The Intel Atom Thread

[Phynaz]

i gave this link as exemple of where
i get my information about current technoilogies

So your source of Intel cost information is 5+ year old research papers that are not written by Intel. At least you admit you "read it on the Internet, so it must be true!"

As Jim Carey would say - "Alrighty then..."

 

[bullzz]

@liahos1 - going by leaks, broadwell isnt doubling the GPU perf (non GT3e version). it might come closer to kaveri but I dont think it beat kaveri at every TDP

I still expect intel to lead in CPU perf but kaveri should come closer. going by current trend, intel isnt really interested in making huge CPU perf increases
ur predictions look way too much

 

[Abwx]

So your source of Intel cost information is 5+ year old research papers that are not written by Intel. At least you admit you "read it on the Internet, so it must be true!"

As Jim Carey would say - "Alrighty then..."

The cost estimation i was relating date from winter
12/13 , it is a note understable by anybody and can
surely be found at advancedsubstratenews.com.

 

[Sweepr]

let me take a stab. kaveri will come out have god awful cpu perf, good but ultimately unplayable gpu perform (relative to a discrete) all in a TDP uncompetitive with intel. amd will talk talk up HSA as the next coming of Christ and claim many many many design wins. and ultimately we may see kaveri in a few budget notebooks on the bargain / special sales section of best buy. sort of like what happened with trinity

Precisely. Then there's also the fact that by the time that you will actually be able to buy a Kaveri-based notebook god knows when in 2014 Broadwell will not be too far from launch. Desktop Kaveri is expected to deliver 1050 Gflops (vs Richland's ~780 Gflops, desktop version), so I'd say they might have 30-40% better graphics performance at the same TDP. An Ivy Bridge-like GPU jump, not even Haswell-like, would put Broadwell in an even more competitive position (vs Kaveri) than Haswell vs Richland right now. And next year most Intel GPUs will probably be a mix of GT2/GT3 (with more expensive Iris Pro GT4 version) and we already know that AMD loves to gimp A4/A6 and sometimes even A8 GPUs, compared to the top A10 models they showcase in reviews.

 

[Vesku]

Yes. Only the final qualification run is accrued as COGS, everything else is OPEX.

Then they could probably sell the current Bay Trail in the $15 range and not quite hit negative. How much die space will Intel save if they push Bay Trail dual cores into that ~$10 space?

 

[blackened23]

Precisely. Then there's also the fact that by the time that you will actually be able to buy a Kaveri-based notebook god knows when in 2014 Broadwell will not be too far from launch. Desktop Kaveri is expected to deliver 1050 Gflops (vs Richland's ~780 Gflops, desktop version), so I'd say they might have 30-40% better graphics performance at the same TDP. An Ivy Bridge-like GPU jump, not even Haswell-like, would put Broadwell in an even more competitive position (vs Kaveri) than Haswell vs Richland right now. And next year most Intel GPUs will probably be a mix of GT2/GT3 (with more expensive Iris Pro GT4 version) and we already know that AMD loves to gimp A4/A6 and sometimes even A8 GPUs, compared to the top A10 models they showcase in reviews.

That's pretty much what I expect from Kaveri. The top of the line LGA desktop version will be a nice SKU with good GPU performance. I think the Richland is a great HTPC APU, I think Kaveri will maintain that for LGA. However, as has always been the case with AMD APUs, expect the mobile offerings to be far less balanced and not nearly as good - and worse than intel across the board in terms of performance "balance" (cpu + gpu). For mobile SKUs, expect disabled turbo and gimped clockspeeds across a wide variety of SKUs, with obvious lack of performance implications. Just the way I predict things to go down.

 

[Vesku]

Even if Kaveri is a really nice design, there will be the OEM adoption problem. Look at all these Bay Trail design wins. OEMs actually had a lackluster presence in the budget back to school PC and notebook market, at least from what I saw in my region, because they were waiting on Bay Trail and only pushing out a very limited amount of AMD designs.

Can't even buy them yet and there's already dozens of soon to be launched Bay Trail products being shown around. Granted a lot of the enthusiasm from the various brands perspective is the potential for sales with light and good battery life Windows x86 tablets. Something Temash has a hard time squeezing into being best suited for 10-11" tablet convertibles or notebooks.

 

[mrmt]

Then they could probably sell the current Bay Trail in the $15 range and not quite hit negative. How much die space will Intel save if they push Bay Trail dual cores into that ~$10 space?

No idea, but... a single module silvermont manufactured in a fully depreciated 22nm node won't be too expensive.

 

[AtenRa]

No idea, but... a single module silvermont manufactured in a fully depreciated 22nm node won't be too expensive.

Dual core Baytrail will be die harvested from the Quad core. But even if they would make a new dual core die, the die size they would save is very small and the mask cost + design etc too high.

 

[Khato]

What you are going to do is offer rebates, marketing money, bonus and so on. This goes straight to a blackbox called SG&A, something that everybody accepts as wildly different from company to company.

Thanks for providing an example of how the accounting portion is actually done My impression had been that it was done otherwise, but again, that's based off the engineering perspective which doesn't always apply to other fields as well as I'd like it to, haha.

 

[Phynaz]

The cost estimation i was relating date from winter
12/13 , it is a note understable by anybody and can
surely be found at advancedsubstratenews.com.

ASN is sponsored by Soitec...The Author’s postings are his own opinions

Yeah, that's an impartial site that has insight into Intel.

 

[Shift_]

The cost estimation i was relating date from winter
12/13 , it is a note understable by anybody and can
surely be found at advancedsubstratenews.com.

Can you please link us to said article? As an engineer you should know facts prevail, and without evidence to back up your claims, anything you say is worthless.

 

[Enigmoid]

Precisely. Then there's also the fact that by the time that you will actually be able to buy a Kaveri-based notebook god knows when in 2014 Broadwell will not be too far from launch. Desktop Kaveri is expected to deliver 1050 Gflops (vs Richland's ~780 Gflops, desktop version), so I'd say they might have 30-40% better graphics performance at the same TDP. An Ivy Bridge-like GPU jump, not even Haswell-like, would put Broadwell in an even more competitive position (vs Kaveri) than Haswell vs Richland right now. And next year most Intel GPUs will probably be a mix of GT2/GT3 (with more expensive Iris Pro GT4 version) and we already know that AMD loves to gimp A4/A6 and sometimes even A8 GPUs, compared to the top A10 models they showcase in reviews.

Richland does not deliver 780 GFLOPS. AMD is including the CPU GFLOPS in that number, a10-5800k delivers 614 GFLOPS and the a10-6800k delivers only 648 GFLOPS. Kaveri is likewise probably including the cpu. Unless they can get around the bandwidth problem then their GLOPS isn't going to mean much either.

 

[Sweepr]

From Intel Earnings Call:

Expect Windows $299 touch enabled Bay Trail powered 2-in-1s by holiday season. Touch adds <$50 to BOM.

At IDF those same 2-in-1s were expected to start at $349.

 

[paperwastage]

From Intel Earnings Call:



At IDF those same 2-in-1s were expected to start at $349.

https://twitter.com/intel/statuses/341800223993110529

well, Intel said $399 back in June 2013

I'd say $299 (the dell venue 8 pro) + keyboard dock(?$50?, no mention of any dock yet... maybe a bluetooth kb + case, like those iPad Keyboard cases) might come close

 

[SlimFan]

At IDF those same 2-in-1s were expected to start at $349.

Intel's own CEO is quoted saying devices could hit $200 back in April, five months before IDF:

Touch devices to drop to $200, says Intel CEO

Do you think that the Windows surcharge is $149 (operating system, dock, keyboard, etc)? If that is true, I'm not quite sure how that would be Intel's fault.

 

[Shivansps]

There are a $250 2955u Haswell Chromebook, I can easily see less than $200 BT devices, but not with W8.

 

[Abwx]

Can you please link us to said article? As an engineer you should know facts prevail, and without evidence to back up your claims, anything you say is worthless.

I wont bother to browse through the site but if you want
you can yourself do the research , this is their 20th edition
print from winter 2012/2103 in PDF format , the relevant
cost estimation article is from Handel Jones , CEO of IBS.

quotes :

Although the real competition is likely to
be between FinFETs and FD-SOI at 20nm,
FinFETs are a new technology (from a highvolume
production perspective), with
significant cost penalties even in Q1/2016.

As shown in this graph, the savings realized by using FD-SOI at the ~20nm node is significant.
Even once FinFETs have matured in Q1/2016, FD-SOI will still offer comparative savings of 50-60%,
depending on die size.

Said graph include Finfets........

 

[liahos1]

http://www.engadget.com/2013/10/16/lenovo-miix2/

new baytrail tablet. I guess they are coming out with two. This one plus the "world's thinnest one".

 

[krumme]

http://www.engadget.com/2013/10/16/lenovo-miix2/

new baytrail tablet. I guess they are coming out with two. This one plus the "world's thinnest one".

Yet another uninteresting tablet. We knew from the start Intel pays everyone and his brother to get their bt into tablets. And they lose loads on mobile right now.
What would be interesting is to know is how is those tablets selling - anyone?

 

[liahos1]

Yet another uninteresting tablet. We knew from the start Intel pays everyone and his brother to get their bt into tablets. And they lose loads on mobile right now.
What would be interesting is to know is how is those tablets selling - anyone?

What makes a tablet interesting to you? At 299 this is a good deal. Maybe some are happy paying 800 dollars for ipad tablets but other people want options.

 

[Arachnotronic]

Yet another uninteresting tablet. We knew from the start Intel pays everyone and his brother to get their bt into tablets. And they lose loads on mobile right now.
What would be interesting is to know is how is those tablets selling - anyone?

They lose loads on mobile because R&D costs are high and sales are still low. When sales go up, the operating loss should quickly narrow.

 

[Arachnotronic]

Yet another uninteresting tablet. We knew from the start Intel pays everyone and his brother to get their bt into tablets. And they lose loads on mobile right now.
What would be interesting is to know is how is those tablets selling - anyone?

Let me guess...if it doesn't say Nexus or have an Apple logo on it, it's uninteresting?

 

[mikk]

http://www.notebookinfo.de/artikel/...3770-fuer-windows-8-und-android-tablets/3995/


As for 3d performance here a new result.

Cinebench R11.5 32-Bit OpenGL – 5,84 fps

For comparison A4-1200 reaches 7.18 fps: http://tweakers.net/productreview/83223/medion-akoya-e1318t.html


So Z3770 19% slower than A4-1200 but considering that Intel doing worse in OpenGL than in directx it is expected. I don't know if 32 bit OpenGL scores in Cinebench 11.5 are worse than in 64 bit though, something to check.

 

[SiliconWars]

Tough one to call Mikk.

What is telling is how much apathy there is for it. Bay Trail is great chip but nobody cares now. You need to be fast in this market. Tablet talk is all but dead now..