Asus Rampage 3 Gene Xeon X5675 overclock?

[DirtyBiker]

Hey Guys I,m hoping to get some advice on the state of my overclock voltages measured from motherboard during stress test for 12 min. are as follows
CPU@ 1.361v
CPU_PLL 1.839v
QPI/DRAM 1.318v
IOH 1.125v
IOH_PCIE 1.520v
ICH 1.120v
ICH_PCIE 1.520v
DRAM 1.525v
some times crashes during gaming or BSOD so I've been upping the CPU and QPI this is where I'm at no crash yet any suggestions are appreciated Thanks

problem is sometimes Cinebench R15 will just stop responding during benchmark or (steam) doom will just crash to the desktop after about 10 minutes of gaming in both cases no error is generated in admin events raised the QPI/DRAM to 1.35v and PLL to 1.85v lowered CPU to 1.5v

update stable voltages so far anyone see any potential issues?
CPU@ 1.35v
CPU_PLL 1.828v
QPI/DRAM 1.356v
IOH 1.127v
IOH_PCIE 1.522v
ICH 1.121v
ICH_PCIE 1.522v
DRAM 1.527v

the actual voltages measured from the motherboard are different then what the bios says

 

[Ruru]

IIRC the PLL should be about 1.88 to 1.9 volts when overclocking. Could it be that your IMC can't keep up with that high RAM clock? I'd try to lower the RAM clock by one step and check does that have any effect.

I don't have the exactly same setup but my 2nd PC has a P6X58D-E and a X5650 and I run that @ 4.2GHz, RAM (3x2 + 3x4 = 18GB) @ 1680. 20x210.

 

[DirtyBiker]

IIRC the PLL should be about 1.88 to 1.9 volts when overclocking. Could it be that your IMC can't keep up with that high RAM clock? I'd try to lower the RAM clock by one step and check does that have any effect.

I don't have the exactly same setup but my 2nd PC has a P6X58D-E and a X5650 and I run that @ 4.2GHz, RAM (3x2 + 3x4 = 18GB) @ 1680. 20x210.

Yes I had the CPU_PLL@ 1.88v but lowered it down as far as I could go, its at 1.828v under load
it seams to be stable now if it crashes again I will try boosting it a little.
I tried lowering ram clock but couldn't get the performance I wanted, can't seem to get this Rampage up to 4.5ghz could just be this X5675.
I also have the P6X58D-E Xenon X6575 but its stable at 4.5ghz 25x with the PLL set to 1.84 1800 on the ram clock I can't confirm the actual voltage like on the rampage because I don't know where to measure voltage from on that board. here's my 4.5Ghz settings on the P6
(update) you may be right I just started upping the PLL doom just crashed to desktop again @ 1.854v no errors Il let you know if it works thanks for the suggestions

 

[DirtyBiker]

IIRC the PLL should be about 1.88 to 1.9 volts when overclocking. Could it be that your IMC can't keep up with that high RAM clock? I'd try to lower the RAM clock by one step and check does that have any effect.

I don't have the exactly same setup but my 2nd PC has a P6X58D-E and a X5650 and I run that @ 4.2GHz, RAM (3x2 + 3x4 = 18GB) @ 1680. 20x210.

Ok I figured it out I just set the motherboard jumper to QPI load line enabled, then set all the voltages to auto, except CPU @ 1.345v and DRAM @ 1.526v, this sets the PLL @ 1.828v. so far I ran multiple Cinebench R15 runs and multiple attempts at finishing the last level of doom no crashes, Woo-hoo she's smooth, here's the numbers measured off the motherboard with a volt meter during stress test
CPU@ 1.349v
CPU_PLL 1.829v
QPI/DRAM 1.354v
IOH 1.127v
IOH_PCIE 1.522v
ICH 1.122v
ICH_PCIE 1.522v
DRAM 1.527v
I may try lowering the CPU voltage as low as it will go and then see if it likes a little more DRAM voltage there 3X8 1833 dims so I'm not really overclocking the memory and any attempts at tightening the timings results in lower benchmark scores I will be using a better 5 heat pipe 120mm fan cooler, when I get the case, its only on my test bench for now using a 3 heat pipe 92mm fan
///////////////////////////////////////////////////////////////////////////////Update//////////////////////////////////////////////////////////////////////////////////
tried lowering CPU voltage slightly cooler temps successful in AIDA64 10 minute stress test 1.343v crashes Cinebench and doom so I upped the voltage to 1.355v which resulted in a drop in QPI/DRAM to 1.347-1.350 measured off motherboard under stress I think this was the problem CPU was starving more testing but the numbers look good and were close to stable

 

[DirtyBiker]

IIRC the PLL should be about 1.88 to 1.9 volts when overclocking. Could it be that your IMC can't keep up with that high RAM clock? I'd try to lower the RAM clock by one step and check does that have any effect.

I don't have the exactly same setup but my 2nd PC has a P6X58D-E and a X5650 and I run that @ 4.2GHz, RAM (3x2 + 3x4 = 18GB) @ 1680. 20x210.

Thanks for your suggestions after a ton of stability testing I figured out the UCLK speed and QPI/DRAM voltage were to high, so I focused on that keeping the QPI/DRAM voltage under 1.3v while keeping the highest stable UCLK speed, 3500 @ 1.298v looks like the sweet spot, I also took your advice and set the PLL to 1.89v. now I'm going to try lowering the CPU and the PLL one at a time for efficiency but the result so far is pretty good
CPU@ 1.385v
CPU_PLL 1.894v
QPI/DRAM 1.298v
IOH 1.126v
IOH_PCIE 1.520v
ICH 1.120v
ICH_PCIE 1.521v
DRAM 1.513v

 

[DirtyBiker]

well a word of caution never set the QPI/DRAM over 1.30v I degraded that CPU it still works but not nearly as good before I set the QPI/DRAM to 1.35v. Two voltages to keep in check when overclocking the X58 Xenon X5675 never use the auto setting for QPI/DRAM or DRAM and never set QPI/DRAM over 1.3v and DRAM over 1.6. and you will have success, good thing is I only paid $12.00 for that CPU bad thing is the prices have doubled.

 

[freeagent]

Are you sure?

I ran my QPI at 1.39v with mems at 1.66v for about a decade and it was still good.. was an E.S. to boot. I also had the LN2 jumper enabled so there was minimal sag. What you may have noticed is mitigations for specter/meltdown. They suck.

 

[DirtyBiker]

Are you sure?

I ran my QPI at 1.39v with mems at 1.66v for about a decade and it was still good.. was an E.S. to boot. I also had the LN2 jumper enabled so there was minimal sag. What you may have noticed is mitigations for specter/meltdown. They suck.

no I disabled specter/meltdown was having game crashes , BSOD and rapid decline of cinebench scores I changed the CPU but with the lower voltages and its much faster the old chip kept wanting higher voltages to remain stable but initially got great scores maybe not a problem on LN2 but this is going to be aircooled 24/7 usage. were you using the Rampage 3 gene with X5675?

 

[RJARRRPCGP]

no I disabled specter/meltdown was having game crashes , BSOD and rapid decline of cinebench scores I changed the CPU but with the lower voltages and its much faster the old chip kept wanting higher voltages to remain stable but initially got great scores maybe not a problem on LN2 but this is going to be aircooled 24/7 usage. were you using the Rampage 3 gene with X5675?

A BSOD, especially "WHEA_UNCORRECTABLE ERROR" with reason being "Internal Timer Error"=Disable C1E, EIST and "turbo core" in the BIOS!

 

[DirtyBiker]

A BSOD, especially "WHEA_UNCORRECTABLE ERROR" with reason being "Internal Timer Error"=Disable C1E, EIST and "turbo core" in the BIOS!

I only have C1E enabled

 

[freeagent]

no I disabled specter/meltdown was having game crashes , BSOD and rapid decline of cinebench scores I changed the CPU but with the lower voltages and its much faster the old chip kept wanting higher voltages to remain stable but initially got great scores maybe not a problem on LN2 but this is going to be aircooled 24/7 usage. were you using the Rampage 3 gene with X5675?

No sorry, I was using R3F with x5690.

 

[Mr.Scott]

QPI @ 1.35 and Vdimm @ 1.68 on my SR-2 with 5675's @ 4.3 for a long long time now. No degradation.

 

[DR4G00N]

well a word of caution never set the QPI/DRAM over 1.30v I degraded that CPU it still works but not nearly as good before I set the QPI/DRAM to 1.35v. Two voltages to keep in check when overclocking the X58 Xenon X5675 never use the auto setting for QPI/DRAM or DRAM and never set QPI/DRAM over 1.3v and DRAM over 1.6. and you will have success, good thing is I only paid $12.00 for that CPU bad thing is the prices have doubled.

I've learned that 1366 can be quite weird when trying to find a stable OC, one day it can be several hour P95 stable and the next it will BSOD every 30mins.
I think your just shooting for too high of an oc off the bat, my X5650 & 5670 do about 4.2 @ 1.35V, 4.5 needs somewhere around 1.45V on both.

FWIW my X5650 has seen up to 1.66V Core, 1.55V VTT & 2.2V Dimm (I don't recommend this of course) all at once and it's still does just the same as when I got it.

 

[DirtyBiker]

I've learned that 1366 can be quite weird when trying to find a stable OC, one day it can be several hour P95 stable and the next it will BSOD every 30mins.
I think your just shooting for too high of an oc off the bat, my X5650 & 5670 do about 4.2 @ 1.35V, 4.5 needs somewhere around 1.45V on both.

FWIW my X5650 has seen up to 1.66V Core, 1.55V VTT & 2.2V Dimm (I don't recommend this of course) all at once and it's still does just the same as when I got it.

How are you measuring voltage? On the rampage I can use a voltmeter to measure from contacts on the Motherboard and the bios can be +or- my qpi/dram is lower then the bios the dram is higher then the bios 1.3325 in the bios gives me 1.295-1.301 depending on load where the dram 1.4972 gives me 1.5 going over these levels or auto settings can push things to high, the CPU will still work but with lower benchmarks, errors and crashes will start to require higher volts then if you never let them get above safe levels my p6x58d-e seems to be a little safer as far as Qpi/dram but the dram can over volt if left on auto if you stick with CPU 1.35, QPI/DRAM 1.30, DRAM 1.6 Max, your going to yeald a higher performance and stable overclock from my experience here's my 4.5GHz settings this would crash or bsod 2 other CPU's that I took over these limits

 

[freeagent]

All of those voltage values I would consider to be low..

Most of those numbers are below what I used to daily, the only one that is higher is your core clock. But voltage wise I would say those are all safe.

 

[silentbogo]

I haven't touched westmere in awhile, but as far as I remember my Rampage II Gene was very finicky about RAM and Uncore.
My "golden sample" x5650 was running daily 3.7-ish GHz at 1.2V during summer, and 4.2GHz the rest of the year, but in both cases it did not like RAM frequencies higher than 1800MHz: had G.skill DDR3-2133 3x4GB kit which could push 2400MHz on Haswell rig, so the limit was definitely with a board itself or a memory controller. I'd get random BSODs or reboots while watching youtube or doing mundane non-intensive stuff.
Don't forget to tune down uncore to approximately 2x RAM frequency. Also I've never even touched QPI or IOH voltages, but it was still good enough to push over 200MHz BCLK for benching and 180+ for daily.
Basically my rule of thumb was to turn everything down to a minimum, and start cranking up that BCLK until the system becomes unstable. Add core voltage and repeat. 1.4V is probably the highest I'd go for daily, and only if you have a decent cooling solution. On air I wouldn't even dare to go over 1.35V. Power draw starts piling up quite fast once you get past 4GHz. When I finally broke 5G barrier, my UPS was screaming in pain, letting me know that my rig with stock GTX750Ti was pulling over 520W!!! This said, the higher you go - the higher is the probability that something else, totally unexpected, is going to hold you back, like PSU or VRM cooling.

 

[DirtyBiker]

I haven't touched westmere in awhile, but as far as I remember my Rampage II Gene was very finicky about RAM and Uncore.
My "golden sample" x5650 was running daily 3.7-ish GHz at 1.2V during summer, and 4.2GHz the rest of the year, but in both cases it did not like RAM frequencies higher than 1800MHz: had G.skill DDR3-2133 3x4GB kit which could push 2400MHz on Haswell rig, so the limit was definitely with a board itself or a memory controller. I'd get random BSODs or reboots while watching youtube or doing mundane non-intensive stuff.
Don't forget to tune down uncore to approximately 2x RAM frequency. Also I've never even touched QPI or IOH voltages, but it was still good enough to push over 200MHz BCLK for benching and 180+ for daily.
Basically my rule of thumb was to turn everything down to a minimum, and start cranking up that BCLK until the system becomes unstable. Add core voltage and repeat. 1.4V is probably the highest I'd go for daily, and only if you have a decent cooling solution. On air I wouldn't even dare to go over 1.35V. Power draw starts piling up quite fast once you get past 4GHz. When I finally broke 5G barrier, my UPS was screaming in pain, letting me know that my rig with stock GTX750Ti was pulling over 520W!!! This said, the higher you go - the higher is the probability that something else, totally unexpected, is going to hold you back, like PSU or VRM cooling.

I have north and south bridge fans with an 80c max threshold set in bios PSU EVGA 650BQ 80+ bronze testing multiple over clocks each focused on maxing out different individual aspects DRAM, OPI, UCLK, CPU frequency DRAM frequency for stability limits and performance on air, but keeping the QPI/DRAM at or under 1.30v and DRAM voltage 1.5v or close to the dims stock voltage, to find the limits of each and which ones yield the best results here's what I have discovered
CPU frequency high as possible under 80c core max temp under stress at or under 1.35v with the following limits
UCLK 3600 max, sweet spot 3500
DRAM 1900-1800 MHz
QPI link 7200 MT/s max
my 4.5 OC yields the best overall results so far

 

[silentbogo]

my 4.5 OC yields the best overall results so far

Sounds about right. I think in my case 4.5G was the line past which OC did more harm than good (clocks went up, but perf went down for some reason, not throttling).
Still did a few benches at 4.7G+, but the best performance in 3DMark and CB15, but results were lower than @4.4G.
I sold my 1366 rig many years ago, but I really miss it. Performance may be lacking on the modern level, but it's the most "fun" platform that ever existed.

 

[DirtyBiker]

I've learned that 1366 can be quite weird when trying to find a stable OC, one day it can be several hour P95 stable and the next it will BSOD every 30mins.
I think your just shooting for too high of an oc off the bat, my X5650 & 5670 do about 4.2 @ 1.35V, 4.5 needs somewhere around 1.45V on both.

FWIW my X5650 has seen up to 1.66V Core, 1.55V VTT & 2.2V Dimm (I don't recommend this of course) all at once and it's still does just the same as when I got it.

I don't use P95 because it puts un realistic stress on the hardware I stick to games and AIDA64 stress test

Sounds about right. I think in my case 4.5G was the line past which OC did more harm than good (clocks went up, but perf went down for some reason, not throttling).
Still did a few benches at 4.7G+, but the best performance in 3DMark and CB15, but results were lower than @4.4G.
I sold my 1366 rig many years ago, but I really miss it. Performance may be lacking on the modern level, but it's the most "fun" platform that ever existed.

ok just got a BSOD while playing doom (driver over ran stack buffer) any suggestions/ I installed the via audio drivers could it be that or voltage related? I'm also thinking it could be memory timings I tightened them up because I'm underclocked on my 1866 3x8 2R dims stock is 10-11-10-27 cr2 I set them to 9-10-9-27 cr1

I haven't touched westmere in awhile, but as far as I remember my Rampage II Gene was very finicky about RAM and Uncore.
My "golden sample" x5650 was running daily 3.7-ish GHz at 1.2V during summer, and 4.2GHz the rest of the year, but in both cases it did not like RAM frequencies higher than 1800MHz: had G.skill DDR3-2133 3x4GB kit which could push 2400MHz on Haswell rig, so the limit was definitely with a board itself or a memory controller. I'd get random BSODs or reboots while watching youtube or doing mundane non-intensive stuff.
Don't forget to tune down uncore to approximately 2x RAM frequency. Also I've never even touched QPI or IOH voltages, but it was still good enough to push over 200MHz BCLK for benching and 180+ for daily.
Basically my rule of thumb was to turn everything down to a minimum, and start cranking up that BCLK until the system becomes unstable. Add core voltage and repeat. 1.4V is probably the highest I'd go for daily, and only if you have a decent cooling solution. On air I wouldn't even dare to go over 1.35V. Power draw starts piling up quite fast once you get past 4GHz. When I finally broke 5G barrier, my UPS was screaming in pain, letting me know that my rig with stock GTX750Ti was pulling over 520W!!! This said, the higher you go - the higher is the probability that something else, totally unexpected, is going to hold you back, like PSU or VRM cooling.

that could be your crashing problem I start by setting the DRAM voltage to 1.5v because I'm use 1866 dims underclocked and QPI/DRAM to 1.3v thats the XMP voltage for these dims, auto settings will overvolt these when overclocking Xenon's, maybe the I7's allowed more voltage I dont know never owned an X58 I7 but I hurt two CPU's IMC doing the same thing it set my QPI/DRAM to 1.38 and DRAM to 1.7v got great bench marks initially but quickly degraded within days of light usage.

DRIVER_OVERRAN_STACK_BUFFER (f7)
A driver has overrun a stack-based buffer. This overrun could potentially
allow a malicious user to gain control of this machine.
DESCRIPTION
A driver overran a stack-based buffer (or local variable) in a way that would
have overwritten the function's return address and jumped back to an arbitrary
address when the function returned. This is the classic "buffer overrun"
hacking attack and the system has been brought down to prevent a malicious user
from gaining complete control of it.
Do a kb to get a stack backtrace -- the last routine on the stack before the
buffer overrun handlers and BugCheck call is the one that overran its local
variable(s).
Arguments:
Arg1: 00002986679b76b0, Actual security check cookie from the stack
Arg2: 000051eae7abb6ee, Expected security check cookie
Arg3: ffffae1518544911, Complement of the expected security check cookie
Arg4: 0000000000000000, zero

Ok I had set my QPI/DRAM to 1.3v in bios but in order to achieve 1.3v to the motherboard it needs to be set at 1.33v tried running full virus scan and it was not completing and disabling on the way out so I reset to defaults in bios scan completed no threats detected which led me to up the QPI/DRAM voltage and lowered the CPU voltage just under 1.35v ran virus scan again works fine no threats so I decided to run another cinebench to test and wow check the results also AIDA64 is causing a strange hard drive access thing that defender is blocking I'm thinking this was the cause of the crash I noticed there is an update available but I keep declining maybe there is a fix will test soon as it updates I will also be using defender as another level of future stability testing

 

[Mr.Scott]

Command rate of 1 is too tight. Switch to 2 and try it again.

 

[DirtyBiker]

Command rate of 1 is too tight. Switch to 2 and try it again.

I think it was the QPI/DRAM voltage bios was set to 1.3v but was only delivering 1.27 so I bumped it up to 1.33v now it peaks at 1.299 and that also gave me a significant boost in cinebench score from 1, 028 to 1033 that also fixed my Windows defender scan

 

[DirtyBiker]

update crashes were due to old 2012 VIA audio driver switched back to standard HD audio driver no more crashing

 

[DirtyBiker]

been playing with the overclock 4.5 is solid, now I got the ram speed up where I like it, settled on 4.47 GHz, QPI speed 3501 MHz DDR3@ 1943, timings 10-11-10 CR1, this allows me to run CPU@ 1.35v (will test lower soon) QPI\VRAM@ 1.30v VRAM@ 1.54v with only a 6 point drop in Cinebench R15 render score, no crash 2 days of benchmarking and gaming, in and out of sleep mode, rock solid. 120-250 fps Fortnite performance mode butter smooth GTX 750 Ti soon to be GTX 1050 Ti

 

[DirtyBiker]

Ok think I found the sweet spot totally stable 2 days benchmarks gaming sleep mode updates no crashes. QPI/VRAM @1.28-1.30v depending on load, VRAM@1.55v overclocking 1866 dims 10-11-10-27 CR1 stock is CR2, CPU@ 1.348v measured with volt meter off motherboard under load, bios voltage numbers are not accurate. a word to the wise never exceed 1.30v QPI/VRAM you will permanently damage the memory controller on the CPU, also never set VRAM to auto or use QPI load line calibration it will overvolt causing permanent damage, CPU load line is ok, never exceed 1.60 on VRAM voltage 1.65 max only if necessary for stability but confirm voltage with volt meter.

 

[freeagent]

a word to the wise never exceed 1.30v QPI/VRAM you will permanently damage the memory controller on the CPU, also never set VRAM to auto or use QPI load line calibration it will overvolt causing permanent damage

No it won’t.