Print Page | Close Window

DDR4 read speed is halfed

Printed From: ASRock.com
Category: Technical Support
Forum Name: Intel Motherboards
Forum Description: Question about ASRock Intel Motherboards
URL: https://forum.asrock.com/forum_posts.asp?TID=3413
Printed Date: 18 Jul 2025 at 1:09pm
Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com


Topic: DDR4 read speed is halfed
Posted By: Zange
Subject: DDR4 read speed is halfed
Date Posted: 10 Sep 2016 at 7:11pm
Hello, i have 1 drr4 Kingston 8 GB DDR4 2666 MHz HyperX FURY (HX426C15FB/8), and in aida mem test i got 10Gb\s read speed, i thought i must get around 20Gb\s, am i right? or i doing something wrong?
Here is my rig and my results




Replies:
Posted By: Xaltar
Date Posted: 10 Sep 2016 at 7:48pm
Difficult to say, you are running in single channel mode so I am not sure just how much that impacts performance. Assuming 100% scaling you should in theory be about 20GB/s with a single DDR4 2666 module but that is assuming a lot.

How do you have the RAM set up in your BIOS, XMP or manual settings?


-------------


Posted By: Zange
Date Posted: 10 Sep 2016 at 8:09pm
i figure it out, after forum search i noticed that someone experienced the same problem and recommend to downgrade bios, after i downgraded to 2.00 i got memory on a full speed, any chance to have an updated bios with my ram working normally?




Posted By: wardog
Date Posted: 11 Sep 2016 at 2:30am
Originally posted by Zange Zange wrote:

i figure it out, after forum search i noticed that someone experienced the same problem and recommend to downgrade bios, after i downgraded to 2.00 i got memory on a full speed,..


Link to this discovered Thread/post please


Posted By: wardog
Date Posted: 11 Sep 2016 at 2:37am
The pics show memory timing differences and a diff in multipliers

I'm about to call foul ...........


Posted By: clubfoot
Date Posted: 11 Sep 2016 at 3:28am
Good catch, the cpu cache latency is also funky...the previous updated UEFI was really bad! It is still low,...why run single channel on a dual channel board? :(


-------------
https://valid.x86.fr/1tkblf" rel="nofollow">


Posted By: wardog
Date Posted: 11 Sep 2016 at 6:49am
Originally posted by clubfoot clubfoot wrote:

,...why run single channel on a dual channel board? :(


Some people that's what their budget allow(s)(ed). I try not to make observations like that in a post for exactly that reason. None of my business, really.


Posted By: Zange
Date Posted: 11 Sep 2016 at 4:15pm
Originally posted by wardog wardog wrote:

Originally posted by Zange Zange wrote:

i figure it out, after forum search i noticed that someone experienced the same problem and recommend to downgrade bios, after i downgraded to 2.00 i got memory on a full speed,..


Link to this discovered Thread/post please

http://forum.asrock.com/forum_posts.asp?TID=1943&PN=2&title=z170-ddr4-2666-low-memory-bandwidth-scores

Originally posted by wardog wardog wrote:

The pics show memory timing differences and a diff in multipliers

I'm about to call foul ...........

i tested ram with various settings, but anyway no luck, but i didnt noticed the multiplier but isnt it a turbo boost?

Originally posted by clubfoot clubfoot wrote:

Good catch, the cpu cache latency is also funky...the previous updated UEFI was really bad! It is still low,...why run single channel on a dual channel board? :(

what with l3 cache? looks normal to me, especially when a ram on a full speed


Posted By: Xaltar
Date Posted: 11 Sep 2016 at 10:35pm
Yes, the different multi is caused by intel's turbo boost. Those numbers look fine now Thumbs Up

I see a lot of people concerned about having the latest BIOS, this isn't always best. In many cases the latest BIOS will not benefit you at all and in some, like yours, it will even hamper performance with a particular hardware config. This is most often due to compatibility updates (RAM, SSDs etc) while your particular RAM (in this case) worked fine on a previous BIOS it ends up not working correctly on the newest one due to changes made to support newer models etc.




-------------


Posted By: Zange
Date Posted: 11 Sep 2016 at 10:47pm
Originally posted by Xaltar Xaltar wrote:

Yes, the different multi is caused by intel's turbo boost. Those numbers look fine now Thumbs Up

I see a lot of people concerned about having the latest BIOS, this isn't always best. In many cases the latest BIOS will not benefit you at all and in some, like yours, it will even hamper performance with a particular hardware config. This is most often due to compatibility updates (RAM, SSDs etc) while your particular RAM (in this case) worked fine on a previous BIOS it ends up not working correctly on the newest one due to changes made to support newer models etc.



is this some kind of a strange politics?, dont you think? i bought my components 7 days ago, and had a headache with the ram speed and etc, the main problem for me is that i cant OC my i5 6500 bcs bios ver. which allows me to do this is ver. 3.00 


Posted By: parsec
Date Posted: 12 Sep 2016 at 1:31pm
Originally posted by Zange Zange wrote:

Originally posted by Xaltar Xaltar wrote:

Yes, the different multi is caused by intel's turbo boost. Those numbers look fine now Thumbs Up

I see a lot of people concerned about having the latest BIOS, this isn't always best. In many cases the latest BIOS will not benefit you at all and in some, like yours, it will even hamper performance with a particular hardware config. This is most often due to compatibility updates (RAM, SSDs etc) while your particular RAM (in this case) worked fine on a previous BIOS it ends up not working correctly on the newest one due to changes made to support newer models etc.



is this some kind of a strange politics?, dont you think? i bought my components 7 days ago, and had a headache with the ram speed and etc, the main problem for me is that i cant OC my i5 6500 bcs bios ver. which allows me to do this is ver. 3.00 


The Sky OC feature is gone, never to return. Intel forced ASRock to remove it, and add BIOS updates that will never allow it to work again. That happened in the early Spring of this year, so is nothing new. If you installed any BIOS version starting at 3.10, Sky OC will not work, even if you had a copy of the 3.00 BIOS. Intel also got Microsoft to issue Windows updates that stops Sky OC from working. You would need a new BIOS chip, with the 3.00 BIOS version, and filter every Windows update to keep the Sky OC killer out.

Sky OC was not perfect. You could not use the Intel graphics with Sky OC. You could not get an accurate VCore reading with Sky OC. It was not like the original system clock over clocking that was discovered by PC enthusiasts, and then designed out of our control by Intel, starting with the Sandy Bridge generation.

We will never again see the type of over clocking that was done with the system/FSB/BCLK clocks from Intel. The processors with that type of architecture are long gone, and now Intel designs the processors to be unable to over clock with the processor system clock. Except those processors that are allowed to be over clocked, the 'K' models, and only on Skylake boards. But it still has some of the drawbacks that Sky OC has.

DDR4 memory is a mess IMO. Over clocking DDR4 is getting worse, not better, given what I experienced with my X99 system. Or I should say, it seems when a new UEFI version improves over clocking for some models of DDR4 memory, other models of DDR4 can't overclock the same anymore.

I have one UEFI version for my X99 Extreme6/3.1 board that let me operate my Crucial Elite DDR4 memory at 2666, 11 10 10 28 1T (that is NOT a typo, 11 10 10 28 1T is correct!) I found a review of this memory that used those settings in their OC test. The good news is I paid 1/4 the price for this memory recently, than when it was first introduced.

Then I updated to a new UEFI version, reset my memory OC, and... no POST. After trying to get it working, I gave up, and also had a memory OC failure corruption of my Win 10 installation. I was able to recover with a Windows Restore point, but I don't trust it, and must install Win 10 again.

Is it strange politics to describe how a new UEFI version may not be better than an earlier one in some way, with certain hardware? No, it is simply the truth. To expect "new" to always be "better", indicates a lack of understanding of how software works, and how difficult it is to get working. I cannot count how many times I've seen the new version of software that was worse, or had new bugs added that the old version did not have.

DDR4 is new, or really immature, only X99 and Skylake PC systems can use it. Those two systems use different processor generations, and need different VCCSA and VCCIO voltages to make their IMCs happy. Memory manufactures don't want to have memory models that can be used on only one system, but can one DDR4 memory model be optimized for both systems? I doubt it.

I pity the UEFI/BIOS programmers that must try and deal with this. Fix one memory model's OC, and you break another. Notice that the DDR4 memory OC preset profiles in the UEFI only work with certain types of memory chips. Why does Intel only guarantee the DDR4 memory speed on Skylake CPUs as 2133? They don't want to deal with it running faster!

I hope AMD is doing their homework on DDR4 and their Zen processor memory controllers.

BTW, whom checks the memory benchmarks in AIDA64, or any other program for accuracy?  How do you check them for accuracy? I never get the same results twice.  That's because this is NOT simple, but in our ignorance of the technical realities, we think it is simple.


-------------
http://valid.x86.fr/48rujh" rel="nofollow">



Print Page | Close Window

Forum Software by Web Wiz Forums® version 12.04 - http://www.webwizforums.com
Copyright ©2001-2021 Web Wiz Ltd. - https://www.webwiz.net