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
The 970 has occupied the P2 price point of $583 since the February refresh for LGA 1366 processors. The 980 replaces it at the same price. If they dropped the 970 to your suggested price range then the $600 price point becomes terrible value and no clued-up enthusiast would pay $200-300 for a +1 multiplier.
The Core i7 990X probably bumped that down.
The Core i7 920/930 sold like hotcakes at ~$280 compared to their faster siblings. Prices scale up ridiculously beyond $300 for Intel. You draw the line at the Core i7 2600K or Core i7 950/960 today.
The Gulftown Core i7 970 would be a great replacement to the aging Nehalem Bloomfields.
Right so you have to offer nice incentives on the $600 and $1,000 price points. Also that they are able to bump the multiplier up one doesn't mean that Westmere yields would be enough for the demand if they were a $300 part, or that Nehalem production is ready to be scrapped.
The 3.6GHz quad Sandy Bridge-E will be a great replacement for them.
Don't get me wrong, I'd love to see $300 6-cores, I just don't think Intel are anywhere near being ready to do that - as evidenced by the Sandy Bridge-E leaked line-up.
Hex-core and an unlocked multiplier become those incentives.
I was under the impression that the entry level Sandy Bridge-E processor was a "locked" hex-core. I fail to recall where I came upon this information. My guts says Xbitlabs.
LGA 1155 + Intel 6 Series is going to offer an unlocked multiplier and much lower platform costs. I would be surprised to see any X79 hit under $250. Even if you are splitting the quad channel RAM on the board.
The base is a quad and is apparently locked, though one would assume that it has 6 cores with two disabled as Westmere quads are. I think you are right on X79 being expensive, all discussion from engineers I've seen points to that too.
I still believe the entry is going to be a hex-core. In its complete state Sandy Bridge-E is an octo-core processor. It seems very excessive to cut that down a quad core outside of certain low power situations.
~$300 might be underestimating the price of entry though.
Doesn't seem to be locked, but "limited".
http://www.anandtech.com/show/4291/additional-details-on-sandy-bridgee-processors-x79-and-lga2011
Ouch, it looks like ~$300 will not be hard to hit then. 6 more clock multipliers is still quite a bit of room to work with on a locked processor. You can overclock the non-K processors as well. You are just limited to 4 more bins. (Bins, multipliers oy!)
The clock speeds are not all that impressive over the current Westmere based Gulftown hex-cores. 32 nm hex-cores have already hit their peak. The Core i7 2600K looks like a better choice if you do not need all those lanes or even waiting for Ivy Bridge to get mainstream PCI-Express 3.0.
Oh and nice article.
Well, think of it this way; if they pushed the clocks to what's possible, what would be left to make the 22nm versions attractive?
With the delay on Sandy Bridge-E, Ivy Bridge is all the more attractive. Think we will even have a chance to see Ivy Bridge based Xeons before the new architecture?Well, think of it this way; if they pushed the clocks to what's possible, what would be left to make the 22nm versions attractive?
Of course not.
What a foolish notion! They will run the SB Xeons first to gain as much profit from the 32nm facilities as possible. Ivy Bridge will follow later, but the # of days between SB and IB could be reduced a bit, assuming they don't run into problems implementing the 22nm process at full scale and the SB-E's don't totally bomb out (don't expect this, but we've not seen anything on quantity pricing yet).
At $885USD for the top E3-1290, I've the impression the top end SB-E5 variants are going to be absolutely cringe-worthy though ...
http://en.expreview.com/2011/07/11/specs-of-intel-xeon-e5-sandy-bridge-ep-cpus-surfaced/18115.html
Eight cores are there for the Xeon parts.
Eight cores are there for the Xeon parts.
First Page Updated!!
CPU:
The Sandy Bridge-E based "Intel Xeon E5"
- E5-1600 Series for Single Socket
- E5-2600 Series for Dual Socket
Known Configuration
- Quad Core / 8 Threads 3.6 GHz 10MB L3
- 6 Core / 12 Threads 3.2 GHz 12MB L3
- 6 Core / 12 Threads 3.3 GHz 15MB L3
- 8 Core / 16 Threads 2.3 GHz 20MB L3
Other Known Specification
- Socket R (2011 pins)
- 2 QPI links
- 40 PCI-E 3 Lanes
- 4 DMI 2.0 Lanes
- Quad-channel DDR3 memory controller
- Max 96 GB DDR3 Per CPU
Chipset:
- Intel X79
- Codenamed Patsburg
- PEG 2x16 up to 4x8
- Intel Rapid Storage Enterprise 3.0
- SATA Total (Max for 6GB/S) 14(10)
- Integrated Gigabit Ethernet MAC (Lewisville PHY)
GPUs:
Possible Candidate (I put the bet on ATI).
- ATI Radeon HD 6970 <- (ATI Replacement for 5870)
- ATI Radeon HD 6870 <- (ATI Replacement for 5770)
100% in 2011 Mac Pro:
- Thunderbolt Port
Other Rumors:
- Late July / Early August Launch along with Lion.
- Stackable / Rackable Newly designs case.
- May be use special Intel CPU
CPU:
The Sandy Bridge-E based "Intel Xeon E5"
- E5-1600 Series for Single Socket
- E5-2600 Series for Dual Socket
Known Configuration
- Quad Core / 8 Threads 3.6 GHz 10MB L3
- 6 Core / 12 Threads 3.2 GHz 12MB L3
- 6 Core / 12 Threads 3.3 GHz 15MB L3
- 8 Core / 16 Threads 2.3 GHz 20MB L3
Other Known Specification
- Socket R (2011 pins)
- 2 QPI links
- 40 PCI-E 3 Lanes
- 4 DMI 2.0 Lanes
- Quad-channel DDR3 memory controller
- Max 96 GB DDR3 Per CPU
Chipset:
- Intel X79
- Codenamed Patsburg
- PEG 2x16 up to 4x8
- Intel Rapid Storage Enterprise 3.0
- SATA Total (Max for 6GB/S) 14(10)
- Integrated Gigabit Ethernet MAC (Lewisville PHY)
GPUs:
Possible Candidate (I put the bet on ATI).
- ATI Radeon HD 6970 <- (ATI Replacement for 5870)
- ATI Radeon HD 6870 <- (ATI Replacement for 5770)
100% in 2011 Mac Pro:
- Thunderbolt Port
Other Rumors:
- Late July / Early August Launch along with Lion.
- Stackable / Rackable Newly designs case.
- May be use special Intel CPU
8 Core / 16 Threads 2.3 GHz 20MB L3
That's an engineering sample, while there may be a CPU with that clock rate, it isn't certain.
Also not sure where you got 96GB per CPU from. It will be at least the same as now which is two quad ranked DIMMs per memory controller, which would give you 8x32GB.
That's an engineering sample, while there may be a CPU with that clock rate, it isn't certain.
Also not sure where you got 96GB per CPU from. It will be at least the same as now which is two quad ranked DIMMs per memory controller, which would give you 8x32GB.
I don't see more than 1x DIMM per channel per socket either, given the physical constraints.
True, but the MP and the Mini have never had the same, either ... "just saying". IMO, it would be an interesting approach if the mini was made into the SP hot rod variant .. and Apple could introduce an A6-based "Mac Micro" under it.
The prior comments about eliminating legacy bottlenecks is what caught my casual eye. Afterall, that was part of the legacy of the G5 back in 2003.
Also, Grand Central Dispatch has been languishing out there in the hinterland for many moons .. a MP with FCPX - and both optimized to exploit it - could potentially be the real story that's yet to break wide open.
-hh
wow @ the low clock speeds. We already have similar clock speed Westmeres, why would anyone pick the new E5 over a current 12 core for example? aside from Sata3 and TB...
I have a feeling the 12 core MP's will still out-performance any of these till the higher clocked DP's come out.
Sent from my iPhone using Tapatalk
I have a feeling the 12 core MP's will still out-performance any of these till the higher clocked DP's come out.
Sent from my iPhone using Tapatalk
Clocks and cores aren't everything... check out this comparison of a Quad Core SB iMac vs a Hex Core Westmere of similar clocks...
That's about a 20% advantage in favor of the SB Quad Core iMac. Now, the iMac didn't rule every test, but it does show that SB has some interesting architectural improvements that are worthy of consideration.
Link to full test results at Bare Feats
oh, so theoratically(?) speaking 2010 hex core 3.x will (would) be ~20% slower than the SB hex 3.x ? that's pretty tempting *if* it will be priced the same... Looking at geekbench scores I still.can't find anything that could outperform the 12cores though.
Don't get me wrong this isn't bad, but I was hoping for a new "king of the hill" MP
thanks for the bare feat article btw!
Ps: just want to add that im talking from my point of view. meaning I use my MP for (professional) editing and after effects. I didnt want to bash any config...
Sent from my iPhone using Tapatalk
I do wonder now, given the number of rumours about a July release means that we have two options now:
1. Apple has got SB-E ridiculously early and will launch July and ship late July.
2. Apple has got SB-E early and will launch July, ship August/Sept.
But I still feel that there won't be a MP refresh until Q3/Q4 this year *shrug*.
Either way my bank balance is ready for a new MP so we will see!
1. Apple has got SB-E ridiculously early and will launch July and ship late July.
2. Apple has got SB-E early and will launch July, ship August/Sept.
But I still feel that there won't be a MP refresh until Q3/Q4 this year *shrug*.
Either way my bank balance is ready for a new MP so we will see!
This result also illustrates that FCPX (under OSX 10.6) doesn't use hex cores very efficiently.
Yes. Just goes to show that FCPX is not as well optimized for multicore as touted. A i7-2600 is close to but not faster than a i7-980 when SW is working correctly. The gains just show that the operations were most likely NOT using all cores with the 2600 in a better space for single thread turbo clock speed.
Same site with actual gaming results.
Kind of not a contest:
http://www.barefeats.com/imac11e.html
Same site with actual gaming results.
Kind of not a contest:
http://www.barefeats.com/imac11e.html