ASRock.com Homepage
Forum Home Forum Home > Technical Support > AMD Motherboards
  New Posts New Posts RSS Feed - PROBLEM OF X370 KILLER WITH 3200Mhz RAM
  FAQ FAQ  Forum Search Search  Events   Register Register  Login Login

PROBLEM OF X370 KILLER WITH 3200Mhz RAM
 Post Reply Post Reply
Author
Message Reverse Sort Order
  Topic Search Topic Search  Topic Options Topic Options
jdhardware View Drop Down
Newbie
Newbie
Avatar

Joined: 22 Mar 2017
Status: Offline
Points: 5 		Post Options Post Options   Thanks (0) Thanks(0)   Quote jdhardware Quote  Post ReplyReply Direct Link To This Post Topic: PROBLEM OF X370 KILLER WITH 3200Mhz RAM
    Posted: 01 Apr 2017 at 10:09pm
Actually it seems that ram speed is pretty damn important for ryzen in gaming because of its 2 module design the connection between both modules (4 cores each) runs at half the ram speed. So you really want to be running 2933 or faster for gaming with larger gains that would expect at 3600mhz. Even if you have to loosen your timings to get there.

it also seems that unless you have samsung memory b die, you won't hit that 2933 speed or above no matter what bios you use, at least not stable. 

The lower score in the benchmark at the same speed can indicate memory instability.

If both boards are x370, same speed and timings the score should be closer than within 15%. I would prime 95/ cinebench test it for an hour and see if it passes those to check stability.

Sometime a little bit more memory voltage can fix it.  1.4volts is ok for DDR4.
Back to Top
Xaltar View Drop Down
Moderator Group
Moderator Group
Avatar

Joined: 16 May 2015
Location: Europe
Status: Offline
Points: 25028 		Post Options Post Options   Thanks (0) Thanks(0)   Quote Xaltar Quote  Post ReplyReply Direct Link To This Post Posted: 31 Mar 2017 at 1:50am
Happy to help Wink

Enjoy your system, Ryzen is looking really good and will keep getting better as the platform matures Thumbs Up
Back to Top
qwased View Drop Down
Newbie
Newbie


Joined: 30 Mar 2017
Status: Offline
Points: 19 		Post Options Post Options   Thanks (0) Thanks(0)   Quote qwased Quote  Post ReplyReply Direct Link To This Post Posted: 31 Mar 2017 at 1:29am
Originally posted by Xaltar Xaltar wrote:

No, what I am saying is that Hynix doesn't do as well in AIDA64 vs Samsung, that does not mean it is slower in real world applications like gaming. AIDA64 tests a very specific workload that isn't fully representative of every possible usage scenario. Bare in mind, Hynix chips could not be rated the same as Samsung ones if they were not performing equally. That particular benchmark favors Samsung chips, just like some games favor Nvida or AMD GPUs.

So basically, ignore AIDA64 and try out some games and other real world applications. Even in AIDA64 your performance isn't bad. You can try tweaking your timings in the UEFI (primary timings is all that is available last I checked with Ryzen). Tighter timings should yield better results in synthetics. 

It isn't uncommon for people to think that benchmarks are a true measure of performance, that is not their purpose. They exist to test a particular workload in exactly the same way across multiple system configurations for comparison purposes. While a RAM kit may perform better in that workload it may well perform worse in another. Hynix and Samsung have different strengths and weaknesses, when all these are averaged out, they perform the same. 

Think of it like 2 drag racing cars:

Car 1 has superior acceleration and takes an early lead but car 2 has a higher top speed and catches up for both cars to finish at the same time. 

This is what I was meaning. 

Got it.Thanks for anwseringBig smile
Back to Top
Xaltar View Drop Down
Moderator Group
Moderator Group
Avatar

Joined: 16 May 2015
Location: Europe
Status: Offline
Points: 25028 		Post Options Post Options   Thanks (0) Thanks(0)   Quote Xaltar Quote  Post ReplyReply Direct Link To This Post Posted: 31 Mar 2017 at 1:27am
No, what I am saying is that Hynix doesn't do as well in AIDA64 vs Samsung, that does not mean it is slower in real world applications like gaming. AIDA64 tests a very specific workload that isn't fully representative of every possible usage scenario. Bare in mind, Hynix chips could not be rated the same as Samsung ones if they were not performing equally. That particular benchmark favors Samsung chips, just like some games favor Nvida or AMD GPUs.

So basically, ignore AIDA64 and try out some games and other real world applications. Even in AIDA64 your performance isn't bad. You can try tweaking your timings in the UEFI (primary timings is all that is available last I checked with Ryzen). Tighter timings should yield better results in synthetics. 

It isn't uncommon for people to think that benchmarks are a true measure of performance, that is not their purpose. They exist to test a particular workload in exactly the same way across multiple system configurations for comparison purposes. While a RAM kit may perform better in that workload it may well perform worse in another. Hynix and Samsung have different strengths and weaknesses, when all these are averaged out, they perform the same. 

Think of it like 2 drag racing cars:

Car 1 has superior acceleration and takes an early lead but car 2 has a higher top speed and catches up for both cars to finish at the same time. 

This is what I was meaning. 
Back to Top
qwased View Drop Down
Newbie
Newbie


Joined: 30 Mar 2017
Status: Offline
Points: 19 		Post Options Post Options   Thanks (0) Thanks(0)   Quote qwased Quote  Post ReplyReply Direct Link To This Post Posted: 31 Mar 2017 at 12:44am
It is my friend's post on chiphell and I ask him what ram does he using.Surprisingly he just use a couple of Asgard 2400mhz stock ram and overclock to 2933 cl18 by himself. So do you think I should try getting a couple of samsung chip ram ? Or maybe it can be solved by updating bios in the future?
Back to Top
Xaltar View Drop Down
Moderator Group
Moderator Group
Avatar

Joined: 16 May 2015
Location: Europe
Status: Offline
Points: 25028 		Post Options Post Options   Thanks (0) Thanks(0)   Quote Xaltar Quote  Post ReplyReply Direct Link To This Post Posted: 31 Mar 2017 at 12:23am
First up, that is from Chip Hell not a "friend" Tongue

Secondly, Aida64 results can vary wildly from one system to another, even with the same settings. It depends on the motherboard and other components. In this instance the Chip Hell screen cap is using different timings on a different board with a different CPU (1700x vs your 1700). We don't know what model RAM it is using or if the individual RAM chips are even the same make. Yours are Hynix, Samsung chips for example tend to be faster than Hynix even at the same settings.

Your results are not bad or wrong is what I am saying here. They are only marginally slower than my DDR4 3600 kit in my Z270 system, again with this particular benchmark results vary quite broadly depending on configuration. Synthetics are just a guide, they often do not reflect actual real world workloads. 
Back to Top
qwased View Drop Down
Newbie
Newbie


Joined: 30 Mar 2017
Status: Offline
Points: 19 		Post Options Post Options   Thanks (0) Thanks(0)   Quote qwased Quote  Post ReplyReply Direct Link To This Post Posted: 30 Mar 2017 at 10:59pm
ASROCK has updated bios of X370 KILLER today, so I updated to 1.93A beta.After that I finally can adjust my ram higher than 2400mhz.But problem still here:I can't set it to run at 3200mhz without bsod

Here is the details of my ram:


When I set it to xmp 3200mhz ,my computer just beep  and restart 5 times and  start up with 2133mhz
Then I set the vddcr_soc voltage to 1.2v and xmp3200 again. I can now log in the windows but bsod in 10 seconds.
So I set it to xmp 2933mhz now and it runs properly. However when I run aida64 ram test,the results are weird:

This is my result : 


But my friend using same 2933mhz ram can get a 15%higher result:


So whats the problem ?
Back to Top
 Post Reply Post Reply
  Share Topic   

Forum Jump Forum Permissions View Drop Down

Forum Software by Web Wiz Forums® version 12.04
Copyright ©2001-2021 Web Wiz Ltd.

This page was generated in 0.688 seconds.