it would be nice to have a "thunderbolt" type port..

rottenmutt

Please post a link to the block diagram.

Not official until Intel spills beans, but this is what is known so far:


 
Just so you know Socket R = LGA2011 and pertaining to SR-X.

Hopefully EVGA doesn't use a pcie bridge chip and uses both the native bridges off both CPUs. Two CPU's would be required to enable all pcie ports, which isn't a big deal as no one will buy this board to only run one CPU. And if they did they probably wont be running quad sli anyway.

do they know of a possible release date yet? or season?

takuhari

do they know of a possible release date yet? or season?

Since SR-X can only use Xeon E5-2600 series processors, there is no point for the board to show up earlier, therefore look for when those Xeons appear to get an idea.   FYI, the Xeon 5500 debuted in March of 2009 while Xeon 5600 and SR-2 arrive on March 2010, so I'd look to this March 2012 for SR-X.

Yeah March of '12 would be nice. Here's a few more pics from VR-Zone:
 
http://vr-zone.com/articl...x-preview/14198-2.html

it looks like a whole bunch of uberness, except for the 8 on one 2 on the other really please just bust it out and make 8 on both it's not like you'll destroy the world with it. *yet*
I already have a case picked out for this bad boy.
and again please please make one with the grey/black combo.
 
side note: xenon 2011 socket could be fun

Is expected any xeon with more than 8 cores for the socket 2011?, it seems that Intel does not have in his roadmap any 10 core or more for this socket.

I've got to say, I'm a huge fan of the concept of the SR- series of motherboards; what I desire in a motherboard is essentially a highly break-out board for all the various features of the platform I'm running - and that is exactly what the SR- series is trying to offer.
 
However, I've got a few nits to pick on the SR-X motherboard, from what I've seen of it, because I don't think it goes far enough.
 
First off, RAM slots: I could care less about aesthetics, but having an asymmetric NUMA configuration is next-to-useless - there is potential for some performance issues those configurations, and naively implemented NUMA-aware software won't really behave nicely.  I'd much rather see 4-DIMMs per socket on both sockets, to free up some board space and minimize trace-lengths - 8 and 16GB DIMMs are readily available, and we haven't even seen LR-DIMMs on the market yet - and those should double capacities at similar prices per GB - memory capacity will hardly be the limit of this platform.
 
Secondly, I/O: Why are we stuck with only dual-GbE LANs and 4 USB3?  Why do we have only 7 PCI-E slots?  The biggest advantage of the SR-X that I can see is in I/O bound or lightly thready workloads that need high configurability and performance.  The Romly-EP platform has FAR too many PCI-E lanes to simply break them all out into slots, even on an HPTX motherboard.  As far as I can tell, if you went with 4DIMMs/socket on both sockets, you'd have room for 8 FHFL PCI-E slots; each CPU socket provides 40 lanes of PCIE3, with one of the sockets offering additional 4 lanes of PCIE2 because it isnt connected to a Patsburg chipset, and then Patsburg itself offers 8 PCIE2 lanes.  From my perspective, the SR- series motherboards are about making the most of what the platform has to offer - and I don't think the current configuration quite achieves that.  I, personally, would like to see 8 PCI-E slots, (4 per socket) configured electrically as PCIE3x8, PCIE3x16, PCIE3x0/x8, and PCIE3x16/x8 for each socket - this would expose all available PCIE3 lanes to the user, and leave room for a PCIE3x8 card even in quad-GPU scenarios.  Then we're left with the matter of what to do with our 12 other PCI-E lanes.  For the PCIE2x4 coming directly off the socket without Patsburg, I'd really like to see at a quad GbE NIC at the very least - something along the lines of Intel's i350-AMT4 controller would do the trick very nicely.  I'd also like to see 2-4 of the lanes from Patsburg going to mini-PCIE slots with mSATA connections - those are nifty little slots that don't take too much board real-estate, and mSATA for SSD caching would be a nice way to utilize the space under the PCI-E cards.  Needless to say, the native SATA and SAS interfaces on Patsburg should all be broken out, but that's already present on the current SR-X.  Pretty much anything else should go to USB3 (or maybe  firewire or thunderbolt to cater to the video-editing crowd), because why the hell not?  This is a "build it that they will come" product model - it's not "because we need it" it's "because someone, somewhere, might want it, and even if they don't we might as well."  Let's take that sentiment and run with it!
 
I wouldn't normally go this far with armchair design suggestions, but this platform is likely to be the absolute peak of the high-end market for both Sandy Bridge EP and Ivy Bridge EP; if this high end has to last us 2 years until Haswell (or, hope against hope, a halfway decent AMD platform :P) it'd better be done in the right way.

massdestructor

Is expected any xeon with more than 8 cores for the socket 2011?, it seems that Intel does not have in his roadmap any 10 core or more for this socket.

I suspect a 12-core variant is possible only with the 22nm shrink to stay within the 150W TDP max, although it is my tentative belief it will be for E5-4600 market early on.  It is the only market where they would need to improve their more cores per dollar ratios. 
 
The reasoning for my thinking is as follows (you should look beyond leaked roadmaps and rumors):

1. The upper end Xeon for enterprise systems in 4-way boards have always had more cores than regular server, let alone desktop.  Intel's first 6-core was a 45nm Penryn part known a Xeon 7400-- before Nehalem.  Intel's first 8-core was a Nehalem-EX known as Xeon 7500 Nehalem part in LGA1567 socket while no such variant arrived in mainstream server, while that same socket saw the Westmere-EX bring 10-core models known as Xeon E7 and the highest-end desktops only saw the 6-core Gulftown.  Intel reserves the highest cores for Xeon multi-processor and i7's won't have it.
2. When Intel created the LGA1567 socket for their upper end Xeons 7500 series, it was on the assumption AMD would stick to monolithic dies like ther first dual-core, quad-core, 6-core and eventually the Bulldozer 8-core.  Intel wasn't counting on AMD finally invoking it a few years back and suddenly made a 12-core and Bulldozer brought a 16-core.  While any one of us can say Intel's parts still beat AMD, Intel doesn't tolerate any advantage AMD can grasp, it is a threat even if we don't think so.  AMD has the potential of selling more cores per dollar in their drop-in compatible LGA1944 while LGA1567 was limited to 10-cores based on the previous idea Intel wouldn't need more to compete.  So to beat AMD, Intel created LGA2011; it is designed for more than ten cores.  How much more is unknown.
3. Just so you know, there is a LGA1356 socket and those parts are known as Xeon E5-2400 series include 8-core models arriving after SR-X and E5-2600's.  Simplified math (ignoring IMC and IPC) comes to 170 pins per core and therefore 11.8 cores can fit in 2011.  I'm sure you can see if the controller and system agent related pins were constant, a 12-core would easily fit.

Would these monster super-core CPus come to SR-X?  I don't know; if there were to come to Ivy-Bridge-EP then sure, but I'm betting they will remain an Ivy Bridge-EX parts.  You did ask about any Xeon in 2011, it is not like SR-X and X79 are the only versions.

listen to people, please get rid of the nvidia bridge chip (no it isn't required for SLI).  Also it would be nice to have another south bridge chip, perhaps with a thunderbolt port.

will the SR-X board be better for gaming then the x79 classified? I am not sure if server processors offer something that gamers need over the normal processors except for possible more l3 cache? Would anyone be able to elaborate?

This isn't a board that will benefit gaming - single-processor Sandy Bridge E based boards have more than enough I/O to satisfy quad GPU solutions, so the presence of extra lanes on the SR-X won't really help.  The point of the SR-X is to enable a CPU monster - Sand Bridge EP processors can have a full 8 cores enabled, and overclocked with two sockets would offer fantastically high performance in a single system image.

I figured as much, so unless it had 4x PCI-E 16x lanes it wouldn't really matter.

Even then it wouldn't matter; current generation GPU's are not constrained by the bandwidth of a PCIE2x8; PCIE3 doubles bandwidth relative to PCIE2.  PCIE3x16 is unlikely to offer an performance difference over PCIE3x8.

n v o


This isn't a board that will benefit gaming - single-processor Sandy Bridge E based boards have more than enough I/O to satisfy quad GPU solutions, so the presence of extra lanes on the SR-X won't really help.  The point of the SR-X is to enable a CPU monster - Sand Bridge EP processors can have a full 8 cores enabled, and overclocked with two sockets would offer fantastically high performance in a single system image.

 
Some of these boards end up as Cuda crunching monsters with as many graphics cards as possible.  I/O does help with the overhead required to manage the processes running.
Please eliminate the bridge chip which will also help with heat and cost.  And as mentioned by others that platform is meant to be a processor monster; therefore, nobody should be running it single socket and if so I'm sure they wont miss the other half of the pci-e slots.

nivekt

farthestkris


evga should stick with either 8 dimms for both or 4 dimms for both, the board looks. disproportionate.

Function over form, I always say. Im sure there is a very good reason for the layout. Besides, with that layout and 8gb DIMMS, the possibility of 96gb of ram exists. Is that not enough?

Lets get wild and shoot for both

Striderstone

will the SR-X board be better for gaming then the x79 classified? I am not sure if server processors offer something that gamers need over the normal processors except for possible more l3 cache? Would anyone be able to elaborate?

1. Server is a use and not a function, meaning there is no such thing as a "server" processor versus a "normal" processor.  They're all the same in terms of the micro-architecture except the difference in the die design that set them apart.  Most of them are still following the x86 instruction set; thus install a Windows/Mac OS and any program that can be installed.  An example non-x86 CPUs would be Intel's Itanium series and the IBM CPU in an Xbox360, neither of which are DIY-user friendly.
2. As for gaming alone, understand that developers don't code games openly to take as many cores as available, there is a limit and right now that varies between 2-4 cores mainly because that's what most people own-- I'm talking about everyone.  That won't change until the very cheapest CPUs becomes quads, but their still dual-cores.  As you can see, unless you get a Core i7-3820 for X79 or a pair of Xeon dual-cores for either SR-2 or SR-X, have more uses.

Consider EVGA's marketing for SR-2:
 

We have literally created a new form factor to fit all the amazing things on one board. Whether you are an extreme gamer, overclocker, power user, workstation user, server admin, folder/cruncher, or just a PC enthusiast; this is the ultimate motherboard. This board will encode your movies, render your images, or even load your games faster than you ever thought possible

 
Gaming isn't the only aspect of appeal.

Has there been any educated estimates as to pricing, I'm going to guess it will be in the 600 dollar range (as I believe , but correct me if I'm wrong , that the SR-2 cost around that)
 
On another note I REALLY hope i can afford this board when It's released , it looks EPIC!
 
Martin

badass1982

Has there been any educated estimates as to pricing

Sort of, even if these were real, consider them to be 1,000 unit bulk MSRP (which is how Intel usually sells processors) and not individual retail.  $600 will limit you to the quads and low-freq 6-cores, you don't even have a shot at the only dual-core model for LGA2011.

I was referring to the price of the BOARD not the CPU's!!!

lehpron

Striderstone

will the SR-X board be better for gaming then the x79 classified? I am not sure if server processors offer something that gamers need over the normal processors except for possible more l3 cache? Would anyone be able to elaborate?

1. Server is a use and not a function, meaning there is no such thing as a "server" processor versus a "normal" processor.  They're all the same in terms of the micro-architecture except the difference in the die design that set them apart.  Most of them are still following the x86 instruction set; thus install a Windows/Mac OS and any program that can be installed.  An example non-x86 CPUs would be Intel's Itanium series and the IBM CPU in an Xbox360, neither of which are DIY-user friendly.
2. As for gaming alone, understand that developers don't code games openly to take as many cores as available, there is a limit and right now that varies between 2-4 cores mainly because that's what most people own-- I'm talking about everyone.  That won't change until the very cheapest CPUs becomes quads, but their still dual-cores.  As you can see, unless you get a Core i7-3820 for X79 or a pair of Xeon dual-cores for either SR-2 or SR-X, have more uses.

Consider for SR-2:

We have literally created a new form factor to fit all the amazing things on one board. Whether you are an extreme gamer, overclocker, power user, workstation user, server admin, folder/cruncher, or just a PC enthusiast; this is the ultimate motherboard. This board will encode your movies, render your images, or even load your games faster than you ever thought possible

Gaming isn't the only aspect of appeal.

Thank you for the information, sorry about wording my question improperly, I do know that making something a "server" is how you use it an what applications are put on it, it was just the simplest way of explaining myself. The information you provided was of great quality and I will put it into consideration

badass1982

I was referring to the price of the BOARD not the CPU's!!!

Calm down, mistakes happen.  Most folks are already interested in the board and ask about CPU cost once they find out it doesn't take i7's.  If the rumor about Asus competing with EVGA this time is true, the price could be lower, but non-enthusiast versions of boards using the same i5520 chipset average about $100 south of SR-2.  Tentatively, search of any rumors on the Intel chipset for their dual LGA2011 (C600 series) and when you find that board cost, add $100.

rottenmutt

Some of these boards end up as Cuda crunching monsters with as many graphics cards as possible.  I/O does help with the overhead required to manage the processes running.
Please eliminate the bridge chip which will also help with heat and cost.  And as mentioned by others that platform is meant to be a processor monster; therefore, nobody should be running it single socket and if so I'm sure they wont miss the other half of the pci-e slots.

I agree completely; what I meant was that, purely for gaming purposes, the extra bandwidth is of marginal benefit; however, there are other ways to use GPUs besides gaming.  For GPGPU, that extra bandwidth is a huge deal.  I agree on the point of the bridge chip.  Kill it NOW.  The Romley-EP platform has 80GBps (unidirectional) of PCIE3 bandwidth - bridge chips are wholly unnecessary given that a fully broken-out board would have 80 lanes to work with.  And unless you go with a true PCIE3 switch (PLX 87xx series is the only one I know of) you won't even gain burst bandwidth, which is the whole point of a PCI-E switch.
 
Honestly, anyone looking for an even bigger bandwidth monster for GPGPU should keep an eye out for Quanta (or someone who OEMs from them); their product roadmaps list a quad-socket LGA-2011 system due sometime in H1 2012; IIRC, it has something like eight PCIE3x16 slots.

Also @EVGA: will this board be compatible with LR-DIMMs? those would boost RAM capacities quite nicely, at the cost of some latency.

 
I am by no means an expert, so be kind to me if I have something completely wrong
 
I am looking at the Wiki page for these Sandy Bridge Server CPUs.
 

 
I see that the memory for some of the CPUs is at DDR3-1333 and some at DDR3-1600.  So, assuming the multipliers are all locked, is it safe to assume the DDR3-1333 Xeons are going to be better overclockers than DDR3-1600 Xeons?
 
If we also assume that the board will only be stable up to 2000 (like the SR-2), then
 
the top 8 core chip 2687W is 3.1 GHz at DDR3-1600, which is probably multiplier 19, so max overclock is 3.8 GHz (19 * 1600 = 3.04 GHz, Intel must be rounding up because 20*1600 is 3.2 GHz).  This chip is already 150W so it may not even overclock as well as this.
 
the top 6 core chip is 2667 but that uses DDR3-1600, so look at the 2640 which is 2.5GHz at DDR3-1333, which is also multiplier 19, which will overclock to 3.8 GHz.  This chip is 95W.
 
I know I'm assuming a lot, but is my math roughly correct?
 
Thanks

I had a link to the Wiki, but it didn't show.  Just search for "Xeon E5-2600 wiki" and it is the first result.

I'll be putting one of these in a Little Devil case. I'll probably use a couple of mid-range octo-core chips once the D stepping is out and upgrade to  high-end Ivy Bridge-EP when they come out.
Mine will be primarily for folding, I'm not sure what gpus I'll be using yet, I'm going to wait for folding benchmarks for the new ranges before I decide that, but there will be 4 of them and they will be the best available :D

they best put 16 slots in total, so 8 per cpu and 4 sata3 6gb/sec.
Than is it a good concurrent for the asus rog motherboard, if you look at the number of 6gb/sec drives you san use

If I was interested in the SR-X (not saying I'm not, heh heh), I would be less concerned about whether it had 12 or 16 DIMM slots and more concerned about whether it officially supported 8GB and 16GB DIMMs. A 16x16GB full complement of RAM would be 256GB. There might be someone out there who can use all that physical RAM, but I'm not sure who that is.

I ran into a memory issue on a machine with 96GB of RAM where the program I was using needed to unroll a for-loop and it hit 100GB of RAM usage and then ran out..