lehpron

beaker7
Still don't get why the 2690 is more expensive than the 2687w.

TDP rating. Unlike enthusiasts, the multi-CPU community has a segment willing to pay more for a CPU that can operate at a lower TDP with the same performance as a higher wattage part; resulting in lower power bill and lower cooling bill, considering the typical Xeon customer purchases dozens to hundreds of these CPUs.

Just so you know, Asus has several variations of dual socket boards because they've been making multi-socket boards for both AMD and Intel for years.  There isn't just one just because that's all enthusiasts care about regarding SR-X. 

Halfdead14

EVGA_JacobF
Its 2x16 and 2x8

What the, since when did this happen? Why can't it do 4 way X16?
EDIT: Okay so I read the last few pages, why aren't you using lanes from the secondary CPU?

They are sort of, 40 lanes per CPU means that 2x16 + 2x 8 = 48 lanes, the other eight has to come from somewhere.  I'm sure the other lanes are going in other slots since we still don't know the configuration, JacobF is just stating that  four dual-slot cards will see dual x16 3.0 and dual x8 3.0.  I'd like to think there are single-slot configurations with more lanes, but I don't know.  We need the configuation to determine what is or isn't possible with one or two CPUs.

I'm sure EVGA made sacrifices just to enable Xeon overclocking in pairs; just as well if lanes matter more than overclocking, then I suggest get a regular dual 2011 by either Supermicro or these by Tyan or wait for Asus and Gigabyte to chime in. 

I have no idea where the extra 8 lanes come from, likely a NF200 chip or similar.... he did say however that the lanes are only pulled from the first CPU though.

Halfdead14

lehpron

beaker7
Still don't get why the 2690 is more expensive than the 2687w.

TDP rating. Unlike enthusiasts, the multi-CPU community has a segment willing to pay more for a CPU that can operate at a lower TDP with the same performance as a higher wattage part; resulting in lower power bill and lower cooling bill, considering the typical Xeon customer purchases dozens to hundreds of these CPUs.

Just so you know, Asus has several variations of dual socket boards because they've been making multi-socket boards for both AMD and Intel for years.  There isn't just one just because that's all enthusiasts care about regarding SR-X. 

Halfdead14

EVGA_JacobF
Its 2x16 and 2x8

What the, since when did this happen? Why can't it do 4 way X16?
EDIT: Okay so I read the last few pages, why aren't you using lanes from the secondary CPU?

They are sort of, 40 lanes per CPU means that 2x16 + 2x 8 = 48 lanes, the other eight has to come from somewhere.  I'm sure the other lanes are going in other slots since we still don't know the configuration, JacobF is just stating that  four dual-slot cards will see dual x16 3.0 and dual x8 3.0.  I'd like to think there are single-slot configurations with more lanes, but I don't know.  We need the configuation to determine what is or isn't possible with one or two CPUs.

I'm sure EVGA made sacrifices just to enable Xeon overclocking in pairs; just as well if lanes matter more than overclocking, then I suggest get a regular dual 2011 by either Supermicro or these by Tyan or wait for Asus and Gigabyte to chime in. 

I have no idea where the extra 8 lanes come from, likely a NF200 chip or similar.... he did say however that the lanes are only pulled from the first CPU though.

 
http://www.evga.com/articles/00668/
The EVGA SR-X uses the Intel® C606 chipset. This chipset has native SATA 6G and SAS.
Only Xeon® E5 Class Socket 2011 CPUs are compatible. It IS NOT compatible with Core i5/Core i7 CPUs. Please visit our CPU Support Page for details.
The EVGA SR-X motherboard supports PCI-E 3.0 performance via a PLX PCI-E 3.0 certified bridge chip. This allows the SR-X to populate all PCI-E 3.0 slots at the maximum bandwidth possible. The PCI-E lane configuration is as follows:

• 1 Card: x16
• 2 Cards: x16 - x16
• 3 Cards: x16 - x16 - x16
• 4 Cards: x16 - x16 - x8 - x8

nateman_doo

Finally they are for sale.

out of stock it seems.

Unfortunately the current CPUs have a locked multiplier so overclocking is limited to say the least.

geort45

LOL, I haven't been able to finish my SR-2 build (well it does work but is missing some cooling stuff and finishing the case) and the SR-X is out already! Daang!

Do the 2011 processors overclock the "same way" as the 1366 did?

Ill quote Jacob on this:
  
Does the board support OC'ing: Yes 
  
Does the board support OC'ing on current CPU's? The answer unfortunately is no. 
  
Current CPU's are multiplier locked, and they don't scale very high on bclock. 
 
http://forums.evga.com/fb.ashx?m=1485672
 
So yes the method is the same, Bclock adjustment, but the yield is not as high.

If Intel doesn't release unlocked CPUs, then I find it hard to buy this type of board. If only it could be overclocked like the SR-2, but they changed that >___>

Oh yeah, Would love to see a HydroCopper water block for the SR-X I know its "not needed" but it makes water cooled systems look beter imo.

No.. the SAS are sata 2.  3Gbs=usless

beaker7

Interesting that Origin is claiming ability to overclock Xeons using SR-X.

link?
 
If it's true they can, then it would be from the baseclock or whatever it's called with a max of 108 I believe.

beaker7

emach

beaker7

Interesting that Origin is claiming ability to overclock Xeons using SR-X.

link?

If it's true they can, then it would be from the baseclock or whatever it's called with a max of 108 I believe.

http://www.originpc.com/g...2-desktop-features.asp

If you go and build-to-order there is an option for overclock which force switches the mobo to SR-X.

I asked the live chat rep what kinds of clocks they were getting and was told that information will be released in a few weeks.

For the amount of money, why not make it a DIY project??? I could easily send 11,000.00+ through them and still not get a quad SLI SR-X

beaker7

ZachA


For the amount of money, why not make it a DIY project??? I could easily send 11,000.00+ through them and still not get a quad SLI SR-X

Of course.  Just pointing out that there must be at least a little to be gained via BCLK if they are offering it.

 
101 bclk wooooo

sokyoku

Thanks for replying jacob. Already read the entire SR-X main page. I have a question that may make you all laugh at me but Ive never used SAS connectors before..  From my understanding I can connect regular SATA devices to the SAS ports and they will work just not reversed.. which is fine. My question is If I connect all SSD's what is the individual bandwith each would recieve?  would it be the equivilant of SATA 3? or SATA 2?  

@ Lephron  the 48 lane thing  Im not certain but I beleieved they discussed this earlier something about the way PCI-E 3.0 works increased the amount of lanes to 1 cpu or something.. so the 48 is still running from 1 cpu.. while the second cpu takes care of all the other system things..   how ever I could be completly wrong and misread the previouse discussion.

 
You will need to get adapters that go from the SAS to SATA ports. Each SAS port can support 4 SATA ports. They are SATA 3G.
 
All PCI-E lanes come from the first CPU socket (right hand one) This is why you can run a single CPU and still have all PCI-E lanes.

So how many people are as bummed out as me?  I was looking forward to the specs specified at CES 2012..   4 way sli x16 on all ...   6g/s sas (had to research it)  I know its small things..   but they where very nice things..  I still plan on finishing a SR-X set up ..  just kinda let down a little i supose.

Jacob, does having all the PCIe slots from one CPU have anything to do with how nVidia restricts SLI certification on Sandy Bridge-EP motherboards? I can understand if having two CPUs supply the lanes for 4-way SLI needs too much crosstalk on the QPI links to make it SLI certified. Or was the choice an EVGA one?
 
As for the 3Gb/s SAS, if you look through the Intel literature, I believe this was a last minute change.
 
See here, a March 2012 Thermal Guide:
http://www.intel.com/cont...set-thermal-guide.html
Go to page 13 and you will see that the SAS was supposed to be 6G.
 
Then see the March 2012 Datasheet:
http://www.intel.com/cont...chipset-datasheet.html
Go to page 37 and the SAS is listed as 3Gb/s.
 
Another interesting thing to note is which generation of PCIe Intel uses for the SAS 4x uplink. Intel gives 3 different generations. On page 43 of the datasheet, they say the port can only operate at 5GT/s (gen 2) or 2.5GT/s (gen 1). Yet below that, they say the expected/supported configuration of the uplink is 4x gen 1. Gen 1 is emphasized on page 95 but then immediately contradicted when they say the expected/supported configuration of the uplink is 4x gen 3. In short, whoever wrote the Intel datasheet really wasn't paying attention. If they are using a link older than gen 3, it would explain why the SAS is only 3Gb/s.

SiriusDragon

I'm also curious about the PCI-E lanes too... I assume they did it for 4-way SLI compatibility...

But they could have powered the secondary slots by the secondary CPU, allowing for extra cards like RAID cards, PCI-E SSDs etc without affecting the GPUs.

While being able to run the board with 1 CPU is nice, it's hardly a selling point.

Assuming unlocked CPUs come out it'll still be a nice board, but the gimping of some of the features makes it slightly less appealing than what it was, it's probably a shame for EVGA too as I guess they were hoping for this to be an epic flagship board.

I don't mind about the RAM slots... I think that's one of the few complaints about the board I don't share. I'd rather have better VRMs and stuff than more RAM capacity.

 
Actually, looking at ASUS board's manual, they have a 16x/8x, 8x, 16x/8x, 8x, 16x, 8x, 16x setup by which the first four slots are powered by the primary CPU and the last three slots are powered by the secondary CPU. Also, they say the board can run 4-way SLI, by this setup, 4-way SLI all at 16x. Now whether having two CPU power the two sets of GPUs causes more or less performance than using 32 lanes from a single CPU as EVGA did (of course expanded by the use of a PLX chip, but those cards are still getting squeezed down a 16x lane) is up for debate. Then again, I think most of the crosstalk between GPUs would be done over the SLI bridge and not the PCIe interface. That would be mostly for communications with the CPU.
 
Anyway, if I were to redesign the SR-X (or a SR-X2), I would do just a few things differently:
1. Use a 16x, 4x, 16x, 4x, 16x, 4x, 16x setup with the first 3 slots coming from the primary CPU and the last 4 slots coming from the secondary CPU. Of course, all could use 16x physical slots. The reason for this is purely from a gamer's perspective, to maximize my bandwidth between my GPUs and the CPUs. As a CUDA rig, 10 8x slots (16x physical) would be more appropriate. If I populate the 4x slots in this scheme, my 16x slots won't suffer.
2. Use the C608 chipset, if for anything, RST3 SAS RAID 5 Support.
3. Perhaps use an Intel X540-AT2 10Gb/s Ethernet Controller. This would highly depend on what the projected adoption rate of 10Gb/s Ethernet was, but considering a Sandy Bridge-EP board from both Tyan ans Supermicro are using it, the adoption rate seems to be improving. Of course I would use an RJ-45 plug and not anything fancy like SFP+. Since the controller uses 8 PCIe 2.0 links, I would let it take all the PCH's links.
4. I would use a TI TSB83AA23 PCI-Firewire chip since C600 series support PCI and the PCIe-Firewire chips are really PCIe-PCI-Firewire chips. Also, I would place 1 Firewire 800Mb/s port on the rear.
5. Then I would connect 4 NEC µPD720201 USB 3.0 controllers to the secondary CPU's 4x PCIe 2.0 link (normally used to connect to the PCH), each one powering 1 rear USB 3.0 port.
6. Perhaps I would move some SATA 3Gb/s ports (native from the chipset) to the rear for eSATA, but I'm not a big fan of external hard drives to begin with.

will the sas ports support raid 5 each so i could create two logical drives out of eight physical drives in raid 5? I know its only sata ii but its my understandig that non ssd's dont saturate sata ii anways?