Sandy Bridge E delayed to 2012

[bulldog147]

Do you think Ivybridge stock speed will be 4.20GHz this time ? (up from 3.40GHz sandybridge)

It would be nice if Intel put all chips 4.20GHz stock speed in 2012! (mean very easy to get overclocking probably up to 5.20GHz!)

 

[paradigm]

It would be nice if Intel put all chips 4.20GHz stock speed in 2012! (mean very easy to get overclocking probably up to 5.20GHz!)

Just because they increase the baseline clock speed for the chips doesn't mean that the available overclocking headroom will increase.

Until the process shrink, all increasing the stock clocks to 4.2Ghz will do is remove overclocking headroom, and artificially increase unit cost to the consumer.

 

[EggCustard]

Like with Ivy Bridge, it seems Sandy Bridge E is delayed to 2012:

http://www.techpowerup.com/148282/Sandy-Bridge-E-Delayed-to-January-2012-Sources.html

To be honest are we all that surprised? I mean is AMD causing mayhem round Intel's feet as it looks down as the world leader? Exactly, AMD can't even get 32nm out the door let alone impress with performance numbers to make Intel finish puffing it's cigar and cut to the chase.

In industry if you do certain things slower you end up saving money, makes perfect sense if there's no viable competition from AMD.

 

[CAT-THE-FIFTH]

To be honest are we all that surprised? I mean is AMD causing mayhem round Intel's feet as it looks down as the world leader? Exactly, AMD can't even get 32nm out the door let alone impress with performance numbers to make Intel finish puffing it's cigar and cut to the chase.

In industry if you do certain things slower you end up saving money, makes perfect sense if there's no viable competition from AMD.

Actually, too many people were believing Intel marketing which is the problem with Ivy Bridge being "delayed" when TBH it probably never was. It was funny when people were actually believing that it would be released only 10 months after Sandy Bridge. On top of this Ivy Bridge is basically a die shrink of Sandy Bridge with 33% more EU.The die is smaller than Sandy Bridge.

Also,as you know it is not AMD who is developing the 32NM SOI process though. Intel uses 32NM bulk processes which are supposed to be cheaper. Intel considered using SOI for their 22NM process butsaid finfets were more cost effective. However, trying to make comparisons between two different processes is really meaningless especially since they are being used for differently designed processors.

The 32NM process AMD is using is being developed by Global Foundries which is part of the SOI consortium which includes IBM,STMicroelectronics and ARM. Global Foundries used to be fully owned by AMD. However,it is only a minority stakeholder as ATIC owns most of it. Global Foundries seems to be expanding at a massive rate and they seem to have some really serious financial backing in the form of ATIC(oil money). It will be interesting to see how much capacity they have in the next year or two.

If anything the AMD CPU devision has made more money this year than for yonks. I expect them to compete very well with their APUs as most CPU sales are not to enthusiasts and it seems according to retail channels they have more or less sold out production. Bulldozer will probably do relatively well as a server and supercomputer CPU too and Cray have already announced systems using it. Of course it probably will also probably be quite strong in many desktop tasks such as encoding and rendering which many workstations are used for. OTH, performance in lightly threaded tasks is unknown.

If anything the main issue is that AMD simply is supply constrained which Intel is not ATM. Intel is much better placed ATM to fulfill greater demand and it has been this case for the last 20 years.

 

[N00bbuilder79]

Despite the delay my next rig will still be built around the 2011 SKT. Lets just hope there are no more delays on top of this one.

 

[EggCustard]

If anything the main issue is that AMD simply is supply constrained which Intel is not ATM. Intel is much better placed ATM to fulfill greater demand and it has been this case for the last 20 years.

Touchwood, and it just makes perfect sense for Intel to take things at a more leisurely pace since if there are problems with SB then it covers them up nicely while if there aren't it just gives a perfect excuse for Intel to do a more economical roll-out, something that will please investors.

Unless I've missed something lately I don't get how AMD can go on for now months without a new CEO named. I said it long ago and I still think there's a lot of mileage left in the following comment, that is I have a serious feeling there's a lot of tension within AMD board room, a lot of pass the parcel, probably finger pointing thus chaos and delay. For Intel it just looks like so from the outside, perhaps no one wants the job knowing what he (or she) will be getting themselves in for.

 

[CAT-THE-FIFTH]

A lot of companies go through turmoil especially since AMD has a rough ride the last few years. However,things are looking to improve for them ATM and if Bulldozer does prove to be decent then I expect things to settle down.

OTH,they have made a profit this year and this is without a CEO. Perhaps,it actually works better for them(joking)??

 

[Madals]

I think it is much more likely the reason it is delayed is due to the current economic climate. People have less money spare so enthusiast chips are in lower demand, lots of people are "skipping" processor generations due to no noticeable gain outside of industry.

It is also worth noting the a large chunk of the enthusiast market are gamers, again, lots of rumours of 2012 being the announcement year for the next generation of consoles which means games will finally start to get a bit more of a stress on hardware. At the moment, there is literally nothing I can think of that you can't run maxed out at 1920x1080 on a I7 930 and GTX480.

These two factors combined really do reduce the demand for a half-upgrade CPU. As stated earlier, I could see it being possible that SBE won't be made and rather Ivy bridge will just cater for all markets.

 

[paradigm]

Why are you so focussed on the "enthusiast" market? As has already been said in this thread, individual component purchases by "enthusiasts" will be a minuscule amount of the market for the x79 platform.

The bulk of revenue will come from OEMs, primarily workstations.

 

[drunkenmaster]

Actually, too many people were believing Intel marketing which is the problem with Ivy Bridge being "delayed" when TBH it probably never was. It was funny when people were actually believing that it would be released only 10 months after Sandy Bridge. On top of this Ivy Bridge is basically a die shrink of Sandy Bridge with 33% more EU.The die is smaller than Sandy Bridge.

Also,as you know it is not AMD who is developing the 32NM SOI process though. Intel uses 32NM bulk processes which are supposed to be cheaper. Intel considered using SOI for their 22NM process butsaid finfets were more cost effective. However, trying to make comparisons between two different processes is really meaningless especially since they are being used for differently designed processors.
.

This is talking abuot Sandybridge E being delayed, a 6(and possibly 8) core version of Sandybridge, its a bigger sandybridge give or take, being out within 10 months is entirely doable and required.

Most of Intel's server platform uses variations of the previous gen hexcore, updating it to Sandybridge-E based stuff = more sales and more performance, when AMD will be hammering it with 16 core Zambezi's in a month or so likely at hugely lower prices.

As for processes, Intel do NOT use a 32nm bulk process, Intel use the latest up to date process tech's, HKMG, strained silicon, etc, etc, etc. Bulk = what TSMC and Glofo will use for gpu's and other bits and bobs, Glofo/Intel use very high end stuff for cpu production.

Ivy was never planned for before Q1 2012, and thats the date thats slipped back to likely late q2 or even q3, but people who were thinking optimistically Jan 1st were probably very wrong to start with, so figure end of March launch moved back 2-4 months. Either way, 32nm launched in 2009, expecting 22nm in 2012 fairly early isn't unrealistic.

End of the day smaller chips at same price = more profit and a large part of Intel's success comes from being early to new processes.

 

[CAT-THE-FIFTH]

This is talking abuot Sandybridge E being delayed, a 6(and possibly 8) core version of Sandybridge, its a bigger sandybridge give or take, being out within 10 months is entirely doable and required.

Most of Intel's server platform uses variations of the previous gen hexcore, updating it to Sandybridge-E based stuff = more sales and more performance, when AMD will be hammering it with 16 core Zambezi's in a month or so likely at hugely lower prices.

As for processes, Intel do NOT use a 32nm bulk process, Intel use the latest up to date process tech's, HKMG, strained silicon, etc, etc, etc. Bulk = what TSMC and Glofo will use for gpu's and other bits and bobs, Glofo/Intel use very high end stuff for cpu production.

Ivy was never planned for before Q1 2012, and thats the date thats slipped back to likely late q2 or even q3, but people who were thinking optimistically Jan 1st were probably very wrong to start with, so figure end of March launch moved back 2-4 months. Either way, 32nm launched in 2009, expecting 22nm in 2012 fairly early isn't unrealistic.

End of the day smaller chips at same price = more profit and a large part of Intel's success comes from being early to new processes.

Intel is using a 32NM bulk process and I started this thread as it was more relevant to some of the desktop users on this forum. I was talking about people thinking 22NM Ivy Bridge being released 10 months after Sandy Bridge. Hints were being made by Intel to various journalists IIRC. I really doubt six or eight core version of the dual channel Sandy Bridge will be released this year.

If you already are developing a quad channel version,ie, Sandy Bridge E then these are the CPUs which are going to be released for the server and high performance workstation market. It makes no sense to make a cheaper socket 1155 six or eight competitor which would steal sales away from socket 2011. The only socket 1155 Xeons are going to rebadged Core i5 and Core i7 chips.

Edit!!

These article state Intel uses a high performance 32NM bulk process:

http://www.realworldtech.com/page.cfm?ArticleID=rwt072109003617&p=5

 

[nightmare99]

I really doubt six or eight core version of the dual channel Sandy Bridge will be released this year.

This is quite right noboddy should expect there to be anything more than quad on the 1155 socket any time soon.

 

[CAT-THE-FIFTH]

The only reason I posted the link is because some forum members were wanting to upgrade to a socket 2011 based desktop computer this year. Since the article is relevant to this I though I would post it.

Probably the server six and eight core Sandy Bridge E processors will be released this year as the leaked roadmaps indicated.

 

[drunkenmaster]

I'm not really sure what the issue is, the roadmap for well over a year has been known to be, late 2011, Sandybridge-e, 32nm 6-8 core, early 2012, ivybridge, mainstream 22nm quad core and lower, late 2012, forget the name kessler/hessler maybe, 22nm octo core mainstream chips. From what I recall Anandtech mentioned Sandybridge-E launching late 2011 in their Sandybridge reviews in January, this has been common information, hence every thread on the forum for a year being Sandy-E end of year, Ivy next year.

Pretty much everything seems to be slipping 2-4 months or so, and at a guess the next launch after Kessler would be 12-16core Kessler-E high end quad channel server/enthusiast chips which would be following their current pattern.

There was never a chance, nor an idea that Ivybridge would be either octo core, nor this year.

As for the question of if Intel's is really "bulk" or not, it really explains it fairly well in that link.

Look at IBM's description for their 32nm bulk vs their 32nm SOI, bulk is for gpu's, consumer products, soc's, other peripherals, its NOT for uber high end chips. TSMC's bulk isn't for making high end stuff, no one elses is.

Most other process designers have 2-3 processes at every level, because they are research companies and foundrys wanting to make everything from calculator chips to high end gpu's.

Intel really design for one customer, and one purpose, hence their process is focused on their chip design and high end cpu design, nothing else. its bulk for them because, simply, they have massive volume and one type most of their volume is high end cpu's. For TSMC/IBM bulk means higher volume gpu, soc, everything else market.

Intel's "bulk" is in no way comparable on price or quality to TSMC's or IBM's.

Look later in the review at the comparison tables.

http://www.realworldtech.com/page.cfm?ArticleID=rwt072109003617&p=11

Intels 32nm "bulk" massively leads the pack, followed by IBM 32 SOI, followed by 45nm SOI, then the have IBM 32nm bulk, alongside TSMC's latest and greatest and Intel's 45nm "bulk".

Intel's processes aren't cheap for any reason other than the massive volume they produce vs AMD.

Essentially bulk(at least how I've always seen it refered to) is what people associate with a cheap high volume version of any given process, Intel's fits that catagory because they don't make multiple versions at every node, because they aren't a foundry. You're also using it to suggest that Intel's process is cheaper, it isn't, Intel just have a very high quality "bulk" process, while most other players have a cheap bulk and a more expensive high end process.

 

[CAT-THE-FIFTH]

I'm not really sure what the issue is, the roadmap for well over a year has been known to be, late 2011, Sandybridge-e, 32nm 6-8 core, early 2012, ivybridge, mainstream 22nm quad core and lower, late 2012, forget the name kessler/hessler maybe, 22nm octo core mainstream chips. From what I recall Anandtech mentioned Sandybridge-E launching late 2011 in their Sandybridge reviews in January, this has been common information, hence every thread on the forum for a year being Sandy-E end of year, Ivy next year.

Pretty much everything seems to be slipping 2-4 months or so, and at a guess the next launch after Kessler would be 12-16core Kessler-E high end quad channel server/enthusiast chips which would be following their current pattern.

There was never a chance, nor an idea that Ivybridge would be either octo core, nor this year.

As for the question of if Intel's is really "bulk" or not, it really explains it fairly well in that link.

Look at IBM's description for their 32nm bulk vs their 32nm SOI, bulk is for gpu's, consumer products, soc's, other peripherals, its NOT for uber high end chips. TSMC's bulk isn't for making high end stuff, no one elses is.

Most other process designers have 2-3 processes at every level, because they are research companies and foundrys wanting to make everything from calculator chips to high end gpu's.

Intel really design for one customer, and one purpose, hence their process is focused on their chip design and high end cpu design, nothing else. its bulk for them because, simply, they have massive volume and one type most of their volume is high end cpu's. For TSMC/IBM bulk means higher volume gpu, soc, everything else market.

Intel's "bulk" is in no way comparable on price or quality to TSMC's or IBM's.

Look later in the review at the comparison tables.

http://www.realworldtech.com/page.cfm?ArticleID=rwt072109003617&p=11

Intels 32nm "bulk" massively leads the pack, followed by IBM 32 SOI, followed by 45nm SOI, then the have IBM 32nm bulk, alongside TSMC's latest and greatest and Intel's 45nm "bulk".

Intel's processes aren't cheap for any reason other than the massive volume they produce vs AMD.

Essentially bulk(at least how I've always seen it refered to) is what people associate with a cheap high volume version of any given process, Intel's fits that catagory because they don't make multiple versions at every node, because they aren't a foundry. You're also using it to suggest that Intel's process is cheaper, it isn't, Intel just have a very high quality "bulk" process, while most other players have a cheap bulk and a more expensive high end process.

Except they are using a bulk process though. That is all that I stated.

Intel also mentioned the reason they were using finfets it was cheaper than using a 22NM SOI process and they also thought it was better suited for low voltage devices.

http://images.anandtech.com/reviews/cpu/intel/22nm/benefitssm.jpg

They estimated that using planar transistors and SOI would cost them 10% more at 22NM whereas finfets would cost them 2% to 3% which means it is cheaper. It is has bean known for a long time that Intel stayed with using bulk(32NM and 45NM too) since it was cheaper for them and they have more experience working with bulk than SOI.

You started going off on another tangent stated that I was implying Intel was using worse technology. I just mentioned it was cheaper.

It is the same with Ivy Bridge. Plenty of people were going on how it would be a massive performance jump in CPU performance despite being a Tick. I stated to many people it would be more focused on additional cost reduction (of course dependent on the yields and other fixed costs) and have the improved IGP which they stated it would have which is more in line with what happened in the past. Guess what?? An Ivy Bridge quad core looks to be 20% smaller than a Sandy Bridge quad core.

On top of this I mentioned that the processes are not even comparable.

However, trying to make comparisons between two different processes is really meaningless especially since they are being used for differently designed processors.

This was on the last page BTW and was my second post in this thread and the opening one I made.

Looking at the die size and number of transistors the current 32NM and 45NM AMD processors are using they have done relatively well TBH. This is supposedly using "worse" and more expensive process technology and worse chip design methods(I am not talking about CPU architecture design) according to the numerous armchair experts on the internet.

BTW,my response was to the following post. Reading the thread from the beginning may help.

To be honest are we all that surprised? I mean is AMD causing mayhem round Intel's feet as it looks down as the world leader? Exactly, AMD can't even get 32nm out the door let alone impress with performance numbers to make Intel finish puffing it's cigar and cut to the chase.

In industry if you do certain things slower you end up saving money, makes perfect sense if there's no viable competition from AMD.

 

[tbh]

Has there been any official confirmation of this?

As far as I am aware the latest Intel roadmaps still show Sandy Bridge-E due for a 2011 release.

 

[clocka]

no there has been no official confirmation.