©ASUSTEK COMPUTER INC. ALL RIGHTS RESERVED.

Zammin · ‎07-13-2018

Hi everyone.*

My RAM is G.Skill Trident Z RGB 3200Mhz 16GB (2x8GB) CL16-18-18-38.*

I have had it running in XMP for a long time now with only the VCCIO tweaked down to 1.15V and it is proven stable in a variety of stress tests and memtest86.

I am now looking to see if I can get a little more performance out of it now that I am done with my CPU overclock. However I am unclear as to whether I am approaching this the right way.

I don't feel ready to mess with timings as I don't understand how to tune them properly, so I'm bumping up the frequency to 3466mhz as the same model ram sells at this speed with the same timings and voltage, this seems like a realistic goal to begin with. I have selected this frequency under DRAM frequency in the BIOS and was able to boot into windows and complete a memory stress test in AIDA64. I have run the cache/mem benchmark in AIDA64 prior to changing the frequency and afterward. My memory read and write speeds went up by a decent bump but my latency stayed the same at 50ns. Does this sound right? Does the latency stay the same because I've left the timings alone?*

I am currently running memtest86 to test for stability but I guess my question is, am I approaching this correctly by simply bumping up the frequency a little bit while AI overclock tuner is still set to XMP? If everything appears stable then is there anything else I should look at?

I saw that my Maximus X Code has memory presets for a range of different DRAM, but it's listed as the actual memory itself (Samsung B die for example) and not the brand and model of the modules. I can't seem to find which of these presets my kit falls under. If anyone knows I would be willing to try out the corresponding preset. *

Nate152 · ‎07-13-2018

Hi Zammin

Raising the frequency is one way to overclock your ram, tightening the timings is another and setting the command rate to 1N are the three main ways to overclocking your ram.

The fastest 3200MHz Trident Z kits are with timings 14-14-14-34 2N, you could try that and if stable then try command rate 1N.

You may need to raise the Dram Voltage, CPU VCCIO Voltage and CPU System Agent Voltage along the way for stability. There's no guarantee you'll get it stable but it's something to try for, tightening the timings should lower the latency.

Absolute Maximums:

Max Dram voltage - 1.50v
Max CPU VCCIO voltage - 1.30v
Max CPU System Agent voltage - 1.35v

I'd probably shoot for:

Dram voltage - 1.45v
CPU VCCIO voltage - 1.25v
CPU System Agent voltage - 1.25v
Dram Frequency - 3200MHz

Primary Timings as listed at the top of Dram Timing Control:

14
14
34
Command Rate 2N

F10 and Enter to save and Exit...

If stable after testing, try Command rate 1N.

ROG Strix 1070 OC Guide - ROG Spatha X Rapid Fire - ROG Falchion Touch Panel - z270 BCLK Overclocking - Xsplit Gamecaster Guide - ROG RX 460 OC Guide

Zammin · ‎07-13-2018

Nate152 wrote:
Hi Zammin

Raising the frequency is one way to overclock your ram, tightening the timings is another and setting the command rate to 1N are the three main ways to overclocking your ram.

The fastest 3200MHz Trident Z kits are with timings 14-14-14-34 2N, you could try that and if stable then try command rate 1N.

You may need to raise the Dram Voltage, CPU VCCIO Voltage and CPU System Agent Voltage along the way for stability. There's no guarantee you'll get it stable but it's something to try for, tightening the timings should lower the latency.

Absolute Maximums:

Max Dram voltage - 1.50v
Max CPU VCCIO voltage - 1.30v
Max CPU System Agent voltage - 1.35v

I'd probably shoot for:

Dram voltage - 1.45v
CPU VCCIO voltage - 1.25v
CPU System Agent voltage - 1.25v
Dram Frequency - 3200MHz

Primary Timings as listed at the top of Dram Timing Control:

14
14
34
Command Rate 2N

F10 and Enter to save and Exit...

If stable after testing, try Command rate 1N.

Hey Nate, thanks once again for your advice.

Although I didn't mention it, the first thing I tried was entering the timings for the faster 3200Mhz kit into the BIOS while in XMP mode, I didn't change anything else though and the system failed to POST. Perhaps I should try those settings again with the voltages you recommended. I'll see how I go with that.

Silent_Scone · ‎07-13-2018

When making changes to memory parameters, I'd recommend ditching Memtest86 and using HCI Memtest Pro or RAM Test

http://hcidesign.com/memtest/

https://www.karhusoftware.com/ramtest/

RAM Test is far easier to configure, so I'd recommend going with this one. Memtest86 isn't going to tell you anything worth knowing beyond a faulty DIMM on newer DDR4 platforms.

If you need any further help dialing down, come back to the thread. It's good to remember that memory timings and frequency are intrinsically related. Be sure to benchmark changes as well as test for stability systematically.

13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090

Zammin · ‎07-13-2018

Silent Scone wrote:
When making changes to memory parameters, I'd recommend ditching Memtest86 and using HCI Memtest Pro or RAM Test

http://hcidesign.com/memtest/

https://www.karhusoftware.com/ramtest/

RAM Test is far easier to configure, so I'd recommend going with this one. Memtest86 isn't going to tell you anything worth knowing beyond a faulty DIMM on newer DDR4 platforms.

If you need any further help dialing down, come back to the thread. It's good to remember that memory timings and frequency are intrinsically related. Be sure to benchmark changes as well as test for stability systematically.

Thanks for advice! I will give HCI MEM Test and Ram Test a go. I had thought Memtest86 was a good one but it's good to know there are better applications out there. I actually already have HCI Mem Test but I cannot test all of my RAM with it as Windows is always using around 2Gb of my available memory. For benchmarking do you think AIDA64's mem/cache benchmark is sufficient?

Cheers

F0x135 · ‎07-13-2018

Watching this thread. I too have a Trident Z RGB kit but 2400mhz 32GB (8x4). Currently OC to 2666 stable @ 1.20v but want to push higher.

In regards to Memory tests, besides the ones listed above, is Memtest64 reliable? I've been using that, Memtest86 & Prime95 blend to stress test.

Silent_Scone · ‎07-13-2018

F0x135 wrote:
Watching this thread. I too have a Trident Z RGB kit but 2400mhz 32GB (8x4). Currently OC to 2666 stable @ 1.20v but want to push higher.

In regards to Memory tests, besides the ones listed above, is Memtest64 reliable? I've been using that, Memtest86 & Prime95 blend to stress test.

If you're referring to the one released by TPU, then it's not nearly as stringent as the other aforementioned tests. It's sometimes possible to pass an hour of MT64 where HCI or RAMTest will flag an error in minutes. It can depend where the instability lies.

Also, you can try Google Stress App test, a better test of DRAM stability than HCI Memtest, however it will require you to either run Linux Mint, or use Bash Terminal within Windows 10 to install the application.

Google stressapp test via Linux Mint (or another compatible Linux disti) is the best memory
stress test available. Google use this stress test to evaluate memory stability of their servers
– nothing more needs to be said about how valid that makes this as a stress test tool.

Install Linux Mint from here: http://www.linuxmint.com/download.php
Install the Google Stress App test from here: http://community.linuxmint.com/software/view/stressapptest
Once installed open “Terminal” and type the following: stressapptest -W -s 3600
This will run the stressapp for one hour. The test will log any errors as it runs.

Instructions for Bash Shell install: https://www.howtogeek.com/265900/everything-you-can-do-with-windows-10s-new-bash-shell/

Zammin wrote:
Thanks for advice! I will give HCI MEM Test and Ram Test a go. I had thought Memtest86 was a good one but it's good to know there are better applications out there. I actually already have HCI Mem Test but I cannot test all of my RAM with it as Windows is always using around 2Gb of my available memory. For benchmarking do you think AIDA64's mem/cache benchmark is sufficient?
Cheers

HCI needs to be correctly configured. See example below, you are required to setup one instance per CPU thread, dividing the amount of availble memory (leave enough for the OS to breathe or risk testing your page filing). HCI consider 1000% to be the 'golden standard' however for larger densities this can be time consuming. Let the test run for 400% coverage.

AIDA64 Memory Benchmark is as good as any a measure, yes.

13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090

F0x135 · ‎07-14-2018

Silent Scone wrote:
If you're referring to the one released by TPU, then it's not nearly as stringent as the other aforementioned tests. It's sometimes possible to pass an hour of MT64 where HCI or RAMTest will flag an error in minutes. It can depend where the instability lies.

Thanks @Silent Scone. Will go ahead and purchase the HCI Pro then.

New to RAM overclocking, have some questions.