Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Apple Corps

macrumors 68030
Original poster
Apr 26, 2003
2,575
542
California
Ivy bridge single cpus will be shipping soon and Intel has yet to get MacPro "compliant" cpus out based on Sandy Bridge. Even allowing for the design problem they announced this feels like a long time.

When Sandy Bridge was firmed up Intel knew that "Sandy Bridge Xeons" were going to be produced - thus whatever design tweaks were needed must have been allowed for so the two cpus could shake hands with each other and the mother board.

Why is this taking so long??
 
A switch from Xeon processors leading Intel's new architecture platform to coming much later.

X58 and 5520 systems continuing to sell well in 2011.

Production issues.

AMD not being competitive at the high end and also having delays.
 
Ask Intel.

What a waste of bandwidth response.

----------

A switch from Xeon processors leading Intel's new architecture platform to coming much later.

X58 and 5520 systems continuing to sell well in 2011.

Production issues.

AMD not being competitive at the high end and also having delays.


The AMD issue is one I have not thought of - the longer the current Xeons sell the more of their development cost is absorbed.
 
I assume it's not much competition and really not a huge demand it seems.
 
What a waste of bandwidth response.

----------




The AMD issue is one I have not thought of - the longer the current Xeons sell the more of their development cost is absorbed.

I am not sure what you're hoping for. No one here can give you a definitive answer and this theorising has already been done to death. My answer is correct, even if you don't like it.
 
I am not sure what you're hoping for. No one here can give you a definitive answer and this theorising has already been done to death. My answer is correct, even if you don't like it.

NO - yours was a snarky response - my focus had been on technical reasons beyond the recently announced one as to why it is takes so long to move to the dual socket configuration when a number of iterations have been done in the past and intel engineers know what they have to do.

The AMD issue was another perspective.
 
Truth be told, Intel have not released the new Xeons yet.

They are native 8-core parts, which are used in the high-end computer segment (although they are salvaged parts with only 6 cores enabled).
 
Baby kitten shortage. They haven't made enough baby kitten sacrifices yet.

NO - yours was a snarky response - my focus had been on technical reasons beyond the recently announced one as to why it is takes so long to move to the dual socket configuration when a number of iterations have been done in the past and intel engineers know what they have to do.

Chip design is complicated. The latest architecture upped the number of memory lanes among other things. Just because they've done dual processor before doesn't mean they can do the exact thing again. They're not duct taping different processor parts together here.

There is also output issues. It could be that the processor is designed, they're just having trouble getting enough good chips out of the plant. That's more unpredictable.
 
A switch from Xeon processors leading Intel's new architecture platform to coming much later.

That's what I've heard: that Intel decided to launch the high-volume Core i7 series before the workstation Xeon series. The only question I have is why is the delay so long in releasing Xeon Sandy Bridge. IIRC, Ivy Bridge chips will be released soon after the launch of Xeon Sandy Bridge. In other words, Xeon is so far behind the initial Sandy Bridge launch, it is about to get leapfrogged.
 
Last edited:
That's what I've heard: that Intel decided to launch the high-volume Core i7 series before the workstation Xeon series. The only question I have is why is the delay so long in releasing Xeon Sandy Bridge. IIRC, Ivy Bridge chips will be released soon after the launch of Xeon Sandy Bridge. In other words, Xeon is so far behind the initial Sandy Bridge launch, it is about to get leapfrogged.

Ivy Bridge isn't far enough ahead of Sandy Bridge where it will matter. Intel won't get any less money.
 
Ivy Bridge isn't far enough ahead of Sandy Bridge where it will matter. Intel won't get any less money.
Ivy Bridge-E Xeons are more than likely out until 2013.

Even the LGA 1155 based one has been delayed a few times. Intel needs to milk these 32nm lines to make up for the lower revenues.
 
That's what I've heard: that Intel decided to launch the high-volume Core i7 series before the workstation Xeon series.

No, it is the lower volume Core i7 series that just launched (those are the same design base as the Xeon; more cores, bigger cache, no integrated graphics). The higher volume Core i7 is the same baseline design as the i3 and i5 (e.g., an integrated graphics , smaller cache, less than or equal to 4 cores ).

The only question I have is why is the delay so long in releasing Xeon Sandy Bridge.

Several reasons.

1. The high volume chips are smaller. ( fewer transistors and smaller dies ). In Intel's Tick-Tock model every "Tick" is a die shrink on a new process. It is easier to get better yields on smaller dies. By placing the smaller chips first, Intel reduces the risks with the launch.

Launch biggest first is would be a better approach to do chest pumping, but it won't reduce risk.

Smaller also means less complexity. Fewer memory channels, less complicated internal bus , fewer PCI-e lanes , less complicated SATA controllers, etc.


2. There is very little market pressure to do so. First, AMD isn't hitting on all cylinders in the server space. Second, server customers tend to be a bit more conservative. They want CPUs and chip sets that have rigorous testing done before deployment.


3. The "plan" was for late Q3 / early Q4 . The plan got screwed by multiple bugs. First the Sandy Bridge chipset had a SATA III bug. (the launch was screwed and probably hiccuped into the Xeon chipset. ). Intel made the dubious move of trying to jam an extra SATA III lanes into the chipset when DMI couldn't handle the bandwidth. The hack they came up with didn't work... again probably negatively impacted schedule. There have been other bugs VT-d along the way too.

PCI-e v3.0 probably hiccuped the schedule also. The "plan" over a year ago was that those certification tests would be finalized by now (10 months after the standard going final seems like a reasonable timeframe). That is running a bit behind also.

Once this all slide into mid November it doesn't make sense to launch Xeon in Dec... so slides to post holidays.


IIRC, Ivy Bridge chips will be released soon after the launch of Xeon Sandy Bridge.

Not really. Xeon Sandy bridge will like come in Jan (maybe Feb). Ivy Bridge won't be around till March April. What Ivy bridge that will be released will be capped at 4 cores and lower memory bandwidth. Except for those myopically focused on single core, in cache performance... they aren't going to be broad spectrum competitive with the Sandy Bridge Xeon line up.


In other words, Xeon is so far behind the initial Sandy Bridge launch, it is about to get leapfrogged.

To a large extent it is Apples and Oranges. As long as the Ivy Bridge Xeon comes out 10-12 months after they are in decent shape.

Intel will likely keep the 4 core Xeon with Ivy Bridge and just clock it high so that the single core , clock chasers have roughly same parity or better than the high volume Ivy Bridge.
 
1. The high volume chips are smaller. ( fewer transistors and smaller dies ). In Intel's Tick-Tock model every "Tick" is a die shrink on a new process. It is easier to get better yields on smaller dies. By placing the smaller chips first, Intel reduces the risks with the launch.

Launch biggest first is would be a better approach to do chest pumping, but it won't reduce risk.

Smaller also means less complexity. Fewer memory channels, less complicated internal bus , fewer PCI-e lanes , less complicated SATA controllers, etc.
People are still under the illusion that they need more cores since quad cores, they have been around since 2006, and at $300 or less from Intel. Gamers want PCI-Express lanes for video cards, not that x16/16 is that much better.

3. The "plan" was for late Q3 / early Q4 . The plan got screwed by multiple bugs. First the Sandy Bridge chipset had a SATA III bug. (the launch was screwed and probably hiccuped into the Xeon chipset. ). Intel made the dubious move of trying to jam an extra SATA III lanes into the chipset when DMI couldn't handle the bandwidth. The hack they came up with didn't work... again probably negatively impacted schedule. There have been other bugs VT-d along the way too.

PCI-e v3.0 probably hiccuped the schedule also. The "plan" over a year ago was that those certification tests would be finalized by now (10 months after the standard going final seems like a reasonable timeframe). That is running a bit behind also.

Once this all slide into mid November it doesn't make sense to launch Xeon in Dec... so slides to post holidays.
I do not believe we ever got a real reason why Patsburg was scaled back so far that it is barely more than the typical Intel 6 Series PCH. DMI was not the limiting factor when they had additional PCI-Express lanes dedicated for SAS. Though adding SATA and SAS onto one controller alone would have made it rather complicated to begin with compared to just pure SATA. VT-d has errata but C2 stepping is coming out in January.

Also we never touched back on this either. AMD's 28nm APUs are dead and HD 7000 is more than likely taking up those wafers. Not to mention they are much higher margin products compared to those APUs but the volume is much lower and the complexity is greater. Give and take there.
 
No, it is the lower volume Core i7 series that just launched (those are the same design base as the Xeon; more cores, bigger cache, no integrated graphics). The higher volume Core i7 is the same baseline design as the i3 and i5 (e.g., an integrated graphics , smaller cache, less than or equal to 4 cores ).

Sorry, I wasn't clear - I was referring to sales volume, not volume in terms of chips per wafer. And I think it's pretty clear that Intel sells far more Core-series than Xeon.

Xeon Sandy bridge will like come in Jan (maybe Feb). Ivy Bridge won't be around till March April.

I think you just proved my point: the gap between Xeon Sandy Bridge and Ivy Bridge (desktop/laptop) is going to be about 1 quarter, as opposed to the nearly 1 year wait for Xeon Sandy Bridge.

However, it's fair to say that Ivy Bridge is a "tock" rather than a "tick", so it's less compelling of an upgrade. Nevertheless, it's a bit weird to release the high-end product that is - by some measure - virtually obsolete at its launch.
 
Last edited:
Baby kitten shortage. They haven't made enough baby kitten sacrifices yet.



Chip design is complicated.

Agreed - Quite a number of years ago in the photolithography era I had a "negative" or mask - amazed at the complexity of the design. I believe they switched to x-ray lithography a number of years ago and not sure of what is used today. The responses from several of you are quite detailed (memory lanes, SATA 3 issues, stepping issues, possible yield issues on the new die shrink...), I had read about a bug but it appears they are working several bugs / holidays / lack of competition considerations also.
 
Baby kitten shortage. They haven't made enough baby kitten sacrifices yet.



Chip design is complicated.

Agreed - Quite a number of years ago in the photolithography era I had a "negative" or mask - amazed at the complexity of the design. I believe they switched to x-ray lithography a number of years ago and not sure of what is used today. The responses from several of you are quite detailed (memory lanes, SATA 3 issues, stepping issues, possible yield issues on the new die shrink...), I had read about a bug but it appears they are working several bugs / holidays / lack of competition considerations also.

I wonder if the circuitry is getting too compact and voltages are jumping, or if this is not the case in desktop processors
 
However, it's fair to say that Ivy Bridge is a "tock" rather than a "tick", so it's less compelling of an upgrade. Nevertheless, it's a bit weird to release the high-end product that is - by some measure - virtually obsolete at its launch.

Ivy Bridge is a refinement of an existing architecture on to a smaller die and that is a tick. It doesn't make any Sandy Bridge-E or EP processor virtually obsolete.

I will grant that high-end IB processors might not make the Core i7 3820/E5-1620 such a compelling choice, but then the high-end Sandy Bridge LGA 1155 processors don't make them one either. There are another two Core i7 and at least nineteen Xeon LGA 2011 processors (we don't know how many 4600 series Xeons there will be) and none of the Ivy Bridge models are any more an alternative than any current LGA 1155 processor would be.
 
Last edited:
Ivy Bridge is a refinement of an existing architecture on to a smaller die and that is a tick. It doesn't make any Sandy Bridge-E or EP processor virtually obsolete.

Thanks, I inadvertently switched the terms tick and tock. However, I still stand behind my comment about being 'virtually obsolete' from day 1. If you're buying a high-end system, do you really want yesterday's technology today? I guess it depends on the needs: some users want stability, but others want to be on the leading edge, and these are often customers with deep pockets.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.