AMD Carrizo Pre-release thread

[mrmt]

But you are also removing the power consumption of the circuitry required to send signals off die to the FCH, and to receive them back.

I don't think the power circuitry to send signal consumes remotely close to the FCH itself. That said, the main point is not whether the added PCH will burden Carrizo power consumption much more than Kaveri, but that the PCH is yet another factor that will bring Carrizo's clocks down.

 

[positivedoppler]

So...Carrizo systems will basically be competing with the tail end of Broadwell/beginning of Skylake at this point.

Looks like it will compete against bw but will beat Skylake to the market.

http://venturebeat.com/2015/01/11/w...net-of-things-chip-market-interview/view-all/

I believe that we have not historically played in commercial. We’ve won the industry’s largest single tender in commercial 18 months ago in India. We won Elitebook book with HP last year. Wait until you see the lineup of commercial platforms I have with Carrizo. It allows us to continue to attack that i3, attack that i5 consumer, and really get to penetrate the commercial market space. We’ll attack graphics. That’s going to be my strategy next.

We’re pleased with the designs we have already with the top customers in the marketplace.

He's almost making it sound like they already have interest or contracts in place for Carrizo. This is vastly different than any approach I've ever seen from AMD. In the past it's been hype, hype, hype, paper launch, months later first demo product, letdown, and finally months later product at store. So in the past when AMD "Launches" a product, it's almost 1/2 year or more for the product to show up on shelf. If What Byrne is hinting can be believed, then AMD has did a complete 180 in terms of the way they introduce product. It's possible they are now hush, hush, hush, and then actual release.

My guess is the first time that we see Carrizo will be on a HP or Lenovo laptop.
If this is true, then in terms of time line, Carrizo where it is now is comparable to when past product from AMD has "Paper Launched" I don't want to detour this into Graphics, but I think he let out more than he should in that quote and I'm fairly certain now that Pirate Island product won't be on the shelf until beginning of 2nd half at the earliest.

 

[el etro]

Where's the hardware H.265 decode on this slide...? Did AMD do the hybrid GPU decode thing and claim that it was done "in hardware" or is this slide just really old?

The slide is old... but i still think many things of it are still right.

 

[Phynaz]

Looks like it will compete against bw but will beat Skylake to the market.

http://venturebeat.com/2015/01/11/w...net-of-things-chip-market-interview/view-all/

I believe that we have not historically played in commercial. We’ve won the industry’s largest single tender in commercial 18 months ago in India. We won Elitebook book with HP last year. Wait until you see the lineup of commercial platforms I have with Carrizo. It allows us to continue to attack that i3, attack that i5 consumer, and really get to penetrate the commercial market space. We’ll attack graphics. That’s going to be my strategy next.

We’re pleased with the designs we have already with the top customers in the marketplace.

He's almost making it sound like they already have interest or contracts in place for Carrizo. This is vastly different than any approach I've ever seen from AMD. In the past it's been hype, hype, hype, paper launch, months later first demo product, letdown, and finally months later product at store. So in the past when AMD "Launches" a product, it's almost 1/2 year or more for the product to show up on shelf. If What Byrne is hinting can be believed, then AMD has did a complete 180 in terms of the way they introduce product. It's possible they are now hush, hush, hush, and then actual release.

My guess is the first time that we see Carrizo will be on a HP or Lenovo laptop.
If this is true, then in terms of time line, Carrizo where it is now is comparable to when past product from AMD has "Paper Launched" I don't want to detour this into Graphics, but I think he let out more than he should in that quote and I'm fairly certain now that Pirate Island product won't be on the shelf until beginning of 2nd half at the earliest.

Except he's he's hyping the product

 

[Abwx]

Intels PCH is made on 32nm for example. If I recall right AMDs FCH is as well.

Seems that there s no tech doc for Kaveri s mobile FCH so not sure but it should be 40nm according to the DT parts node IIRC, so they have some little power margin improvement left when shrinking the FCH to 28nm.

But you are also removing the power consumption of the circuitry required to send signals off die to the FCH, and to receive them back.

I sincerely doubt that such a small integrated chipset will make much difference to the power consumption. It's not like it's handling 4 way Crossfire.

Does a port between a CPU and an external FCH require that much power.?.

Max power usage is 10% of the total TDP for a Kabini/Beema FCH according to AMD, this point to 1.5W at most.

 

[ShintaiDK]

FCH/PCH is connected via a PCIe x4 link. Also known as DMI or A-Link Express.

So it cant be all crazy.

 

[Abwx]

AMD FCH like A85X and A88X are made at 65nm.
Also, only Broadwell PCH is made at 32nm. Older chipsets are made with 45nm.

Thought it was 40nm as i posted above, quite possible that it s 65nm but it would be backward in respect of some AM3+ chipsets that are 55nm.

 

[frozentundra123456]

Except he's he's hyping the product

Exactly. The proof is if/when the products come out and whether they are premium designs or just cheap low end models.

 

[el etro]

I still bet Carrizo will be the better product AMD made of Bulldozer architecture. Can't compete with Intel on the single-threaded race, but will be much more decent than anterior competitors.

 

[mrmt]

I still bet Carrizo will be the better product AMD made of Bulldozer architecture. Can't compete with Intel on the single-threaded race, but will be much more decent than anterior competitors.

I'm betting on the opposite. Carrizo will be the swan song of the AMD consumer CPU division.

 

[Shehriazad]

I'm betting on the opposite. Carrizo will be the swan song of the AMD consumer CPU division.

I don't think Carrizo will have any huge impact either way even if it ends up being successful.

If anything... Zen will make or break the consumer CPU market for AMD. (assuming it's still happening)


Since we don't even know if Carrizo is still even happening for Desktops or higher end solutions...the impact is likely to be low either way...unless they manage to sell that thing to emerging markets like they were trying with AM1. Our own markets are saturated with Intel and Qualcomm. I might be terribly wrong...but I think for now the emerging markets are the only thing that AMD can still target without trying to fight windmills.

 

[NTMBK]

I'm betting on the opposite. Carrizo will be the swan song of the AMD consumer CPU division.

They're designing a new x86 core. They launching products targeting Android for the very first time. They're going to keep making consumer parts.

 

[mrmt]

They're designing a new x86 core. They launching products targeting Android for the very first time. They're going to keep making consumer parts.

Given that AMD has no mobile IP, a very poor balance sheet and actually 0 android software development, a position that makes Nvidia Tegra appear to be a strong product, my forecast for the TAM of AMD-powered Android devices is 5 devices for the entire product cycle: the three people left at AMDZone plus two from some dark conner of the internet.

AMD is cutting down to the bone their sales force and their marketing team. Whatever is left of it after 2015 won't be even close to have the same presence in the channel and with OEMs they used to have. Even if Zen is barely competitive, it won't reach many markets because of lack of sales infrastructure.

 

[cbn]

Since we don't even know if Carrizo is still even happening for Desktops or higher end solutions...the impact is likely to be low either way...unless they manage to sell that thing to emerging markets like they were trying with AM1. Our own markets are saturated with Intel and Qualcomm. I might be terribly wrong...but I think for now the emerging markets are the only thing that AMD can still target without trying to fight windmills.

Regarding emerging markets, I don't see Kaveri doing so great in that area. Same goes for some type of BGA Carrizo desktop. (Carrizo is too large of a die and too specialized to be cheap)

I think to compete in emerging markets, AMD needs to provide more cpu and much less iGPU at a lower price.

 

[Gideon]

No Carrizo for desktops

http://techreport.com/news/27652/amd-carrizo-will-face-off-with-broadwell-u-wont-land-in-desktops

Dissapointing, for sure, but IMHO considering their business situation, the only viable solution. Forcing this into FM2, as initially planned, or designing a totally new socket/mobo isn't a wise choice. Especially as 65W TDP would have been max, and there is no way it would have been anything much more than a sidegrade from Kaveri (in perf only not perf/watt perspective).

Edit:

Though a rant, this is also worth mentioning:
http://techreport.com/news/27652/am...ll-u-wont-land-in-desktops?post=874394#874394

I also remember that discussion clearly. Now it's seems that the opposite is true and AMD will lose much of the desktop APU business to Broadwell-K + EDRAM as-well. Certainly not all the market, but cost as the only argument to go AMD in more and more niches is getting pretty tiresome. It's been here since 2006.

I'm not happy for such news, but just getting super-tired of fanboys who keep asking for 4+ module AM3 chips and non-stop blaming AMD for going mobile. AMD is stuck with low-power bulk processes for the foreseeable future. Designing smaller mobile parts with less cores and simpler uncores IS the only realistic thing to do with their current R&D budget.

AMD cancelled the 10-core Piledrivers and new chipsets as they already did the math for them in 2011:
http://www.pcgamer.com/amds-10-core-piledriver-chips-revealed/

Now they decided the same for new FM2 chips, somewhere in the last year. As Carrizo is otherwise designed as a SoC, I definitely agree with that approach. It's Highly unlikely that developing a Frankenstein FM2 Carrizo with a Northbridge, hampered by the high density libraries etc, would have covered the engineering costs, and it seems that AMD agrees with me.

I just really hope that the DO manage to do the mobile stuff right. A 4-core 512 shader Carrizo @ 15W with Tonga Color Compression would actually be very competitive with Broadwell U (even the Iris version). I also really hope they beefed the Carrizo-L GPU to at least 256 shaders. This has been the Achilles Heel of the Beema lineup. Together they would make a much more compelling platform. 10 times better at least than the current terrible 17" suck'o'balls Beema laptop designs.

 

[Shehriazad]

No Carrizo for desktops

Again this dreaded "news" pops up.

There is absolutely no source or reasoning given why they would know that...they're also the ONLY site to report that Carrizo won't happen for Desktop at all.

Sure it's possible...but imho they are just trying to make their headline more worthy of viral news crap.

 

[ShintaiDK]

Again this dreaded "news" pops up.

There is absolutely no source or reasoning given why they would know that...they're also the ONLY site to report that Carrizo won't happen for Desktop at all.

Sure it's possible...but imho they are just trying to make their headline more worthy of viral news crap.

Really?

AMD shared some new details about its upcoming Carrizo and Carrizo-L APUs at the Consumer Electronics Show this week. Among the new info: AMD has no plans to offer Carrizo as a socketed chip for desktop PCs.

Looks to me it came from AMD.

 

[Abwx]

Though a rant, this is also worth mentioning:

You deem worth mentionning the post of what looks to be a rabid anti AMD shill.?..

Anyway kudos to Techreport, they manage to extract some money from their "pay to troll" scheme...

 

[Abwx]

Again this dreaded "news" pops up.

There is absolutely no source or reasoning given why they would know that...they're also the ONLY site to report that Carrizo won't happen for Desktop at all.

Sure it's possible...but imho they are just trying to make their headline more worthy of viral news crap.

AMD said that the product will be launched as BGA only, from this all the rest are speculations.

 

[Alatar]

AMD said that the product will be launched as BGA only, from this all the rest are speculations.

That essentially means no desktop.

No enthusiast parts, no unlocked parts, no high end parts, no fancy mobos, no upgradeable systems etc.

Yes some company can make a BGA board for desktop/HTPC like they do with baytrails etc. but when people talk about mobile only they mean BGA only.

 

[Shehriazad]

Even if it comes with delays...they better be releasing something else for FM2+ sometime this year...

I know that their financials suck...but releasing a new platform that was supposed to last until 2016...and then only releasing a single chip for it?

 

[NTMBK]

Even if it comes with delays...they better be releasing something else for FM2+ sometime this year...

I know that their financials suck...but releasing a new platform that was supposed to last until 2016...and then only releasing a single chip for it?

Enjoy your Kaveri Refresh.

 

[Shehriazad]

Oh and yea...AMD just posted literally the same info from above given on another site to their Facebook...so I guess I have to accept that Desktop Carrizo is officially dead.

 

[CHADBOGA]

Oh and yea...AMD just posted literally the same info from above given on another site to their Facebook...so I guess I have to accept that Desktop Carrizo is officially dead.

Do you have a link to this?

When I go to their Facebook page I can't see it.

 

[NTMBK]

Do you have a link to this?

When I go to their Facebook page I can't see it.

They posted a link to a Kitguru article (http://www.kitguru.net/laptops/note...-carrizo-based-laptop-prototypes-at-ces-2015/) with this line: "Previously it was reported that the “Carrizo” APUs will not make it to mainstream and high-performance desktops, but will be aimed at notebooks and all-in-one PCs."