AMD Ryzen Fabric Clock 1467Mhz (DDR4 2933Mhz) vs Fabric Clock 1933Mhz (DDR4 3866Mhz)in couple games

Joined
Feb 23, 2009
Messages
948
EDIT: Jan 22 2020 Added 1920x1080 for those who thought first test was useless at 2560x1440


BF5 and GTAV 1920x1080






Just want to say all results are within margin of error and are the same for both resolutions 1920x1080 and 2560x1440 and will be the same for 98% of the PC games people play.

Want the best out of your system ram,fine tune Ram .I do not play at 1280x720 or 1920x1080. If you have normal gear tune your ram and put cash toward better GPU.

Tested a old game and new game at 1920x1080 and 2560 x1440 on AMD Ryzen 3800X.
The test is to show that tuned timings for your DDR4 Ram is more important than actual fabric clock on AMD Ryzen.

So for the people building a Ryzen system and can google or have a little knowledge, it may be advisable to pick a good value ram and tune up that ram or for the people with DDR4 3000Mhz ram there is no need to waste money buying DDR4 3800Mhz ram,thinking your gaming on AMD Ryzen system will be better.

I just could not be bothered testing a bunch of games ,as I have done this many times and results are always the same.

So I set CL12 DDR4 2933 Mhz vs CL16 DDR4 3866Mhz both have tight ram timings .
The memory clock (mclk), the memory controller clock (uclk), and the infinity fabric clock (fclk) are all synced for best results 1:1:1.
You can also run Ram at 1:1:2 and still have great results.

Of course I do not care what you do with your ram,I am just trying to help out someone who needs the help by showing actual video and how it does work when PC gaming.

Since I am not using a capture card and just Nvidia shadowplay with all task running,outcome may be different than with a capture card.

Computer and video editor used
♦ CPU - AMD 3800X With MasterLiquid Lite ML240L RGB AIO
♦ GPU - Nvidia RTX 2080
♦ RAM - G.Skill Trident Z 16GB DDR4(F4-4000C18D-16GTZ) (2x8)
♦ Mobo - MSI X470 - Gaming Plus(E7B79AMS.AA0-AMD ComboPI1.0.0.3)
♦ SSD - NVME SSD Intel 512GB
♦ DSP - LG 27" 4K UHD 5ms GTG IPS LED FreeSync Gaming Monitor (27UD59P-B.AUS) - Black
♦ PSU - Antec High Current Pro 1200W
VIDEO INFORMATION
► FPS Monitoring : MSI Afterburner/RTSS
► Gameplay Recorder : Nvidia Shadowplay
► Edit Videos : VSDC Free Video Editor http://www.videosoftdev.com/
2560x1440 video only
RDR2
Sleeping dogs 4:20
Timings/Voltages 6:05
 
Last edited:

tangoseal

[H]ardForum Junkie
Joined
Dec 18, 2010
Messages
8,272
For anyone interested.All Timings /Voltages end of video
Tested a old game and new game at 2560 x1440 on AMD Ryzen 3800X.

The test is to show that tuned timings for your DDR4 Ram is more important than actual fabric clock on AMD Ryzen.

I just could not be bothered testing a bunch of games ,as I have done this many times and results are always the same.

So I set CL12 DDR4 2933 Mhz vs CL16 DDR4 3866Mhz both have tight ram timings .

The memory clock (mclk), the memory controller clock (uclk), and the infinity fabric clock (fclk) are all synced for best results.

Want the best out of your system ram,fine tune Ram .
I do not play at 1280x720 or 1920x1080,nor should people with best CPU's and GPU's.If you have normal gear tune ram and put cash toward better GPU.

RDR2
Sleeping dogs 4:20
Timings/Voltages 6:05

Computer used
♦ CPU - AMD 3800X With MasterLiquid Lite ML240L RGB AIO
♦ GPU - Nvidia RTX 2080
♦ RAM - G.Skill Trident Z 16GB DDR4(F4-4000C18D-16GTZ) (2x8)
♦ Mobo - MSI X470 - Gaming Plus(E7B79AMS.AA0-AMD ComboPI1.0.0.3)
♦ SSD - NVME SSD Intel 512GB
♦ DSP - LG 27" 4K UHD 5ms GTG IPS LED FreeSync Gaming Monitor (27UD59P-B.AUS) - Black
♦ PSU - Antec High Current Pro 1200W
I play at 1080p 240hz and 3440x1440 at 100hz. Tuned ram makes a big differnce at 1080p for sure.
 

selarom

n00b
Joined
Jun 12, 2017
Messages
8
Those results are amazing. Good job dude. Have you tried decoupling FCLK and make it run faster while keeping RAM at 2933CL12? or perhaps another speed?
 
Joined
Feb 23, 2009
Messages
948
Thanks not sure if amazing though.
Yes I actually did back in July 2019 for the first time.Same results really but that was on different CPU 3600X. Also tested 2133/2400/2933/3200/3733/4000/4200 Ram all the same results but tested all at 2560x1440,I can add those test below with timings.
Did not test AMD 3800X with 2933Mhz Ram with FCLK at 1933Mhz and UCLK at 1467,not thought to test that because I tested it so many times in the past and it made no real difference.
Also even with the test I posted,some people are so confused and do not the believe the test,that it is confusing to me with some of the stuff people say and come up with is so strange to me.Basically tried to keep these test simple as i could,I suppose.

Ryzen 3600X Tested Tight Ram Timings 2133/2400/2933/3200/3733/4000/4200
♦ 2133Mhz (16GB)CL10-10-10-10-21
♦ 2400mhz (16GB)CL10-11-11-11-21
♦ 2933mhz (16GB)CL12-14-13-13-26
♦ 3200mhz (16GB)CL14-14-14-14-28
♦ 3733mhz (16GB)CL16-17-16-16-34
♦ 4000mhz (16GB)CL16-18-17-17-36 -(memory clock 2000x2/fabric clock 1800x2/memory controller 1000x2)
♦ 4200mhz (16GB)CL16-18-17-17-36 -(memory clock 2100x2/fabric clock 1800x2/memory controller 1050x2)


 
  • Like
Reactions: Mega6
like this

Mega6

2[H]4U
Joined
Aug 13, 2017
Messages
2,745
3733 MHz matching IF speed is most efficient. Nice work, confirming AMDs findings. Not sure it's worth the price of the ram at this point, but definitely something to shoot for.
 
Last edited:

Jamie Marsala

Limp Gawd
Joined
Mar 9, 2016
Messages
191
3733 MHz matching IF speed is most efficient. Nice work, confirming AMDs findings. Not sure it's worth the price of the ram at this point, but definitely something to shoot for.
Just a quick question. But why is 3733Mhz better than say 3800Mhz? I run my Ram at 3800 with IF at 1900. Works fine. I have not tried to run it at 3733 because it is great at 3800. Is 3733 somehow a sweet spot or something?
 

sirmonkey1985

[H]ard|DCer of the Month - July 2010
Joined
Sep 13, 2008
Messages
21,763
Just a quick question. But why is 3733Mhz better than say 3800Mhz? I run my Ram at 3800 with IF at 1900. Works fine. I have not tried to run it at 3733 because it is great at 3800. Is 3733 somehow a sweet spot or something?
less stress on the IMC. running 1900 can potentially limit your max boost which would be the primary reason for it. plus some boards just don't like running at 1900.

Those results are amazing. Good job dude. Have you tried decoupling FCLK and make it run faster while keeping RAM at 2933CL12? or perhaps another speed?
only real hit effected by decoupling is latency if that matters to you..
 
Last edited:

Mega6

2[H]4U
Joined
Aug 13, 2017
Messages
2,745
Just a quick question. But why is 3733Mhz better than say 3800Mhz? I run my Ram at 3800 with IF at 1900. Works fine. I have not tried to run it at 3733 because it is great at 3800. Is 3733 somehow a sweet spot or something?
3733 is exact sync with IF stock speed. this is where IF works best and most efficient.
 
Last edited:
Joined
Feb 23, 2009
Messages
948
3733 is exact sync with IF stock speed. this is where IF works best and most efficient.
So I have the memory clock (mclk), the memory controller clock (uclk), and the infinity fabric clock (fclk) and are all synced for best results 1:1:1. @ 3866Mhz.
Are you suggesting I should lower to 3733Mhz for working better and most efficient or is this just what you read or tested for best results.
4k link for those who can not see screenshot
https://i.postimg.cc/SmczLjBJ/Desktop-Screenshot-2020-01-25-13-26-47-43.png

Desktop-Screenshot-2020-01-25-13-26-47-43.png
 

Mega6

2[H]4U
Joined
Aug 13, 2017
Messages
2,745
So I have the memory clock (mclk), the memory controller clock (uclk), and the infinity fabric clock (fclk) and are all synced for best results 1:1:1. @ 3866Mhz.
Are you suggesting I should lower to 3733Mhz for working better and most efficient or is this just what you read or tested for best results.
4k link for those who can not see screenshot
https://i.postimg.cc/SmczLjBJ/Desktop-Screenshot-2020-01-25-13-26-47-43.png

View attachment 218268
1:1:1 should be the best at the same memory settings. Memory timings come into play here, so that is the trade-off. A better way to put it is 3733 / 1:1:1 is a good place to START. Once you break the fclk:mclk:uclk 1:1:1 performance degrades.
 
Last edited:
Top