Print Page | Close Window

Overcocking the x370 killer SLI. Help!

Printed From: ASRock.com
Category: Technical Support
Forum Name: AMD Motherboards
Forum Description: Question about ASRock AMD motherboards
URL: https://forum.asrock.com/forum_posts.asp?TID=5579
Printed Date: 10 May 2024 at 9:25am
Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com


Topic: Overcocking the x370 killer SLI. Help!
Posted By: Drain_Bamaged
Subject: Overcocking the x370 killer SLI. Help!
Date Posted: 14 Jul 2017 at 4:58am
" rel="nofollow - Hi everyone. 

New here. Please be kind.

I am having trouble with my x370 Killer SLI. RAM is Corsair Vengeance LPX 3000mhz. It seems I can overclock the CPU ( 1600) no problem IF I leave RAM at default ( 2133mhz). But I hit a wall of 3.7ghz with RAM at 2800mhz.

I can get 3.9 with modest (fixed) voltage with RAM at 2133. But if I try for such overclocks and so much as touch the ram it won't get into Windows and just continually reboots.
I have newest BIOS and Windows is updated.

Can anyone assist with this RAM issue? How do you even adjust timings? On the OCtweaking page of bios, it will only "load xmp" profiles, and ignores changes in timing value. On the other page called "ram overclocking" or something similar, I'm not sure what to make of it. What would a value of 1400mhz represent? 
How do I change timings if this is the issue? How do I overclock the CPU and use faster memory speed?

Thank you.



Replies:
Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 6:45am
Ok scratch my comment about the ram overclocking page. It's speaking in double data rate language. 1500mhz is 3000 mhz. This is the page to tweak timings. Not that this helps me. The same behaviour is displayed. In fact my PC jsut failed to boot at 2800mhz ram...loaded into Windows, looked at CPUZ and it's 2133! This platform is not ready man. It's so immature and flaky.

I am getting absolutely astonishingly awful Realbench scores and overclocking is really tricky. I need to learn how to do Pstate overclocking and I don't understand why cpu clock speed and ram speed interfere with each other.


Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 8:51am
" rel="nofollow - No one around?


Posted By: AlbinoRhino
Date Posted: 14 Jul 2017 at 9:55am
" rel="nofollow -
Originally posted by Drain_Bamaged Drain_Bamaged wrote:

[URL=][/URL]No one around?


Try 2933 or 3066.

3000 gives Ryzen heartburn.


Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 9:57am
I understand that. In fact 2933 is the only closest speed available. But it just boot loops if I select 2933 and save.

I had 2800 but lost it being stable somehow ( every attempt at 2800 is a boot loop now).

The weirdest part for me is I cant over clock the CPU and have Ram anywhere near it's rated speed. It's like as soon as I try overclocking, the flaky ass memory controller throws a fit.


Posted By: wardog
Date Posted: 14 Jul 2017 at 10:03am
I still haven't seen a request for the mem Kit P/N.

Please go to the below link and install the Freeware version of Thaiphoon Burner. Then post here the capture generated from the Read function.

http://www.softnology.biz/files.html" rel="nofollow - http://www.softnology.biz/files.html


Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 10:11am
Memory Module
Manufacturer
Corsair
Part Number
CMK16GX4M2B3000C15
Serial Number
00000000h
JEDEC DIMM Label
8GB 1Rx8 PC4-2133P-UA0-10
Architecture
DDR4 SDRAM UDIMM
Speed Grade
DDR4-2133P downbin
Capacity
8 GB (8 components)
Organization
1024M x64 (1 rank)
Register Manufacturer
N/A
Register Model
N/A
Manufacturing Date
Undefined
Manufacturing Location
Taiwan
Revision / Raw Card
0000h / A0 (8 layers)
DRAM Components
Manufacturer
Hynix
Part Number
H5AN8G8NAFR-TFC
Package
Standard Monolithic 78-ball FBGA
Die Density / Count
8 Gb A-die (21 nm) / 1 die
Composition
1024M x8 (64M x8 x 16 banks)
Clock Frequency
1067 MHz (0.938 ns)
Minimum Timing Delays
15-15-15-36-50
Read Latencies Supported
16T, 15T, 14T, 13T, 12T, 11T, 10T
Supply Voltage
1.20 V
XMP Certified
1499 MHz / 15-17-17-35-52 / 1.35 V
XMP Extreme
Not programmed
SPD Revision
1.0 / January 2014
XMP Revision
2.0 / December 2013
Frequency CAS RCD RP RAS RC FAW RRDS RRDL CCDL
1067 MHz 16 15 15 36 50 23 4 6 6
1067 MHz 15 15 15 36 50 23 4 6 6
933 MHz 14 13 14 31 44 20 4 5 5
933 MHz 13 13 14 31 44 20 4 5 5
800 MHz 12 11 12 27 38 17 3 5 5
800 MHz 11 11 12 27 38 17 3 5 5
667 MHz 10 9 10 22 31 14 3 4 4
667 MHz 9 9 10 22 31 14 3 4 4
Frequency CAS RCD RP RAS RC FAW RRDS RRDL
1499 MHz 15 17 17 35 52 33 6 8


Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 10:24am
Got it!!

At least the RAM instability /clock speed issue. I changed procODT to 48. With this change I got 2933!

Now I need to learn how to do Pstate overclocking. I really don't understand it but all I know is I dont want to run my CPU at a fixed voltage.


Posted By: wardog
Date Posted: 14 Jul 2017 at 11:58am
Originally posted by Drain_Bamaged Drain_Bamaged wrote:

Got it!!

At least the RAM instability /clock speed issue. I changed procODT to 48. With this change I got 2933!

Now I need to learn how to do Pstate overclocking. I really don't understand it but all I know is I dont want to run my CPU at a fixed voltage.


Now change to my recommended procODT of 53.3 if you haven't yet.

Going "lower' on the procODT value is defeating what you are attempting to do there.

Lover values entered there only serve to further limit the mem. ie: akin to lowering CAS/CL


Posted By: wardog
Date Posted: 14 Jul 2017 at 12:00pm
Originally posted by wardog wardog wrote:

ie: akin to lowering CAS/CL


Clarification

As in lowering CL will lower your odds at attaining to obtain their rated speed.


Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 12:02pm
" rel="nofollow -
Originally posted by wardog wardog wrote:

Originally posted by Drain_Bamaged Drain_Bamaged wrote:

Got it!!

At least the RAM instability /clock speed issue. I changed procODT to 48. With this change I got 2933!

Now I need to learn how to do Pstate overclocking. I really don't understand it but all I know is I dont want to run my CPU at a fixed voltage.


Now change to my recommended procODT of 53.3 if you haven't yet.

Going "lower' on the procODT value is defeating what you are attempting to do there.

Lover values entered there only serve to further limit the mem. ie: akin to lowering CAS/CL
way ahead of ya XD. Learned 53.3 is default and AMD recommends 50-60 so I had already gone ahead and did that.

Thank you very much for commenting.

Now...are you able to lend a hand with overclocking? I know my 1600 can do 3.8 at 1.35v. I would like to use Pstate overclocking. What is the process? Do I use off set vcore? if so, do I take default vcore ( something like 1.24v) and enter a offset that would make it 1.35v total? What about the Pstates page?

My tries at overclocking left me wondering why clock speed stays at max and never goes down at idle ( which I believe is where Pstates come in?

Thank you Wardog.


Posted By: wardog
Date Posted: 14 Jul 2017 at 12:08pm
" rel="nofollow - DB

Have you tried bumping SoC just a pinch to 1.1 or 1.175, DRAM Voltage up an additional +0.025 to 1.375, procODT to 53.3, and surely MEM_VTT to 0.690 or maybe even 0.700 yet?

That advice there of mine has helped many Users. That knowledge is gleaned from days and days of mem testing here the 6 Kits of DDR4 I have.

Also, consider that a "starting point". You may very well want to lower those v's as testing there shows/allows for..





Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 12:10pm
" rel="nofollow - Yes, I have come across that info. SOC is at 1.1. As far as Dram voltage, I (believe?) my RAM instability /unwillingness to clock at 2933 was due to procODT as 1.35 seems to be fine.
VTT is half of Dram voltage automatically. Not sure what benefit raising that offers? I know NOTHING about VTT though.


Posted By: wardog
Date Posted: 14 Jul 2017 at 12:18pm
p=state OC'ing is hampered in BIOSes containing AGESA 1.0.0.6.

It's not th AGESA yet rather the internals of BIOSes that have .6. That's the reference point I'm using.

Tonight ASRock is beginning to release new BIOSes with AGESA 1.0.0.6a.

When 3.00 is released ror your board, remember FIRST to Reset to UEFI Deaults, Save and Exit, THEN and ONLY then flash as this is major BIOS re-write/update.


Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 12:19pm
Originally posted by wardog wardog wrote:

p=state OC'ing is hampered in BIOSes containing AGESA 1.0.0.6.

It's not th AGESA yet rather the internals of BIOSes that have .6. That's the reference point I'm using.

Tonight ASRock is beginning to release new BIOSes with AGESA 1.0.0.6a.

When 3.00 is released ror your board, remember FIRST to Reset to UEFI Deaults, Save and Exit, THEN and ONLY then flash as this is major BIOS re-write/update.

Thank you. Very good to know.


Posted By: wardog
Date Posted: 14 Jul 2017 at 12:19pm
" rel="nofollow - VTT is the memory controller(IMC) voltage. AMDs past(??) weak link.


Posted By: wardog
Date Posted: 14 Jul 2017 at 12:22pm
And FYI. There is actually no "Default' procODT value as it is derived from many variables present.


Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 12:35pm
" rel="nofollow - Ugh. Why is it sitting at 1500mhz? I have 1.36v atm with level 3 LLC ( 1.35vcore in bios), and 3800mhz is set in bios....

Coolnquiet is off, global cstates are off. What the heck causes this?


Posted By: wardog
Date Posted: 14 Jul 2017 at 12:43pm
" rel="nofollow - Where are you seeing this 1500?

Are you sure that 1500 doesn't represent 3000 in DDR speak?


Posted By: wardog
Date Posted: 14 Jul 2017 at 12:45pm
I've NEVER seen mem on Ryzen post at an actual 1500.


Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 12:51pm
I meant CPU clock speed. was a 1500mhz. Solved by using offset vcore ( a reply on reddit told me there is a bug with using fixed vcore and offset fixes it. I scored like....420 in Cinebench CPU test with this "bug".

Even with out this "bug" I havnt solved miserable benchmark scores. at 3.8ghz and 2900mhz ram I'm scoring 90,000 system score in Realbench. 1255 Cinebench CPU score.

I just don't get it lol. 


Posted By: Drain_Bamaged
Date Posted: 14 Jul 2017 at 12:55pm
Also, totally incorrect sensor readings.  example: Realbench shows 20. some odd for "CPU usage" and zero for temperature. HWinfo64 shows 0.994 for VID, etc. A lot of incorrect numbers on a lot of software.

I still say Ryzen wasn't ready.



Posted By: wardog
Date Posted: 14 Jul 2017 at 1:16pm
Make damn sure you are using the latest versions available of monitoring apps.

Many had issue with proper reporting concerning Ryzen in previous revisions.



Too. ONLY use one monitoring app at a time.

The chip responsible does NOT at all like more than one app accessing it. Been that way for decades now.


Posted By: Drain_Bamaged
Date Posted: 15 Jul 2017 at 1:25am
" rel="nofollow - Well, 3.9ghz "stable" ( can't 100% confirm yet but appears so) @ 1.375v.

I am watching
https://www.youtube.com/watch?v=52Tw-wcT7o4
For Pstate overclocking as well as trying to understand
https://hardforum.com/threads/ryzen-pstate-overclocking-method-calculation-and-calculator.1928648/  's spread sheet. The Hardcopy guide is easy to follow but then loses me at Vcore. "Change your offset voltage by adding or subtracting"

adding or subtracting....what?

Also, this article contradicts the video. In the video he says put it all on auto, and then you go input numbers for Pstate which will controll Vcore I presume. The article says  to " change your offset" ( in a very vague manner I might add) and shows pictures doing it.

So, A little confused.
My default vcore is 1.2375. so offset would be + 0.14. But do I leave vcore on auto...or use offset? Contradicting info dang it.
I tried following the video. at idle, CPU is at 1500mhz, doesnt change under load ( this is bthe bug again with using fixed vcore)and  vcore is at a static 1.375...so much for Pstates. Also there is no option for minimum proc frequency in the AMD balanced power plan.

Edit: I learn AMD disables Pstates when you overclock. WTF is this? No really....WTF is this? That's absurd.




Posted By: Drain_Bamaged
Date Posted: 15 Jul 2017 at 2:34am
" rel="nofollow - No really, my mind is blown right now. AMD disables Pstates when you overclock? What kind of goofy crap is that? 


Posted By: Drain_Bamaged
Date Posted: 15 Jul 2017 at 4:46am
OK scratch that, all is well, sorted it out. 
1.375v, 3900mhz for load, dropping to between 1.225 and 1.237 and 1500mhz to 2400mhz-ish across the cores now as I type this. I followed
https://linustechtips.com/main/topic/769094-guide-to-p-state-variable-frequency-overclocking-on-the-crosshair-vi-hero/" rel="nofollow - https://linustechtips.com/main/topic/769094-guide-to-p-state-variable-frequency-overclocking-on-the-crosshair-vi-hero/
and ONLY set Psate zero., went back to OCTweak page and set my offset vcore. Booted back into windows and AMD power plan had the option to set min proc frequency.

I've now done everything I wanted to do with this chip and board! Naturally I wonder if it can do 4ghz but I don't see a point. It will bring with it a need for a lot more vcore and heat.

Thank you again Wardog.


Posted By: AlbinoRhino
Date Posted: 15 Jul 2017 at 8:57am
" rel="nofollow - What LLC level are you using?


Posted By: Drain_Bamaged
Date Posted: 15 Jul 2017 at 9:24am
lvl 1 for SOC. Havnt decided on 2 or 3 for Vcore. Feel free to tell me otherwise.


Posted By: Drain_Bamaged
Date Posted: 02 Aug 2017 at 6:52pm
I ditched the crappy Corsair LPX and got some Gskill FlareX. Leaving CPU at 1.3v, 3800mhz for now. I want to dial in a Pstate overclock.


Posted By: Pencha
Date Posted: 04 Aug 2017 at 2:43am
Same here. I cant get stable 3.9 because mbo has big v drop. If i start playing with load line calibration multiplier get stuck at 15.5 (1547MHz).
Asrock fix this please.
I found a lot of buggs on this MBO.


Posted By: Drain_Bamaged
Date Posted: 04 Aug 2017 at 2:45am
LLC should not be causing any sort of multiplier bug... I have a killer SLI and don't experience this. I use level 1 for SOC and 2 for CPU.  The only reason I havnt gone for 3.9ghz is I hesitate as it needs 1.37v or more ( not 100% sure) and only 1.3v for 3.8ghz. I don't think it matters and 3.8 is so easy to attain.

What does cause a multiplier bug is using fixed Vcore.


Posted By: Pencha
Date Posted: 04 Aug 2017 at 6:25am
" rel="nofollow - Any solution for that?


Posted By: Drain_Bamaged
Date Posted: 04 Aug 2017 at 6:26am
Originally posted by Pencha Pencha wrote:

" rel="nofollow - Any solution for that?
Solution for wut? Your issue you blame on LLC? It's not LLC.


Posted By: parsec
Date Posted: 04 Aug 2017 at 8:53am
Originally posted by Pencha Pencha wrote:

" rel="nofollow - Any solution for that?


As Drain_Bamaged said previously in this thread, using Offset Voltage.


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


Posted By: Pencha
Date Posted: 04 Aug 2017 at 3:38pm
Tnx  guyz  ;)



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