Richland & Kabini rumours

[AtenRa]

Wait and see, wait and see! We don't know if these things will pan out- I can see a combination of 2 out of the 3 being true, but I'm not convinced they'll pull off all three. Let's hope they can (and that Intel can catch up too).

I will have to clear up about the Half Power Consumption i have mentioned.
It is for a DUAL CORE Kabini at around ~1.9GHz. That will have 25%+ performance over Brazos(1.7GHz).

 

[Nemesis 1]

I don't deal with phone chips, so I don't know what iGPU will they use. But Bay Trail is a tablet, mobile, and desktop Atom. There will be also an industrial version for embedded systems.
The chip will use 4 execution units, so it will not really an HD Graphics 4000 iGPU (with 16 EUs). The architecture is the same.

Thats nice only 4 EU. I like reasoning . the 22nm will stay with imagination on the phones . The Tablets tho I just can't buy the 4 EU story . The 10 watt haswell tablet chip will use 40 EUs and it uses only 10- watts. So if its all the same I not buying into the 4 EU thing you mentioned . Yes I seen it written as you said it . But the Intel roadmap I linked to is the only offical roadmap I have seen that comes from intel all others are photochopped work and have nothing to do with intels roadmaps . See you guys care about tablets as do I but phones is where I want to see intel In the dang phones everthing else will take care of itself

 

[Nemesis 1]

Nemesis sorry for asking this obvious (but kind of redundant) question but why are you here posting in this topic? I really can't see the reason. You don't bring anything valuable to the discussion and you only either flame others (directly) or post offtopic stuff in Kabini thread(atom this,atom that etc.).

I hope some moderator reads your posts since it's really becoming tiresome and all can agree on this,I'm sure.

Oh yes . I sure the mods will see . Really It was not I who brought intel into this topic . The moment you start comparing to intel you open the door for rebuttle. I am sure you would love to hang another bann to your list of people YOU engaged threw your hype of AMD products. What we here at AT understand is this . We read the forums over at XS . we have seen what you and Frank have done with your PHI hype and your BD hype . Now your tring to sell a product again befor its released with HYPE. The AT guys for the most part been long timers here . We know A bit more than you want to except . As I said Info is great . THE HYPE not so much .

 

[SiliconWars]

Oh yes . I sure the mods will see . Really It was not I who brought intel into this topic . The moment you start comparing to intel you open the door for rebuttle. I am sure you would love to hang another bann to your list of people YOU engaged threw your hype of AMD products. What we here at AT understand is this . We read the forums over at XS . we have seen what you and Frank have done with your PHI hype and your BD hype . Now your tring to sell a product again befor its released with HYPE. The AT guys for the most part been long timers here . We know A bit more than you want to except . As I said Info is great . THE HYPE not so much .

I'm sure nobody at AT or anywhere else has the first idea what you are going on about.

 

[grimpr]

Just ignore him or make jokes, he's a poor good guy with a passion on AMD, on medication and living in the 90's.

 

[Idontcare]

Temash tablets will literally put PCs in your fingertips

That's a messed up mental image D:

LOL, Better that versus it being a PITA

Just ignore him or make jokes, he's a poor good guy with a passion on AMD, on medication and living in the 90's.

From circa 2009:

Nemesis is a riddle wrapped in a mystery inside an enigma, which is then wrapped in bacon, deep fried, slathered in butter, turkey, and more bacon, deep-fried again, and ready to cause heart failure in just about anyone that attempts to digest him thereafter :biggrin:

 

[NTMBK]

wrapped in bacon, deep fried, slathered in butter, turkey, and more bacon, deep-fried again

Mmmm, sounds like Christmas.

 

[ShintaiDK]

Mmmm, sounds like Christmas.

Sounds like a heart attack and a blood clough!

 

[NTMBK]

Sounds like a heart attack and a blood clough!

I live in Scotland, the two aren't far apart :thumbsup:

 

[ShintaiDK]

I live in Scotland, the two aren't far apart :thumbsup:

Ohh, the great people behind the deep fried marsbar!

 

[NTMBK]

Ohh, the great people behind the deep fried marsbar!

And the deep fried pizza, and the deep fried haggis, and deep fried cheeseburger, and...

 

[Idontcare]

...and the deep fried haggis...

D:

I suppose I should try it before I knock it, but...shudders at haggis (while happily willing to enjoy sushi )

 

[mrmt]

Yes, with half the power consumption, 25%+ higher performance for the CPU cores and at least 50% higher performance for the iGPU over Brazos, Intel doesnt have a chance.

Did AMD GPUs got a 50% reduction in power consumption? No, they didn't. That 50% number is a PR slide, not an actual product. If you are expecting this I would suggest a bit of salt, but I showing restrain when hyping AMD products is too much for you.

Well, Temash Win8 tablets will be the only product in the market that you would be able to play current DX-11 games, use your PC applications and more. Temash tablets will literally put PCs in your fingertips

As of now there are 0 DirectX 11 games for tablets and you can bet that there won't be for quite some time. The ecosystem isn't ready, and nobody isn't going to develop bleeding edge games for AMD systems, so what you are saying here is a non-issue.

Kabini will have two more cores and much better iGPU than Ivy Celeron/Pentiums have. Kabini will also be much cheaper than those Intel CPUs. It seams that 2013 will be very difficult for Intel in the low end.

There is a reason why Thuban and Bulldozer cannot beat Core and the reason is IPC. Core clocks will be similar but IPC will be a lot higher. It may win one benchmark here and there in very specific cases, but for most cases Core will be simply better.

I never said that Core won't go head to head with Kabini. What I said is that it will check on upper market brackets like cheap 15 inch notebooks. Kabini won't face competition at the bottom (11, 12 inch cheap notebooks) until Silvermont.

We have already seen how great Brazos fared against ATOM and low budget Celerons/Pentiums, it is only logical to expect that Kabini will do only better with half the TDP, 25% higher CPU performance and 50%+ iGPU performance over Brazos. And all that, almost in the same die size of Brazos.

You really have no clue, do you? Did you check AMD gross margins as of lately? Did you notice the drop? Did you notice that they lost market share at the bottom market? Did you notice that AMD could not sell Brazos in notebooks and had to resort to selling in desktops? No? probably you didn't, but let me tell you what is happening:

AMD is losing the bottom market, there is no demand for Brazos or Trinity even with the current cratered prices. All that talk about new markets is simply an admission that AMD cannot offer any kind of competition to Intel. You would know that if you bothered yourself into looking at AMD financials and not at their marketing slides.

Will Kabini fundamentally change this? No. Kabini will be a very nice product, it will improve on something already good. But Kabini simply isn't enough. Intel is tick-tocking both product lines, and both product lines will soon have a far smaller gap between them, be it on price, performance or power consumption. Whatever market Kabini has now will be under pressure from both Intel lines.

AMD big core line didn't buckle with Conroe or Nehalen, it buckled with cumulative pressure from a lot of short release cycles that AMD could not match. This is *exactly* what is starting to happen with Atom. When Brazos launched AMD did get performance and efficiency crown, now Intel reached efficiency levels that AMD cannot touch, and 6 months after Kabini launch it may reach similar performance levels on top of that efficiency. Where do you think the things will go three generations from here, with AMD spread thin between phasing out big core, designing an update to Kabini and their ARM chip?

And how bad the situation can turn to AMD? With big core AMD had both the market share and gross margin to lose, so they could crawl until today, but Kabini market is different. The margin is low, and the volumes are nothing compared to big core, so what we will probably see is a much faster decline if tick-tock show its magic.

 

[Haserath]

AMD should charge higher margins for Kabini while they have a chance. Increasing margins 20-30% will increase product price by, what, <10%? That's where AMD is really failing, low margins but barely any product price difference.

 

[CTho9305]

There is a limit on how low you can go with pricing, and AMD is already too close to this limit. They are already in the sub-40% territory, if they go to the sub-35% they are done for, unless we see a complete overhaul of their cost structure (mostly personnel).

Having this in mind, do you think AMD can get 35%+ margins competing against 14nm Core and 14nm Atom in 2014? I don't think they can.

Are you accounting for the fact that AMD's margins are likely heavily influenced by selling giant Bulldozers for peanuts? The margins can be very different across the product lineup... you can make a lot of money on small dies even at low prices. If you were thinking about the Bulldozer-family stuff (Richland? I don't know what that is) then sorry - I assumed you were writing in the context of the low-power APUs.

 

[Haserath]

Yes i have read it thx

Well, either way(VLIW4 or GCN) it will destroy everything Intel have(iGPU) or will have in the next 2 or more years in the 9W and bellow.

Also, just imagine Tablets playing DX-11 games(1024x768) like DIRT, Civ 5 and more. With USB3/HDMI and the ability to use any Windows device like mouse, keyboards, Printers etc, and use your Windows applications.

You could have your PC with you 24/7 at the size of a 10-11" tablet

A 384 GCN core with real bandwidth barely plays those games. 1/3 of that means slide-show.

I've been able to use an android tablet and now I own an ipad. Besides 'hardcore' gaming and 'real work,' which can be done. IPad is fantastic comparatively. I don't need a fancy, full blown OS on a device that doesn't have the performance.

So AMD will be on a "half baked" 20nm, but intels 14nm will be all singing, all dancing? Do you have anything to back this up?

The rumors of 20nm being poor and Intel always delivering...

 

[AtenRa]

And the deep fried pizza, and the deep fried haggis, and deep fried cheeseburger, and...

I have tried a deep fried pizza once in Rutherglen Glasgow and that was the first and the last time i did. I could try Irn-Bru again but not Deep Fried Pizza.

There was a small takeaway in Duke street (East End) having Deep Fried Burgers with chips at only 1.2 pounds from 11:00am until 12:30pm. I used to eat one every week with tons of vinegar :biggrin:

 

[AtenRa]

D:

I suppose I should try it before I knock it, but...shudders at haggis (while happily willing to enjoy sushi )

I wouldn't be so brave as you trying a deep fried haggis.

ps: I would suggest you try smoked(peat) Whiskey instead

 

[pablo87]

Long story short, mrmt is right. As good as Kabini may be, the structural problems facing AMD remain.

It doesn't take a PHD to figure out that 15MM processors per quarter at $20 profit per doesn't cover the overhead incl. interest, even if AMD manages to get $150MM gross profit contribution from dGPU and royalties per quarter.

Unfortunately, that's the tip of the iceberg.

The real problem that rears its head again and again is the bloody WSA. The more Kabini's AMD sells, the more money they lose. Not because the profit per processor on Kabini is less (its probably more all things considered - much smaller die size, lower packaging cost, better positioning, etc..), but because they have to pay Global Foundries whether they take parts from them or not.

Imagine 100% of AMD's 15MM processors shipped per quarter are Kabini's fabbed at TSMC. Assume AMD's ASP on these is $40 and cost incl. packaging is $20. Gross Profit is $20 x 15MM = the afore mentioned $300MM. But then AMD still has to pay GF 275MM per quarter on top of that.

Imagine another scenario where all those Kabini's are fabbed at GF and yields and costs are the same as TSMC. There is still a guestimated shortfall on the WSA of $100MM per quarter (play around with cost per wafer, yields and Kabini die size and you'll see). That's because Kabini's die size is too small to meet the annual take or pay commitment under the WSA alone.

It creates a very strange dichotomy that AMD has no financial incentive to sell Kabini's or to improve upon it, even though its the more demanded, better and more cost effective design. No wonder Seifert didn't want to order more Brazos. And no wonder they took a write-off on LLano's - they placed those orders to meet the WSA even though the demand wasn't there...

Obviouses to me are:

- port Kabini to Gloflo asap
- make Kabini's successor bigger die size (as long as TDP constraint can be met).
- cancel Kaveri since a bigger Kabini will sit right on top of it, and it would reduce engineering cost further.
- negotiate to further reduce the annual take or pay in 2014
- GPU at GF

The problem with the 5th point is that its probably the most risky yet probably becomes the 2nd priority because AMD refuses to make the difficult decisions involved in points 2 to 4.

Another point is ATI's overhead per quarter used to be $170MM (which included chipset btw). If Kabini is all you are designing on the CPU front, the overhead of the combined entity should be a lot lower than $450MM per quarter. In other words, AMD would probably be way better off being ATI + Kabini design team.

 

[mrmt]

Are you accounting for the fact that AMD's margins are likely heavily influenced by selling giant Bulldozers for peanuts? The margins can be very different across the product lineup... you can make a lot of money on small dies even at low prices. If you were thinking about the Bulldozer-family stuff (Richland? I don't know what that is) then sorry - I assumed you were writing in the context of the low-power APUs.

I did account for those facts but I don't see Brazos margins well above 45% for that matter. Brazos margins should be around 40%. But even if your assumption was correct it wouldn't change much of the facts I stated in the previous post:

Brazos/Kabini cash flows are too small to generate such a huge amount of cash to withstand tick-tock for two or three generations, but the product will be very good in the short term, and will be a much better bet for AMD's money than Kaveri or the Excavator derivative.

 

[mrmt]

The real problem that rears its head again and again is the bloody WSA. The more Kabini's AMD sells, the more money they lose. Not because the profit per processor on Kabini is less (its probably more all things considered - much smaller die size, lower packaging cost, better positioning, etc..), but because they have to pay Global Foundries whether they take parts from them or not.

Didn't Kumar stated that Kabini will move from TSMC to GLF? It will be very interesting to compare frequencies and yields of both 28nm process. For some reason I think AMD won't be happy with this switch. :whiste:

 

[NTMBK]

Brazos/Kabini cash flows are too small to generate such a huge amount of cash to withstand tick-tock for two or three generations, but the product will be very good in the short term, and will be a much better bet for AMD's money than Kaveri or the Excavator derivative.

Yeah, the window for Kabini success is fairly narrow. I'd expect it to do very well against Clover Trail and still be competitive against Haswell and Valleyview- but once Intel is on 14nm Atom and Broadwell, it's going to struggle. It all depends on how fast AMD can iterate Kabini, and how fast their foundry partners move.

 

[mrmt]

Yeah, the window for Kabini success is fairly narrow. I'd expect it to do very well against Clover Trail and still be competitive against Haswell and Valleyview- but once Intel is on 14nm Atom and Broadwell, it's going to struggle. It all depends on how fast AMD can iterate Kabini, and how fast their foundry partners move.

Not only the foundry partner. AMD already declared that they are going for a slower adoption rate of bleeding edge nodes. As much as it appears suicide, that's what you do when your foundry partner is your loan shark. AMD will move to another node only when the economic conditions are right for them, not only when the partner is ready.

 

[NostaSeronx]

Kabini/Temash =
84,000,000 chips / per month theoretical max

50% = amount for Kabini and Temash
20% = assumed amount for AMD

8,400,000 chips / per month assumed
Sucessful Metric Achieved chips: 55%

3,780,000 chips / per month assumed
Net cost: $8 to 10 per chip
Sale for: $90 - 120 range

$340,200,000 / per month gained ($99)

Assume a year is nine months of sales

9 * $340,200,000 = $3,061,800,000

To break even they only need to make $3,300,000,000. Kabini and Temash will be very successful in comparison to VIA's new 28-nm monolithic quad-core and Intel's 32-nm/22-nm Atoms.

This is if Kabini/Temash is being made at GlobalFoundries... if they are being made at TSMC expect $4 billion for the year.

 

[mrmt]

3,780,000 chips / per month assumed
Net cost: $8 to 10 per chip
Sale for: $90 - 120 range

Brazos currently amounts for 40% of AMD *shipments*, not revenues, as Bulldozer/Trinity have a higher ASPs. So even if Brazos amounts for half of the shipments, it won't amount for half of AMD revenues by a significant margin. Even if they did

But you should really review your revenue assumptions, as those you brought here are really wild. AMD ASP is Q312 is below $60 and average cost per unit is around $36. There isn't a chance in hell that AMD will be able to buy Kabini for $10 and sell for $90.

Ed: AMD sales cryptic sales forecasts for 2013 points to something between 4.5 and 5.0 billion, of which 1.2 billion will be GPUs and some 200-300 million from Seamicro. You are expecting more from Kabini than AMD expects from their entire CPU division.