maybe it is desirable to have such an SKU. What matters most are the products at $200/$300 price points.

Actually, I think AMD will go as high as $500-$600 with their halo product, but not $1k. Yet. If they establish themselves as a performance leader (or something close to that), they'll explore boutique SKUs.

AMD should aggressively target the $200/230 Core i5 to $300/350 Core i7 Kabylake market. Take a nice % of this market and they can make tons of money.

You are basically correct. Though I am surprised by how many 5820ks Intel has sold.

Noctua offers free upgrade-kits for AMD Ryzen platform
http://www.guru3d.com/news-story/noctua-offers-free-upgrade-kits-for-amd-ryzen-platform.html

Kind of sucks that they require proof-of-purchase for an AM4 motherboard. I guess they don't want people reselling the kits that don't otherwise need them.

Anyone who wants the AM4 mounting hardware today is going to have to pay ~8 euros for it. Might be worth it to avoid the wait.

 

[AtenRa]

So now you add today .

The "TODAY" was in the first post, pay attention next time

ZEN dies at 200mm2 TODAY could be close to 1.8x more expensive than 300mm2 Bulldozer dies at 28nm BUT,

If cheapest ZEN die price is $199, then its 2.5x times higher than cheapest Bulldozer FX4350 at $80.

4C 8T ZEN at $199 will outsell Bulldozer 2M 4T FX4350 by more than 2x
Add higher profit per die and you still make more money than selling Bulldozer dies at $80

 

[ShintaiDK]

The DIY is the same market the FX4350 is sold, and what i said is that 4C 8T will outsell the FX4350. I dont really care how large the DIY market is, but we do know Intel sold 4x more $350
CPUs than $195 CPUs in DIY market..

So again we are not talking about the entire CPU market but only for the DIY and in this market AMD will easily outsell the FX4350 with RYZEN.

So goalpost move again?

By DIY you even have to limit it to Newegg. How is the trend in the rest of the world? Specially Asia. The US market is rather small these days.

 

[dacostafilipe]

Desktop RYZEN will aim the DIY retail market

I'm not so certain about that.

I see a lot of potential for the 8C/16T in Workstations and the small server market will love it. Depending on how great XFR works, it could also be an interesting alternative for pre-build gaming rigs (ex Alienware).

Let's hope we get enough CPUs to even build our own PCs ...

 

[ShintaiDK]

The "TODAY" was in the first post, pay attention next time

No, your first post about it didn't include today in relation to the FX4350. It only included that Ryzen dies would cost 1.8x more than FX dies to manufactor.

 

[cytg111]

Why you do this?
Ryzen is starting in the server space and moving on from there.
The DIY market, while i have no numbers, I am sure is miniscule.

1. Servers
2. APU's and OEMs
3. Mobile
4. ??
5. Profit.

Anything we get inbetween as DIY'ers is scraps - as its allways been (with exception of HEDT).
Ryzen hasnt even released yet and you are discussing adoptation numbers in the scraps-area as a monetary success criteria for AMD

 

[AtenRa]

You are basically correct. Though I am surprised by how many 5820ks Intel has sold.

Those are only DIY volume from Newegg, its not surprising that enthusiasts will want 6Core High-End Desktop CPUs. Its why 6700K was selling 4x more than second Core i5 CPU in that list.