5960x Overclock help !

killswitch73 · 12 Jan 2016 at 02:22

just built a gaming system .

spec

5960x water cooled (2 x 480mm rads)
2 x 980ti sc water cooled on same loop as cpu
16gb 3000mhz ram
MSI godlike mobo
EVGA 1600 P2

I have a problem trying to get the best out of overclocking my pc .

the most stable overclock i can get so far is 4.25ghz

i can sometimes i can get 4.375 if not stress testing and just benching things like firestrike . and this is using the MSI control center with 34x125mhz and is fine until i use OCCT to stress it . i also got 4.5ghz through same method all on stock volts . but now it blue screens when i do 4.5ghz using MSI software so that 4.5ghz was a 1 off and got a score with firestrike and didnt crash .

with the 4.25ghz its on stock volts and is 33x125mhz and is rock solid and in CPUz its reading 1.29v. and OCCT and aida64 reads that my temps max out at 61c when understress for 2 hours . all from bios and no software

my problem is when i try and go further from the bios adding Volts to the cpu . if i go 1.35v it crashes on OCCT , and so does 1.375 and 1.4v . infact adding any volts even on 4.25 @ 1.30v crashes and so does 4.25 @ 1.35v .

my main concern is my temps get a max of 70c which is on core 4 at 4.5ghz @ 1.4v for about 30 mins then bluescreens .

i can post fine , i even posted fine @ 4.6ghz @ 1.4v

I'm not the best overclocker out there but wouldn't it not blue screen when i go over the temps ?

im now starting to think my power supply is a problem . but a EVGA Supanova 1600w P2 . is plenty .

using prime95 is a no go as it crashes right after i start it @4.375ghz and over .

could this be a really bad chip ?

if it helps if i add volts on my graphics cards it also freezes when trying to bench with things like valley and firestrike . i think the best stable overclock is 1412mhz . soon as i notch up the power it also crashes . with temps not going over 49c .

this is whats making me think i have a dodgy PSU .

hope this has made sense as im tired and gooing to bed lol

killswitch73 · 12 Jan 2016 at 02:31

just to add i have already tried the ram at 2133mhz and strap at 100mhz without XMp, with XMP the strap goes to 125mhz .

tyler_jrb · 12 Jan 2016 at 05:49

Ensure that in the bios the max long duration TDP is not set to auto. Ensure its set to about 250w. That will ensure the CPU doesn't get choked by power constraints. Should be in CPU power management.

Other than that get those GPUs flashed

, if you want to get big clocks with them. Just make sure to read the clocks out of afterburner as fire strike gives lower clock readings.

It could be the chip itself has hit its wall where any extra voltage doesn't make a difference. But unless its a really bad chip I doubt it.

killswitch73 · 12 Jan 2016 at 09:01

My TDP is set to auto yes ..I read somewhere else it needed this but didn't know where it was ..I did find it yesterday .can't do anything until tomorrow or late tonight as working today damn .what I will say is that this chip I've got did 4.4ghz stable on the Alienware board with only 1x 120mm crappy OEM closed loop . Has to be something to do with power .

hadri · 12 Jan 2016 at 09:49

I was about to say I gave up on OCCT following advice elsewhere and use the ASUS ROG stress testing tool - but that's prime95 based. It quite possibly could be power, but what catches my eye is your ram speed of 3000mhz. I know some early z99 boards had big issues with that ram speed and people had to alter the c-state power management to get it stable until MB bios's were updated. Are you using the latest bois?

Noobish question, but what's wrong with the strap being 125? My ram has a OC XMP profile that uses a strap of 127.1 (weird number I know) so I just use that as I couldn't be bothered to OC the ram and I adjusted the base clock to suit

killswitch73 · 12 Jan 2016 at 10:37

The ram is not realy the problem as even if I don't use the xmp profile and just have the ram at 2133 and have a strap of basic 100MHz. The ram in this board is fine as I emailed MSI and they say it's perfectly fine as they say it's the best ram for the board and can go 3200mhz with overclock but not interested in overclocking the ram . It's not juse the cpu tbough as the gpus are lacking also specially when they are both under water and max temps on graphics reaches 49c ..I don't think I've seen it above 54c on games and that's with fans and pump only at 30% . But benching the graphics past 1412mhz is none existing .for a start I can't get anywhere near my old score of 28k on fire strike using the same chip and 3x gtx980 reference and on air ..I can only get 24k what in have now with faster ram and 2x 980ti and cpu is the same cpu that was in my alienware ..The Alienware was limited for cooling as it for very hot in the 80s for graphics . All the loss of score is from the graphics ..surly 3x 980 reference cards can't gain 4k points over the 2x 980ti sc and under water and 40% cooler ..

killswitch73 · 12 Jan 2016 at 10:44

I think Tyler is right about the TDP though as I don't think it's getting enough power . But the graphics cards is another question as maybe I'm expecting to much from the cards ..

hadri · 12 Jan 2016 at 10:55

Ah that certainly eliminates the ram as an issue! Out of curiosity, are you seeing any benefit of x2 980ti's over the x3 980s? My 980's are sc, do you think it's worth the switch?

killswitch73 · 12 Jan 2016 at 12:31

hadri said:
Ah that certainly eliminates the ram as an issue! Out of curiosity, are you seeing any benefit of x2 980ti's over the x3 980s? My 980's are sc, do you think it's worth the switch?

Yes ..as 3 x gpus not really optimised in some games and can lower frames in other too ..The 2 x980tI I have a good increase becuase of the vram ..They ran my 4k screens at around 20 frames better off and even more on older games ..project cars, grand theft auto , witcher 3 , battlefield all had a greater increase ..I'm now on 3x 27in 2560x1440p screens with a res of 7680x1440 and my 980tis have come to life ..project cars im getting over 100fps on ultra settings. Only game to bring it to its knee's is still crysis 3 ..where I get about 70-80 with a few settings turned down .

Nickolp1974 · 12 Jan 2016 at 15:19

Stick to 100 strap, set mems to 2666mhz and check for profiles in mem section and set ram voltage and also sa volts to 1.06v, set x43 multi and 1.3v and test and not with OCCT! Asus real bench or something like, if bsod make a note of code, 101, 125, 9c etc etc post back

Edit: and about 1.9v on input, you wont need to mess with any other settings at these speeds!

RockoSRI · 12 Jan 2016 at 15:28

Nickolp1974 said:
Stick to 100 strap, set mems to 2666mhz and check for profiles in mem section and set ram voltage and also sa volts to 1.06v, set x43 multi and 1.3v and test and not with OCCT! Asus real bench or something like, if bsod make a note of code, 101, 125, 9c etc etc post back

Edit: and about 1.9v on input, you wont need to mess with any other settings at these speeds!

This.

I'm at 4.5ghz(45x100) on an i7 5820k with 1.3v and 1.9 input. Asus Real Bench and XTU I have used to stress test. Both tools worked well

tyler_jrb · 12 Jan 2016 at 17:52

I have both my EVGA SC's flashed with a 1.25v bios. Allows them to do +210 on the core clock or about 1500mhz on each. I can put a 1.28 bios where they will do 1550 but they throw out a lot more heat. A flashed bios also allows for 121% power where stock is quite low.

The TDP should be in CPU power management but if it isn't just go through all the bios until you find max longevity power or along those lines. Auto is stock TDP or 140w and it won't exceed that amount. Over clocked its likely the chip is pushing more than that, you just need to put in a higher figure so it can use more power if required.

killswitch73 · 12 Jan 2016 at 18:06

Nickolp1974 said:
Stick to 100 strap, set mems to 2666mhz and check for profiles in mem section and set ram voltage and also sa volts to 1.06v, set x43 multi and 1.3v and test and not with OCCT! Asus real bench or something like, if bsod make a note of code, 101, 125, 9c etc etc post back

Edit: and about 1.9v on input, you wont need to mess with any other settings at these speeds!

Ive already done this with the memory and didn't work and where would be the error codes im getting sentences on the blue screen in windows 10 ..like watch-dog checker or something like that ..it's all to do with failed overclock .

I need to put the TDP up to 250w as it's on auto and this could be the problem and of course OCCT ..and will try the asus stress tester .. it's defo a Power problem i have and it has to be the TDP as its mighty strange that its crasshing less then 65c .when I did the ram though I only set it to default 2133 and didnt add the volts .I had 1.9 on input also but some reviews out there for my board suggested 20 or even 21 input . But I will do what you suggested in case I have missed something ..and will report back .

killswitch73 · 12 Jan 2016 at 18:14

tyler_jrb said:
I have both my EVGA SC's flashed with a 1.25v bios. Allows them to do +210 on the core clock or about 1500mhz on each. I can put a 1.28 bios where they will do 1550 but they throw out a lot more heat. A flashed bios also allows for 121% power where stock is quite low.

The TDP should be in CPU power management but if it isn't just go through all the bios until you find max longevity power or along those lines. Auto is stock TDP or 140w and it won't exceed that amount. Over clocked its likely the chip is pushing more than that, you just need to put in a higher figure so it can use more power if required.

I found where the TDP was yesterday but didn't mess with it ..I remember the alienware had that tdp on 250w on standard and greyed out in the motherboard ..so it has to be that mate .it seems it crashes with no power .and guess what motherboards Alienware use ..MSI boards as the layout is near exact same and so is the overclocking software they used .

I've got a few things to try now ..will try a memory test also ..im using the Overclockers Avexir memory with lifetime warranty ..I got spare crucial 2400mhz ram I can try also ..hoping it's not a bad board as it cost more than the ROG ..should have gone ROG lol

killswitch73 · 12 Jan 2016 at 18:20

I can get 175 core and 350 memory on any more core and it bums out and freezes ..but it's not a blue screen it's a driver crash ..I've gone back to 358 drivers for now and will test in all in next few days ..determined to get my 4.5ghz day to day clock and maybe 4.7ghz bench ..This is why I built it ..but I suppose having a bit more knowledge would help .

Nickolp1974 · 12 Jan 2016 at 21:28

Try a program called whocrashed, analize on next boot and it will tell you bsod code, you will NOT need to adjust tdp settings etc for those clocks, auto is fine. I never needed to running 5.6ghz.
Yes testing your memory is a good idea, set everything to defaults then set xmp and run memtest

killswitch73 · 13 Jan 2016 at 02:26

Nickolp1974 said:
Try a program called whocrashed, analize on next boot and it will tell you bsod code, you will NOT need to adjust tdp settings etc for those clocks, auto is fine. I never needed to running 5.6ghz.
Yes testing your memory is a good idea, set everything to defaults then set xmp and run memtest

Computer name: DESKTOP-TBR0D3F
Windows version: Windows 10 , 10.0, build: 10586
Windows dir: C:\Windows
Hardware: MS-7883, MSI, X99A GODLIKE GAMING (MS-7883)
CPU: GenuineIntel Intel(R) Core(TM) i7-5960X CPU @ 3.00GHz Intel586, level: 6
16 logical processors, active mask: 65535
RAM: 17068859392 bytes total

--------------------------------------------------------------------------------
Crash Dump Analysis
--------------------------------------------------------------------------------

Crash dump directory: C:\Windows\Minidump

Crash dumps are enabled on your computer.

On Mon 11/01/2016 18:52:17 GMT your computer crashed
crash dump file: C:\Windows\Minidump\011116-11437-01.dmp
This was probably caused by the following module: hal.dll (hal!HalBugCheckSystem+0xCF)
Bugcheck code: 0x124 (0x0, 0xFFFFE000C0EC1028, 0xBF800000, 0x124)
Error: WHEA_UNCORRECTABLE_ERROR
file path: C:\Windows\system32\hal.dll
product: Microsoft® Windows® Operating System
company: Microsoft Corporation
description: Hardware Abstraction Layer DLL
Bug check description: This bug check indicates that a fatal hardware error has occurred. This bug check uses the error data that is provided by the Windows Hardware Error Architecture (WHEA).
This is likely to be caused by a hardware problem problem. This problem might be caused by a thermal issue.
The crash took place in a standard Microsoft module. Your system configuration may be incorrect. Possibly this problem is caused by another driver on your system that cannot be identified at this time.

On Mon 11/01/2016 18:41:49 GMT your computer crashed
crash dump file: C:\Windows\Minidump\011116-11484-01.dmp
This was probably caused by the following module: ntkrnlmp.exe (nt!KeBugCheckEx+0x0)
Bugcheck code: 0x4A (0x633046A4, 0x2, 0x0, 0xFFFFD00025D2AEC0)
Error: IRQL_GT_ZERO_AT_SYSTEM_SERVICE
Bug check description: This indicates that a thread is returning to user mode from a system call when its IRQL is still above PASSIVE_LEVEL.
The crash took place in the Windows kernel. Possibly this problem is caused by another driver that cannot be identified at this time.

On Sun 10/01/2016 19:44:32 GMT your computer crashed
crash dump file: C:\Windows\Minidump\011016-10421-01.dmp
This was probably caused by the following module: hal.dll (hal!HalBugCheckSystem+0xCF)
Bugcheck code: 0x124 (0x0, 0xFFFFE000C6AC1028, 0xBF800000, 0x124)
Error: WHEA_UNCORRECTABLE_ERROR
file path: C:\Windows\system32\hal.dll
product: Microsoft® Windows® Operating System
company: Microsoft Corporation
description: Hardware Abstraction Layer DLL
Bug check description: This bug check indicates that a fatal hardware error has occurred. This bug check uses the error data that is provided by the Windows Hardware Error Architecture (WHEA).
This is likely to be caused by a hardware problem problem. This problem might be caused by a thermal issue.
The crash took place in a standard Microsoft module. Your system configuration may be incorrect. Possibly this problem is caused by another driver on your system that cannot be identified at this time.

On Sun 10/01/2016 16:53:13 GMT your computer crashed
crash dump file: C:\Windows\Minidump\011016-9968-01.dmp
This was probably caused by the following module: hal.dll (hal!HalBugCheckSystem+0xCF)
Bugcheck code: 0x124 (0x0, 0xFFFFE0008B4E2028, 0xBF800000, 0x124)
Error: WHEA_UNCORRECTABLE_ERROR
file path: C:\Windows\system32\hal.dll
product: Microsoft® Windows® Operating System
company: Microsoft Corporation
description: Hardware Abstraction Layer DLL
Bug check description: This bug check indicates that a fatal hardware error has occurred. This bug check uses the error data that is provided by the Windows Hardware Error Architecture (WHEA).
This is likely to be caused by a hardware problem problem. This problem might be caused by a thermal issue.
The crash took place in a standard Microsoft module. Your system configuration may be incorrect. Possibly this problem is caused by another driver on your system that cannot be identified at this time.

On Sun 10/01/2016 15:56:21 GMT your computer crashed
crash dump file: C:\Windows\Minidump\011016-10515-01.dmp
This was probably caused by the following module: hal.dll (hal!HalBugCheckSystem+0xCF)
Bugcheck code: 0x124 (0x0, 0xFFFFE0014AAE2028, 0xBF800000, 0x124)
Error: WHEA_UNCORRECTABLE_ERROR
file path: C:\Windows\system32\hal.dll
product: Microsoft® Windows® Operating System
company: Microsoft Corporation
description: Hardware Abstraction Layer DLL
Bug check description: This bug check indicates that a fatal hardware error has occurred. This bug check uses the error data that is provided by the Windows Hardware Error Architecture (WHEA).
This is likely to be caused by a hardware problem problem. This problem might be caused by a thermal issue.
The crash took place in a standard Microsoft module. Your system configuration may be incorrect. Possibly this problem is caused by another driver on your system that cannot be identified at this time.

On Sun 27/12/2015 13:24:41 GMT your computer crashed
crash dump file: C:\Windows\Minidump\122715-9828-01.dmp
This was probably caused by the following module: ntkrnlmp.exe (nt!KeBugCheckEx+0x0)
Bugcheck code: 0xA (0x8, 0x2, 0x0, 0xFFFFF80345ED995F)
Error: IRQL_NOT_LESS_OR_EQUAL
Bug check description: This indicates that Microsoft Windows or a kernel-mode driver accessed paged memory at DISPATCH_LEVEL or above.
This appears to be a typical software driver bug and is not likely to be caused by a hardware problem.
The crash took place in the Windows kernel. Possibly this problem is caused by another driver that cannot be identified at this time.

Nickolp1974 · 13 Jan 2016 at 08:11

So mainly 124 bsods, tend to be vcore related, what i would do. Find power options for your board and set LLC or whatever its called on your board to around 75%, this maybe not hugely important at this stage, it may even be worth setting to max just to test then drop it back after, your voltage may be drooping to much.

Find out your cache volts in the monitor section of the bios and add a few clicks more manually, so if its 1.15v put it at 1.2v and test, you will be fine to around 1.3-1.35v but shouldnt need nowhere near that. Increasing cache volts can increase stability to your main cpu clock.
If the above fails then add more vcore or drop your clocks, it does appear sadly that you didn't do great in the silicone lottery.

Still do check your memory though, check under XMP, if it changes to 125 strap set back to 100 and set mem freq to 3000, 1.35v(ram) and set SA too 1.1v

killswitch73 · 13 Jan 2016 at 09:57

Nickolp1974 said:
So mainly 124 bsods, tend to be vcore related, what i would do. Find power options for your board and set LLC or whatever its called on your board to around 75%, this maybe not hugely important at this stage, it may even be worth setting to max just to test then drop it back after, your voltage may be drooping to much.

Find out your cache volts in the monitor section of the bios and add a few clicks more manually, so if its 1.15v put it at 1.2v and test, you will be fine to around 1.3-1.35v but shouldnt need nowhere near that. Increasing cache volts can increase stability to your main cpu clock.
If the above fails then add more vcore or drop your clocks, it does appear sadly that you didn't do great in the silicone lottery.

Still do check your memory though, check under XMP, if it changes to 125 strap set back to 100 and set mem freq to 3000, 1.35v(ram) and set SA too 1.1v

Is that the CPU ring voltage the cache voltage ?

This is what my layout looks like .

http://www.guru3d.com/articles-pages/msi-x99a-godlike-motherboard-review,22.html

So we have a better understanding of my board and that I don't fry anything .

Can do all testing later this afternoon and will also do the strap back to 100 .

Also if I remember correctly my memory frequency I can only choose 2666 or 3200 if I have a strap of 100 ..strange there there is no 3000 freq .

faxfan · 13 Jan 2016 at 10:02

I doubt it that's power, I have an Asus X99 deluxe with the following -

Corsair AX 860
5930K @ 125Blck x 34 - voltage 1.248 (auto)
64GB 2666 @ 2800*
Titan X EVGA SC @ 1360MHz

I found the memory caused the most issues when overclocking when set to 2666 (the official rate) it caused really stability problems.