Overclocking ASUS Rampage Extreme X48 and 16GB RAM

[silverspaceship]

Hi.
I didn't want to necropost, so I created a new thread. I have an ASUS Rampage Extreme motherboard and decided to test it with 16GB of RAM. I saw on forums that theoretically it can support 16GB, even though according to the specs it should only support 8 GB. For this, I purchased and installed 4 sticks of Corsair Dominator GT CMT8GX3M2A2000C9, each 4GB, version 5.11 (all identical). After that, the motherboard booted into Windows without any issues, and all 16GB were successfully recognized.

Additionally, I ran tests using HCI Memtest and TestMem5 for several hours to ensure everything was working stably (all tests passed without errors). All of this was done with stock configurations, without any overclocking. However, on the RAM modules I purchased, the rated speed is 2000 MHz with timings 9-10-9-27 at 1.65 V, so I tried to run them using an XMP profile at 2000 MHz. But no matter how hard I tried, I couldn't get 2000 MHz in this configuration; I tried just simply enabling XMP, as well as manually setting the values in the BIOS. I managed to get 500 FSB without any problems, but the memory maxed out at 1800 MHz (if I recall it correctly at 1.9V).

So my question is: what is the problem with not being able to achieve 2000 MHz? Is it an issue with the memory, or is it because all 4 slots are populated, or should I adjust the voltage further? I'm just new to overclocking, so I apologize if this is some kind of dumb question 😅 (so not going to break world records, I just want to figure it out).
CPU – Q9550;
BIOS – Latest;

[TerraRaptor]

Some boards struggle to do 2000 even with 2x2GB. Chipset is limiting factor. Making 4x4GB running at 1800 is already quite strong. Best memory for 4x4 should actually be rev4.13 dominators (d-die based) if you care of performance, yet 2000 is likely impossible to be rock stable 

[silverspaceship]

35 minutes ago, TerraRaptor said:

Some boards struggle to do 2000 even with 2x2GB. Chipset is limiting factor. Making 4x4GB running at 1800 is already quite strong. Best memory for 4x4 should actually be rev4.13 dominators (d-die based) if you care of performance, yet 2000 is likely impossible to be rock stable 

Thanks, I'll try again with just 2 sticks instead of 4, see if I can get 2000 MHz in this configuration. I'll also try to look for ver4.13 Dominators, but I suspect they'll be hard to find, especially 4 sticks.

[silverspaceship]

Recently, I managed to get a QX9770 and decided to play around with it on my Rampage Extreme, and here’s what I noticed. If I run a Prime95 stress test with large FFTs – it fails. Same with the OCCT memory test, it detects errors. The system is running at stock settings, I only set standard (minimal) values for CPU PLL, SB, and disabled DRAM read/write controls and some other default things. All other BIOS settings are on auto.
I also tried different RAM – G.Skill 3x2GB and another set of Corsair Dominators 2x4GB – but the issue still persists, although I’m sure those sticks are in good condition.

Then I tried lowering the FSB clock from the default 400 MHz to 333 MHz, and the problems disappeared. Prime95 and OCCT now pass without issues.

What could be the problem?
The Fujitsu caps are fine – I checked them all.
I tried different DIMM slots on the motherboard (A2, B2) and (A1, B1) – that didn’t help. Though one 4GB stick seems to work fine at 400 FSB.
I also tried different DRAM clocks from the default 1066 MHz up to 1600 MHz – that also doesn’t change anything.

Is this an issue with BIOS settings, is the motherboard dying, or is it something to do with the quad core CPU?

[TheQuentincc]

FYI I tried with my DDR3 2100+ 6-7-6 capable REX (using Elpida Hyper) 2Gbit Micron M-die (4GB/stick, 2200+ 7-7-7 capable) but I wasn't able to clock any higher than 1850 in single channel configuration. I guess it just hate high density IC.

[silverspaceship]

8 minutes ago, TheQuentincc said:

FYI I tried with my DDR3 2100+ 6-7-6 capable REX (using Elpida Hyper) 2Gbit Micron M-die (4GB/stick, 2200+ 7-7-7 capable) but I wasn't able to clock any higher than 1850 in single channel configuration. I guess it just hate high density IC.

What about dual-channel? Is it only going to get worse?
However, I still don't understand why I can't stably pass stress tests even with 1066 MHz at 400 FSB.

[TheQuentincc]

I didn't tried dual channel cause I was aiming DDR3 2000

[TerraRaptor]

2 hours ago, silverspaceship said:

What about dual-channel? Is it only going to get worse?
However, I still don't understand why I can't stably pass stress tests even with 1066 MHz at 400 FSB.

Because nb is at high load. Try bumping nb voltage to 1.45v. Some boards may need to be re-pasted as nb thermal paste is likely dry af.

[silverspaceship]

22 minutes ago, TerraRaptor said:

Because nb is at high load. Try bumping nb voltage to 1.45v. Some boards may need to be re-pasted as nb thermal paste is likely dry af.

Thx. As soon as I received it I replaced all the thermal pads and thermal paste on the motherboard. Although I did not scalp the NB. The NB temperatures are on average 50C and a maximum of 55.
I tried gradually increasing the voltage on the NB up to 1.45V, but it didn't help. Should I increase the voltage further?

[TerraRaptor]

Try bumping vtt to 1.35v

 

Check voltage values in hw monitoring page in bios - are these in line with what was actually set?

[silverspaceship]

21 hours ago, TerraRaptor said:

Try bumping vtt to 1.35v

 

Check voltage values in hw monitoring page in bios - are these in line with what was actually set?

I want to thank you for pointing me in the right direction. I tried setting Vtt to 1.35V and NB to 1.45V, and that helped—Prime95 and OCCT stress tests passed successfully.

After that, I gradually lowered the NB voltage and Vtt. It seems I’ve found a sweet spot for 4x4GB DDR3 at 1066 MHz and a 400 FSB clock. My results are:  
- FSB Termination Voltage (Vtt): 1.28566V  
- North Bridge Voltage: 1.39172V

The BIOS Voltage Monitor shows values slightly lower than those set:  
- FSB Termination Voltage (Vtt): 1.277V  
- North Bridge Voltage: 1.369V

With these settings, all the stress tests continue to pass successfully. I’m not sure if I can go any lower. When I reduce Vtt to 1.272V, the large FFT test in Prime95 stops passing. Also (I’m not sure if this really had any effect), but in the BIOS settings I changed EPU II Phase Control from Auto to Full Phase. It seems that after this adjustment, the motherboard became more stable at lower Vtt and NB settings.

I would also like to ask you, if I want to increase the RAM frequency from 1066 to 1600 at the same 400 FSB, for example, will I need to adjust Vtt and NB again? Or if I have already set them up and it is working stably, then there is no need to touch them?

Edited April 13 by silverspaceship

[TerraRaptor]

You will likely need to adjust these going higher frequency and tighter timings (if minimized for lowest values before). Nb volts mostly depend on ram frequency and timings, vtt volts depend on fsb (though to some extent both of these voltages may affect stability of fsb and ram).