x58 xeon x5675 overclocking help needed with high speed mem

[oinkypig]

qpi/vtt at 1.35 is definitely not fatal. iN FACT, too low of qpi/vtt was one reason my overclock to 4.8Ghz was unstable all the damn time. if i found the x58 xeon overclocking thread about all settings i will post it for you. No one has brought up what is your cooling solution? give that a read. https://www.techpowerup.com/forums/threads/overclocking-the-x58-a-practical-guide.108526/
Also your running way to high a voltage for lower than 4.4Ghz. I was stable at 1.32v @with 4.3Ghz and i still thought that was a tad high. I needed 1.45-1.5v for 4.6Ghz and 1.55v for 4.8Ghz. Neither were done without a good air cooler keeping temps down too.https://www.techpowerup.com/forums/attachments/untitled-png.90192/
go into hwbot(32nm westmeres)-look through cne 11.5, cne15, cne20 benchmarks-pick the ranking apart and search for images posted by those users that show what settings the used to achieve the overclocks. check out x5650,x5660,x5670,x5680,x5690 chips as well for more rankings since they are based on the same westemere precessors.
cne 11.5 rankings
Cne 15 rankings
cne 2 rankings

 

[agent_x007]

^Source : LINK.

It's no surprise your chip degraded when you put 1,65V of Vcore through it on daily basis (max. Intel spec 1,4V).
1,35V VTT or QPI/DRAM is 100% within specs for 32nm stuff.
Also, in most cases, it's about difference between voltages (like DRAM vs. IMC over 0,5V), not absolute values themselves that are deadly.

 

[Regeneration]

Gulftown/Westmere real limits:

QPI/VTT above 1.3v and/or DDR above 1.64v will degrade the IMC in 24/7.

Vcore above 1.325v will degrade the core. LLC needs to be taken in consideration.

Tcase: 83c

Some chips are more resilient than others.

 

[oinkypig]

1.64v on a chip doing 4.2ghz is less heat output than a chip at 1.5voltz doing 4.6ghz.
Frequency and heat output are linear in terms of raw performance demands on these chips. It is going to push more out the more that you push
A x5650 is not binned any differently than a x5675 or x5690 for any reason other than price gouging. They are the same exact chips manufactured under the same exact processing with mulitplier limitations. The OP needs to stop running 25x on a cpu that is supposed to be only 24/7 stable on a 23x platform. If you think This platform is a total bust/go get a ryzen

Yeah as @Regeneration said above don't waste your time&money on another(X) CPU the result will most likely be identical....instead of that if you have money to waste it will be much better to buy 1 more memory stick to complete your triple memory channel....maybe the BEST solution for you is to sell this ram(or return/change)and get other this time on 1866Mhz instead and that will probably work without any problems...GL

 

[PolRoger]

If you think This platform is a total bust/go get a ryzen

I'm not sure that X58 is now a total bust?? But it is getting rather "long in the tooth". If this were my only setup I think a total upgrade to a new Ryzen platform would be worth considering.

I still enjoy "putzen around" with various older generation computer tech and I recently acquired several 2x8GB DDR3 kits to be able to run 4x8GB quad channel on my X79 platform. I had never tried 8GB DIMMs on X58 before so I thought I'd give it a try...

I can't seem to get triple channel with 8GB DIMM to train/post at much higher speeds compared to some of my 3X4GB and 3X2GB kits...

3X8GB @1944C9:

 

[Zyll Goliat]

1.64v on a chip doing 4.2ghz is less heat output than a chip at 1.5voltz doing 4.6ghz.
Frequency and heat output are linear in terms of raw performance demands on these chips. It is going to push more out the more that you push
A x5650 is not binned any differently than a x5675 or x5690 for any reason other than price gouging. They are the same exact chips manufactured under the same exact processing with mulitplier limitations. The OP needs to stop running 25x on a cpu that is supposed to be only 24/7 stable on a 23x platform. If you think This platform is a total bust/go get a ryzen

WTF are you talking about???Where did I ever said that this platform is total bust???Did you read at all my post above???
If anything I always LOVE X58 platform and I always go for BEST price/performance setups for myself and until recently I was using X58 with 6-core Xeon so I know what I am talking about.......

 

[sebaskitty]

hey i'm putting a quick 2 or 3 day pause on this oc.

a friend gave me another x5675 chip and i gave his my bad performer as he is using them in non oc servers..


but i'm doing some case modding aswel as swapping my aio for a nh-d15.

the cpu wil be adhered using liquid metal once i've validated it's peformance and i'm hitting thermal limits


here is a snipped of my case modding atm

 

[sebaskitty]

hmm i might have figured out why my pc was so unstable

1 i got a new chip. the chip booted 5ghz in windows at 1.58v core and everything else at the max recommended

2 my corsair psu is failing

the voltages are quite low

and when connected to a test psu both the voltages are higher and vcore run higher

 

[freeagent]

My x58 is hard on power supplies. I usually feed it a new 850w unit every three or four years. Max I can all core bench mine is 4800 1.6v, it’s getting harder to run high speeds though. Not sure if it’s the culmination of all the mitigation’s, degradation, or a combination of the two.. as I’ve been running this CPU since 2010 I think. I am leaning more towards mitigation though. Also find it hard to run tight timings too, mainly anything CAS6 faster than 750mhz and higher speed CAS7 closer to 850-900mhz. So I gave up and am running a mixed set of CAS9.

 

[Tomgang]

My x58 is hard on power supplies. I usually feed it a new 850w unit every three or four years. Max I can all core bench mine is 4800 1.6v, it’s getting harder to run high speeds though. Not sure if it’s the culmination of all the mitigation’s, degradation, or a combination of the two.. as I’ve been running this CPU since 2010 I think. I am leaning more towards mitigation though. Also find it hard to run tight timings too, mainly anything CAS6 faster than 750mhz and higher speed CAS7 closer to 850-900mhz. So I gave up and am running a mixed set of CAS9.

I can't say about psu, but replacing one every 3-4 years sounds very odd to me. I have used the same psu since my system whas new back in May 2009. So the psu I have, has just passed it's 11 year of run time. Well it is a 1500 watt psu, but it has not had a easy life either. Starting out with a gtx 285 3 way sli setup and a I7 920 all overclock to all it cut handle and else the psu has handle sli setup with I7 920 and a i7 980x right up to when I got my gtx gtx 1080 TI that is the first card only running a single card. My psu has run X58 oc setup and sli for at least 8 years of its life time and still going strong.

Replacing a psu every 3-4 years sounds to me you are buying cheap PSU'S, you are unlucky or simply buying to small a psu for the wattage your system can pull.

 

[sebaskitty]

My x58 is hard on power supplies. I usually feed it a new 850w unit every three or four years. Max I can all core bench mine is 4800 1.6v, it’s getting harder to run high speeds though. Not sure if it’s the culmination of all the mitigation’s, degradation, or a combination of the two.. as I’ve been running this CPU since 2010 I think. I am leaning more towards mitigation though. Also find it hard to run tight timings too, mainly anything CAS6 faster than 750mhz and higher speed CAS7 closer to 850-900mhz. So I gave up and am running a mixed set of CAS9.

this corsair is a rm650x that i bought in december 2017. i thought it would've been good. but nope.

i tossed in a 8 or so year old 80 plus kolink 700 watt psu in it for the time before buying a new unit. but so far 12v is 1v hgiher than the corsair underload.

these xeons love power currently have a |stable" 4.6 ghz overclock. stil running tests but so far so good.

also noticed getting 150 cinebench points more by just swapping the psu so thats a win

 

[freeagent]

My 24 pin ATX is a little worn. If volts are low try moving that around a bit.. do it in the bios so you can see the effects in real time

I did run a 650 on this thing for a few months, overclocking wasn't a good idea, and that was at a time when the rig had a 970 instead of a 980. Even at the settings in my sig Real Bench will pull a little more than 500w. Such a pig.

 

[sebaskitty]

ahh i see.

i want to grab a 850watt unit when i can. it should be enough for this rig. and the ryzen rig i want to build next year.

what brand do you recommend?

currently i'm working on overclocking my ram. currently tightening the timings on a 1600mhz ram clock

wish i had triple channel right now

 

[Edwired]

It quite interesting i have the x5675 on asus p6x58d-e @ 1.264v with turbo enabled it running according to taskmanager 3.86ghz but boosts to 4.20ghz all cores. Ram is running at 1680mhz 9-9-9-24-110-t2 As i have antec hcg 850w it one of those with modular version seems to hold up since only an odd time it does something funky and a simple reboot fixes it

 

[freeagent]

View attachment 158721
^Source : LINK.

It's no surprise your chip degraded when you put 1,65V of Vcore through it on daily basis (max. Intel spec 1,4V).
1,35V VTT or QPI/DRAM is 100% within specs for 32nm stuff.
Also, in most cases, it's about difference between voltages (like DRAM vs. IMC over 0,5V), not absolute values themselves that are deadly.

This is new. They used to give max vcc as 1.55v. And right there they list vtta and d as 1.4v, so you wont fry your mem controller going passed 1.35v. I gave mine 1.39 for years and its fine. I also wont give mine more than 1.6vcc for too long. A couple of quick sessions and that's it. Normally I keep the vcore within its stock voltage rage, but I use the i7 side for 1.375v instead of 1.35v because its the same shit in a different pile. And 100MHz, because I am greedy

 

[steen]

the xeon w3670 i s the cheapst W chip with a unlocked multiplier right? i can see a few for sale for 20$ on aliexpress

Nope, only W3680/90 have unlocked multipliers. W3670 can overclock well but also has locked lower QPI/IMC multis, so BCLK is the only way to overclock. That's why it's cheap. May as well use a W5675 because it doesn't matter that it's base CLK is lower when overclocking, both have 26x turbo multi. There's more variability within SKUs anyway.

It quite interesting i have the x5675 on asus p6x58d-e @ 1.264v with turbo enabled it running according to taskmanager 3.86ghz but boosts to 4.20ghz all cores. Ram is running at 1680mhz 9-9-9-24-110-t2 As i have antec hcg 850w it one of those with modular version seems to hold up since only an odd time it does something funky and a simple reboot fixes it

I ran 3x HP Z400 W3680 @ 3.8-4GHz until fairly recently on the OEM Delta 475W PSUs. All super stable. Not bad being mild OC for 6yrs given mediocre motherboard VRM.

 

[Mr.Scott]

 

[Regeneration]

Didn't know FCKGW-RHQQ2-YXRKT-8TG6W-2B7Q8 is also working on Windows 7.

 

[freeagent]

View attachment 162346

Didn't know FCKGW-RHQQ2-YXRKT-8TG6W-2B7Q8 is also working on Windows 7.

The devil made it his own

 

[lexluthermiester]

cpu core-1.45-1.5v
cpu pll-1.88v
qpi-1.35v
dram-1.66v

Those voltages are all way too high for long-term daily use.
CPU core should never top 1.4v
CPU PLL should never top 1.825v
QPI PLL should never top 1.25v
DRAM should never, EVER top 1.65v(you will swiftly burn out the IMC with voltage that high)

qpi/vtt at 1.35 is definitely not fatal.

Oh yes it is. Never run it that high on a chip you care about.

 

[FAH addict]

Mine GA-X58-UD5 is stable with L5640 @3600 MHz cores 3200 MHz uncore with Uncore voltage=QPI/Vtt =1.30 V, RAM 1600 MHz @1.66V @bios/1.64V real.
If I want higher than 3600 MHz at cores - Uncore voltage should be QPI/Vtt = 1.34-1.36 V (1.36 for higher than 3800 MHz cores) for raising uncore clock to (Cores clock minus 2xBCLK) or to (Cores clock minus 4xBCLK)MHz, RAM could be clocked up to 2000 MHz with something like 10-11-10-30, but mixing 2 sets of different sticks is tricky thing and would need playing with secondary timings if mixed RAM stick's clock higher than 1600 MHz.
RAM for 1.66V should have "1.65 V" printed on it paper labels and should be cooled with some fan above it for stability because 1.65V sticks really hot.
BSODs from my initial X58 OC were mostly from too little QPI/Vtt or from RAM overheating at 1.66V @bios.
Overall Vcore higher than 1.35-1.36 at Westmeres is too much and platform become sort of nuclear plant with 1.42 Vcore.

 

[lexluthermiester]

1.65v ram should be avoided as IMC degradation is a very real problem. 1.5v should be the goal. It should also be noted that the differences in RAM performance between DDR3 1333, 1600 and 1866 are almost within the statistical margin of error for most benchmarks, games and programs because as the speed goes up, the timings have to go up with them.

Using DDR3 1600 at 1.5v 9-9-9-27 timings is the sweetspot for performance on the X58 platform, as general rule.

 

[agent_x007]

@lexluthermiester It's not about absolute voltage, it's the difference between DRAM and IMC that's deadly when it's over 0,5V (in both ways +/-).

 

[lexluthermiester]

@lexluthermiester It's not about absolute voltage, it's the difference between DRAM and IMC that's deadly when it's over 0,5V (in both ways +/-).

While that's true, my experience has been that it's better to stay away from the 1.65v limit, even accounting for natural variances. Intel themselves stressed the importance of staying away from that limit.

 

[Regeneration]

@lexluthermiester It's not about absolute voltage, it's the difference between DRAM and IMC that's deadly when it's over 0,5V (in both ways +/-).

That's unconfirmed rumor that some people spread. Makes absolutely no sense. It's hard not to be in 0.5v range.