I would talk to your source because your samples are way behind.
Also, who says faster has anything to do with it? Or AMD for that matter? This is simple cost of production economics.
1). They already dumped $6B into a new fab earlier this year
2). 32nm is 50% smaller than 45nm. Similarly, cheaper.
Is it possible? Most likely it is, but the cost would be astronomical...
All rumours aside, I still don't expect to see Clarksdale on the shelves until Q1 2010...and nothing I've seen so far has convinced me otherwise. We might see a few chips shipped here and there in late Q4, but extremely low quantities...
BTW, the relevance is that we're talking about 1.8 million Clarksdale chips per month in 2010...I can't imagine the yields will be good enough by Q3 for that, but it certainly is possible.
taltamir
Lifer
- Mar 21, 2004
- 13,576
- 6
- 76
i don't see why it will be astronomically expensive OR impossible.
Manufacturing-wise, they just have to replace one chip with another, they can certainly do it.
Sales-wise they just have to price it right.
unless you are doubting their ability to create a new 32nm node process and get it to the point of producing 10% of their current total capacity? in which case I also disagree, it is not gonna be astronomically expensive or difficult.
Now if intel said they will increase their sales TENFOLD by the end of the year... id be skeptical of their ability to manufacture or sell that much. But "replacing 10% of our current product with next gen product while still shipping 90% of last gen product" is an extremely conservative and simple procedure.
Manufacturing-wise, they just have to replace one chip with another, they can certainly do it.
Sales-wise they just have to price it right.
unless you are doubting their ability to create a new 32nm node process and get it to the point of producing 10% of their current total capacity? in which case I also disagree, it is not gonna be astronomically expensive or difficult.
Now if intel said they will increase their sales TENFOLD by the end of the year... id be skeptical of their ability to manufacture or sell that much. But "replacing 10% of our current product with next gen product while still shipping 90% of last gen product" is an extremely conservative and simple procedure.
I think you need a quick brief on what it takes to ramp a production chip...IDC will be MUCH better at that than I.
For now, what I am saying is that the initial chips of any new manufacturing node or new design require time to mature into high enough yields to make them economically viable. You don't just "replace one chip with another"...
I'll make a quick stab at explaining, and maybe IDC can clarify my many mistakes...
When you produce a wafer, the first ones are invariably flawed in some way...maybe the design needs to be changed, there are too many defects, or the doping mix needs to be tweaked.
Each test wafer takes ~2-3 months to create and then test, corrections are made, then the next 2-3 months of creation and testing. Much of this time is mitigated by trying many different solutions at once (multiple test wafers created simultaneously), however it almost always takes at least 2-3 turns (3 month creation/test period) for yields to become viable.
You may have seen IDC and myself mention AMD's system of APC? That is a revolutionary system that automatically tweaks the test wafers (or even production wafers) for optimal production. It is extremely efficient, and is the main reason AMD has been even slightly competitive these past 8 years.
Intel uses their large number of Fabs and the "Copy Everything" method as it suits their business model better.
What you said was "unless you are doubting their ability to create a new 32nm node process and get it to the point of producing 10% of their current total capacity?"
I am doubting that they can do that by Q3 while maintaining anywhere near their current profit margin
There is more than one definition of "hits". I personally am more interested in getting a laptop with Arrandale, and so if I can get a macbook pro or a Thinkpad with an Arrandale, for a reasonable price(ie. less than a 1100 for the thinkpad and less than $1300 for the MBP) they get to say "Told you so". Shipping out a couple of review samples or releasing at pathetic clockspeeds(see: phenomI) doesn't really count as a launch in my book.
I would love to get a 32nm MBP with a free iPod Touch "3GS" with a student discount by the end of the year, and so I really hope you are wrong, but I am skeptical.
I would love to get a 32nm MBP with a free iPod Touch "3GS" with a student discount by the end of the year, and so I really hope you are wrong, but I am skeptical.
Yeah, basically the key point here is that regardless the yields, the node, the maturity, cycle-time, etc, all that matters is the gross margins and unless the gross margins of 32nm product is equivalent (or superior) to the gross margins of the existing 45nm sku's being replaced/supplanted then it makes little business sense to ramp volume shipments to double-digit percentage levels.
Single digit percentage levels makes some sense even if the margins aren't entirely there as there is a bit of marketing leverage to be maximized as well as continuing to substantiate the shareholder's existing sentiment that they own a piece of an unchallenged technology leader.
(Not telling you anything new, just expanding on the line of thinking for the benefit of other readers)
ilkhan
Golden Member
- Jul 21, 2006
- 1,117
- 1
- 0
Im with drizek, Im waiting for a dell latitude e4310 with an arrandale. Be that november or feb, nothing else is even in my sights. But I do think they'll release in volume in mid november.
They're prepping 3 fabs for volume (plus the already going research fab (D1D?)), we'll see clarkdale/arrandale as soon as the first is capable of profit. They'll expand volume and add xeon chips when the rest come online.
They're prepping 3 fabs for volume (plus the already going research fab (D1D?)), we'll see clarkdale/arrandale as soon as the first is capable of profit. They'll expand volume and add xeon chips when the rest come online.
We've all seen the graphs for Intel's cross-over rates with past 90nm->65nm and 65nm->45nm and I think it is safe to say it would be atypical and unusual for Intel to pull of a node release and ramp of that scale.
Sure they can do it, they have the money and the manpower, but it would be without precedence and this fact alone bears consideration when evaluating the FUD factor of any rumors regarding volume ramps when they get to this level of "wow, that would be impressive" magnitude.
The difference this time is that intel is doing a die shrink with midrange CPUs, not high end ones. Before, the new process would be prohibitively expensive for several months, but it appears that this wont be the case this time around. Another thing is that this time around their new CPUs will actually be significantly faster than the previous generation, if nehalem is any indication. You are also going to have a lot of people coming out to buy new computers with Snow Leopard and Windows 7, so they should be seeing higher than average sales. 32nm also seems to have gone better than previous shrinks from what intel is saying, so if that is true then they will be able to meet demand.
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,785
- 136
Remember guys, this is a dual core Clarkdale. These CPUs will be shipped to majority of the system builders and dual cores still remain most popular out of all. The combination of potentially a powerful IGP capable of competing against AMD/Nvidia's then-current IGP, low power and cool running core, and a very powerful CPU core will prove to be a very compelling choice.
I'm beginning to think that Clarkdale is like Atom in that it is not a shared CPU across all platforms with different SKU's depending on binning, rather a very small die with high yields and very low manufacturing cost.
Couple the low cost with the low power of 32nm and the 45nm IGP and Clarkdale will set new performance per $$ benchmarks and use less power than current generation Atoms. The new Pine Trail Atoms will regain the power title when they get the 45nm IGP in 2010.
Clarkdale will be the greenest CPU Intel has ever produced and will sell PC's like hotcakes, especially with its introduction with Win 7.
ilkhan
Golden Member
- Jul 21, 2006
- 1,117
- 1
- 0
Clarkdale uses far more power than atom does. Its a power efficient chip, but its not an atom. (If you count the NB, itd be close, perhaps. But thats gone soon enough anyway.) An ULV arrandale is in the same power range as an atom+965(?) solution, but not any clarkdale.
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,785
- 136
Not possible. Process tech doesn't help that much. Die size increase may be 50% but power consumption reduction usually lies only 30-40% reduction per generation.
Cool fact, pardon the pun
The Atom N270 used in Netbooks are rated at 1.1V 3A at max power operation. The Core 2 Solo ULV at 45nm are rated at 1.1V 3.0A at Deep Sleep and SuperLFM.
TDP-wise, Atom N270 has a 2.5W TDP at full load while Core 2 Solo ULV at 45nm has 2.1W at AutoHalt, Stop Grant Power.
Now that's for a previous generation single core which is also a ULV.
http://www.brightsideofnews.co...ols-core-i7-to-507-ghz
scroll down, they were playing with Wolfdales in their hotel room, and apparently it looks good. Built in auto overclock seems pretty cool too, I am glad they are actually taking advantage of it.
I can't figure out what they mean though in the last paragraph about IGPs. To me it seems like instead of leaving the field open, intel is cutting off amd and nvidia. I am assuming that the intel solution is going to end up being a lot cheaper than red/green. Also, what exactly does nvidia have to offer in the IGP space to go against AMD? NV9400 or something new? Can AMD use HD4200 in laptops?
Is Core i5 Mobile going to be Arrandale or Clarkdale based?
scroll down, they were playing with Wolfdales in their hotel room, and apparently it looks good. Built in auto overclock seems pretty cool too, I am glad they are actually taking advantage of it.
I can't figure out what they mean though in the last paragraph about IGPs. To me it seems like instead of leaving the field open, intel is cutting off amd and nvidia. I am assuming that the intel solution is going to end up being a lot cheaper than red/green. Also, what exactly does nvidia have to offer in the IGP space to go against AMD? NV9400 or something new? Can AMD use HD4200 in laptops?
Is Core i5 Mobile going to be Arrandale or Clarkdale based?
ilkhan
Golden Member
- Jul 21, 2006
- 1,117
- 1
- 0
I presume you meant westmere in their hotel room.
45nm quad core *clarkfield* chips won't have any way to use an intel GPU, it'll require a seperate GPU, either nV or AMD. Same as Lynnfields will require on the desktop (although on the desktop you have expansion slots). Although, either can of course have a seperate GPU core on its on on the motherboard. It just won't be "integrated" to the NB (as we've seen until now) or CPU (like with the dual cores).
45nm quad core *clarkfield* chips won't have any way to use an intel GPU, it'll require a seperate GPU, either nV or AMD. Same as Lynnfields will require on the desktop (although on the desktop you have expansion slots). Although, either can of course have a seperate GPU core on its on on the motherboard. It just won't be "integrated" to the NB (as we've seen until now) or CPU (like with the dual cores).
I figured they would divide Wolfdale into Clarkdale and Arrandale (and Gulftown), but I guess that would have made too much sense. Add to that changing the names of current Penryn Core 2s to Core i5/3 just makes it all so confusing. Wikipedia set it all straight though.
Do we have any idea what hte clockspeeds for Arrandale are going to be? The Clarkdale ones on wiki seem pretty low at 1.6-2.0
This statment,
"Intel's Capella notebook platform will be the home of 45nm quad-core octa-thread Lynnfield processors and 32nm Clarkdales [dual-die CPU: 32nm dual-core quad-thread CPU with 45nm integrated GPU]"
He is confusing Clarkdale with Arrandale, correct? And Lynnfield is actually Clarksfield?
And he forgot the "l" in Calpella. I regret reading/linking to that article.
Do we have any idea what hte clockspeeds for Arrandale are going to be? The Clarkdale ones on wiki seem pretty low at 1.6-2.0
This statment,
"Intel's Capella notebook platform will be the home of 45nm quad-core octa-thread Lynnfield processors and 32nm Clarkdales [dual-die CPU: 32nm dual-core quad-thread CPU with 45nm integrated GPU]"
He is confusing Clarkdale with Arrandale, correct? And Lynnfield is actually Clarksfield?
And he forgot the "l" in Calpella. I regret reading/linking to that article.
TRENDING THREADS
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 10K
-
Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)- Started by Vattila
- Replies: 13K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 7K
-
Facebook
Twitter