Thursday, November 17th 2011

SB-E: Enthusiast Full 8 Core Dual Socket Monsters On The Way Early 2012

The latest Sandy Bridge-E 6 core processors have just been released, to excellent reviews. However, the architecture is designed for 8 cores, so these current i7-3960X & i7-3930K processors actually contain those 8 cores, but with two turned off in order to enable them to fit within a manageable 130 W power envelope. Hence there's quite a bit more potential to be released and soon. Therefore, anyone looking to invest in the premium-priced SB-E platform right now, should note that these processors are at the initial C1 stepping and have the VT-d hardware virtualization issue and PCI-E 3.0 compatibility uncertainty. The VT-d problem will be a real show stopper where hardware acceleration of a virtual machine is a must, so it shouldn't be ignored.

VR-ZONE brings us news that the fully unlocked SB-E 8 core chips will be released as the long awaited Xeon E5 family of processors, which will be built on the C2 stepping, solving the above issues. However, being 8 core, these will be very power hungry indeed, consuming around 150 W at just 3 GHz with all 8 cores active and 20 MB of L3 cache. At 2.5 GHz though, the new processors are expected to fit within the 95 W power envelope.
Being "Xeon" processors, they are primarily intended to be used in multi-socket configurations in servers and very high end desktops for businesses that need this kind of raw power. Of course, the other market that these E5 processors are aimed at, are the hardcore PC enthusiasts who have very deep pockets and want the ultimate power in their desktops at any price. It's rumoured that such people might be using these powerhouse systems for more than just playing Sudoku, doing a bit of word processing and browsing the internet, but these are unconfirmed at the time of publication.

Later on in 2012, Ivy Bridge-E is expected to be released, which will be an optical shrink of SB-E, while also using Intel's new 22 nm tri-gate transistors. This will bring significant performance improvements, including an increase to 10 cores and 25 MB L3 cache. TechPowerUp has more details on this generation, here. So, with these new high wattage processors eventually becoming more mainstream, could we see the return of the ill-fated BTX (Balanced Technology Extended) form factor case, introduced by Intel around seven years ago? Only time will tell.

For more detail on these SB-E & IB-E processors, hop on over to VR-ZONE.
Add your own comment

51 Comments on SB-E: Enthusiast Full 8 Core Dual Socket Monsters On The Way Early 2012

#1
HumanSmoke
mastrdrver said:
W3680 (I think that's the one.)

They are sold as "locked" but you can move the multiplier over the "limit".
Not sure about that one, but some of the server Xeons (X5xxx) certainly aren't multi-locked in the conventional sense.
Couple of examples ( 2P capable) Westmere-EP's
Xeon X5677 (quad, nominal 26x multi)
and
Xeon X5690 (hex, nominal 26x multi)
Xeon X5687 (quad, nominal 27x multi)
Posted on Reply
#2
Unregistered
HumanSmoke said:
Not sure about that one, but some of the server Xeons (X5xxx) certainly aren't multi-locked in the conventional sense.
Couple of examples ( 2P capable) Westmere-EP's
Xeon X5677 (quad, nominal 26x multi)
and
Xeon X5690 (hex, nominal 26x multi)
Xeon X5687 (quad, nominal 27x multi)
Yes. Since they (Westmere-EX) are much of the same thing as an extreme Gulftown like a 990x; but with dual QPI. Some chips had better bins than their desktop equivalents. And were also much more expensive cause of it. For example, the X5675 can OC to 4.7-5 Ghz range easier than an i7 970, while doing it at 95W TDP instead of 130. ;)

15th Warlock said:
I don't get your point, the SR-3 is clearly being marketed as an enthusiast board, that will support two server/workstation processors (namely SB-EP Xeons), it's not being marketed as a server/workstation board, hence I referred to it as an exception.
Boards aren't the point, the chip is. It needs an unlocked multi to OC regardless of which board you plug it into. So if the SR-3 can take it and OC, then a desktop X78 should also be able to do the same as long as it's supported (as pointed before, similar to using Xeon's on X58).

15th Warlock said:
As for the quality of EVGA's motherboards, I have never bought one of their boards, I still have an old GTX285 rocking in my HTPC, but I cannot vouch for the quality of their more recent graphic cards either, I didn't know their products were known because of their low quality... :ohwell:
They used to be top of the line. Both the Classy 4-way X58 and the P55 Classy 200 were some of the best boards in their class. As well as their 285's. Things went down when their Europe mobo division broke (since they're a small company that mainly sell GPU's). One of the guys went Biostar, Shamino went Asus and a few others went Sapphire to build the Pure Black X58. It happened in August of last year.
#3
NC37
15th Warlock said:
I have to agree with btarunr on this one, the title to this thread isn't entirely accurate, although the SR-3 might be an exception to the rule, these CPUs are not targeted to enthusiasts, but to work on a workstation/server configuration, not even Intel has ever referred to Xeons as enthusiast processors, even though a select few use these processors at home.
A select few and a bunch of Mac users. Remember, Apple thinks consumers need Xeons in their towers. Course that helps them justify the premium I guess. I admit it should be an option for the professionals, but I know more non pro people with Mac Pros than I do otherwise. They don't need Xeons.
Posted on Reply
#4
Jizzler
FordGT90Concept said:
Like I said in a previous thread, Intel can't release a Xeon chip with VT-d borked. They'll piss off a large chunk of the Xeon market.

Intel apparently has no problem with releasing a Core i7 chip with VT-d broken because most Core i7 users will never use it anyway.
Sorry, I didn't CnP enough. That quote was referring to the currently released SB-E chips. Was wondering if this part of the OP was still true:

Qubit
Therefore, anyone looking to invest in the premium-priced SB-E platform right now, should note that these processors are at the initial C1 stepping and have the VT-d hardware virtualization issue and PCI-E 3.0 compatibility uncertainty.
Ars and others are now reporting that VT-d is supported based on the Ark entry, but I haven't seen any word from users or not saying it was enabled/working.

Posted on Reply
#5
qubit
Overclocked quantum bit
Yes, it's 'supported', but due to a bug in the processor, the hardware acceleration part doesn't work.
Posted on Reply
#6
Hayder_Master
btarunr said:
No, these are not enthusiast products. These are 2P server/WS.
right, but don't forget there is EVGA SR3 out there.
Posted on Reply
#7
qubit
Overclocked quantum bit
Hayder_Master said:
right, but don't forget there is EVGA SR3 out there.
That SR3 mobo, as reported by TPU. :cool:
Posted on Reply
#8
Breit
It's rumoured that such people might be using these powerhouse systems for more than just playing Sudoku, doing a bit of word processing and browsing the internet, but these are unconfirmed at the time of publication.
made my day. :cool:

cheers!
Posted on Reply
#9
Breit
HumanSmoke said:
Not sure about that one, but some of the server Xeons (X5xxx) certainly aren't multi-locked in the conventional sense.
Couple of examples ( 2P capable) Westmere-EP's
Xeon X5677 (quad, nominal 26x multi)
and
Xeon X5690 (hex, nominal 26x multi)
Xeon X5687 (quad, nominal 27x multi)
your'e right when you say 'they aren't locked', but thats also true even for the lowest core i7-9xx parts. the point is that the westmere(-ex)s have a lowest multi of 12 (regardless of processor type/number) AND they have a highest multi, which in case you haven't an extreme edition cpu, is exactly the multi they are rated at (including turbo bins of course) and changing the multi is allowed only in that range! so only extreme edition cpus are allowed to overclock via multi and as far as i know there aren't any xeon extreme edition parts (yet?)!

the last extreme edition xeon i know of was the intel core 2 extreme qx9775 (socket-771) back in the days (good old asus z7s, don't miss you. ;))...
Posted on Reply
#10
repman244
John Doe said:
Xeon equivalents of i7 920 and Yorkfield chips were made with enthuasiests in mind.
Xeon chips are always made for workstations/servers etc. due to support for ECC RAM and those which run 2P have dual QPI. Their purpose is not for the enthusiast, they just use them because they are cherry picked (low power/TDP).
Posted on Reply
#12
Super XP
Is there a benefit in using a nice ASUS Server motherboard for either a SB-EP or a Interlagos setup for say a Gaming Rig or games are just not designed right to really benefit from the extra cache, memory and cores?
Posted on Reply
#13
Wile E
Power User
John Doe said:
Which exception? It's not like eVGA is getting special treatment from Intel...

for a fact, their late offerings have been a joke in regards to quality and/or performance with decade old Copper ring-choke's. After having lost Shamino and rest of the crew, they haven't done anything at all. They're wiping the floors (as a small company as well) and nobody cares. Except for fanboys on their forum, whom I got banned for. Obviously suggesting people Twin-Frozr's and constantly bashing them wasn't a good idea. lol. But that's another story.

That said, the SR-3 can't OC on it's own. It needs unlocked multi's. Not a "single or dual-QPI?" situation like it was with X58. If it technically can OC, then a desktop board that supports the chip (like on X58) would OC as well.
The SB-E platform allows clocking via base clock manipulation. No need for an unlocked multi.
Posted on Reply
#15
btarunr
Editor & Senior Moderator
Hayder_Master said:
right, but don't forget there is EVGA SR3 out there.
...which is not going to unlock your Xeon.
Posted on Reply
#16
Wile E
Power User
btarunr said:
...which is not going to unlock your Xeon.
Which isn't needed to clock on 2011.
Posted on Reply
#17
btarunr
Editor & Senior Moderator
Wile E said:
Which isn't needed to clock on 2011.
Then your SR3 is as good as a cheapo 2P workstation board from Tyan or Supermicro. SnB-E responds to BClk increase, but it's nowhere as responsive as Westmere/Nehalem were. That's one of the reasons Intel started its LGA2011 lineup off with two unlocked Core chips.

Unless Intel does something similar to Core 2 Extreme QX9775 (2P-ready Core chips for Skulltrail), SR3's future is bleak.
Posted on Reply
#18
Wile E
Power User
btarunr said:
Then your SR3 is as good as a cheapo 2P workstation board from Tyan or Supermicro. SnB-E responds to BClk increase, but it's nowhere as responsive as Westmere/Bloomfield were. That's one of the reason Intel started its LGA2011 lineup off with two unlocked Core chips.
Tyan and Supermicro aren't going to give you the voltage options the SR3 will, let alone the PCIe layout. Besides, even the locked SB-E chips have multi options above their turbo frequency multi.

And the argument that Intel started off with unlocked chips on this platform is kinda moot, as they always release unlocked chips with a new enthusiast platform.
Posted on Reply
#19
btarunr
Editor & Senior Moderator
Wile E said:
Tyan and Supermicro aren't going to give you the voltage options the SR3 will, let alone the PCIe layout.
That wasn't my point. Despite all the world's voltage options, LGA2011 chips aren't going too far with just BClk-tuned overclocking (even if voltage-assisted). They still need upwards unlocked multiplier to achieve any "big" overclocked speeds.

Wile E said:
Besides, even the locked SB-E chips have multi options above their turbo frequency multi.
Locked Sandy Bridge-E chips (like i7-3820) will, because they are advertised "limited unlocked". Locked Sandy Bridge-EP Xeon chips will NOT. They aren't advertised to have absolutely any performance tuning capabilities to begin with. They will not have even a single notch of multi above its max Turbo multiplier.

Sandy Bridge-EP will be the only chips you can run on SR3. The topic concerns SnB-EP, not SnB-E.

Wile E said:
And the argument that Intel started off with unlocked chips on this platform is kinda moot, as they always release unlocked chips with a new enthusiast platform.
That wasn't my point, either. My point was that Intel launched LGA2011 series with only >$500 unlocked chips, excluding the locked quad-core (i7-3820). That's because 3820 doesn't go too far with just its BClk, and it makes i7-2700K look like a better option for its price range.

You will still be able to buy i7-3820 in February, by then you'll know what a fail SnB-E is with BClk overclocking. It will respond to BClk overclocking, but nowhere near as effectively as "locked" Westmere Xeon/Core or Nehalem Xeon/Core processors did.
Posted on Reply
#20
Wile E
Power User
btarunr said:
That wasn't my point. Despite all the world's voltage options, LGA2011 chips aren't going too far with just BClk-tuned overclocking (even if voltage-assisted). They still need upwards unlocked multiplier to achieve any "big" overclocked speeds.



Locked Sandy Bridge-E chips (like i7-3820) will, because they are advertised "limited unlocked". Locked Sandy Bridge-EP Xeon chips will NOT. They aren't advertised to have absolutely any performance tuning capabilities to begin with. They will not have even a single notch of multi above its max Turbo multiplier.

Sandy Bridge-EP will be the only chips you can run on SR3. The topic concerns SnB-EP, not SnB-E. If qubit doesn't give this article a complete overhaul by today (as we discussed yesterday), we'll pull it off.



That wasn't my point, either. My point was that Intel launched LGA2011 series with only >$500 unlocked chips, excluding the locked quad-core (i7-3820). That's because 3820 doesn't go too far with just its BClk, and it makes i7-2700K look like a better option for its price range.

You will still be able to buy i7-3820 in February, by then you'll know what a fail SnB-E is with BClk overclocking. It will respond to BClk overclocking, but nowhere near as effectively as "locked" Westmere Xeon/Core or Nehalem Xeon/Core processors did.
OK, I'll bite. What makes you believe this?
Posted on Reply
#21
btarunr
Editor & Senior Moderator
Wile E said:
OK, I'll bite. What makes you believe this?
Google your way through base clock overclocking of SnB-E keeping multiplier constant. I won't google them for you.
Posted on Reply
#22
Wile E
Power User
btarunr said:
Google your way through base clock overclocking of SnB-E keeping multiplier constant. I won't google them for you.
Who's to say this will remain a constant?

EDIT: And I'll pick up with this conversation with you tomorrow. Gnite all.
Posted on Reply
#23
Breit
i'll bet sr3 have an option to change cpu bclk gear ratio (maybe even fine-tunable, just a guess), while tyan/supermicro will not offer such thing in a thousand years...

so that alone with limited bclk adjustments in +/- 10mhz range and adjusting multi downwards gives at least some overclocking potential.

lets say you have a max. availabe turbo-multi of 39 (assuming 3.9ghz max turbo freq. on high end xeons) AND you are able to change the bclk to max +10mhz AND your cpu is stable at a gear ratio of 1.66, then you come in at 176mhz bclk on the cpu alone, wich is a little over 6.8ghz.

further assuming your xeon is a 8-core variant, you are ending up with a 16-core 32-thread monster at 6.8 ghz AND 8 channels of ddr3 goodness... i mean WTF?
even if your xeons 'only' work at a gear ratio of 1.25, you come in at 5265mhz on the core assuming this +10mhz on the bclk.
Posted on Reply
#24
Wile E
Power User
btarunr said:
Google your way through base clock overclocking of SnB-E keeping multiplier constant. I won't google them for you.
Throw me a bone. All I've found was an Anand article. What phrase did you google?
Posted on Reply
#25
mastrdrver
Breit said:
your'e right when you say 'they aren't locked', but thats also true even for the lowest core i7-9xx parts. the point is that the westmere(-ex)s have a lowest multi of 12 (regardless of processor type/number) AND they have a highest multi, which in case you haven't an extreme edition cpu, is exactly the multi they are rated at (including turbo bins of course) and changing the multi is allowed only in that range! so only extreme edition cpus are allowed to overclock via multi and as far as i know there aren't any xeon extreme edition parts (yet?)!
Incorrect at least for the w3680.
Posted on Reply
Add your own comment