Xeon L5639 Overclocking on X58

[ClockHound]

well apparentelly i was using some of the shittiest most generic thermal paste in the world, so i went out and bought some some ic diamond which lowered my temperatures to 75c max during prime 95 4ghz 1.35vcore from 85-90c, so that was a great help still idling around 35-40c tho, but im using the evo version.

edit: hovering right at 90c at 4.2 and 4.4ghz on prime 95, so i think its an issue with pushing past 1.35v on my processor as i have 4 cores that run hot while the other 2 run 10-15c less

Well...that's exactly what I was gonna say - better TIM! ;-)

10-15c spread between cores is not normal. 5c sure, but have to wonder if:

a) The base of the EVO and the IHS are not meeting as flush possible and why the IC diamond is really helping to 'fill' the gap.

b) The mount pressure is uneven, either from defective mounting hardware (most likely) or unequal torque applied by the user (unlikely)

c) The IHS is wonky and the EVO is fine. That would suck...but there are ways to deal with that. ;-)

So....which cores are the cool ones that get all the cool temps? And the not-so-cool hot cores?

Can you post the temps by core?

0 - ??c
1 - ??c
etc...

We can then compare the core die layout to the EVO base and see if there is an obvious (mis)connection.

Or you could just go get a Kotetsu/Mugen 4/DH-15/14 and spoil the troubleshooting fun.

Gotta admit, this is getting better by degrees. ;-)

 

[jforce321]

Well...that's exactly what I was gonna say - better TIM! ;-)

10-15c spread between cores is not normal. 5c sure, but have to wonder if:

a) The base of the EVO and the IHS are not meeting as flush possible and why the IC diamond is really helping to 'fill' the gap.

b) The mount pressure is uneven, either from defective mounting hardware (most likely) or unequal torque applied by the user (unlikely)

c) The IHS is wonky and the EVO is fine. That would suck...but there are ways to deal with that. ;-)

So....which cores are the cool ones that get all the cool temps? And the not-so-cool hot cores?

Can you post the temps by core?

0 - ??c
1 - ??c
etc...

We can then compare the core die layout to the EVO base and see if there is an obvious (mis)connection.

Or you could just go get a Kotetsu/Mugen 4/DH-15/14 and spoil the troubleshooting fun.

Gotta admit, this is getting better by degrees. ;-)

from cpuid hwmonitor

core 0: 82
core 1: 76
core 2: 71
core 8: 73
core 9: 81
core 10: 83

assuming its like that due to hyperthreading.

 

[ClockHound]

The separation of the hotter cores appears normal - the two hottest are on the outside of the die - inner two get the benefit of having some lower temp silicon between them.

What's not normal, to me, is the 12c difference. At load my inner/outer core is difference is 4-5c at load.

We need to find an image of an EVO base and 56xx IHS and see if it's that the EVO can't fully cover the hottest parts of the IHS. Or your EVO is slightly/largely out-of-spec.

Maybe other Xeoniacs can chime in with their core temp spread at load and their various coolers.

Or we could both read this thread from page 1 and take notes. That was fun when it was only 30 pages....now, not so much....

 

[pouki67]

@jforce 321 : Hi guys. I just read your posts and want to ad an information. I don't know why (i'm a noob) but the temps go way higher when coupled with high memory frequency. Under 1600mhz it's better for me.

I have the same issue temps have 10 degree difference and it is the same cores than yours. I've heard some guys removing the IHS but seems to be dangerous.
I use my old noctua nhd14 and can't go higher than 4ghz with 1.36 Vcore on my x5650. It goes way too hot after that. Oh, and i use a basic arctic silver 5 thermal paste.
I stop at 4ghz it's better than my old i7 920. He was running at 3.8ghz.
More cores, less power consumption, same heat

Sorry for my english, i'm french. (yeah it's not a good excuse)

 

[jforce321]

I went behind my motherboard and tightened up the evo assembly for what i could but it seems that the two pieces of metal on the inside of the tension holder for the cpu block off the edges of the evo a tiny bit which may be helping in increasing the temperatures, but i might have to go back and untighten some stuff since my memory is having trouble registering which makes me think i put the tension way too high after readjusting the nuts behind the board. The memory modules have always been kind of finicky like that, but it seemed with the quick testing i did that i was able to reduce my prime 95 at 4ghz from 75c to 70c on the outer cores which brought it closer to the 5c difference, so i believe it helped a bit. sometimes tho i just feel like buying a 4x3gb ddr3 1600 1.5v kit, instead of my 6x2gb 1.65v kits which cause me a bit of trouble after all these years.

edit: i had to rush to work after fixing the evo nuts so i only got to take a quick temp reading on prime 95 before i had to run off, so sadly i cant do any physical work on the computer for the next 9 hours

 

[Mk pt]

What's not normal, to me, is the 12c difference. At load my inner/outer core is difference is 4-5c at load.

X5650@ 4GHz [1.34/1.35(?) vcore | 1.341 qpi v] after a fast/simple V-ray render:

5ºc difference.
cooled by old NH-D14
mboard: P6T dlx v2 with original P6T Dlx v2 BIOS

Unfortunately the P6T doesn't have offset voltage settings. Either you type in what you want or set it on automatic..

P6T Deluxe v2 doesn't alow offset voltage too.
Doesn't alow offset voltage neither can keep 21/22x 'turbo multiplier' on heavy load in 6 cores..

But now I 'flashed' P6X58D [premium i think] BIOS and now I have offset voltage..
Still, doesn't alow 22x multi on heavy load on all cores.

With a P6T WS bios I could keep 22x multiplier all time/full load in 6 cores, but I'll lose offset voltage.

I prefer offset voltage for 24/7 use.


Warning:
Flashing BIOS from different Mboards may 'burn' your Mboard.
I'm not responsible if you burn your board.
I only read and tried this procedure for P6T Deluxe v2, I don't know if it works with different Mboards.

 

[jforce321]

We need to find an image of an EVO base and 56xx IHS and see if it's that the EVO can't fully cover the hottest parts of the IHS. Or your EVO is slightly/largely out-of-spec.

thats how the evo sits on the chip, seems that the small metal flaps on the inner part of the tension assembly block off the sides a bit.

 

[multifrag]

I'm running 2xE5540 with intel motherboard(s5520sc) Want to upgrade to 2x X5650 and maybe OC them. Any suggestions on the motherboard would be very appreciated.

 

[Mk pt]

..
thats how the evo sits on the chip, seems that the small metal flaps on the inner part of the tension assembly block off the sides a bit.

you really need to clean that cooler and mboard heatsinks.


My NH-D14 needs a cleaning too, but it's not as bad as yours.
Yesterday I cleaned fan filters from my atx case, but I didn't clean the cooler.
Cleaning the cooler means removing and changing thermal paste, aka lot of work.

 

[Mk pt]

I'm running 2xE5540 with intel motherboard(s5520sc) Want to upgrade to 2x X5650 and maybe OC them. Any suggestions on the motherboard would be very appreciated.

You can't OC in a dual socket Intel mboard.

The only X58 dual socket mboard that can OC is EVGA Classified SR-2.
And that's a expensive mboard in ebay.

Maybe dual X5670 or dual X5675 is better, they have higher frequency.


To use E56xx or X56xx you'll need to check which BIOS your board is using.

Intel® Xeon® Processor 5600 Series
The server board has to be at BIOS 45 minimum to boot with Intel® Xeon® Processor 5600 Series installed. Once booted, an upgrade to BIOS 48 or later must be done for the processor to run properly and the new features to become available.

From tested processor list @ support for that board

This means you'll have to update/have BIOS 45 before changing to a E/X56xx cpu, otherwise the board won't boot.

 

[pouki67]

Just go down from 1.368v to 1.352v (cpuid software) and the system seem to be stable on 4ghz with the x5650. I've done a mistake in the previous post, i have an old NH-D12 not the 14 one.

@ MK PT : The max temp you show under 60 degree is your max temp with Hyperthreading on ? If it is, there's a gape between the NH-D12 and NH-D14. I reach 80 with Prime95 Blend test HT on (ambiant room 19 degree).
I have a p6x58D-premium with the last official BIOS 1501. This motherboard allows me to use 22-23x multiplier. 22x on all cores and 23x is a turbo on max 2 cores. The offset can be used too.

 

[Mk pt]

@ MK PT : The max temp you show under 60 degree is your max temp with Hyperthreading on ? If it is, there's a gape between the NH-D12 and NH-D14. I reach 80 with Prime95 Blend test HT on (ambiant room 19 degree).
I have a p6x58D-premium with the last official BIOS 1501. This motherboard allows me to use 22-23x multiplier. 22x on all cores and 23x is a turbo on max 2 cores. The offset can be used too.

It's under 60ºc with HT on but it's on rendering software [V-ray], after a simple/fast render.
Only 4min.10s to render. It was only a simple/fast test - I wasn't testing stability.

In Prime95 or another heavy stability test I think temperature with same oc would be 70/low 70's º.
Prime95 and other stability test [OCCT cpu/Linpack, ..] put an extreme load over cpu, so extreme that your cpu will never get close of those temps in every day software.

There's a gap between NH-D12 NH-U12P and NH-D14, but is not >20º.
Comparison here. With higher oc the gap will be a bit higher - NH-D14 [and NH-D15] need a real hot cpu to show what's capable of, but never with >20º compating with a decent 12cm fan tower cooler like the U12P





Now I flashed the P6X58D [premium I think] BIOS and I'm testing offset voltage.
Tried 4.13GHz with v1.35-something but I think it's not stable yet.
Today or tomorow I'll try to reach a stable 4.1/4.15GHz oc.

-------------


You have a P6X58D premium board, mine is a flashed P6T Deluxe v2.
I've read in a different forum [from my country] feedback that a P6T Deluxe v2 flashed with P6X58 BIOS couldn't keep 22 multiplier in heavy loads.
Flashing P6X58D BIOS gives me [on a P6T Dlx v2] offset voltage, but it can't keep 22x multiplier in load.
Only with a P6T WS BIOS on my P5T Dlx v2 I can have 22x multiplier all time, but with P6T WS BIOS I'll lose offset voltage.

It's nice to have offset voltage.
200/205/208 BLCK x20 It's not difficult to have, and I don't want to get more than 4.1/4.2GHz. Because temps and power consumption.


But one day I'll get a X5675 to try some crazy benchmarks runs.

 

[pouki67]

@ Mk pt : ok i haven't well understand. You're right 4.1 or 4.2GHz is cool. Even with 4GHz I'm steel happy. Twelve cores at that frequency is overpower. Average 20/30 % from I7 5820 stock with his 15mo cache.

 

[jforce321]

i reduced my ram from 6x2gb to 4x2gb and it seems to of lowered the temperatures by a good amount im wondering if having all the ram slots filled was forcing alot of stress on the cpu.

core 0: 80
core 1: 77
core 2: 70
core 8: 71
core 9: 78
core 10: 82

during prime 95 using cpuid hwmonitor

 

[ClockHound]

Yeah...more slots means more work for the memory controller.

Pretty sure if we ran these guys with 0 sticks, we'd hit 5Ghz easy. ;-)

jforce321, are you using 1.65V ram?

I'm running 3 sticks of 4GB 1.5V - at 4.2 I use these memory cntlr settings:
1.2125v VTT
1.82v PLL
1.10v IOH

What have you got the uncore set to? Maybe post a screenshot of all your settings.

Here's a 12Gig set of low voltage (1.25V):
http://www.newegg.com/Product/Produc...82E16820231466

Kinda wish I'd gone that route...maybe on the 5675 build. ;-)

 

[jforce321]

Yeah...more slots means more work for the memory controller.

Pretty sure if we ran these guys with 0 sticks, we'd hit 5Ghz easy. ;-)

jforce321, are you using 1.65V ram?

I'm running 3 sticks of 4GB 1.5V - at 4.2 I use these memory cntlr settings:
1.2125v VTT
1.82v PLL
1.10v IOH

What have you got the uncore set to? Maybe post a screenshot of all your settings.

Here's a 12Gig set of low voltage (1.25V):
http://www.newegg.com/Product/Produc...82E16820231466

Kinda wish I'd gone that route...maybe on the 5675 build. ;-)

yeah im running 1.65v in all the slots which is why i think pushing the frequency really raises the temperatures, tho i think my board only supports at lowest 1.5v ram from what the website said since i have a p6t.

but my 3 settings are similar to the ones you posted.

 

[jforce321]

for some reason now my computer will only recognize 8 out of 12gb on the board, even though cpuid shows all of it. it seems to be an issue with slots 3 and 4, I'm wondering if i messed up something

 

[Burpo]

Tightening CPU cooler down too tightly is known to cause such issues with X58. Whether it's cpu related or motherboard is the question.. I had an Asus P6X58D Premium do the same thing to me. My kid's still gaming on it w/8Gb instead of 12.

 

[jforce321]

Tightening CPU cooler down too tightly is known to cause such issues with X58. Whether it's cpu related or motherboard is the question.. I had an Asus P6X58D Premium do the same thing to me. My kid's still gaming on it w/8Gb instead of 12.

well i tightened the back nuts on my cooler while it was still hooked in on the front, so i think that the pressure from that might of messed it up

I'm an idiot.

But do you know of any way to fix it or if itll hurt my gaming too much to have to use dual channel config over tri channel.

Also I dont know what the dual channel config is for a p6t se so i just have a set of each type of ram i had in slots 1/2 and 5/6.

 

[hasu]

for some reason now my computer will only recognize 8 out of 12gb on the board, even though cpuid shows all of it. it seems to be an issue with slots 3 and 4, I'm wondering if i messed up something

I had a similar issue with a Gene. After resetting (clear CMOS and remove battery and put it back), it detected more slots (4 out of 6 slots, iirc). To me that happened even with default RAM voltage. In the back of my mind I think I saw RAM discrepancy right from the beginning when I filled all the 6 slots, but I had other issues to deal with for that board so I did not give much attention to it. My testing was not conclusive as I was mostly using 32bit Vista and a 275 W SFF PSU. When I send it back the seller said it worked for him with the original RAM that came with it.

Are you using 64bit OS? Try resetting the bios and use all default settings. Try enabling and disabling XMS memory settings. Another option is to try to flash the latest bios, may be?

Unless you get an inexpensive motherboard it is hard to experiment much with it!

 

[hasu]

Tightening CPU cooler down too tightly is known to cause such issues with X58. Whether it's cpu related or motherboard is the question.. I had an Asus P6X58D Premium do the same thing to me. My kid's still gaming on it w/8Gb instead of 12.

Is that a manufacturing/design flaw? Will loosening of cooler help?

 

[jforce321]

I had a similar issue with a Gene. After resetting (clear CMOS and remove battery and put it back), it detected more slots (4 out of 6 slots, iirc). To me that happened even with default RAM voltage. In the back of my mind I think I saw RAM discrepancy right from the beginning when I filled all the 6 slots, but I had other issues to deal with for that board so I did not give much attention to it. My testing was not conclusive as I was mostly using 32bit Vista and a 275 W SFF PSU. When I send it back the seller said it worked for him with the original RAM that came with it.

Are you using 64bit OS? Try resetting the bios and use all default settings. Try enabling and disabling XMS memory settings. Another option is to try to flash the latest bios, may be?

Unless you get an inexpensive motherboard it is hard to experiment much with it!

I'm using 64 bit windows 8.1, but im already on the latest bio's but i could reinstall that i guess and see what happens, otherwise i did try some overclock settings from a video without xmp but i saw the whole 12gb while doing that, so ill try going back to default and messing around with xmp settings as well. currently at work so i cant do anything to my computer atm.

 

[hasu]

I'm using 64 bit windows 8.1, but im already on the latest bio's but i could reinstall that i guess and see what happens, otherwise i did try some overclock settings from a video without xmp but i saw the whole 12gb while doing that, so ill try going back to default and messing around with xmp settings as well. currently at work so i cant do anything to my computer atm.

(You are right it is XMP not XMS!)

If it is already on the latest BIOS, then re-flashing may not have any effect. Are you sure your memory is still good? I think 1.65v is too high for 1.5v memory. Have you tried the memory itself on another machine? I think someone else also had a similar issue here in this thread. Not sure if he could fix it (I did not follow up with that). I would never recommend more than a mild over clocking centered around BCLK. (but that is just me!)

 

[jforce321]

(You are right it is XMP not XMS!)

If it is already on the latest BIOS, then re-flashing may not have any effect. Are you sure your memory is still good? I think 1.65v is too high for 1.5v memory. Have you tried the memory itself on another machine? I think someone else also had a similar issue here in this thread. Not sure if he could fix it (I did not follow up with that). I would never recommend more than a mild over clocking centered around BCLK. (but that is just me!)

its some of the earlier i7 tri channel memory that was rated at 1.65v so its not over spec with what i was doing. the memory is still good because i can switch it between all the slots, its just slots 3 and 4 that wont work with whatever ram i put in them, tho cpuid can see them in windows tho bio's and windows cannot so the motherboard is registering them through cpuid but they cant be read somehow.

 

[Burpo]

Don't believe CPU ID (or any other software). If it's not showing up in bios, it's not usable..