Russian460
Ok, thanks for the info gordan79! Just to check then, I should be able to run two of the ct8g3ersld4160b sticks to start, right?

Yes, that will work fine. Be aware that with just 2 DIMMs you will only have 1/3 of the possible memory bandwidth since you will only be using one of the 3 memory channels on each CPU.
 

Russian460
I will PM you to exchange contact info once the forums allow me to do that.

spilart
can i send you an MP for bios information please ?

No need to PM - if you have any SR-2 related questions ask here and I'll try to help. The info may be useful to others.

spilart
OK, thanks gordan
i want to get seetings bios to turn Stable my x5690 on air at 4.0 or 4.2 Ghz
 

Tis ar my bios settings for 4.6Ghz,If not mentiond its on auto or on stock
 
CPU
 
C1E : disabled
MAX cpuid : disabled
Intel VR : enabled
Intel VT-D : enabled
Corency support : disabled
Execute-disable BIT : enabled
Active core : ALL
A20M : disabled
Speed step : disabled
C-state tech : disabled
Acpi T state : disabled
CPU : 200
PCIE : 100
CPU multi : 23
QPI link : full-speed
QPI frequency :
Memory frequency : DDR-1333
MCH strap : auto
CPU uncore : auto
 
Voltages
 
Vdroop control : whitout vdroop
Boot Vcore : 1.41875v
Eventual Vcore : 1.41875v
Boot cpu0 VTT : 1.35v
Eventual VTT : 1.35v
Dimm voltage : 1.65v
 
Signal tweaks
QPI 0 : -85
QPI 1 : -16
 
max fsb for bord and cpu for the moment 220Mhz " VTT @ 1.457V "
http://hwbot.org/submission/2709740_skulstation_reference_clock_classified_sr_2_219.96_mhz
max cpu speed for the moment +5.3 GHz
http://hwbot.org/submission/2710396_skulstation_cpu_frequency_xeon_x5690_5318.52_mhz
 
for more +5G benches go to http://hwbot.org/user/skulstation/
 

What ar your temps?
Whit VTT @ 1.35V you can try to run 200
Whit my cpu 4941.60Mhz 27 * 183 @ 1.646V i hit 80c° while running cinebench r15.
Whit the cpu 4823.79 24 * 200.99 @ 1.541V i hit 71c° while running cinebench r11.5.
While running  super pi 32M whit cpu 4999.73 25 * 200 @ 1.588V i hit 54c°.
so if your temps letting you and you ar willing to run more Vcore ther is alot room to play whit.

whit the settings thad i postet i can run 4.6Ghz on air close to 85c° while benching.
i think you gone need more than 1.30Vcore to go above 4.2Ghz. and 1.4Vcore for +-4.5ghz or more.
ceep in mind if you fry your cpu i am not responsibal :)
but my cpu is more than 3 years old an still running like day 1
 
 

Pushing past 1.35V core is, IMO, a waste of time, and I would be very surprised if you are running properly stable at above 180 bclk. I'm running 4GHz at 1.35V (70C max T-junction using HR-22 air coolers).
 
QPI doesn't OC well at all, and minimum QPI setting at 177MHz hits the maximum everything is rated for. Beyond that everything will begin to fall apart pretty quickly. Beyond about 166 bclk you will need to drop uncore multiplier from 20x. You might get away with 19x for 177, or you might need to drop it to 18x.
 
If you think you are stable much beyond 4GHz or 180 bclk it is almost certainly because you aren't testing particularly strenuously. My stability measure is 24 hours of cpuminer --benchmark and 24 hours of OCCT. You will notice that at around 180 bclk the clock speeds start to spike all over the place. That is a certain sign of instability onset. At that point even if OCCT and cpuminer aren't detecting errors you'll notice that the performance actually drops, so pushing that far is counter productive.

Stable means stable under long term harshest software load that can be produced. What you are describing is essentially benchmark stable. You also don't know whether you are getting errors in your data as a conequence of OC-ing. It may not matter for gaming and benchmarks, but it is almost certain you are getting audio distortions in your video above 180 bclk. The fact that you don't care doesn't make it stable.

I never said 4.2 is stably achievable. Then again, I never tried watercooling, thermo-electrics or any other kind of super cooling. I'm running 4GHz on air with HR-22 air coolers. Those extra 5% hardly seem worth pursuing.

The difference between a good heatsink (e.g. Noctua with decent fans) and the HR-22s that I'm using is about 30C (I bull**** you not). Mine were hitting similar temperatures with a more conventional cooler and they weren't stable at 4.0GHz (by my definition of stability at least). With the HR-22s I cannot get them to get them past 70C even using cpuminer, which is much, much harsher than cinebench. I transcoded most of my DVD collection into H264 so I can Plex it to my Chromecasts in DirectPlay mode in a massively parallel job that kept all CPU cores saturated with 0% idle time. It took best part of a month! The CPUs never touched more than 65C under that load with the HR-22s.
 
Look here:
http://ark.intel.com/products/52576/Intel-Xeon-Processor-X5690-12M-Cache-3_46-GHz-6_40-GTs-Intel-QPI
Intel spec says the Tcase should not exceed 78.5C
 
The temperatures various sensors will be showing are Tjunction rather than Tcase. Tjunction can be up to 15C higher. I try to make sure that the Tjunction never exceeds the spec for Tcase, just to make sure I have a margin for error. If you want to follow the same rule of thumb, you need to drop your CPU temperatures by about 20C, which means monster heatsinks or water cooling.

correction:
GPU pending: 2x970(80) SLI?

:)
delarisch suggested SR2-Wndws10 installing may be "problematical", can't imagine why, SSD pending.
Lack of need for ECP v3 understood, yes I'm expecting to review posts as to 5670 clocks, on air especially.
Hoping to stay very cool regarding ambient temp error margins.
Was my understanding SR2 clocking required non-ECC RAM, obviously a misconception.
96GB ECC with 4GHz CPU was not originally an option to consider, and on air! Really that's a marvel!
More pictures of the under-stuffed Elysium(s) would be helpful.
The only way I'd get into real liquid cooling would be an "inexpensive" used rig.
How about Peltier cooling on the chipset and RAM?
Can I wire up an HPTX-sized Peltier mat for the backside of the board or what?
Thanks for listening.
If I was Intel I would never give my Tesla-GPU adversary another opportunity to clock the Xeon.
 
 
 

Gordon ty,
You are so very much ahead of me, my clocking exp. is minimal, very modest on some i7s.
It's good you post it for the others who may find it useful, but QPI I know it between the CPU's, little more.
I have not seen SR2 BIOS yet, I'm not testing any limits as the boards are becoming irreplaceable.
5690 is for testing limits, I chose 5670 95w very cool and very discounted, defray cost second SR2, backup. 
Your earlier reference to 5670 parameters will become relevant to me after I decide on a SSD and GPUs.
I'll look Thermalright HR-22 but likely use Corsair H70 CPU water-cooler I'm ready with, and installed before.
When A2 board in hand be buying maybe pair of EVGA or NVidia 970 GPU cards, suggestions or comments?
Various Linux OS possibilities, may try Hackingtosh adventure, but Wndws10 must fit to SSD, no problem?
And how many CUDA cores does anybody want to buy me? NVM that's just talk. 
Thinking I'm nearly ready to unbox Xigmatek Elysium, like those others using.
  
Considering the attention to detail which would go into producing an EVGA SR2,
Considering the attention EVGA would likely pay to it's prospective enthusiast users,
Useful or not, it seems remarkably inconsiderate to have not included ECP v3 detail in all cases.
I understand warranty expired on both my "new" SR2 examples.
None the less "Thanks EVGA for treating your customers like idiots!"
Ha, that's a good one, ha! I should look at my box to see if it's listed or not.
 
Gordon? On Amazon is an SR2 in a Cooler-Master Cosmos with 5670 and water cooling, Oregon US,
and auction due to expire shortly for third time. I could pick it up but don't need the three monitors, don't need the very deluxe Radeon cards, I let it go a third time now. Bought second backup board instead.
Would you care to discuss pros and cons of that configuration?
Bye now, GTG
   
 
 

Looked at my A1 box, it lists ECP v3 included, it better be there.
Looked at few Thermalright HR22 reviews, I'd be afraid to hang it off my board in the Elysium case.
Found pic of top mounted small radiators in Elysium case.
No need trouble over that Cosmos build, it was sold last week but he relisted it again, idk why.
No response necessary, thanks for the info. 

Have had SR2 A1 new since 9/2010.
Have acquired since Elysium case, 48(4x(3x4)GB Corsair Vengeance 1600, 2xCorsair H70 CPU liquid cooler,
and last month 2xXeon 5670, next week backup board arrives SR2 A2 "new".
Next maybe Mushkin Reactor 1Tb SSD SATA3 , PSU opts. include 2x1000w PC Pwr. & Cooling, 950w also.
As one H70 cooler used, cooling & power opts. are not final.
Availability of 1366 CPU's and lack of availability of SR2s decided now or never to build.
What to do with the 7xPCIe options? Consider the expense.
Compare: (not Tesla K80  4992 cores 24Gb $5000, it's not a graphics card) (I'm not made of money)
1a) 2x EVGA GTX 980 ACX 2x1024 cores 2x4Gb $1200 (which 4 slots?), +physX EVGA 460 or better 
1b) 3x EVGA GTX 980 & etc.
2a) 2x PNY Quadro K4200 2x1344 cores 2x4Gb $1550+ 2x(single-slot), + physX EVGA 460 or better 
2b) 3x PNY
2c) 4x PNY Quadro K4200 4x1344 cores 4x4Gb $3100+ 4x(single-slot), ...
3a) unreleased 2xTesla K8 2x1536 cores 2x8Gb 2x? 2x(single-slot), ... 
3b) unreleased 3xTesla K8 3x1536 cores 3x8Gb 2x? 3x(single-slot), ...
3c) ...
4) (not an option), visit Colfax (Sunnyvale, Ca.) use configurator to max. rack-mount, ~$80,000
Summary: SR2 A2 5670 Xeon +clock,
1) 2x EVGA 970? (4 slots plus physX)
2a) 2x EVGA 980, physX 460+ (4 slots plus physX) 
...
2c) 2x EVGA 980, physX Tesla K8 (4 slots plus physX)
3a) 2x PNY K4200, physX ... (2 slots plus physX)
...
3c) 4x PNY K4200, physX ... (4 slots plus physX)
4) unreleased Tesla K8 ... 
!:) hurry up and wait.
 
If using ?SLI 4x single-slot PNY K4200  and single slot Tesla K8 physX, maybe Creative SB Recon, & etc.
how should I slot the cards please?
SR2 maybe not be big bad wolf-fest but even at stock setts it should be a Super Record.  
    

By the way also please,
 Are you trying to tell me that of two new EVGA SR2 I should expect a better than vanishing chance of less than perfection? "Thanks a lot EVGA for treating your customers like ..."
 Really it is astounding how in both hardware and software consumer electronics these days, in their incessant
rush to market, the monopolists force consumers to do their product testing for them. I would not be surprised if PNY went the way of BFX the day I bought my fourth card. If you want to understand who builds computers in America I'd suggest "The Computer" by Goldstein. The monopolies exist in spite of anti-trust regulation for the good of "The People". Everybody else gets scraps. It's just the nature of the beast. People keep the best for themselves and get what they can for the rest, "Money" by Galbraith.
 So what exactly are the odds 2 out of 2 $600 "enthusiast" EVGA motherboards are dysfunctional as they come out of the anti-static wrap? Bring on the data miners.  Who is the (:p)' super-computer we should ask? I'd bet the Bureau of Standards uses a top-notch Intel-NVidia-Oracle  and could figure it out if there was some incentive. Further insight into this dilemma is to be had from "Information Technology Standardization" by Carl Cargill. You may rest assured it is conspiratorial and has been since AT&T gave rise to Lucent. What ever did become of Penzias and the cosmic creationist point of view? Last I recall he was attending Oracle's convention with Ellison. Lucent, becoming the most widely held stock corporation in only ten years, has been very quite of late, maybe because they were treating their customers like idiots and/or Hubble's telescope  was more than a little nearsighted.
 Anybody be interested in a signed copy of Mathmatica3? Nah, sorry, silly question. How about cellular automata and "A New Kind of Science." 'Could be Wolfram Research can give us the answer, but for how many x86 cores. Maybe I should just ask my new jolly green giant of a friend (no not Cargill De-icing), that "Jade" and "Emerald" crusher of computational trivia, EVGA-NVidia-Intel-Radeon. 
(Hint: Richard Feynman pointed out if you launch 24 rockets and the last explodes, chances of the next launch exploding would be about 1 in 24, not one in a million billion according to Ed Regis, author of "Who Got Einstein's Office".
 Pass the scraps and rage on, I have my "Road to Reality" by Roger "the beautiful" Penrose,
'nuf said
  
     
   

OOPS! misquoted, very sorry, "Thanks EVGA for treating us like ..." maybe better