AMD Confirms, Zen On Track For Q4 2016 Availability On High-End Desktops

[raghu78]

Exactly,everybody remembers when AMD brought out phenomx6 and intel countered with their own 6cores...oh wait,no no but when AMD brought out their 8core FX intel did counter with their own...wait no what?
But finally when AMD brought out their APUs with very strong (entry) level gaming graphics intel countered...

No still Intel keeps doing what they think is right,improving their designs with each generation instead of jumping from one thing to another in desperate hope of something finally "sticking" .

Any skylake CPU gives you much more than any simmilar sandy bridge CPU so more value at the same price.

There are many 2500k/ 2600k users happily enjoying 5 Ghz with water cooling and see no such huge improvement to jump ship. Intel has found higher clocks more difficult to achieve at smaller nodes. The overall gain in single thread performance when you take into account that sandy clocked better than every subsequent design is negligible or not worth the change of an entire CPU+MB+RAM setup. You are better served investing that money into faster GPUs where the improvement is massive.

20% market share in servers is a pipe dream.

As always only time will tell whether AMD can gain share or not. But one thing is sure for AMD to thrive they need atleast 20% market share in servers. AMD has no business case if they are not consistently profitable and able to repay their huge debts. AMD has only 3 years before a huge chunk of that debt (around USD 500 million) comes up for repayment.

 

[mrmt]

Have you forgotten what Intel did to distort the market? And, for all we know, what they're still doing? I really have to wonder what the landscape would look like if AMD hadn't been damaged this way.

Which is not to discount the corrupt SOBs who gutted the company from the inside of course, nor the overvaluation of the ATI merger. It's amazing the company is still going after that, especially with the WSA agreement.

AMD made a lot of money and gained a lot of market share despite of Intel practices, so even if the practices were harmful to AMD they were not decisive for the success or failure of the company or its products. But AMD's own management team managed to squander the profits and debt in order to pursue foolish ventures like Bulldozer, cash out ATI shareholders or APU, this is the kind of harm beyond anything Intel could do to AMD. Give another billion dollars to people like Hector Ruiz, Dirk Meyer or Rory Read and the only thing you'll get is yet another explanation on why their plan didn't work and a bigger bill to foot.

That said, Zen will win or fail by its own merits, not by Intel's disposition.

 

[TheELF]

There are many 2500k/ 2600k users happily enjoying 5 Ghz with water cooling and see no such huge improvement to jump ship. Intel has found higher clocks more difficult to achieve at smaller nodes. The overall gain in single thread performance when you take into account that sandy clocked better than every subsequent design is negligible or not worth the change of an entire CPU+MB+RAM setup.

We are not talking about the same thing...nobody said anything about if it is worth to upgrade or not,but since you brought it up you need to run the 2500/2600 at 5Ghz with the added cost of water cooling and power consumption (as small as it might be) to get something close to skylake with stock air cooling? How is that not higher value?

 

[Dresdenboy]

AMD made a lot of money and gained a lot of market share despite of Intel practices, so even if the practices were harmful to AMD they were not decisive for the success or failure of the company or its products. But AMD's own management team managed to squander the profits and debt in order to pursue foolish ventures like Bulldozer, cash out ATI shareholders or APU, this is the kind of harm beyond anything Intel could do to AMD. Give another billion dollars to people like Hector Ruiz, Dirk Meyer or Rory Read and the only thing you'll get is yet another explanation on why their plan didn't work and a bigger bill to foot.

That said, Zen will win or fail by its own merits, not by Intel's disposition.

Not to forget that they didn't exactly get a good deal with GF+WSA due to the already critical situation.

The problem with such strategies is, that they are speculative per se. There are always unknowns of different dimensions, and there is no book containing information about how to do such a deal. Nope, banks and investors are the consultants a company gets, not AT forum members.

As many like to point out, what went wrong in the past (this is easy, hindsight bias), we might speculate, what AAPL, ARMH, INTC, AMD, NVDA, and the others should do next. How good will we be at this game? How about this: a lot of information and signs needed to detect a planned attack on Pearl Harbour were available long enough before the attack, just not nicely connected. Looking at them now might give the impression, that this was easy to see. Wrong, hindsight bias at work.

 

[Arachnotronic]

As always only time will tell whether AMD can gain share or not. But one thing is sure for AMD to thrive they need atleast 20% market share in servers. AMD has no business case if they are not consistently profitable and able to repay their huge debts. AMD has only 3 years before a huge chunk of that debt (around USD 500 million) comes up for repayment.

AMD's opex is at an annual run rate of ~$1.6 billion. At 35% gross profit margin, they need annual revenue of around $4.6 billion to break-even.

They don't need 20% of the server market to hit that.

 

[DrMrLordX]

Taking into account AMD's reputation at present, don't you think it would be a better decision to gradually move to an Intel pricing level? The worst thing is to release at $999 and within months, drop to $500-600. This destroys brand image.

Surely with the apparent small size of Zen, they can sell at $500-600 and substantially improve their margins.

I agree, but I am not the one making decisions. They put the FX-9590 out at $900 or so. If Summit Ridge is all that and a bag of chips, and in short supply, it will launch at high prices. AMD knows that if they sell a short-supplied chip for which there is demand with an MSRP of $500, that e-tailers will jack that up to $700 or more just on supply/demand alone. So AMD will be thinking, "we need to adjust MSRP to match what the e-tailers will set as a price ceiling for a rare chip". The price might come down in 2017 as the Zen lineup diversifies and there's an increase in production.

Even if we take the 40% IPC claims over Excavator (which would 50% to 60% over Piledriver) as being even closer to 30% to 40% over Piledriver,its still about Ivy Bridge level,which is respectable.

It's going to be very difficult to juxtapose "IPC" between Zen and Excavator, since one is SMT and one is CMT, which leads to potentially-different thread scaling. If Zen winds up scaling like Haswell/Skylake with thread count, 8c/16t Summit Ridge will look artificially quick on workloads spawning 8 or fewer demanding threads, which is not the behavior typical of Construction cores.

People are forgetting that the cores in the Bulldozer line have probably only got to around K10 level with Excavator

Sorry, but that's just flat-out wrong. Even my Streamroller can whip K10 chips in legacy code like SuperPi 32m. When you start using software with more-modern ISA extensions, SR and XV run circles around aged Deneb chips. The extent to which XV can beat Deneb in y-cruncher is insane.

I expect the situation to be like with the Phenom II X6 and Lynnfield Core i5 chips,except AMD probably will have paid a bit closer attention to power consumption this time.

8c/16t Summit Ridge is said to have a 95w TDP. I do not think power consumption will be a big problem, at least not at stock settings.

Personally i wouldnt sell Construction Core ZEN dies with less than 6 Cores.

I don't think AMD is going to sell Construction core Zen dies at all, unless they've secretly been listening to NostaSeronX.

Have you forgotten what Intel did to distort the market?

Most of the regulars remember. The problem is that some folks apparently never saw anything wrong with market distortion. It's just marketing folks . . . right? Right.

Case in point: Intel still hasn't faced any penalty for what they did outside the EU.

And, for all we know, what they're still doing?

The only documented evidence of recent "market distortion" has been what they did in the tablet arena, and AMD has not been the primary target of those efforts. AMD did suffer some collateral damage.

I really have to wonder what the landscape would look like if AMD hadn't been damaged this way.

With Ruiz at the helm . . . maybe not much better.

Which is not to discount the corrupt SOBs who gutted the company from the inside of course,

Ruiz and who else?

It's amazing the company is still going after that, especially with the WSA agreement.

And who is to blame for the WSA? Well, lots of people actually, but mostly it's . . . Hector!

Dude,the 4C/8T is going to be the same huge chip as the 8C/16T(unless they do an MCM) - how the heck are they going to make margins on such a chip??

Harvested dice are essentially "free". You have x% yield, and the good chips go into the price brackets you set for them according to how they bin. Check. Then you either toss the failed silicon, or you gimp it so that it'll pass validation, and you can set pretty much any price you want for it. The only costs associated with those chips is the extra work to validate the chip (if any) and the inherent costs of carrying yet another SKU.

AMD loves to harvest failed dice. Expect 4c and 6c Summit Ridge, probably a few months after launch.

Besides, Zen is designed to be configured in 4c clusters anyway. If AMD wants to launch Summit Ridge with a native 4c/8t variant, the design is ripe for scaling down to that core count. What I would not expect is native 6c.

 

[mrmt]

Not to forget that they didn't exactly get a good deal with GF+WSA due to the already critical situation.

To go asset-heavy and build a new fab was Ruiz decision as soon as he got to the company, and to burn all the cash they had and get some debt was also Ruiz decision (Meyer was COO at the time too), so their critical situation was not a matter of Höherer Gewalt but one of the consequences of a chain of failed strategic decisions, so we shouldn't really pity them. You reap whatever you sow.

As many like to point out, what went wrong in the past (this is easy, hindsight bias), we might speculate, what AAPL, ARMH, INTC, AMD, NVDA, and the others should do next. How good will we be at this game? How about this: a lot of information and signs needed to detect a planned attack on Pearl Harbour were available long enough before the attack, just not nicely connected. Looking at them now might give the impression, that this was easy to see. Wrong, hindsight bias at work.

My bias tells me you are smart enough to know the difference between reaching a conclusion from a single data point (Pearl Harbor) and reaching a conclusion from multiple datapoint (asset-lite, Bulldozer, APU, Globalfoundries, the whole smash).

My bias also tells me that you know that despite a lot of companies, including all you mentioned in your post, facing the same speculative scenarios and decisions they managed to make the right ones, or less damaging ones, and didn't go as close to bankruptcy as AMD went and that distinction is clear as water for you.

What my bias can't get is why you are throwing this smoking screen to defend a very incompetent management team. It's not a single datapoint that makes this management team incompetent but the sum of all data available and the measurements of the results they got to their shareholders. And btw ultimately this analysis will be biased, as the only "victory" that matters for AMD is the commercial/financial victory.

 

[mrmt]

Ruiz and who else?

It's easy to put all the blame on Ruiz shoulders, and while he must bear a lot of it, he wasn't alone making the decisions. For example, just two names, Dirk Meyer was COO by the time of ATI acquisition and Bruce Claflin (current chairman) was BoD member since 2003 and watched all of the screws ups while sitting in his chair. If you dig a little you'll see a few names that went through a lot of turbulent periods at AMD, and had a say on these decisions.

I also think that the BoD and executive managers in general have a very different and deformed understanding of the market they are supposed to master, but this streak of bad decisions is usually more related to deficiencies in their corporate structure (internal controls, strategy, engineering and marketing systems not generating or transmitting correct data to decision makers) than the twisted or incorrect understanding of a single BoD member, no matter how powerful he is. Their greatest failure lay in not remedying these deficiencies and not in any individual decision they made.

You don't kill a company of AMD's size with one bad decision, it takes a lot of bad decisions and time to kill one.

 

[raghu78]

AMD's opex is at an annual run rate of ~$1.6 billion. At 35% gross profit margin, they need annual revenue of around $4.6 billion to break-even.

They don't need 20% of the server market to hit that.

I said to thrive they need 20% of the server market. AMD cannot succeed with breakeven. They need consistent and sustained profitability. They have massive debt to repay (USD 2+ billion). AMD will soon die if they don't find a way to generate cash from profits to repay debts. The challenge is to get to breakeven first and then become consistently profitable and repay the debt.

 

[raghu78]

It's easy to put all the blame on Ruiz shoulders, and while he must bear a lot of it, he wasn't alone making the decisions. For example, just two names, Dirk Meyer was COO by the time of ATI acquisition and Bruce Claflin (current chairman) was BoD member since 2003 and watched all of the screws ups while sitting in his chair. If you dig a little you'll see a few names that went through a lot of turbulent periods at AMD, and had a say on these decisions.

I also think that the BoD and executive managers in general have a very different and deformed understanding of the market they are supposed to master, but this streak of bad decisions is usually more related to deficiencies in their corporate structure (internal controls, strategy, engineering and marketing systems not generating or transmitting correct data to decision makers) than the twisted or incorrect understanding of a single BoD member, no matter how powerful he is. Their greatest failure lay in not remedying these deficiencies and not in any individual decision they made.

You don't kill a company of AMD's size with one bad decision, it takes a lot of bad decisions and time to kill one.

You might not kill with one decision. But you definitely can set in motion a chain of events with a bad decision. The overpaying for ATI acquisition had a ripple effect. It deprived AMD of cash (the lifeblood of any company), saddled it with huge debt and finally forced them to sell their fabs. To make things worse AMD's execution faltered and Intel was back with a vengeance and maintained their dominance thereafter as AMD went from bad to disastrous with Bulldozer. It was the perfect storm and AMD never recovered.

 

[mrmt]

You might not kill with one decision. But you definitely can set in motion a chain of events with a bad decision. The overpaying for ATI acquisition had a ripple effect. It deprived AMD of cash (the lifeblood of any company), saddled it with huge debt and finally forced them to sell their fabs. To make things worse AMD's execution faltered and Intel was back with a vengeance and maintained their dominance thereafter as AMD went from bad to disastrous with Bulldozer. It was the perfect storm and AMD never recovered.

I agree that the ATI acquisition, especially the way it was structured was a very bad business decision, but not the only one. What if AMD at the beginning of the last decade didn't build the Dresden factory and instead deepened the alliance with IBM? That would mean at least 2.5 billion in their balance sheet, a lot less working capital and a management team and engineering resources much less distracted from the development of good products.

 

[IntelUser2000]

We are not talking about the same thing...nobody said anything about if it is worth to upgrade or not,but since you brought it up you need to run the 2500/2600 at 5Ghz with the added cost of water cooling and power consumption (as small as it might be) to get something close to skylake with stock air cooling? How is that not higher value?

Higher value = no

It would be true for *new* system purchase, not someone with a 2500/2600.

1. If you want responsiveness*: 35% over 2600K is negligible. 2600K is very very fast. Maybe you want to spend on an SSD if you don't have one
2. For gaming, you'd want a video card upgrade first. The same $500 spent on a 6700K+Mobo+DDR4 gives you 2x+ performance
3. Multi-threaded, you'd want an -E setup. If you already do, go back to #1, and #2

*My upgrades with responsiveness felt:

1. Celeron D to a Core 2 Duo: Somewhat
2. WD Raptor HDD to a X25-M: Little more
3. Core 2 2.4GHz to Core i5 661 3.5GHz, 60% IPC improvements resulted in really unleashing the X25-M, more felt than HDD to SSD using Core 2

Responsiveness not felt
1. Core i5 661 to Core i7 2600K: Not even a little bit. Understandable, because the single thread improvement is only 20%

As such I still have the 2600K. It'll likely last until I see 40-50% gains in single thread. I assume 7nm tock would do it. Actually what I'd really want is a DDRx + 3DXP DIMM system which allows near instant boot and near instant program loading and scans(spyware/adware/scandisk, etc). Before that 2600K should break down first.

 

[redzo]

AMD Confirms, Zen On Track For Q4 2016 Availability On AMD High-End Desktops

I believe I've said this before about the DC memory design. Has anything changed?

Since Zen is built/deployed only around a dual channel memory subsystem isn't it logical to assume that it will be competing against intel's mid-range platform.

While AMD may call it high performance / high end core design it seems to fit this "high-end" description only regarding their own product stack. They are totally silent about market positioning against their competitor. Or did I miss anything?

This erroneous approach will only exponentially increase people's expectation and it is only going to be toxic for the final product at launch.

 

[AtenRa]

AMD Confirms, Zen On Track For Q4 2016 Availability On AMD High-End Desktops

I believe I've said this before about the DC memory design. Has anything changed?

Since Zen is built/deployed only around a dual channel memory subsystem isn't it logical to assume that it will be competing against intel's mid-range platform.

While AMD may call it high performance / high end core design it seems to fit this "high-end" description only regarding their own product stack. They are totally silent about market positioning against their competitor. Or did I miss anything?

This erroneous approach will only exponentially increase people's expectation and it is only going to be toxic for the final product at launch.

The information we have talks about 95W TDP max, if you believe they will compete against Intel HEDT 130-150W TDP SKUs be my quest.

 

[TheELF]

Higher value = no

It would be true for *new* system purchase, not someone with a 2500/2600.

Who talked about someone who already has a CPU?
Only you!

A CPU having better value or not does not depend on you owning it or not.

Does it make sense for someone today to buy the 2600k instead of a skylake?
No it doesn't because the skylake is much better value today.

 

[Arachnotronic]

Who talked about someone who already has a CPU?
Only you!

A CPU having better value or not does not depend on you owning it or not.

Does it make sense for someone today to buy the 2600k instead of a skylake?
No it doesn't because the skylake is much better value today.

Mind boggling, isn't it?

I think the anti-Intel crowd really grasps at straws these days.

 

[AtenRa]

It's true. What I find so funny is that Intel has been delivering most enthusiasts on this forum great hardware year-after-year and, despite the basically complete lack of competition from AMD, has actually given people more value for their $ with each generation.

Yet, AMD is the "good guy" in their eyes while Intel is the "bad guy"

I didnt want to derail the thread but giving 5-10% higher IPC for the last 5 generations is not what i call "more value for the money with each generation". Especially not with Skylake, that is more expensive (even without counting the lack of heat-sink) than the previous generations and only gives 5-10% higher performance.

 

[ShintaiDK]

Mind boggling, isn't it?

I think the anti-Intel crowd really grasps at straws these days.

So true :thumbsup:

 

[ShintaiDK]

I believe I've said this before about the DC memory design. Has anything changed?

Since Zen is built/deployed only around a dual channel memory subsystem isn't it logical to assume that it will be competing against intel's mid-range platform.

Its dual channel and a x16 PCIe from the CPU. Its essentially identical to the LGA115x build up. The only difference is that the APUs seems to have x8 PCIe only. At least Bristol Ridge.

 

[AtenRa]

I said to thrive they need 20% of the server market. AMD cannot succeed with breakeven. They need consistent and sustained profitability. They have massive debt to repay (USD 2+ billion). AMD will soon die if they don't find a way to generate cash from profits to repay debts. The challenge is to get to breakeven first and then become consistently profitable and repay the debt.

2015 server market was almost 55B US.

Intel x86 Server Chip revenue was close to 15B.

10% will be 1.5B, with 45% margins AMD could make a 650M profit.

That is more than enough to repay the debt of 600M by May 2019, can they win 10% of the server market with ZEN by the end of 2018 ?? I believe they could.

Edit: ehm i knew something was way off, 55B is the entire server market.

 

[Arachnotronic]

2015 server market was almost 55B US.

Server chip market is not $55 billion. Do try to get your facts straight, thanks.

 

[ShintaiDK]

Lol, 55B server chip market. Maybe in 2030

 

[AtenRa]

Yea my bad, fixed that.

 

[TheELF]

I didnt want to derail the thread but giving 5-10% higher IPC for the last 5 generations is not what i call "more value for the money with each generation". Especially not with Skylake, that is more expensive (even without counting the lack of heat-sink) than the previous generations and only gives 5-10% higher performance.

Yes it only gives 5-10% higher performance with lower frequencies and lower power consumption!!! And it has a higher powered igpu inside the same power envelope.

I know a lot of people don't care about all of that,but computing is on it's way to become not only portable but wearable and skylake is a big step towards that direction.

ALSO how come HMB is a big deal feature and everybody should praise it but ddr4 is just a waste of good silicone?

 

[TheELF]

The information we have talks about 95W TDP max, if you believe they will compete against Intel HEDT 130-150W TDP SKUs be my quest.

Yes but is that 95W for one module? (4c/8t)
I mean the haswell i7 (4c/8t) already has 84W so it woudn't be so far fetched.