Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Mac Pro 2011 "Sandy Bridge E" everything we know
- Thread starter Amethyst
- Start date
- Sort by reaction score
I'd tend to agree.
It's been what, 6 months since the 2600K was released, and another 3 months before SB-E is released.
Bigger/faster caches, small improvements here and there I wouldn't be surprised by.
I wouldn't be surprised if there is no difference too!
Bad news everyone, possible delayed until Q1 2012
http://vr-zone.com/articles/sandy-b...Men,+Geek,+Nerd,+Overclockers+and+Enthusiast)
http://vr-zone.com/articles/sandy-b...Men,+Geek,+Nerd,+Overclockers+and+Enthusiast)
LGA2011 parts originally slated for Q3 where shifted to Q4 2011, so it's not entirely unexpected due to the additional complexity of these parts (they've never developed a mass production CPU this complex before).
Sad, but there are realities to deal with as complexity increases.
The July/August delusional kids aren't going to like this news. Wonder what they'll post about next.
It is still possible to see another bump from slight newer Gulftown offerings. Not that a single clockspeed multiplier is much to right home about. The rest of the platform is much bigger concern.
X79 also supports LGA 1366 too. Maybe that is a the magic of this update.
The differences are QPI links for DP/MP and and ECC RAM support. The performance is identical at the same clock speeds. It all comes from the same wafer.
Stay calm and carry on.
With Lion in the pipeline and the Mac Pro community getting all preemptive about a new wonder box with space age CPUs, I'd like you all to sit back and reflect for a minute.
The arms race Apple is fighting with its consumers is not one the community can afford to be rushed, just let them bring out the new toys exactly as fast as they want/need/can afford to, anything rushed by Intel will be buggy/unstable/rubbish anything rushed by Apple same as dat.
Just chug along on your machines till the next gen is good and ready, your current beast will be obsolete soon enough.
My 3,1 will do me another couple of years, it has got faster over time not slower thanks to upgrades; just because they have things twice as quick doesn't make it any slower than it is.
This is mind over matter, if you don't mind, it don't matter.
With Lion in the pipeline and the Mac Pro community getting all preemptive about a new wonder box with space age CPUs, I'd like you all to sit back and reflect for a minute.
The arms race Apple is fighting with its consumers is not one the community can afford to be rushed, just let them bring out the new toys exactly as fast as they want/need/can afford to, anything rushed by Intel will be buggy/unstable/rubbish anything rushed by Apple same as dat.
Just chug along on your machines till the next gen is good and ready, your current beast will be obsolete soon enough.
My 3,1 will do me another couple of years, it has got faster over time not slower thanks to upgrades; just because they have things twice as quick doesn't make it any slower than it is.
This is mind over matter, if you don't mind, it don't matter.
Time will tell which ones were correct....
The "july/august kids" or the "not before Q4 super-pseudo-experts".
Who cry louder is not always right....
Last edited:
Let's not forget that a large percentage of MacPro users are the people Apple just apparently pissed off with FCX...
I wouldn't put money on it, but I wouldn't be surprised if Intel was pushing back consumer parts to ramp production for Apple's order now. The whole "Custom" CPU thing is obviously not Intel being a Fab for Apple designed parts. However Intel CEO did say they would be interested in building custom parts based on Intel Architecture for large clients. So more likely Apple has been working with Intel to deliver parts that may offer a different feature set based on Mac Hardware. Another thing that might potentially be custom fab'ed for Apple could be a chipset.
Some examples would be:
A Core i5 part without the integrated GFx
A Xeon part with a different core clock & multiplier & a non-production amount of cache's.
A chipset with a non-production amount of QPI's or QPI rates, PCIe lanes, etc.
So if Apple knows that OSX will never use more than 5MB of L3 cache (just an example, albeit not a very real one) then they could have chips made with 5MB of L3 cache rather than 15, 20 or 30 that would cost more.
I still think a "Mid 2011" refresh is coming soon but I'm thinking the MBA will be the darling that Apple brings to the July dance. Then the Mac Mini / Mac Pro will follow in the August timeframe.
I wouldn't put money on it, but I wouldn't be surprised if Intel was pushing back consumer parts to ramp production for Apple's order now. The whole "Custom" CPU thing is obviously not Intel being a Fab for Apple designed parts. However Intel CEO did say they would be interested in building custom parts based on Intel Architecture for large clients. So more likely Apple has been working with Intel to deliver parts that may offer a different feature set based on Mac Hardware. Another thing that might potentially be custom fab'ed for Apple could be a chipset.
Some examples would be:
A Core i5 part without the integrated GFx
A Xeon part with a different core clock & multiplier & a non-production amount of cache's.
A chipset with a non-production amount of QPI's or QPI rates, PCIe lanes, etc.
So if Apple knows that OSX will never use more than 5MB of L3 cache (just an example, albeit not a very real one) then they could have chips made with 5MB of L3 cache rather than 15, 20 or 30 that would cost more.
I still think a "Mid 2011" refresh is coming soon but I'm thinking the MBA will be the darling that Apple brings to the July dance. Then the Mac Mini / Mac Pro will follow in the August timeframe.
Well the pseudo experts are actually just pragmatic people making a probable guess based on the following facts:
- Intel has never produced custom processors for the Mac Pro before
or any other Mac device I believe.
- The X79 chipset is not ready yet, again now we need a custom chipset and Intel processor. Which makes the rumored August release even more unlikely.
- Sandy Bridge-E is on the rodmap for Q4 2011. If judging on past trends apple will release computers using the said processors after the PC industry gets it.
So what's more probable the below or the above:
Intel making custom processors or giving access to Sandy Bridge-E three month early before the Wintel market (pissing all PC enthusiasts off) as well as getting a custom/modified X79 chipset which isn't finished yet! So yeah a Mac Pro with parts that are not even finished yet or haven't passed much testing is very likely for August.
Rationale says Q1/likely Q2 2012 for new Mac Pro.
Enthusiast processors (Xeon should be similar):
HC 3.3GHz 15M
HC 3.2GHz 12M
QC 3.6GHz 10M
If the GPUs Apple gives us are any indication we will get the 6970 and the 6870.
Patsburg Chipset (Xeon version of X79)
40 PCIe 3.0 lanes (!!!), yes, that mean if OS X supported Crossfire and the MP had a beefy enough PSU you could have 4x6970s running at full throttle, 60FPS on six monitors anyone? Because Patsburg can split PCIe to 4x8 and PCIe 3.0 is ~2x the bandwidth of PCIe 2.0 the GPUs will have full bandwidth.
Since it is quad channel we will keep 4xDIMMs/CPU (I don't think Intel will go higher in channels in the near future).
There are four Patsburg varients, A, B, D, T (???)
Shared: up to 14 USB 2.0 ports, four SATA 3Gbps ports, two SATA 6Gbps ports
A Varient: four extra SATA 6Gbps ports via expander (probable)
B Varient: same as A except SAS instead of SATA via expander (unlikely for use)
D Varient: eight extra SATA/SAS 6Gbps ports via expander
T Varient: D plus RAID5 with nonvolatile SRAM cache (unlikely for use)
HC 3.3GHz 15M
HC 3.2GHz 12M
QC 3.6GHz 10M
If the GPUs Apple gives us are any indication we will get the 6970 and the 6870.
Patsburg Chipset (Xeon version of X79)
40 PCIe 3.0 lanes (!!!), yes, that mean if OS X supported Crossfire and the MP had a beefy enough PSU you could have 4x6970s running at full throttle, 60FPS on six monitors anyone? Because Patsburg can split PCIe to 4x8 and PCIe 3.0 is ~2x the bandwidth of PCIe 2.0 the GPUs will have full bandwidth.
Since it is quad channel we will keep 4xDIMMs/CPU (I don't think Intel will go higher in channels in the near future).
There are four Patsburg varients, A, B, D, T (???)
Shared: up to 14 USB 2.0 ports, four SATA 3Gbps ports, two SATA 6Gbps ports
A Varient: four extra SATA 6Gbps ports via expander (probable)
B Varient: same as A except SAS instead of SATA via expander (unlikely for use)
D Varient: eight extra SATA/SAS 6Gbps ports via expander
T Varient: D plus RAID5 with nonvolatile SRAM cache (unlikely for use)
1 These is an example "custom" processor from intel for mac
- Intel Core 2 Duo P7500 and P7700 in First MacBook Air
- Intel Core 2 Duo E8135 and E8235 in iMac Early 2008
2 These is an example x79 motherboard from various vendor.
- http://www.anandtech.com/show/4404/computex-2011-twelve-x79-motherboards
- http://www.ninjalane.com/articles/general_information/sapphire-tw_visit/page4.aspx
- http://motherboardnews.com/2011/05/30/x79-motherboards-galore-at-computex-2011-pre-show/
3 Apple has an early access to Intel CPUs/Chipsets many times.
- Z68 in iMac 1 month before release.
- Xeon "Harpertown" Bus 1600 in Mac Pro 1 month before release.
Correct.
What they used for the original Air was listed as a custom chip, but it was actually a design Intel designed for what they saw as a new market segment. Unfortunately, when the initially went around to system vendors, they all turned it down. Then Apple asked about a CPU for what became the Air later on, and Intel took that existing design (was just sitting on a shelf), updated it a bit to speed things up, and voila - the Air was born.
Actually, there's recent rumors that have indicated that the SB-E has been pushed back to Q1 2012. It's not been confirmed by Intel yet, but it makes sense IMO due to the additional complexity (2011 pins is one heck of a hint on this). Example article.
Unfortunately, this isn't correct.
The X79 is still PCIe 2.0, and it even has the same lane count as Nehalem/Westmere chipsets (X58 and 5520). You end up with 36 lanes for PCIe slots, of which Apple uses 32 for Slots 1 and 2 (each a 16x lane dedicated slot), and 4x that are shared for both Slots 3 and 4 (sharing done via a PCIe switch).
But it is designed with a way to reduce the existing DMI bottleneck in the current chipsets by using an additional 4x PCIe lanes for storage bandwidth.
Makes sense, given the inclusion of a SATA III controller (10 ports @ this spec, and another 4x at SATA II).
It is funny how people tend to believe only rumors that suit their theory...
There have been some rumors about an August release of the MacPro, the "not before 4Q" people has been quite fast to mark that as rubbish...
Now we have an unconfirmed rumor about an alleged postponing of SB xeons, and suddenly that became a dogma.
Either one believes in rumors or one doesn't, since without a confirmation from Intel, all of them are pure speculations.
Apple could even release a non-xeon Sandy Bridge MacPro for the Single Processor base model....
As i said, let's wait and see, but do not mark the opposite group as "kids"....
Last edited:
And where is PCIe 3.0 listed in that slide (portion I was primarily responding to)?
It's neither listed as a bullet point on the left side, or in the Key Features section on the right.
The marketing department would have made sure PCIe 3.0 was mentioned if it were there (what they've done historically).
There's more to it than just cherry picking what fits the theory of the SB-E 16 y/oparts not reaching Apple's hands 6 months or more early, and this is the part that seems to be ignored by the "It will be here in August" crowd.
This difference comes down to experience in my observations over the years. Some of us have seen a number of announcements that didn't meet the original release dates, and were delayed on more than a few occasions (not just Intel either), and in some cases, never actually released (aka vaporware in both software and hardware).
In my particular case, I'm a Computer Hardware Engineer and deal with the types of issues that cause such delays. One major factor, is the common tendency to under-estimate the development time. Another is complexity (takes more time to complete the hardware verification phases of design = working out the bugs).
Think of it this way; past versions of 771 pins weren't as complex as the 2011 pin parts Intel is working on now. The pin count has more than doubled in ~ the last 3 years (last new 771 pin parts were the 2008 Harpertowns). It's not just more pins, but there's more transistors as well. All of which means there's a higher probability of bugs in the initial design that need to be sorted, and this takes more time to sort out (should be common sense if you stop and think about it for even a moment).
I never used the word "kids"... You're attributing this to the wrong person.
I can understand where the sentiment comes from, but again, I see the basis for that comment as lack of experience (which there is plenty of scientific evidence to support that younger generations have less than older ones).
Where I've seen the discussions get out of hand (and why some of these threads have gotten a bit out of hand), is how some of the August crowd have posted an "in your face" attitude within comments (very confrontational rather than logical). So some of us will recall 12 - 20 year olds for example, as that's more common amongst this particular age group than say the 40 - 60 year old crowd.
There are eight PCI Express 2.0 lanes, SNB-E PCI Express is different.
PCI Express 3.0
PCI_E3
Last edited by a moderator:
If you look at all the boards that have been shown they are all 3.0, that was how it became known. I guess Intel haven't really said anything official yet so that's all we have to go on.
Ah, OK. Nice catch.
I was going off of the slide that had been presented for X79, not board images (only saw one, and it wasn't close enough to see the PCB legend for PCIe 3.0 - LGA2011 socket was definitely visible though).
Nice to see that there are some PCIe 3.0 lanes, as Intel's chipsets have been a bit disappointing to me recently (X58/5520 on).
The H68 definitely comes to mind... I haven't gone and dug for board images (just saw the one mentioned above).
Nice to see they did though.
The H68/Z68 wasn't wonderful IMO, and had me concerned that they were trying to cut costs on the chipset for LGA2011 (had the impression they only addressed the storage bottleneck to a good extent <added SATA III and dealt with the DMI bandwidth issue from the previous ICH's>, and included the ICH on the same die, but stopped there).What I don't get, is why that wasn't highligted/emphasized by Intel with the X79 leaks, as it would have generated a lot of enthusiasm for the socket.
What I don't get, is why that wasn't highligted/emphasized by Intel with the X79 leaks, as it would have generated a lot of enthusiasm for the socket.
Maybe because the controller is on the CPU it is on other slides that weren't leaked? The processor slide was really brief IIRC.