SR-2 Optimization Thread

Hi Guys,

I am trying to give my CPU's (x5650) and other push, at the moment I am running at 190x @ 22 (4.180Mhz)
but I am finding it a bit hot, I have been using MIBW and Shimano's templates and I hit Low 70's at peak and I was wondering if anyone has found some cooler settings?

Henrik
 
tived said:
Hi Guys,

I am trying to give my CPU's (x5650) and other push, at the moment I am running at 190x @ 22 (4.180Mhz)
but I am finding it a bit hot, I have been using MIBW and Shimano's templates and I hit Low 70's at peak and I was wondering if anyone has found some cooler settings?

Henrik
Click to expand...
that determination will be pretty specific to your CPUs and your cooling. You can simply try to shave a bit if your vcore off and see if it remains stable.
 
jebo_4jc said:
that determination will be pretty specific to your CPUs and your cooling. You can simply try to shave a bit if your vcore off and see if it remains stable.
Click to expand...

Vcore, cooling (better HSF) and your ambient temps around the rig.

Those will be the biggest factors. Out of the two rigs I have, 68C is the best I get out of mine under 100% load 24/7
 
Last edited:
Small problem here - Trying to get this SR2 running...reports both cpu's in bios but in Windows cpu # selection is greyed out in cpu-z as well as Eleet. Both cpu jumpers are to the left on the mb...what am I missing? :confused:

EDIT: Figured it out...must use Win 7 Pro or Ultimate. :rolleyes:


noprocessorselection.jpg
 
Last edited:
tived said:
Hi Guys,

I am trying to give my CPU's (x5650) and other push, at the moment I am running at 190x @ 22 (4.180Mhz)
but I am finding it a bit hot, I have been using MIBW and Shimano's templates and I hit Low 70's at peak and I was wondering if anyone has found some cooler settings?

Henrik
Click to expand...

Hi mate,

I try to not exceed mid 70s for any great length of time.

try 4.1ghz - 187 x 22

CPU0 - 1.375 to 1.38750v vcore
CPU1 - 1.36250 to 1.3750v vcore

1.35 to 1.375v vtt on both.
 
PSA: It's summer time

To that end, I'm working on dialing in my voltages which were apparently very sloppily set some time during the winter when I was indifferent to the heat my CPUs were producing. A week ago I was set to 1.4 VTT and 1.36 vcore on both of my L5640s at 200x18 = 3.6ghz.

Now I'm testing 1.35 VTT and 1.325 vcore on CPU 0 (which results in 1.329 in Windows) and 1.3375 vcore on CPU 1 (which results in 1.317 in Windows).

I recall some conversation about differing voltages on the CPUs, but that vdroop on CPU 1 is massive, and exactly the opposite of what CPU 0 is doing. This is with vdroop off by the way.

What are other people using for voltages (especially L5640 users)? Would there be interest in another of Jebo's favorite Google spreadsheets to track our SR2 setups?
 
Anywhere from 1.325 to 1.400....it really depends on the chips, and I am serious on that much of a spread, with L5640s. I usually erred on the over-volted side since heat isn't an issue in my basement, but it is significantly cooler down there without the SR-2s running... :)
 
I got different BIOS and software sensor readings between the two CPU voltages, but using a multimeter I always measured them the same, so I set them the same in BIOS. With good airflow I've been able to equalize the temps on the two procs too.

I had to do a lot of voltage tweaking on the X5650s (mine are kinda slow) but the L5640s were pretty easy to OC so I haven't messed with them much, and they've been stable for months.
 
musky said:
Anywhere from 1.325 to 1.400....it really depends on the chips, and I am serious on that much of a spread, with L5640s. I usually erred on the over-volted side since heat isn't an issue in my basement, but it is significantly cooler down there without the SR-2s running... :)
Click to expand...

I'm at 1.375 vtt, 1.325v bios vcore (1.315v in windows) for both cpu's at 3.5ghz, these settings have been folding super bigadv for about a month. My temps as reported in Windows are 63c on the hottest core, however for some reason cpu0 socket massively over reports its temp so the little lcd reads mid 70's.
 
R-Type said:
for some reason cpu0 socket massively over reports its temp so the little lcd reads mid 70's.
Click to expand...
Another quirk of this board. It seems there are many.

I have continued to drop voltages. I'm down to vcore of 1.318 and 1.337 in BIOS, 1.317 and 1.317 in windows. VTT is set to 1.325 and 1.35 in BIOS, and I'm getting 1.328 and 1.338 in Windows

Oh and I've shaved about 20W off my power draw so far, as measured by a kill a watt.
 
Is this thread linked in our sticky guides? If not, it should be!
 
My SR-2 with dual hex ES flavor has underclocked itself. My permission to underclock was not granted.

This weekend I would like to OC the beast and run some linux. I have no clue how to do this.

Any help is appreciated. Thanks.
 
Hi, I don't know if this has been posted but I was having a ton of issues with overclocks and the x5650 after switching from my 5620's which seemed to love anything you threw at them.

After playing with my 5650's I had settled at 4.1 for a while but decided to mess around with it to see if I could get 4.3. After many failed attempts I went back to my settings for 4.1 but they would not boot. I kept clearing and retrying but nothing seemed to work.
So I found a thread about a guy on another forum whos overclock just stopped working one day and he couldn't get it to work for the life of him. So he unplugged the board and pulled the battery and after a day tried it again and everything worked.
So I decided to try this myself but I only let it set for 3 hours. and what do you know, it worked again and I was able to hit 4.3 with no issues.

After seeing this I decided to mess with the bios about 20 times and do a bunch of cmos clears to see if I had any issues with my old settings, and what do you know they came back.
So I did the same thing to test it again and everything is back to working order once again. So if you are having issues I would really recommend trying it and seeing if it helps.

I do have a new issue though if anyone can help.
Under IBT and most applications my computer will no longer hit 22x but only 21x multiplier.
Maybe I am missing something simple here but do you guys have any suggestions?
 
wow that's interesting. I had a x5650 for a while, and it didn't always like being overclocked either.

do you have speedstep and c-states enabled?
 
I have speedstep and turbo on, but c-state is off.
From what I have seen most people run c-state off, or maybe I have just been reading settings for to long
 


Finally got the SR-2 back up with H2O. I need some higher multiplier chips, 3.6GHz is not enough. I wish I could keep up with the top 20 but my power bill says otherwise.
 
I am in the process of a new SR-2 build. I got it basically assembled and it is booting. I am having an issue installing/configuring its memory. I have 4 4GB stick, When in install 1 in slot 1& 3 in both banks, the BIOS only sees/reports 12GB. If I install 1 in slot 1 in both banks, the BIOS only reports 4GB. If I install 1 in slot 1 & 3 of the CPU0 bank, only 4GB is reported.

Can anyone tell me which slots and banks I need to put the memory to get all 16GB to report/show?
 
I use the red slots, but you really need to just buy an additional two sticks of ram as the performance penalty is certainly noticable when running only dual channel. You should not even consider handicapping an SR-2 build for lack of $40 of ram.:rolleyes:
 
Thanks guys. Until I can get 2 more sticks, should I put 1 in slot 1, 3, and 5 of the CPU0 bank?
 
Blowstroker

Reseat the CPUs and try again. I had that happen to me when i had an SR-2 and that fixed it.
 
Well I reseated the CPUs and I think I have made one step back.... kind of. Now the BIOS reports only 8GB with 4 4GB sticks installed. I have them in slots 1 & 3 in both banks ( the first two red ones closest to each CPU ).

I finally fired up windows and ran CPU-Z. It reports seeing 16GB with slots 1 to 4 each having 4GB installed, which is correct.

I am now thinking it might be a BIOS issue and the ES chips I have. Is there any way to determine what version of BIOS is loade? I am thinking I have a standard vanilla version and is working by luck. I have tracked own the custom A54 BIOS musky here has created. Should I try to flash that version or does any have any other ideas to try?

Thanks,
Keith
 
Check on how your mother board is flexing/bending. It has caused memory to disappear/reappear on several of ours.

Try laying your case on its side.
 
If there was flex in the motherboard, why would it not show up in the BIOS, but show up in CPU-Z? How would I check if there was flex in the board? Unfortunately putting it on its side is not an option with the water cooling setup I have :(
 
Last edited:
Well I'm be :eek: Turns out to be flex. I loosen up the waterblocks some and boing..... 16GB is now posting. thanks for pointing that out jebo_4jc, I would have not thought of that.
 
Like I said, several of us encountered this strange issue. Glad you identified it.
 
If you are booting into an OS, the modded bios is not going to help you.

The now-infamous "disappearing memory" issues with SR-2s is still somewhat of a mystery. Board flex is one cause - turn you case on its side and see if it all picks up. Others include incorrect memory settings. I had a set of OCZ Gold that would run perfectly at CAS 8 "auto" settings, but dropped 2-3 sticks trying to run at CAS 9. Try different memory speeds - DDR-800 or DDR-1066 are probably the safest. Try higher or lower MCH Strap settings, higher is normally safer. Loosen or tighten memory timings - looser should be more likely to work, but I did have the one case described above where it was not. Check your actual memory voltage - all of the SR-2s I have owned required different memory voltage settings for the two banks for them to run the actual target voltage, and they all needed to be set under the target voltage (1.62-1.63V was common to get 1.65V actual.) It may just be that the board does not like the memory you have. I had a set of OCZ Platinum that worked fine in every other board I tried, but would not work at all in an SR-2. Kingston HyperX worked well for me, and R-Type at least had 6 x 1GB sticks for sale in the DC FS/FT forum.
 
First round of overclocking:

Basic system
2 x E5625 ES Xeons
Waterblocks on CPUs & Chipset
4 x 4Gb DDR3-1333 ECC/Reg Crucial
CoolerMaster Gold Pro 1200w
2 x 750 SATA Drives
GTX 460

//------------------------------------------------------------------

Bios settings

Frquency/Voltage Control
CPU Frequency - 166
PCIE Frequency - 100
CPU Multiplier - 18 ( max )
QPI Frequency - Auto (4.800GT)
Memory Frequency - DDR-1066
MCH Strap - 1333

CPU Configuration screen
Max CPUID Limit - Disabled
Coherency Support - Disabled
A20M - Disabled
Speedstep - Disabled
C-State - Disabled
Everything else is Enabled

Memory screen
tCL - 9
tRCD - 9
tRP - 9
tRAS - 20
tWR - Auto (10)
tRFC - Auto (74)
tWTR - Auto (5)
tRRD - Auto (4)
tRTP - Auto (5)
tFAW - Auto (20)
Command Rate - 1T
Everythign else is Auto

Voltage
VDroop (both) - Without VDroop
Vcore (all) - 1.3500
Vtt (all) - 1.3500
Memory CPU0 - 1.65
Memory CPU1 - 1.65
IOH - 1.400
Everything else is Auto

//----------------------------------------------

Setup has been stable for a 3 hours running 'fah6 -smp 24'.

- 6997 takes 1m 23s per percent
- 6991 takes 1m 20s per percent

Temps have been in the low 30s under full load. At stock speeds, temps where in the low 202 to upper teens. One temp has me a little concerned, but I have no idea what it is. At stock speeds it was about 40C and now is about 65C. Anyone know what that one is?

I am currently testing the next level 3.2Ghz :)

//=======================================================================

OK, I bumped the Core Frequency up to 177 to get 3.2Ghz and left everything the same. System has been up running about 3.5 hours. Core temps are in the mid to upper 30s. The unknown temp( from above ) has reached about 70C to 72C or that is what lm_sensors is reporting. If something was/is that hot, I would expect it to be putting out a bit of heat. Most of the things I have checked are cool or cold. The CPU and chipset blocks are cold to the touch( got to love 3C ambient temps to dissipate the cooleant :D ), memory is cool( 60mm fan on each bank ), random place on the board. Until I can figure out what that temp is, I will be keeping it at 3.0Ghz.

Currently it is chewing on a 6903 WU with percent times at 31m 30s. This walks all over my old quad Opteron setup ( 4 x 8389 DDR2-667 ) which did this WU round 48m per percent. Once this WU gets down I will drop down to smp WU to help out with the backlog :)

//=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-==-=-

Update 11/21/11:

I am starting to think the high temp sensor reading is BS. One thing I noticed as I started up the FAH client, pretty radically. This sensor starts out about 38C at idle. As the client starts up it changes as the processing goes up and down, the sensors changes to. It ranges from 38C to 68C, but swings one way to the other within a second. I am not smart, but I can not think of anything that has a thermal constant to change 30C in a second. I think it is now time to see how far past 3.2Ghz I can get it.
 
Last edited:
With a small VTT bump and changing the MCH strap to 1600, I would expect those settings to take you up to 200blk without issue. Assuming you are getting the standard 1x ES turbo, this would be 3.8ghz.
 
R-Type said:
With a small VTT bump and changing the MCH strap to 1600, I would expect those settings to take you up to 200blk without issue. Assuming you are getting the standard 1x ES turbo, this would be 3.8ghz.
Click to expand...

If I change the MCH strap to 1600, how will that affect the DDR3-1333 memory? Will it try to overclock it to 1600Mhz?
 
402blownstroker said:
If I change the MCH strap to 1600, how will that affect the DDR3-1333 memory? Will it try to overclock it to 1600Mhz?
Click to expand...

No, your memory frequency is dictated by multiplier assigned with the memory frequency setting. Typically setting memory at 1066 and then upping the bus to 200blk gives you an effective memory frequency of 1600mhz. If your ram is only 1333, I think you will have to set it to 800mhz in bios to avoid overclocking it.

The MCH strap applies timings for the northbridge IOH chip to operate at, you want to set this to whatever your effective memory speed after overclocking will be.
 
While this isn't a common usage scenario around here, I thought I would share a lesson I learned after fighting my SR-2 daily driver build for nearly a week.

NUMA MUST be enabled if you use more than one GPU, if it is disabled you will POST loop at code 2A until you want to stab somebody.
http://www.realredraider.com/vbulletin/showthread.php?p=179930

I was ready to blame the AMD gpu's for this, turns out NUMA does have a reason for being on Windows.
 
In a very quick glance through, I've already learned so many things to keep me from having to shoot myself, I usually drive myself to the brink assuming I'm doing something WRONG, when the facts are that a bit of knowledge can be so very useful.

On my folding forum (KitGuruFolders!), we have a spreadsheet at goo-goo-docs, where we list our gear, and all can see if the combinations help them in setup or new acquisitions.

Maybe you all here could gain from such a table. We let everyone have at it and I try to keep a fresh copy, if someone toasts the orig.

I still have most of mine in the mail, but it will be comprised of two E5645/2.4Gs, with two 212 EVOs, (hope they clear the board!), 6 x 4Gb Crucial 1333, a 750 Gold Corsair, (I was only going to start with my GTX 460 SE, but now I have a couple of budget GTX 560s on the way.

The Blackhawk Ultra claims to have room for a second PSU for us air coolers, maybe I can put another up topside if needed. I'll not overclock the GPUs, but the CPUs will get a bump.

Bruce Hinton

[glad you'se guys are here!]
 
I am having some trouble with having new memory detected. I currently have 4x4GB installed in dual channel. When I add 2 more 4GB sticks for triple channel, only 20GB is reported. This is in slots 1, 3, and 5 ( all the Red ones ). I have also tried adding 4 more sticks to populate slots 1, 2, 3, and 4 for dual channel. With 4 slots populated, I can not get the system to post. I have verified several times the DIMMs are seated correctly.

Any ideas to get the memory to be picked up?
 
Check to see whether your motherboard is flexing. Put the case on its side if needed.
 
I do not think it is that. I have laid the case on its side and even loosened a lot of the mounting hardware around the slots. I have isolated to the 5th slot of CPU2. If there truly was flex wouldn't the other slots in the bank be affected to?
 
402blownstroker said:
I do not think it is that. I have laid the case on its side and even loosened a lot of the mounting hardware around the slots. I have isolated to the 5th slot of CPU2. If there truly was flex wouldn't the other slots in the bank be affected to?
Click to expand...

Have you tried the black slots?
 
You must log in or register to reply here.
Back
Top