[SOLVED] Gigabyte P45 Overclocking

prince_xaine

Commendable
Feb 3, 2018
948
20
1,465
109
Specifications:

Motherboard: Gigabyte GA-EP45-UD3R
Processor: Intel Xeon X5470 @4.00 GHz
Memory: 16 GB DDR2 @830 MHz
Graphics: Gigabyte Nvidia Geforce GTX 1060
Storage: Samsung 850 Evo 500GB
Storage: Western Digital Blue 500GB
Power: Rosewill Glacier 700W 80+ Bronze


More details:
Load-Line Calibration: ON
Virtualization: OFF
Spread Spectrum: OFF
CPU Voltage: 1.38v
RAM Voltage: 1.80v
FSB: 400 MHz
CPU Mult: 10x
CPU Fine Control: OFF

I have also turned off just about any other cpu setting that I deemed unnecessary, such as thermal throttling, stepping, etc.

The issue:
No matter what settings I place above, it is not stable. I can set the CPU to a 333 FSB, with just about any configuration of memory, including different on-hand sticks of memory and it still crashes in games. I was able to run Prime95 small FFTs for several hours without issue, with a maximum temperature of about 70C. As for the memory, I ran memtest86 for a few hours overnight and was also met with no errors. The memory isn't even overclocked as it is rated for 800 MHz. I adjusted the timings and tried increasing the voltage but the issue persisted.

I have tried following old threads and adjusting some of the more 'discrete' voltage controls, but they either make no difference or make the situation worse. Again, I must stress that the system crashes whether overclocked or not, and I have tried things on the software end as well, such as Windows 7 and Windows 10, removing Page File, Updating drivers, microcode, etc. I have tried using a different CPU (As the X5470 is obviously outside of spec) but the issue persisted. The memory is configured to be directly double the speed of the FSB, which is ideal based on old threads, but any setting I place here again still causes crashes. I have also disabled all motherboard features I do not use.

I have tried each individual slot of memory, I have tried resetting the CMOS and just about every other 'simple' fix that you could think of. I don't need a common solution here. I have tried and failed with those. I am looking for an outside the box solution, but you can run simple things by me just in case I missed something (or forgot to put it here).

When it doesn't crash, it runs perfectly. I get over 100 FPS in DOOM at ultra settings. This PC is a monster for it's age. But I can't figure out what might be causing the instability. I must also note the instability usually only occurs in games. Yes this points to drivers, rogue software conflicts, missing updates, etc. But all of these things have been met on a clean install of Windows. Drivers installed directly from manufacturer's webpage.

I am out of ideas. Would also like to point out that this machine crashed whilst editing this post with a memory_management BSOD, again, the memory isn't overclocked.
 
Last edited:
Just to add before, replace thermal compund beneath the MCH as they are "old"
Mch core voltage max is 1.36/1.38 as Ive used. I would not suggest to use 2.4-2.6V on ram without heatsink.
My cpu voltage was high just for benchmark test not for long operation.
Im just not sure which voltage for fsb is max safe but you can google them.
To hit 1:1 ratio, dor instance ram speed is 900Mhz, your fsb has to be 455mhz.
Sorry for spamming this post, had to edit and add things between just to make sure i do not lose typing progress as it happens to my phone it refreshes and i lose it.


I would suggest you to lower the ram divider to the lowest possible ratio, then as you increase the fsb you have to lower the multiplier for the cpu to get higher fsb.

The thing you aim for is ram divider to hit 1:1 ratio, soo no cpu or ram lags behind each other. That would be 800Mhz in your instance, soo you can start off with 533Mhz or 400.
As you increase fsb you will increase your ram speed and hit target of 800Mhz sooner or later.

As for load line calibration at high overclocks you wont need it as you need the voltage drop.

You need to increase mch (aka north bridge) voltage, Ram voltage max safe is 2.1V,
Play around, motherboard has indicator of the sade voltage as indicating from blue to yellow and final red.

If i find a guide that ive used for my Q6600 3.6ghz at 450Fsb with stock cooler, OCZ 1066->1200Mhz 5-5-5-18-1T (first 3 are sum of the 4th fyi, or 6-6-6-18 1T which are a bit more relaxed timings), with UD3R-P45.






 
Last edited:

prince_xaine

Commendable
Feb 3, 2018
948
20
1,465
109
Do you have a guide I can follow to replace the MCH thermal compound? Also does it matter what kind of paste I use? I have some Thermal Grizzly, Arctic Silver 5 and also some Noctua paste.
 
Last edited:

prince_xaine

Commendable
Feb 3, 2018
948
20
1,465
109
Okay so I applied new paste to the MCH and overclocked it to 1.36v, my memory is at 1.900v, my cpu is at 1.380v and I am still getting crashes. I also ran memtest86 for 24 hours with no errors, and prime95 for 8 hours with no errors. I am at a loss. lol
 

prince_xaine

Commendable
Feb 3, 2018
948
20
1,465
109
So just an update...I thoroughly cleaned the motherboard, removed everything from the chassis and started from scratch (even removed a USB 3.0 Header and replaced it with a USB 2.0 Header - just because the motherboard can't use it), then reassembled and now it seems to be working fine. Was just averaging above 100 FPS in DOOM, ran a stable 60 FPS in Naruto Shippuden: Ultimate Ninja Storm 4 and Dark Souls 3. All of these games were played at ultra settings. This machine doesn't feel that old anymore. The only thing I can say is one of the fans appear to be making noise, and I don't think the Noctua NH-D14 CPU Cooler can keep up with the heat of the processor, though maybe this could be due to overall poor airflow in the case (it does have cheap fans after all.)
 
Reactions: RememberThe5th

ASK THE COMMUNITY