Zen hasn't taped out yet

[Arachnotronic]

None of what Kumar said makes sense and flies in the face of everything Su has said up till this point.

From Q2 to Q3 Su has talked about a dual tapeout (Q2) and multiple tapeouts over both foundries (Q3). It just doesn't make sense for Zen not to have taped out.

Alt: Kumar is talking about Zen APU. That's about the only thing that makes sense.

He specifically referred to HEDT, not maintream APUs.

 

[Hans de Vries]

This thread started exactly 1 year (365 days) after this one:

http://forums.anandtech.com/showthread.php?t=2411975

So K12 will be on 28nm and there isn't a single mention to Zen in the last two conferences.
It seems that K12 and Zen are indeed low cost, scope constrained projects
that won't have any relevance at all on the consumer and server market.

Is this going to be a yearly tradition ()

 

[imported_ats]

He said that they have no enginering sample, and it s logical since those wont be available before a few months, so i really dont see what is new that would had deserved a new thread.

Currently what they have at hand are partly or fully functional dies that are used to test the different blocks or even the full chip but wich will be slightly redesigned to definitly fix the caracteristics.

Um, ES are any non-RS dies. There is no barrier for function for an ES. In fact, an ES can be completely non-functional.

 

[jhu]

Holy crap, how can someone be so emotionally invested in AMD that they believe in a forum dream more than they believe the company CFO??

Zen doesn't exist. Check back in a year.

Truthiness. If it feels right, then it must be.

 

[imported_ats]

I don't know where these time after tapeout estimates come from. Maybe desktop CPUs have some much slower or stretched out processes. In mobile from tapeout to mass production is about 7 months, of which is 3 months to get silicon back from the fab and the rest is just validation and characterisation. We're talking about high-end FinFET SoCs here.

You are also talking about: it runs our 1 version of the OS and a couple built in applications without any shockingly visible errors so ship it.

Real CPU validation is a bit more in depth with things like: does it run all legacy OSes without error, does it run all versions of that legacy OSes without error, does it do that across different OSes, and across different applications (LOTS AND LOTS of applications). Then does it work with all these hardware addons, etc.

single purpose socs might get released in 4 months after silicon, but it is only because they've had a fraction of the validation, and in my experience, it certainly shows up in device stability.

 

[Abwx]

Um, ES are any non-RS dies. There is no barrier for function for an ES. In fact, an ES can be completely non-functional.

Not the ES that are sent to OEMs a few months before industrial production begin, and here we have the (non official) info that such samples will be avilable only in early 2016, for now they only have test silicon with non definitive microcode.

I repost the slide so everybody can understand how the industry work :

The definitive stepping is projected for april 2016, it would be impossible to have the industrial production starting in august 2016 if the very first test silicon was available only in april, unless of course one believe in april s fools...

Saying otherwise is aknowledgment that one is totaly cluless about how much work is required between a first tape out and industrial production of a CPU, a six months delay would be never heard of...

Notice that industrial production must start in early august for the chips getting out of the fab in late september..

 

[imported_ats]

Not the ES that are sent to OEMs a few months before industrial production begin, and here we have the (non official) info that such samples will be avilable only in early 2016, for now they only have test silicon with non definitive microcode.

ES are sent to partners generally as soon as possible and have little to nothing to do with when production begins. Its not at all uncommon for partners to receive A0/A1 ES samples to support their own work (bios, platforms, et al).

And I don't need some damn slide to tell me how the industry works since I've been in the industry. ES is simply ANY chip that isn't PRQ'd.

 

[dark zero]

Also see the other quote:



So they say a couple of months until tape-out, but then they say it's 12-24 months for product shipment. Not sure if there's anything wrong with my math, but given 2015 is almost over, I don't see how they could be claiming 2016 launch...

In the best case, we might see only the highest tier of Zen at 2016 and at not the performance we are expecting.
And in the worst case, Zen would be an Aborted Product since past 2017, investors won't want to maintain a dead uARCH. Unless AMD is sold to the Chinese.

July 2010 -> October 2011 launch means ~15 months.

If Zen based HEDT CPUs tape out in February 2016, then I'd expect availability in March/April 2017.

Doesn't give AMD too much breathing room before SKL-E hits, it would seem.

We didn't have Broadwell-E, so, SKL-E will take even longer.... hoping that there is not supply issues, if that so, definately 14nm is the cursed generation of Intel and AMD in HEDT.

 

[Abwx]

ES are sent to partners generally as soon as possible and have little to nothing to do with production begins. Its not at all uncommon for partners to receive A0/A1 ES samples to support their own work (bios, platforms, et al).


And I don't need some damn slide to tell me how the industry works since I've been in the industry. ES is simply ANY chip that isn't PRQ'd.

ES sent to OEMs are definitive chips with definitive microcode, OEMs have no time to waste with chips that are potentialy faulty and are not supposed to do the debbugging work...

As for Bioses you are completely out of track as well, the bios is defined by AMD and no one else, the OEMs can eventualy remove a few features but they are limited by what is allowed by AMD microcode, rest is as i said aknowledgment that you are among the ones that do not understand how this industry work, otherwise you wouldnt do such erroneous statements.

 

[burninatortech4]

So AM4 motherboards almost a year before the Zen HEDT platform? Sheesh.

 

[itsmydamnation]

Holy crap, how can someone be so emotionally invested in AMD that they believe in a forum dream more than they believe the company CFO??

Zen doesn't exist. Check back in a year.

Holy crap, how can someone(not just specifically you) be so emotionally invested in bashing AMD that they ignore information from the CFO that doesn't suite there agenda??

lets Occam's razor this:

assumption: CFO knows more about finance then he does EE, eg financial statements should have a higher weighting then technical statements.

Zen was a clean sheet design that started few years ago. We are in the final figure of executing

As a project working towards the final milestone. Would you think AMD expects first silicon to be in the "final figure of execution" given they know full well how many revisions might be needed?

the milestone that you want hear us talk about is Zen tapping out, which should be over the next several months And then putting samples in the hands of our customer

The taping out of silicon, this is the statement open to interpretation and thus we need to look at the other information supplied.

then starting portfolio of revenue in 2017.

key word being portfolio, ie launch of Zen APU, Zen Server and high end desktop

And by the way, because we have this reuse approach for cores

reiterating the above

But the key is tapping out in the next several months, samples and customers for the validation of the product over the 2016 time frame and then the revenue ramp happening in 2016.

Customers will be validating Zen in 2016 and Zen will start selling in 2016.

I think the key is getting through 2016 continue to stabilize the computing and graphics business, commercial, professional graphics embedded and even the high-end desktops here at Zen Core should all be accretive from a margin standpoint.

Specifically stating that that Zen based highend desktop will affect margin and thus revenue during 2016.


Putting all this together, considering how long once ES/final silicon is in hand it takes to bring it to market I think we can say the phrase "Zen tapping out, which should be over the next several months And then putting samples in the hands of our customer" is referring to ES/final that AMD will send to partners/customers.

if I had to guess Zen that ships in Servers will need a new stepping/revision compared to desktop and take significantly longer to validate. This will be because of the distributed memory / SOC interconnect on the server chips that wont be used for desktop, thus the big difference in the expected shipping dates.

 

[LTC8K6]

It will be a miracle if Zen chips are for sale in 2016, imo.

I want this miracle, though. Intel needs a push.

 

[Vesku]

Ignoring what exactly is meant by tapeout, they don't expect to ship for revenue until beginning of 2017. So my expectations haven't changed, at best a few 10s of thousands of units of a high end desktop Zen SKU at end of 2016 everything else showing up in channel in 2017.

 

[imported_ats]

ES sent to OEMs are definitive chips with definitive microcode, OEMs have no time to waste with chips that are potentialy faulty and are not supposed to do the debbugging work...

OEMs are regularly sent chips with known faults and known workarounds so that they can do work on other areas. Microcode is generally done prior to tapeout and the only updates it will receive are bug workarounds.

As for Bioses you are completely out of track as well, the bios is defined by AMD and no one else, the OEMs can eventualy remove a few features but they are limited by what is allowed by AMD microcode, rest is as i said aknowledgment that you are among the ones that do not understand how this industry work, otherwise you wouldnt do such erroneous statements.

Bios is modified routinely by vendors. The only thing that AMD defines and no one else does is the CSR interface/definitions and security measures. Pretty much everything else can be either reference code or customer specific code or any combination in between. Apparently you live in bizarro world where all bios look and function exactly the same...

As for how this industry works, my name is on numerous patents and I was one of the lead architects for QPI, so I think I've got just a bit more of a clue.

 

[VirtualLarry]

Ignoring what exactly is meant by tapeout, they don't expect to ship for revenue until beginning of 2017. So my expectations haven't changed, at best a few 10s of thousands of units of a high end desktop Zen SKU at end of 2016 everything else showing up in channel in 2017.

Delayed for mass availability... much like SKL? (Where are the SKL Celerons?)

 

[krumme]

Nothing have changed.
What matters is yield at shipping time not to say perf and perf/watt.

That is - if what matters for us is price / perf. Still i asume thats the case. As enthusiast waiting for a little compettition in monopoly wonderland for 10 years, 3 months delay for better yield or perf is better than some TTM goal.

 

[frozentundra123456]

Wow, nothing like spinning a delay as a good thing.

 

[Gikaseixas]

This thread started exactly 1 year (365 days) after this one:

http://forums.anandtech.com/showthread.php?t=2411975



Is this going to be a yearly tradition ()

That's why I try not to be part of these discussions.
AMD Bashing Team (ABT) cheers for anything negative about AMD as if this will benefit the end user. Fellas, GPU's are advancing much faster due to competition.

Ofcourse i'm not blind and I see AMD f...ups too but some people here just can't help themselves.

 

[Erenhardt]

Wow, nothing like spinning a delay as a good thing.

LOL! What delay?

 

[coercitiv]

Wow, nothing like spinning a delay as a good thing.

Have you read the transcript?

 

[tential]

Holy crap, how can someone(not just specifically you) be so emotionally invested in bashing AMD that they ignore information from the CFO that doesn't suite there agenda??

lets Occam's razor this:

1.assumption: CFO knows more about finance then he does EE, eg financial statements should have a higher weighting then technical statements.


2.As a project working towards the final milestone. Would you think AMD expects first silicon to be in the "final figure of execution" given they know full well how many revisions might be needed?


3.The taping out of silicon, this is the statement open to interpretation and thus we need to look at the other information supplied.


4.key word being portfolio, ie launch of Zen APU, Zen Server and high end desktop


5.reiterating the above


6.Customers will be validating Zen in 2016 and Zen will start selling in 2016.


7.Specifically stating that that Zen based highend desktop will affect margin and thus revenue during 2016.


Putting all this together, considering how long once ES/final silicon is in hand it takes to bring it to market I think we can say the phrase "Zen tapping out, which should be over the next several months And then putting samples in the hands of our customer" is referring to ES/final that AMD will send to partners/customers.

if I had to guess Zen that ships in Servers will need a new stepping/revision compared to desktop and take significantly longer to validate. This will be because of the distributed memory / SOC interconnect on the server chips that wont be used for desktop, thus the big difference in the expected shipping dates.

Occam's Razor
"Among competing hypotheses, the one with the fewest assumptions should be selected."

The CFO, "Zen is taping out in the next 3 months".
Zen is taping out in the next 3 months.

Zero assumptions.

 

[AtenRa]

The CFO, "Zen is taping out in the next 3 months".
Zen is taping out in the next 3 months.

Zero assumptions.

That is not what the CFO said, read again the transcript

Zen was a clean sheet design that started few years ago. We are in the final figure of executing and the milestone that you want hear us talk about is Zen tapping out, which should be over the next several months.

and

But the key is tapping out in the next several months, samples and customers for the validation of the product over the 2016 time frame and then the revenue ramp happening in 2016.

Now, if they will ship in late 2016 (revenue ramp 2016) and release in Q1 2017 is possible but im expecting a late Q4 2016 release with low availability at first.

Edit: Ohh and one last thing,

The question was about Servers not Desktop, so Devinder Kumar may have only talked about SERVER parts

Hans Mosesmann - Raymond James & Associates.
Okay, fair enough. Got a few more topics to get to before we end here. So, as Zen, the new x86-based processor; high-end. It's your return potentially into the data center/server area. What's the timing? And what kind of share do you guys…?

 

[bystander36]

That's why I try not to be part of these discussions.
AMD Bashing Team (ABT) cheers for anything negative about AMD as if this will benefit the end user. Fellas, GPU's are advancing much faster due to competition.

Ofcourse i'm not blind and I see AMD f...ups too but some people here just can't help themselves.

While competition can help, CPU's have a lot harder time advancing due to their serial nature. With GPU's, they can just throw more cores into a GPU and performance increases. CPU's don't have such luxury.

 

[beginner99]

I hope this is wrong, Zen has already taped out and will delivery. Honestly I'm a big critic of AMD and currently it makes 0 sense to buy any AMD CPU/APU. (It does make sense for GPUs because better performance/$).

Still given AMDs track record, I have doubts, huge doubts Zen will arrive when it actually still matters. And this is IMHO the problem. It has to either delivery on performance or be cheap. The first we can guess. Harwell single-threaded is absolute maximum. So if it releases just 6 month before Skylake-E why should I upgrade to Zen instead of waiting for Skylake-E? We can also pretty much estimate performance of Sky lake-E and lowest tier will almost certainly have 8 cores as well. Only reason is if Zen is mucho cheapo. And with Radeon 300 Series we saw AMD doesn't want to be the cheapo brand anymore. So my hopes are pretty low AMD will deliver on performance/$.

 

[itsmydamnation]

Occam's Razor
"Among competing hypotheses, the one with the fewest assumptions should be selected."

The CFO, "Zen is taping out in the next 3 months".
Zen is taping out in the next 3 months.

Zero assumptions.

Occams Razor does not allow you to ignore evidence, if you don't ignore evidence the assumptions you have made are:

CFO doesn't know his own numbers
AMD has awesome magic and can go from first silicon to shipping product in 6 months
CFO (and all of AMD previous time frames for Zen) has unrealistic expectations on time to market
CFO is lying about amd forcasts
CFO knows the technical business better than the financial