Hello,
I am running an SR-2 with 4 x GTX 780's.
 
Are there any issues with maximum PCI-E lanes ? I have read some i7 chips only support a limited number of lanes...
 
I am running with 2 x X5650 (Xeon)
 
Thanks !

Here is the manual:
https://www.evga.com/supp.../files/270-WS-W555.pdf

Page 33 says that it is designed to support up to four x16 or x8 graphics cards. Pretty vague actually regarding the lane limitations.

Here's another link with more information:
http://www.evga.com/suppo...ewfaq.aspx?faqid=59283

W555 – Classified SR-2
PCIe slot 1: 16x (8x if slot 2 is filled)
PCIe slot 2: 8x
PCIe slot 3: 16x (8x if slot 4 is filled)
PCIe slot 4: 8x
PCIe slot 5: 16x (8x if slot 6 is filled)
PCIe slot 6: 8x
PCIe slot 7: 16x

Hope this helps.

So, somewhere between 56 and 64 lanes max if I understand correctly. I think the max is 56 lanes and I think slot 7 will be 8x with four cards, but the documentation is a bit confusing.

If you already know, why are you asking?

It's only x16 on 1, 3, 5, and 7 in certain cirumstances.

It could be 8x, 8x, 8x, and 16x with four video cards in worst case. It could be 16x, 8x, 8x, and 16x. It could be all sorts of different combinations. You need to make your question much clearer if you want a better response. It depends on if your video cards are single-slot cards (liquid cooled?) and how many cards you plan to put in the other slots. If they are dual slot video cards and you aren't using other slots, and you already know the answer, why did you ask? Just to make fun of people who answered?

If you are only using 4 graphics cards in slots 1, 3, 5, and 7 and are using no other cards, all four cards will have x16 wide PCI-E 2.0 with those CPUs. But you already knew that. So, the point of asking?

@ty_ger07 You misunderstood ? I am not making fun ?! I didn't know the pci-e bandwidth stuff at all and I appreciate your help.
 
@cuda-dude Nice ! thanks ! what is 16x v2 for all the cards ?

Oh, I misunderstood. I thought you said you already read the manual front to back and already knew that the four cards would run at x16.

most ram slot issues of not detecting the ram end up being cpu not correctly seated or bent pins in the cpu socket. Can also be the ram itself though so it is easy to switch sticks around also.

@gordan79 by 'see the slot' do you mean CPU-Z gives data on it (or that the slot pull down is available)?
it's definitely not seeing all the ram on the SPD page, however the total is correct on the Memory page.
 
It might be possible some cpu pins are bend, I had some cables pushing against the cpu heatsink, might have screwed with it...
 
I'm running the oldest bios, 2010. It's been ok for me...

After running CPU-Z. I tried putting all the ram back in and it works !
So probably a loose ram or loose cpu thing.
Thank you for the help !

A GTX 780 (reference version) has TDP of 230W, so 4x230W=920W _just_ for the GPUs.
Xeon X5650s have a TDP of 95W each, 2x95=190W (assuming stock clocks and voltages, no overclocking).
Just between those two: 920W+190W=1110W.
The motherboard itself doesn't appear to have a total TDP for all the components isted, but given all the power conversion circuitry, ICH10 and a pair of Nvidia NF200 PCIe brodges (those can get quite hot), I'm not sure I would want to budget a _minimum_ of 100W for the motherboard itself.
So call that a 1210W _minimum_, without any margin for error, without overclocking, without disks, and without any other peripherals connected (e.g. via USB).
Then consider that you should always scale the PSU so that you never have to push it past 80% for optimal efficiency: 1210W/0.8, and you are already at 1512.5W
 
And that is without overclocking anything in the machine. Once you start adding disks and OC-ing, even relatively conservatively, it wouldn't be surprising to see the total according to above calculations hit 2000W.
 
The system in my signature is running with a CoolerMaster 1500W PSU, and I wouldn't want to try adding an extra pair of GPUs to it without getting a bigger PSU. Also note that decent PSUs over 1500W are quite difficult to find, as no good manufacturers make them (e.g. EVGA, CoolerMaster, PC Power and Cooling). And believe me, you don't want to have that much expensive hardware hanging off a 3rd rate PSU manufacturer (and there's precious little choice between the above named good PSU manufacturers and the 3rd rate ones you shouldn't be touching with a barge pole). For example, a friend's machine crossed my desk the other day that was a complete write-off after the ****ty Tagan PSU in it fried everything.

I'm enjoying my EVGA Supernova 1600 P2 in my rig.
 
If you are using multiple PSU, you split the load. 
 
Putting water pumps, non-boot drives and such on the second PSU.  This of course means you need to add a switch to start the auxiliary PSU before you start the main PSU/PC. 
 
Your wall outlet had better be up to the task - probably looking at a dedicated circuit buy then as 1600W will pull up to 13.3 Amps at 120V.

Thank you again, very interesting.
So, if I want to add, say 2 gtx 980ti to my setup (see below) for instance, I can add a second 1000w psu and it will be enough, right?
But I'm afraid of screwing up my system by playing the electronic apprentice, do you know any good l inks on how to perform this safely?
 

boumay
So, if I want to add, say 2 gtx 980ti to my setup (see below) for instance, I can add a second 1000w psu and it will be enough, right?
But I'm afraid of screwing up my system by playing the electronic apprentice, do you know any good l inks on how to perform this safely?

Google it, what you'll find is going to be no worse than what I can find using the same methods. :)
The rig in my sig is comfortably running using a single 1500W PSU. My recommendation is to just buy a bigger PSU, until you have no choice but to run two (unless it's for redundancy). EVGA do make some 1600W models.
 

Cool GTX

gordan79

Cool GTX
Putting water pumps, non-boot drives and such on the second PSU.  This of course means you need to add a switch to start the auxiliary PSU before you start the main PSU/PC. 

 
I would normally put the water pumps on the same circuit powering the device that is being cooled. That way you cannot get into a situation where one PSU fails and some hardware stays running without it's cooling still running.
 

Though I follow the logic in your statement; WP pull quite a bit of juice.  In a 2 CPU & 4 GPU setup, most would run two separate loops with 1 or 2 pumps per loop. At 24W (2A) per pump for a D5 pump it may be too much for the main PSU.
 
With software and hardware monitoring of CPU & GPU,  auto-shutdown is not only feasible but generally the best practices standard.
 
Again I agree if the main PSU has the capacity to support your pump(s) use it.

I'm more saying that you should group whatever pump is running the cooling for a device should be on the same loop as the PSU, i.e. have mobo+CPU pump run of the main PSU, and the GPUs and the pump for that loop run off the second loop. That way if a PSU fails, both the thermal load and the cooling for that load will fail at the same time, and thus nothing gets fried, even if the rest of the machine stays running. It's always a good idea to design things so that in the worst case they fail safe.
 
Particularly important since PSUs are the second most likely component to fail in a computer (after spinning rust hard disks, for at least some brands of hard disks).

As far as performance is concerned, you are wasting your time and money on running 4 GPUs anyway. The reason I have two in my machine because the 2nd one is passed to my wife's virtual gaming workstation. She games at 2560x1600, I use the Linux host and game a 3840x2400. Neither of us are feeling an overwhelming need for a faster GPU. I'm pondering getting a 980Ti, but only because I'm thinking about upgrading to a 5K monitor, and my 780Ti doesn't have dual DP outputs required to drive it.
 
If you are planning to use 4 GPUs for virtualization (quad-seat virtual workstation), I'd be inclined to under-clock and under-volt the CPUs and GPUs until it all fits comfortably into the power envelope of the biggest single PSU you can find.
 
Also, 2000W PSUs do, in fact, exist. For example:
https://www.overclockers....pply-bl-ca-031-sf.html

It's the only sanely priced 2000W PSU I am aware of, that is not to say I am making any statements as to it's quality. I have never owned a PSU of that brand, nor have I seen proper technical reviews of it.
 
That's not to say that there aren't plenty of reviews of it out there, what I mean to say is that nearly all reviews of PSUs are completely useless. They are based around somebody putting it in a machine and then feeling clever when they say "Hey, I put all this hardware on it and it didn't blow up and the machine didn't crash in the hour I spent 'testing' it." The bast majority of reviews are far from worth the electrons that were displaced to download the page. In fact, they are worse than useless because they imply something may or may not be a good product without any meaningful analysis having been performed on it.
 
If you are looking for PSU reviews, unless the load was generated by a proper, calibrated, industrial grade load tester and the voltages and ripples were tested with an oscilloscope at a good range of loads (say, every 10% between 0% and 110% of it's rated load), then the review contains no useful information at all; it's not a technical review, it's pure marketing.

PCI-E lanes were on the motherboard rather than integrated into the CPU. Socket 1366 is the last generation they didn't integrate the PCI-E lanes.
 
The 5520 chipset had 32 lanes of PCI-E 2.0
http://ark.intel.com/products/36783/Intel-5520-IO-Hub
 
Then you have 2 NF200 controllers that add another 32 lanes.
http://www.legitreviews.com/evga-classified-super-record-2-sr-2-motherboard-review_1437/2
 
You have 64 lanes in all.