SR-2 Optimization Thread

[chapel]

I want to pass along a good deal on Intel hexcore chips.
X5660 socket 1366 3067mhz turbo in Linux makes bigadv at stock speed.
http://www.ebay.com/itm/151163971481?_trksid=p2055119.m1438.l2649&ssPageName=STRK:MEBIDX:IT

I have an Asus board that I had to park when Stanford cut return times, then I found these.
I bought 2 for 400$ I offered 175$ but 200 is as low as he will go.
Shipped fast and work real well. 170Kish PPD

If this is not a good place to post this, sorry.
Move or delete if needed.
200 for a good hexcore 1366 s great, they were 500 just a couple months ago.

guess I'm asking too much for my X5647...

 

[Bill1024]

Upcoming changes to bigadv threshold

Moderators: Site Moderators, PandeGroup
1 post • Page 1 of 1
Upcoming changes to bigadv threshold

Postby kasson » Today, 10:43 am
We have a policy of periodically re-evaluating the bigadv program, including the threshold required to run bigadv projects.
It is the intent of bigadv to match large and resource-intensive work units with some of the most powerful machines used by FAH donors. This "most powerful" line naturally advances with computing power. To date, bigadv has been a CPU-based program, and with increasing numbers of CPU cores and power of those cores, we have decided to lay out a roadmap of bigadv threshold changes for the next several months.

Feb 17 (two months from today): bigadv threshold will become 24 cores
Apr 17 (four months from today): bigadv threshold will become 32 cores

We want to give advance notice of these changes, and we recognize that change is not always welcome or comfortable. We should also emphasize that the science performed by donor machines is valuable in all parts of the FAH project, and part of the change in bigadv threshold is because we would like to encourage moderately powerful machines to help boost the capabilities of non-bigadv SMP projects where we do a lot of this science.

We also recognize that core count is not the most robust metric of machine capability, but given our current infrastructure it is the most straightforward surrogate to evaluate.

Thank you once again for your generous participation in the Folding@Home project!

 

[dpoverlord]

Hey guys sorry for opening up this old thread but was wondering if anyone could help me.

I just bought a Xeon X5660 from a i7-930 and was under the impression it would perform better. Right off the bat I am getting a higher overclock (4,5GHZ vs 4.2ghz on the 930), but it does not seem to be performing better.

Then moved the multiplier to X23 and got to 4.5ghz.

The problem is bench wise I do not see a difference with this chip to the 930.

Any idea what settings I have that are wrong?

I mistakenly posted in another Xeon thread thinking I made my own :-(

I just installed a Xeon 5660 and have it running at my customized i7-930 settings (Had a golden chip did about 4.4ghz). I bough it after reading the OCN thread about the Xeons killing the 930's

http://www.overclock.net/t/1461359/...5639-benchmarks-inside-longest-post-ever/0_70

Really would love some help on getting the most out of this chip.

Off the bat, the good news is that Windows loaded fine with these. Bad news is I have the Clock set to 21. Knowing this I should change it to 23. Can you guys look at my settings to give me your opinion?

These are my settings right now:

2. > Advanced Frequency Settings [Press Enter]
CPU Clock Ratio............................:21
CPU Frequency ........................(GHz):4.03GHZ 192x21
> Advanced CPU Core Features [Press Enter]
Intel(R) Turbo Boost Tech..................:Disabled <<<<yes this is right
SHOULD THIS BE DISABLED?
CPU Cores Enabled..........................: ALL
CPU Multi Threading........................: ENABLED
CPU Enhanced Halt (C1E)....................:DISABLED just while your benching for a stable over clock, when stable this can be Enabled

C3/C6/C7 State Support.....................:Disabled this always stays disabled, even after a stable o/c or even if your running stock

CPU Thermal Monitor........................:Auto <<<<change to Enabled
CPU EIST Function..........................:DISABLED just while your benching for a stable over clock, when stable this can be Enabled

Bi-Directional PROCHOT.....................:Enabled
QPI Clock Ratio............................: X36
QPI Link Speed........................(GHz): 6.91GHZ
Uncore Clock Ratio.........................:X13
Uncore Frequency......................(MHz): 2496Mhz
>>>>> Standard Clock Control
Base Clock(BCLK) Control...................:Enabled
BCLK Frequency(MHz)........................:192 <<<<change 191
Extreme Memory Profile(X.M.P.).............:Disabled
System Memory Multiplier (SPD).............:6
SHOULD THIS BE 8?
Memory Frequency(MHz)......................:1066 1152
PCI Express Frequency(MHz).................Auto: <<<<change 100 this can be left to auto but some use the 100mhz as that's what auto will run @ I have mine set to 101mhz as its more stable on my o/c GPU

>>>>> Advanced Clock Control
CPU Clock Drive (mV).......................:800mV
PCI Express Clock Drive (mV)...............:900mV
CPU Clock Skew (ps)........................:0
IOH Clock Skew (ps)........................:0
> Advanced Memory Settings [Press Enter]
Extreme Memory Profile(X.M.P.).............:Disabled
System Memory Multiplier (SPD).............:6
SHOULD THIS BE 8?
Memory Frequency(MHz)......................:1066. 1152
Performance Enhance........................:Standard <<<< this is fine
SHOULD I HAVE THIS SET FOR TURBO OR EXTREME?!
DRAM Timing Selectable (SPD)...............:Quick <<<<this is fine
Should this be AUTO
Profile DDR Voltage........................:1.5V
Profile QPI Voltage........................:1.175V
Channel Interleaving.......................: 6 AUTO
Rank Interleaving..........................: 4 AUTO
>>>>> Channel A, B, & C
> Channel A, B, & C Timing Settings [Press Enter]
>>>>> Channel A, B, & C Standard Timing Control
Cas Latency Time...........................:7 AUTO
tRCD.......................................:7 AUTO
tRP........................................:7 AUTO
tRAS.......................................:20 AUTO
>>>>> Channel A, B, & C Advanced Timing Control
tRC........................................: 27 AUTO
tRRD ......................................:4 AUTO
tWTR ......................................:4 AUTO
tWR........................................:8 AUTO
tWTP.......................................:19 AUTO
tWL........................................:7 AUTO
tRFC.......................................:60 AUTO
tRTP.......................................:4 AUTO
tFAW.......................................:16 AUTO
Command Rate (CMD) ........................:1 AUTO
>>>>> Channel A, B, & C Misc Timing Control
B2B CAS Delay..............................: - AUTO
Round Trip Latency.........................:58 AUTO
> Channel A, B, & C Turnaround Setting [Press Enter]
>>>>> Channel A, B, & C Reads Followed By Reads
Different DIMMs............................:6 AUTO
Different Ranks............................:5 AUTO
On The Same Rank...........................:1 AUTO
>>>>> Channel A, B, & C Writes Followed By Writes
Different DIMMs............................:6 AUTO
Different Ranks............................:6 AUTO
On The Same Rank...........................:1 AUTO
> Advanced Voltage Settings [Press Enter]
>>> CPU
LoadLine Calibration.......................: AUTO <<<<change to Level 2
CPU Vcore..................................:1.31250V
xDynamic Vcore(DVID).......................: 0.0 AUTO
QPI/VTT Voltage............................:1.335V
CPU PLL....................................:1.8V
>>> MCH/ICH
PCIE.......................................:AUTO
QPI PLL....................................:1.200V
IOH Core...................................:AUTO
ICH I/O....................................:AUTO
ICH Core...................................:1.2V
>>> Dram
DRAM Voltage...............................:1.5V
DRAM Termination...........................:AUTO
Ch-A Data VRef.............................:AUTO
Ch-B Data VRef.............................:AUTO
Ch-C Data VRef.............................:AUTO
Ch-A Address VRef..........................:AUTO
Ch-B Address VRef..........................:AUTO
Ch-C Address VRef..........................:AUTO

> Miscellaneous Settings [Press Enter]
Isochronous Support........................: Enabled
Virtualization Technology..................: Enabled
Edited by kev8792 - 8/1/13 at 7:35am

 

[Deleted member 82943]

is this watercooled?

 

[tived]

I think you just need to disable to toggling, just can't remember how

all the best

Henrik

 

[W.Feather]

This has absolutely nothing to do with the EVGA SR-2, please make a thread for your question in the Overclocking and Cooling section where it would be more relevant.

 

[Tiberizzle]

Is 1.456V set VCore (1.43-ish measured) with Vdroop off too high for 24/7 on x5650s? Is there a overvoltage breakdown or extremely accelerated degradation due to VCore at some point for the 32nm chips, or is the main concern thermals? Has anyone run with high or very high VCore on these chips for long enough to experience the failure modes or anecdotally validate voltages >1.45V (or >1.4V for that matter)?

I get 71C max load temp water cooled at 200BCLK x20 with turbo on with two x5650s in my SR-2, 4.4 all core and 4.6 4-core both seem totally stable, and most cores 5-10 degrees lower than the hottest, but it's not long term stable with any less VCore than the 1.456V set point. I am thinking the temperatures are pretty good so it doesn't really matter, VTT within the maximum at 1.4V and most other voltages are at default.

If the 1.45-1.55V range is relatively usable with its use being advised against due to practically solvable thermal issues rather than instant death socket slagging fireworks issues, I might have to order another radiator and some higher multiplier chips

edit: honestly I'd settle for instant chip death at excessive voltage, as long as it died quietly without harming anything else these Xeons are getting cheap...

 

[Deleted member 82943]

1.465v seems way high, though I am on air and anything close to 1.3v sends my chips to 80C+

 

[=Gunslinger=]

@ Tiberizzle,

Which are the key voltages you've set to hold 200 bclock stability?

I also agree 1.46 vcore seems pretty high for 24/7 use.

 

[Tiberizzle]

Vtt 1.4, uncore multiplier 16x, and memory settings (in my case DDR-800 multiplier, DDR-1600 MCH strap auto defaults, 1.65V) seemed the most relevant to high BCLK stability

One of the CPUs has 3x 8GB DIMMs and I am pretty sure they are responsible for the uncore / memory setting sensitivity, at higher memory settings some or all of the 8GB DIMMs fail to detect quite frequently

Vcore 1.456 was necessary to get it working at 200 BCLK turbo, it was stable at <=1.4V Vcore without turbo (at 185-200 BCLK 20x multi)

IOH and ICH core voltages may have helped clean up some kernel log noise and fruity USB behavior

 

[Deleted member 82943]

i mean your temps seem fine but the voltage just seems way high for long term 24/7 use

 

[=Gunslinger=]

Thanks for the tips, will try them out this weekend.

 

[geok1ng]

Here is situation:

2x E5606
KVR1333D3Q8R9S/8G

I have 12 sticks of this 8GB Registered quad ranked ECC.

Booting 48GB in triple channel OK. Works like a charm with 1600MHz strap.

when going with 96Gb / 12 sticks, same old problem- cold boot after clear CMOS shows 96 GB, no boot after entering BIOS.

There are enough reports of successful 96 GB on the SR-2 for me to believe that it may be possible. But each and every report i found lacks a complete description of BIOS settings, so i ask fellows [H]arders for:

CPU you are using to boot 96GB- most reports come from high end CPUs with DDR 1333 support- i have a couple of X5675s coming, so i would save a lot of effort if no one proved a boot with DDR 1066 CPUs.

memory you are using with PN. Registered ECC is the most common memory reported at 96GB, my PN was based on 2 reports. But turns out that quad ranked is the least used memory type at 96GB. single and dual rank are more common.

each and every setting in BIOS I have tried loosening the memory timings, using higher strap, increasing vDIMM and vTT, among other things.

One can easily lose many days trying to find the secret sauce for 96GB on this mobo, so any help is greatly appreciated.

 

[W.Feather]

To my knowledge the only way for 96gb to work is by using ECC memory (Which im too cheap to buy 96 gb of ), and to sacrifice your first born ..... Most people that have 96gb dont have a successful boot every time (From my memory atleast)....

 

[Deleted member 82943]

false, for me at least. I had 96GB of DDR3 1866 non ECC running on the latest bios of my SR2 and two x5660s, and booted fine every time.

 

[geok1ng]

XS SR-2 thread has this:

I am crazy? is it running at stock 1333 DDR3 timings and voltages with non-registered non-ecc RAM

update, screenshot proves nothing, as one can boot into windows using F2 after a CMOS reset, also the PN of that memory is for registered ECC dual ranked DDR3, just like mine.

 

[geok1ng]

Yesterday i received a couple of X5675s to finally replace the E5606s. Booted OK, but even after loading defaults the PS2 keyboard behaves erratically on BIOS screen. It passed a few hours of prime95 without errors, so i am at a loss with this odd behaviour, since the keyboard works perfectly inside windows.

This may be covered somewhere else here, but i did not experienced problems using a PSU with a single 8pin EPS cable and going just with a 6pin PCIE to power the second CPU socket. It will be another month before i receive my adapter, should i wait or proceed with OC tweaking?

 

[W.Feather]

With how finicky the SR-2 is, I personally would not look to overclock the board without full power going to the board.

 

[geok1ng]

Update on my build: fro some weird reason the mobo will not boot with both 8-pin EPS connectors plugged, but i managed to boot with 2x 6pin-PCIe + 1x8-pin EPS.

The new X5675s are weird: one CPU socket idles at 60ºC and the other at 40ºC. I was expecting at most 12ºC gap between CPUs, and this is using Indigo Xtreme TIM. Is that normal?

 

[402blownstroker]

One socket is know to run hotter than the other, but I do not remember which one right off hand. I think I had a diff of 6C to 7C. The temps at idel for the one at 60C seems very wrong. Most likely the heatsink is not seated correctly.

 

[geok1ng]

I tried to increase mounting pressure on the Kraken x40, which with the crappy plastic backplate meant inserting metal plates between the motherboard and backplate. it was no use. i tried mnually press the block over the CPu and watch temps, again no result.

So i am back to mounting bench, removing Indigo Xtreme TIM, which may not be compatible with the Kraken X40. Indigo site only mentions the Corsair h50 as incompatible with a vague reference to :

Heat sinks/waterblocks with no spring loading such as the Corsair h50:
Spring clamping mechanisms (compression, clip springs, etc) are required to maintain pressure (on Indigo Xtreme) as bondline thins out during reflow.

Which is a shame because the other CPU with a Nepton 280L is idling less than 6ºC above ambient.

update, problem solved, replaced Indigo with plain thermal paste and load temps 65ºC.

here is what happened

Turns out Krakens do not apply enough pressure to force the metal to spread during reflow. The Nepton works perfectly.

 

[geok1ng]

After much grieve i am finally able to make Turbo bisn work on the SR-2 , but the criteria for turbo to actually kick in still elude me.

my X5675s turbo to 26x single core; 25x all cores when i set VTT to 3.5/3.75 ( bootup/eventual) and Vcore to 1.306/1,31 and CPU multi to 15x.

But i lose my turbo if i put more juice on VTT or Vcore or lower the multi to 14/13/12x.

Lowering the multi is desirable to decrease idle power usage, and upping the voltage is desirable to try for higher bCLKs.

performance is better at 25x 177 with TURBO than at 23x195 without TURBO, and heat generated in the undercloked template is much lower.

This is becoming an e-peen problem because i am less than 3% from the top world scores on 2 chess benchmarks

A second issue is that the systems boots and benchs OK with C-state T-state enabled, but random crashes at idle. What do you use for sucessfull turbo + c-state?

 

[fiscap]

I'm having an issue with fully populating all (12) DIMM slots with 48GB of G.SKILL DD3-1333 4GBx12 (CL9-9-9-24) memory modules. I've run with 24GB of (4GBx6) for the last year without issue and thought that doubling the RAM with the exact model modules would be a no-brainer, but I can't POST with all 12 slots populated.

I've tried a little bit of troubleshooting to confirm the new modules are good by replacing the original 6 modules with the new ones, running memtest86, and running for a week inside the OS - no issues at all. I've also been able to run without issue with both 8 and 10 slots populated, running with 32GB and 40GB respectively. However with all 12 slots populated, no POST and the cycling of 69 00 E8 02 error codes.

I've got the system running extremely stable, so I don't want to mess around with the OC settings too much. Any suggestions on what I should attempt first to successfully run with all (12) banks populated would be appreciated.

To provide some additional information, here are the specs on what I'm running:

Dual Xeon X5650 - OC'ed to 4.0GHz
(12) G.SKILL DD3-1333 PC3-10666 4GB modules (F3-10666CL9T2-24GBR - two kits)
Single AMD Radeon HD 6970
H2O cooling on CPU's only

I've primarily used may_i_be_worthy's post as an overclocking guide - http://hardforum.com/showpost.php?p=1036223866&postcount=142

Here are some screenshots for my current settings:

 

[geok1ng]

back off you overclock, select 2:6 memory divider= ddr3 800 and try 1.4V VTT.

if all else fails, try 6x8GB ECC RAM, like me.

 

[geok1ng]

And i urge you to consider going turbo, as your CPU can do 22x on all cores

 

[geok1ng]

Mission acomplished, the SR-2 once again owns both records on chess benchmarks:

 

[fiscap]

Thanks for the suggestions geok1ng. A couple of quick questions, where in the BIOS does it allow you to set the memory divider, so that I can try switching from 2:8 to 2:6 as you suggest? Also, any thoughts on why my config would work just fine OC'ed with 24GB, 32GB and 40GB, but choke on all 12 slots populated to 48GB?

 

[geok1ng]

Lower you BCLK. One does not run 200 BCLK with 6 memorys modules on the first try.

memory frequency. DDR 800= 2:6 divider; DDr 1066 = 2:8 divider and DDR 1333= 2:10 divider.

I believe you can run DDR 800 with 1600 MCH strap, but if fails to boot, try 1333 strap.

upping memory voltages usually does NOT help, but increasing VTT does. 1.35 to 1.4V is where the money is.

Finally you can try tweaking the CPU uncore frequency. Try 16x (2133) and increase until 20x. With a slower uncore you gain some headroom to avoid going above 1.4V VTT.

Much is said about missing memory modules. but every time i boot with a "missed" memory module it means 2 things:
-My setting is not stable, at least not "10 passes of Intel Burnt test" stable.
-A slower memory frequency, looser memory timings and higher VTT helps me more than messing around with CPU cooler screws.

Just give it a try and show us how your X5650 behaves with Turbo ON.

 

[Shadowmeph]

is it normal for 0CPU and 1 cpu core to run 15-20c warmer then the other?
My L5639 cpus are running at 3.3Ghz and according to open hardware monitor one is at 33c and the other cpu is at 53 and I am idle

 

[Shadowmeph]

I am having a kind of issue with determining which is the correct cpu reading, when I use cpuid I get a reading of 3.4 ghz but then sometimes it jumps up to just over 4 ghz when I am pretty much idle but when I am running the evga eleet at the same time it shows I am at 3.2 ghz and it might jump up to 3.4ghz I am pretty sure that both cpuid and evga eleet at reading the same cpu.

 

[geok1ng]

TMonitor gives you a real look at processor clocks, but does not sees all 12 cores. you can see where and how turbo is kicking in.

 

[chapel]

What's a good price to pay for an SR-2 these days? I'm still seeing them in the $800 range

 

[402blownstroker]

What's a good price to pay for an SR-2 these days? I'm still seeing them in the $800 range

Basic bare boards go from $350 to $400. Ones with Waterblocks go from the $450 to $500. $800 is defintiely a Craigslist price.

 

[Shadowmeph]

I picked up mine with 12GB ram and L5639 chips for 450CAD

 

[chapel]

Basic bare boards go from $350 to $400. Ones with Waterblocks go from the $450 to $500. $800 is defintiely a Craigslist price.

ebay... but I get it

 

[geok1ng]

Some months ago i asked how to reach 1-2 cores max turbo bin and still keep c3/c6 state stable. turns out that when you enable c3/c6 wil also unlock the 1-2 cores max turbo bin. but such setting requires more Vcore, as it implies that you will reach 1 more multiplier for single thread services. After pumping up Vcore a notch i am happy to report that C3 make a real difference in power consumption and the SR2 nows run much cooler at idle than before with c3/c6 disabled. Under Windows 7 suspend mode is broken, and if you select Suspend the PC will only come back after a hard reset, kind of defeating the purpose of suspending in the first place.