AMD Carrizo Pre-release thread

[mrmt]

AMD just needs to strengthen it's ties to some strategic partners and start getting more design wins.

That would be a real problem. No OEM can build an entire line up with AMD processors only.

 

[monstercameron]

That would be a real problem. No OEM can build an entire line up with AMD processors only.

Why is that? also dont bring up the cost structure FUD.

 

[NostaSeronx]

I don't think Carrizo will clock lower than Kaveri if they're using SOI

FD-SOI has the potential for high TDP performance. In the case of Carrizo, it would be used for the lower leakage current. Rather than for a higher clock rate as SOI is low power, low heat, and low cost.

In IoT which was publicly full FinFET for past two years;
http://i.imgur.com/Lk0vFCX.jpg

 

[mrmt]

Why is that? also dont bring up the cost structure FUD.

I already showed you some things about the cost structure, on why Kaveri isn't really cost competitive against Core because it is basically double the die size, which means all things equal that it costs more than 100% to manufacture.. But let's assume that AMD would be able to swallow very low margins and beat Intel to the punch in terms of price to our hypothetical strategic OEM.

Just look at the AMD line up. It's mostly focused on low end offers, meaning that our strategic OEM would be restricted to low end SKUs.

Also every one of the top OEMs are betting on 2-1 convertibles, which AMD have essentially 0 share on this market, mostly because of high power consumption/battery life. Our strategic OEM also would be shut of this market.

In order to get high end chips and cost effective 2-1, OEMs must go to Intel, and you don't please your main supplier by straightening relationship with its main competitor.

 

[monstercameron]

I already showed you some things about the cost structure, on why Kaveri isn't really cost competitive against Core because it is basically double the die size, which means all things equal that it costs more than 100% to manufacture.. But let's assume that AMD would be able to swallow very low margins and beat Intel to the punch in terms of price to our hypothetical strategic OEM.

Just look at the AMD line up. It's mostly focused on low end offers, meaning that our strategic OEM would be restricted to low end SKUs.

Also every one of the top OEMs are betting on 2-1 convertibles, which AMD have essentially 0 share on this market, mostly because of high power consumption/battery life. Our strategic OEM also would be shut of this market.

In order to get high end chips and cost effective 2-1, OEMs must go to Intel, and you don't please your main supplier by straightening relationship with its main competitor.

ok, on the die size issue, do you have any sources backing you up on how much it costs for amd to put out a larger die vs intel? wafer costs, production costs, margins etc?

ok here is a bit of a disconnect, you claim that AMD only offers low end skus then go on to say that amd doesnt have any solutions for 2-1 convertibles. Convertibles mostly use lowend skus like baytrail, or at best middle of the pack skus like haswell u and broadwell y, both parts that beema and mullins compete with in terms of performance, price, power and bom costs.

 

[Shehriazad]

All this talk is going deep into the guesstimation zone.

I guess we get to see first results in december when the first mobile chips are officially introduced to us (and not just some engineering samples).

And just mayyyybe we get to draw some better conclusions from that ^^;

 

[Abwx]

All this talk is going deep into the guesstimation zone.

I guess we get to see first results in december when the first mobile chips are officially introduced to us (and not just some engineering samples).

And just mayyyybe we get to draw some better conclusions from that ^^;

AMD s chips specs are published and are official since 6 months, who is talking of ES.?.

 

[frozentundra123456]

I already showed you some things about the cost structure, on why Kaveri isn't really cost competitive against Core because it is basically double the die size, which means all things equal that it costs more than 100% to manufacture.. But let's assume that AMD would be able to swallow very low margins and beat Intel to the punch in terms of price to our hypothetical strategic OEM.

Just look at the AMD line up. It's mostly focused on low end offers, meaning that our strategic OEM would be restricted to low end SKUs.

Also every one of the top OEMs are betting on 2-1 convertibles, which AMD have essentially 0 share on this market, mostly because of high power consumption/battery life. Our strategic OEM also would be shut of this market.

In order to get high end chips and cost effective 2-1, OEMs must go to Intel, and you don't please your main supplier by straightening relationship with its main competitor.

I dont understand your reasoning that in order to get a design win AMD would have to supply chips across the entire product line. I dont really know what would stop an OEM from making an AMD product to fill a particular niche and still sell intel chips in a different market segment. Now whether AMD has a compelling chip in any particular segment is another question. I could see an AMD APU on a low end gaming laptop for instance if they could solve their bandwidth problems and offer decent performance at a cheaper price than intel plus discrete. In fact, I would love to get my grandson a 500 to 600 dollar light gaming laptop, but just dont feel the performance is quite there yet.

Unfortunately, I am expecting Carrizo not to be a compelling improvement over Kaveri. I also wonder when we will see real retail availability, since it is still almost impossible to find Kaveri laptops, although I did see one advertised from a local store, either Best Buy or Microcenter, cant remember for sure which.

 

[Shehriazad]

AMD s chips specs are published and are official since 6 months, who is talking of ES.?.

Oh, all the specs are published?


Then kindly tell me the name of the Carrizo FM2+ Desktop chip, its mhz clocks for the highest model, all tech upgrades it does or does not have aboard and how much better the IPC is compared to Kaveri.

Because that is what I consider to be the actual specs...and not the transistor count or which method was used in the 28nm process....we can't even deny or confirm if HBM is going to be on the chip and everything is still rumor based/we got some sketchy Sisoft screenshots.



*cough*

 

[Abwx]

Oh, all the specs are published?

Then kindly tell me the name of the Carrizo FM2+ Desktop chip, its mhz clocks for the highest model, all tech upgrades it does or does not have aboard and how much better the IPC is compared to Kaveri.

Because that is what I consider to be the actual specs...and not the transistor count or which method was used in the 28nm process....we can't even deny or confirm if HBM is going to be on the chip and everything is still rumor based/we got some sketchy Sisoft screenshots.

*cough*

I thought you were answering to the post above yours, sorry, as for Carrizo, well, all the known infos are more or less summarized in some previous posts and slides.

 

[NostaSeronx]

My guesses for top end Carrizo.

XV; 3.9+ GHz(base) - 4.4+ GHz(boost) @ 65W // SR; 3.7 GHz(base) - 4.0 GHz(boost) @ 95W
CR; 512 @ 900+ MHz // KV; 512 @ 720 MHz

Carrizo in general over Kaveri, hopes;
CR; 4K @ 60p decode // KV; 1080 @ 60p decode
CR; 4K @ 30p quality encode // KV; 1080 @ 30p quality encode
CR; ACP2 - 32-bit // KV; ACP - 16-bit or 24-bit
CR; HSA 1.0 - IOMMU v3 // KV; FSA 0.85 - IOMMU v2.5
CR; Enhanced Tonga/Iceland CUs // KV; Enhanced Hawaii/Bonaire CUs
CR; Hi Perf HSA Bus(Ring-based Cache Coherent Hypertransport) // KV; FSA P<->P Bus(Garlic, Onion, Onion+)
CR; Adaptive Voltage & Frequency Scaling // KV; Dynamic Voltage & Adaptive Frequency Scaling
CR; Integrated Voltage Regulation // KV; External Voltage Regulation

FM2+ limitations;
- Waterton iFCH disabled
- PCIe 3.0 GPP disabled
- DDR4 disabled
- HDMI 2.0 and DisplayPort 1.3 support disabled

--- http://semiaccurate.com/forums/showpost.php?p=224726&postcount=893 ---

The FS2 mention is important
http://www.anandtech.com/show/7945/...oard-preview-asus-asrock-gigabyte-msi-biostar
http://images.anandtech.com/doci/7945/4 AM1 Platform.png

 

[Shehriazad]

My guesses for top end Carrizo.

XV; 3.9+ GHz(base)

I would totally love seeing that, althrough I unfortunately already kind of doubt that this will be the case.

XV 3.9 base would finally be way faster than the Richland at base clocks and spit on Kaveri.

My personal guess is that it's gonna be 3.4GHZ base.... with a 10% IPC increase and thus be once again just as fast as Richland and Kaveris top models.

Although they are regularly updating their tech on the FM2(+) socket they don't seem to have an actual interest in beating their older chips on the CPU side even though they easily could.

Why they didn't release the 860K at 4.0 Base and 4.2 boost is still beyond me. Would easily stay within specs and the 860K has lower base volt requirements, anyway.


But we shall see....maybe XV cores will end up being a surprise even if the few pointers we have look like they're yet another iterative change.....all I want is an actual performance increase ffs XD.

 

[mrmt]

mullins compete with in terms of performance, price, power and bom costs.

Does it? Then why Mullins was shunned by OEMs, despite being that good?

 

[mrmt]

I dont understand your reasoning that in order to get a design win AMD would have to supply chips across the entire product line.

For that point you are making, e.g., selling AMD in a few market brackets along with Intel products? There isn't any point to make, especially because every OEM does that already. Most OEMs have bottom of the barrel SKUs powered by AMD processors, and a few such as HP have actual descent notebook with AMD inside.

But I assumed that monstercameron meant by strengthening relations with OEM somehow getting an OEM to give preferential treatment to AMD chips, and if that were to happen both problems I mentioned (the cost problem, and the Intel relationship problem) would become relevant.

As for Carrizo, only HBM parts will make the part remotely relevant for OEMs, if they can't get them to work Broadwell 14nm parts will eat them for breakfast.

 

[Abwx]

Does it? Then why Mullins was shunned by OEMs, despite being that good?

Cost structure, a 25-30$ chip cant compete against a chip that is half as good but wich is free and 70$ subsided, notice that being free wouldnt be enough to be choosen over a Mullins..

 

[maarten12100]

I already showed you some things about the cost structure, on why Kaveri isn't really cost competitive against Core because it is basically double the die size, which means all things equal that it costs more than 100% to manufacture.. But let's assume that AMD would be able to swallow very low margins and beat Intel to the punch in terms of price to our hypothetical strategic OEM.

Just look at the AMD line up. It's mostly focused on low end offers, meaning that our strategic OEM would be restricted to low end SKUs.

Also every one of the top OEMs are betting on 2-1 convertibles, which AMD have essentially 0 share on this market, mostly because of high power consumption/battery life. Our strategic OEM also would be shut of this market.

In order to get high end chips and cost effective 2-1, OEMs must go to Intel, and you don't please your main supplier by straightening relationship with its main competitor.

High power consumption you do realize that Kaveri ULV has the same or lower power consumption than Intel's i3 and i5 ULV Haswell offerings. Go check notebookcheck the Kaveri notebook is a obviously better choice actually even Beema is a better choice.

Also AMD has a take or pay agreement so they have to buy chips or pay a fine. Also GF is manufacturing cheaper than Intel I suppose. Clearly Intel can just discount their chips as much as is needed.

I dont understand your reasoning that in order to get a design win AMD would have to supply chips across the entire product line. I dont really know what would stop an OEM from making an AMD product to fill a particular niche and still sell intel chips in a different market segment. Now whether AMD has a compelling chip in any particular segment is another question. I could see an AMD APU on a low end gaming laptop for instance if they could solve their bandwidth problems and offer decent performance at a cheaper price than intel plus discrete. In fact, I would love to get my grandson a 500 to 600 dollar light gaming laptop, but just dont feel the performance is quite there yet.

Unfortunately, I am expecting Carrizo not to be a compelling improvement over Kaveri. I also wonder when we will see real retail availability, since it is still almost impossible to find Kaveri laptops, although I did see one advertised from a local store, either Best Buy or Microcenter, cant remember for sure which.

Intel has an extremely strong graps on the American market so you will not see high end Kaveri notebooks and would have to settle for ones with a crappy screen and crappy everything. Sad but true

Does it? Then why Mullins was shunned by OEMs, despite being that good?

Baytrail chips were cheaper actually Intel is/was pouring so much money into that they may have gotten money and the design of the notebook for using it. See the chinese rockchip firm.

Also Beema is in many laptops here in Europe and it is the exact same die

Cost structure, a 25-30$ chip cant compete against a chip that is half as good but wich is free and 70$ subsided, notice that being free wouldnt be enough to be choosen over a Mullins..

This

 

[Gideon]

Cost structure, a 25-30$ chip cant compete against a chip that is half as good but wich is free and 70$ subsided, notice that being free wouldnt be enough to be choosen over a Mullins..

WHAT ? I don't really know anything about the business but can you elaborate what does it mean? Are you honestly saying that Intel give Bay Trails away for free AND pays 70$ on top of that ? I highly doubt that's the case and IMO claims like that require proof.

 

[maarten12100]

WHAT ? I don't really know anything about the business but can you elaborate what does it mean? Are you honestly saying that Intel give Bay Trails away for free AND pays 70$ on top of that ? I highly doubt that's the case and IMO claims like that require proof.

"With contra revenue, Intel is paying tablet makers to cover the additional bill of materials (BOM) costs. “This is not a price reduction; it’s truly a BOM cost equalizer,” Krzanich said. It will also pay for “non-recurring engineering” costs, which means it will cover the cost of taking an ARM tablet design and porting it to an Intel chip."

So they will pay everything needed to make the design equal to the cheap ARM SoCs in terms of cost additionally they pay for all the R&D needed to design the product. Companies basically make money while spending nothing.

 

[ShintaiDK]

Cost structure, a 25-30$ chip cant compete against a chip that is half as good but wich is free and 70$ subsided, notice that being free wouldnt be enough to be choosen over a Mullins..

You make up random nonsense again.

Also Even without BT, Mullins wouldnt be picked either, because then there is ARM.

Always someone else but AMD to blame...

 

[Sweepr]

You make up random nonsense again.

Also Even without BT, Mullins wouldnt be picked either, because then there is ARM.

Always someone else but AMD to blame...

If Mullins was that good companies like MSI and other loyal AMD partners would have launched tablets by now and people would buy them even if they cost more than BT alternatives. Just look at Nexus 9, iPad Air 2 and future Core-M >10'' premium tablets, there's always space for better performing products. Problem is, perhaps Mullins is not nearly as good in terms of performance and battery life when packed inside thin/light/cheap 7-8'' tablets like Bay Trail or ARM SoCs and on top of that it costs quite a lot more than BT, but this idea might hurt the fanboys thinking it's the second coming of Christ.

 

[monstercameron]

Does it? Then why Mullins was shunned by OEMs, despite being that good?

Mullins has 1 design win - which is pitiful- that shows it can compete, hp stream 14.

 

[III-V]

So they will pay everything needed to make the design equal to the cheap ARM SoCs in terms of cost additionally they pay for all the R&D needed to design the product. Companies basically make money while spending nothing.

Your conclusions do not follow from the statements you've quoted. Again, it's an equalizer. OEMs would be making similar margins on comparable ARM devices, except they'd be stuck with Windows RT.

 

[ShintaiDK]

Mullins has 1 design win - which is pitiful- that shows it can compete, hp stream 14.

A laptop, entirely different market as well. And no contrarevenue to compete against.

 

[monstercameron]

A laptop, entirely different market as well. And no contrarevenue to compete against.

At this point , even with competitive products amd doesn't stand a chance with oems. Intel is doing something to incentivize them to use the super crappy celeron branded atoms in every single product category. It can't be Bom costs or power or availability or performance.

Intel must be buttering up up oems for them to drop amd so hard right after they all flooded the market with bobcat craptops and netbooks.

 

[Erenhardt]

At this point , even with competitive products amd doesn't stand a chance with oems. Intel is doing something to incentivize them to use the super crappy celeron branded atoms in every single product category. It can't be Bom costs or power or availability or performance.

Intel must be buttering up up oems for them to drop amd so hard right after they all flooded the market with bobcat craptops and netbooks.

I think you don't give intel enough credit. They are the powerhouse of the industry. OEM's need to listen to them if they want to have their business operating under the same rules as their (oem's) opponents.

Intel have power to make amd disappear from the shelves with a snap of a fingers.

Chicken and egg problem.
There are no available kaveri notebooks because people don'y buy them?
Or
People don't buy them because there are no kaveri notebooks on shelves?