The Corsair User Forums  

Go Back   The Corsair User Forums > Corsair Product Discussion > Memory

Reply
 
Thread Tools Search this Thread Rate Thread Display Modes
  #1  
Old 04-29-2019, 07:39 PM
A Computer Guy A Computer Guy is online now
Registered User
A Computer Guy's PC Specs
 
Join Date: Jan 2019
Posts: 130
POST ID # = 1003018
A Computer Guy Reputation: 10
Default Possibly some good news for Ryzen users...

If you have an ASRock X470 Master SLI/AC you may get improved memory compatibility with the P3.20 version BIOS/UEFI update including AGESA 0.0.7.2.

Running CMK16GX4M2A2400C16 (2 kits) 16-16-16-39-55-1T, 2400MHz @1.2v

Old P1.9 BIOS/UEFI with AGESA 1.0.0.6
I was not able to use the original timings when overclocked and occasionally memory retraining would kick-in again on boot every now and then.
20-20-20-48-68-1T, 2933MHz @1.35v, SOC@0.920v (oc stable, 4 sticks)

New P3.20 BIOS/UEFI with AGESA 0.0.7.2
I was able to use the original timings while overclocked and the memory retraining problem is now gone.
16-16-16-39-55-1T, 2933MHz @1.35v, SOC@0.920v (oc stable, 4 sticks)

Did anyone else notice any memory improvements with their Ryzen boards with AGESA 0.0.7.2 updates?
Reply With Quote


  #2  
Old 04-30-2019, 12:53 PM
Corsair Rick's Avatar
Corsair Rick Corsair Rick is offline
Corsair Employee
Corsair Rick's PC Specs
 
Join Date: Feb 2016
Posts: 274
POST ID # = 1003094
Corsair Rick Reputation: 10
Default

Hello A Computer Guy,

That's great to hear! I'll definitely keep that in mind and mention it when needed. Thanks!
Reply With Quote


  #3  
Old 05-22-2019, 11:33 PM
Gronar Gronar is offline
Registered User
Gronar's PC Specs
 
Join Date: Mar 2019
Location: Round Rock, Tejas
Posts: 105
POST ID # = 1005429
Gronar Reputation: 10
Default

Quote:
Originally Posted by A Computer Guy View Post
If you have an ASRock X470 Master SLI/AC you may get improved memory compatibility with the P3.20 version BIOS/UEFI update including AGESA 0.0.7.2.

Running CMK16GX4M2A2400C16 (2 kits) 16-16-16-39-55-1T, 2400MHz @1.2v

Old P1.9 BIOS/UEFI with AGESA 1.0.0.6
I was not able to use the original timings when overclocked and occasionally memory retraining would kick-in again on boot every now and then.
20-20-20-48-68-1T, 2933MHz @1.35v, SOC@0.920v (oc stable, 4 sticks)

New P3.20 BIOS/UEFI with AGESA 0.0.7.2
I was able to use the original timings while overclocked and the memory retraining problem is now gone.
16-16-16-39-55-1T, 2933MHz @1.35v, SOC@0.920v (oc stable, 4 sticks)

Did anyone else notice any memory improvements with their Ryzen boards with AGESA 0.0.7.2 updates?
I'm still waiting on Asus to post the BIOS that was announced on the day you posted this, 4/29. Still can't get mine stable at 3200 but can run at 3066,
Reply With Quote


  #4  
Old 05-23-2019, 08:52 PM
A Computer Guy A Computer Guy is online now
Registered User
A Computer Guy's PC Specs
 
Join Date: Jan 2019
Posts: 130
POST ID # = 1005524
A Computer Guy Reputation: 10
Default

Quote:
Originally Posted by Gronar View Post
I'm still waiting on Asus to post the BIOS that was announced on the day you posted this, 4/29. Still can't get mine stable at 3200 but can run at 3066,
I think you said in a different post you had all four slots filled?

Frequency becomes more limited when all four slots are filled from the load on the memory controller. It may be very difficult to reach DDR4-3200 in that case. For many the limit is DDR4-2933 when all four slots are filled.
Reply With Quote


  #5  
Old 05-23-2019, 10:19 PM
A Computer Guy A Computer Guy is online now
Registered User
A Computer Guy's PC Specs
 
Join Date: Jan 2019
Posts: 130
POST ID # = 1005534
A Computer Guy Reputation: 10
Default

Quote:
Originally Posted by A Computer Guy View Post
If you have an ASRock X470 Master SLI/AC you may get improved memory compatibility with the P3.20 version BIOS/UEFI update including AGESA 0.0.7.2.

Running CMK16GX4M2A2400C16 (2 kits) 16-16-16-39-55-1T, 2400MHz @1.2v

Old P1.9 BIOS/UEFI with AGESA 1.0.0.6
I was not able to use the original timings when overclocked and occasionally memory retraining would kick-in again on boot every now and then.
20-20-20-48-68-1T, 2933MHz @1.35v, SOC@0.920v (oc stable, 4 sticks)

New P3.20 BIOS/UEFI with AGESA 0.0.7.2
I was able to use the original timings while overclocked and the memory retraining problem is now gone.
16-16-16-39-55-1T, 2933MHz @1.35v, SOC@0.920v (oc stable, 4 sticks)

Did anyone else notice any memory improvements with their Ryzen boards with AGESA 0.0.7.2 updates?

If anyone is interested here are instructions for building a simple spreadsheet to calculate memory timings.

Columns/Formulas:

Column A: "Old Freq."
Column B: "Description"
Column C: "Old Timing"
Column D: "Old NS" (format to 4 decimals)
- formula: "=(1/(A2/2)*C2)*1000"
Column E: "New Freq."
Column F: "Calc. Timing" (format to 4 decimals)
- formula: "=D2*(E2/2)/1000"
Column G: "New Timing" (format to zero decimals)
- formula: "=ROUND(F2, 0)"
Column H: "Try Timing"
Column I: "Try NS"
- formula: "=(1/(E2/2)*H2)*1000"
Column J: "Delta NS"
- formula: "=I2-D2"

Instructions:
- Fill out the columns A, B and C for frequency, description, and timings. (info from your memory specifications)
- The "Old NS" will fill in with timings in nanoseconds.
- You should be able to compare this to your memory specification nanosecond timings.
- Fill out the "New Freq." column with the desired frequency for your overclock.
- The "New Timing" column should approximate timings that you may be able to use at the higher frequency.
- From there you can try reducing the timings and see the the change in nanoseconds.

Sample:

Code:
-----|-------|-----|----------|------|----------|-----|-----|-----------
A    | B     | C   | D        | E    | F        | G   | H   | I
-----|-------|-----|----------|------|----------|-----|-----|-----------
2400 | CAS   |  16 |  13.3333 | 2933 |  19.5533 |  20 | 14  |   9.5465
2400 | RDC   |  16 |  13.3333 | 2933 |  19.5533 |  20 | 17  |  11.5922
2400 | RP    |  16 |  13.3333 | 2933 |  19.5533 |  20 | 17  |  11.5922
2400 | RAS   |  39 |  32.5000 | 2933 |  47.6613 |  48 | 39  |  26.5939
2400 | RC    |  55 |  45.8333 | 2933 |  67.2146 |  67 | 55  |  37.5043
2400 | tRFC1 | 420 | 350.0000 | 2933 | 513.2750 | 513 | 420 | 286.3962

Last edited by A Computer Guy; 05-23-2019 at 10:27 PM.
Reply With Quote


  #6  
Old 05-24-2019, 07:36 AM
GWNewman1 GWNewman1 is offline
Registered User
GWNewman1's PC Specs
 
Join Date: Aug 2018
Posts: 14
POST ID # = 1005568
GWNewman1 Reputation: 10
Default

Quote:
Originally Posted by A Computer Guy View Post
I think you said in a different post you had all four slots filled?

Frequency becomes more limited when all four slots are filled from the load on the memory controller. It may be very difficult to reach DDR4-3200 in that case. For many the limit is DDR4-2933 when all four slots are filled.
Ran into this with my 2700x on my Strix X470 - it was godawful picky on running 32gb 3200 c14 kits with 4x8gb. I went through 3 of 4 kits before I wound up replacing it with a Crosshair VII and it worked like a charm with my Vengeance Pro kit. The Strix would fail memtest left and right with Windows being sketchy under load. RMA'd it back to Asus.
Reply With Quote


  #7  
Old 05-24-2019, 09:13 AM
Baio Baio is offline
Registered User
Baio's PC Specs
 
Join Date: Oct 2014
Posts: 71
POST ID # = 1005579
Baio Reputation: 10
Default

Asus CrossHair Hero VII WiFi and 2700x: with the latest BIOS 2304 I can run my G.Skill @3600 CL16 (target speed) for the first time.

Baio
Reply With Quote


  #8  
Old 06-13-2019, 11:19 PM
A Computer Guy A Computer Guy is online now
Registered User
A Computer Guy's PC Specs
 
Join Date: Jan 2019
Posts: 130
POST ID # = 1007924
A Computer Guy Reputation: 10
Default

Well some bad news.

If you have an ASRock X470 Master SLI/AC you may get worse memory overclock with the P3.24 version BIOS/UEFI update including AGESA 1.0.0.1.

This BIOS/UEFI update completely destroyed my overclock.

was on 16-16-16-39-55-1T, 2933MHz @1.35v, SOC@0.920v (oc stable, 4 sticks)
back to 20-20-20-48-68-1T and returned memory training problems even at this setting.

Beware.

Last edited by A Computer Guy; 06-13-2019 at 11:21 PM.
Reply With Quote


  #9  
Old 06-13-2019, 11:39 PM
Gronar Gronar is offline
Registered User
Gronar's PC Specs
 
Join Date: Mar 2019
Location: Round Rock, Tejas
Posts: 105
POST ID # = 1007926
Gronar Reputation: 10
Default

Quote:
Originally Posted by A Computer Guy View Post
Well some bad news.

If you have an ASRock X470 Master SLI/AC you may get worse memory overclock with the P3.24 version BIOS/UEFI update including AGESA 1.0.0.1.

This BIOS/UEFI update completely destroyed my overclock.

was on 16-16-16-39-55-1T, 2933MHz @1.35v, SOC@0.920v (oc stable, 4 sticks)
back to 20-20-20-48-68-1T and returned memory training problems even at this setting.

Beware.
Pretty much what happened to me too after getting the BIOS update I wanted. I could run the 4 sticks at 3000 before that, now I'm reduced to 2800. Going to be replacing this TUF board soon I hope.
Reply With Quote


  #10  
Old 06-14-2019, 12:00 AM
A Computer Guy A Computer Guy is online now
Registered User
A Computer Guy's PC Specs
 
Join Date: Jan 2019
Posts: 130
POST ID # = 1007929
A Computer Guy Reputation: 10
Default

Quote:
Originally Posted by Gronar View Post
Pretty much what happened to me too after getting the BIOS update I wanted. I could run the 4 sticks at 3000 before that, now I'm reduced to 2800. Going to be replacing this TUF board soon I hope.
I'm hoping the updates that come out soon for the 3000 series will help but not banking on it to improve 2000 series. I'm thinking of going back to 2 sticks to see how the situation has changed there.

Saw a youtube video explaining how daisy chain memory configuration impacts performance and overclocking potential so the situation is really just the worst case scenario with 4 sticks. Basically a performance vs. capacity situation.
Reply With Quote


  #11  
Old 06-14-2019, 08:12 AM
Gronar Gronar is offline
Registered User
Gronar's PC Specs
 
Join Date: Mar 2019
Location: Round Rock, Tejas
Posts: 105
POST ID # = 1007961
Gronar Reputation: 10
Default

Quote:
Originally Posted by A Computer Guy View Post
I'm hoping the updates that come out soon for the 3000 series will help but not banking on it to improve 2000 series. I'm thinking of going back to 2 sticks to see how the situation has changed there.

Saw a youtube video explaining how daisy chain memory configuration impacts performance and overclocking potential so the situation is really just the worst case scenario with 4 sticks. Basically a performance vs. capacity situation.
I can run two sticks at 3400 with no issues at all, haven't tried to push it further than that yet. Right now though I've gone to quantity over speed, running the full 32GB.
Reply With Quote


  #12  
Old 06-16-2019, 03:06 AM
A Computer Guy A Computer Guy is online now
Registered User
A Computer Guy's PC Specs
 
Join Date: Jan 2019
Posts: 130
POST ID # = 1008141
A Computer Guy Reputation: 10
Default

Quote:
Originally Posted by Gronar View Post
I can run two sticks at 3400 with no issues at all, haven't tried to push it further than that yet. Right now though I've gone to quantity over speed, running the full 32GB.
I got a CMW32GX4M2C3200C16 (ver 4.32, 2 sticks, dual rank, 32GB) kit working at 3200. Will post results in the next few days.
Reply With Quote


  #13  
Old 06-16-2019, 11:21 PM
A Computer Guy A Computer Guy is online now
Registered User
A Computer Guy's PC Specs
 
Join Date: Jan 2019
Posts: 130
POST ID # = 1008247
A Computer Guy Reputation: 10
Default

So it seems I found a nice middle ground with CMW32GX4M2C3200C16 where I have 32GB (2 sticks, dual rank) that runs at DDR4-3200 with a Ryzen 5 2600.

If I underclock it to DDR4-2933 the XMP timings entered manually work with no problem.

If I match the clock to DDR4-3200 this is where things got difficult as initially it wouldn't post 99% of the time.

I figured out when I was attempting to apply the settings from DRAM Calculator for Ryzen 1.5.1 to the BIOS/UEFI there was some additional technique involved that I had not been doing before when dealing with my Vengeance LPX sticks where I had significantly limited success.

Trying to input and apply all the settings all a once from DRAM Calculator into the ASRock BIOS/UEFI was not working and I found splitting up the saves allowed me to finally post at the higher frequency.
For example:
1) Set to DDR4-2133 with AUTO timings, save, reboot
2) Set procODT, RTT, and CAD_BUS options, save, reboot (this is what finally allowed me to post above DDR4-3200)
3) Set voltages for DRAM and SOC, save, reboot (for ASRock be sure to use the option "SOC OVERCLOCK VID" not that SOC offset option!)
4) Increase frequency, save, reboot
5) Set timings, save, reboot
6) Test and adjust

To prepare for entering timings I first prepared a worksheet from the SPD information with the memory timings in nanoseconds. I would use the nanosecond timings with the debug profile option in DRAM Calculator to calculate a safe JEDEC like profile for my target frequency.

Next this information is used to create another worksheet to calculate all timings that may work at the higher XMP frequency and their translation into nanoseconds. I would use these timings in nanoseconds with the debug profile option in DRAM Calculator to calculate a safe XMP profile for my target frequency.

With the worksheets and DRAM Calculator profiles I reconciled what timings to use and test with and additional settings.

To make a long story short here are the timings that worked for me.
Code:
DDR4-3200
 18 - tCL   (from XMP 16+2)
 18 - tRCD  (from XMP)
 18 - tRP   (from XMP)
 36 - tRAS  (from XMP)
 58 - tRC   (from XMP 54+4)
  6 - tRRDS (from XMP)
  9 - tRRDL (from XMP)
 36 - tFAW  (from XMP)
  4 - tWTRS
 12 - tWTRL
 12 - tWR
  6 - tRDRDSCL
  6 - tWRWRSCL
560 - tRFC  (calculated from SPD data) 
416 - tRFC2 (calculated from SPD data)
256 - tRFC4 (calculated from SPD data)
 18 - tCWL
 10 - tRTP
  6 - tRDWR
  3 - tWRRD
  1 - tWRWRSC
  7 - tWRWRSD
  7 - tWRWRDD
  1 - tRDRDSC
  5 - tRDRDSD
  5 - tRDRDDD
  8 - tCKE

DRAM = 1.35v
SOC = 1.012v (SOC OVERCLOCK VID = 54)
Gear Down Mode = Enabled
Power Down Mode = Disabled
BGS = Disabled
BGS-ALT = Enabled
ProcODT = 60 ohms
RTT_NOM = RZQ/7(34)
RTT_WR = RZQ/3(80)
RTT_PARK = RZQ/1(240)
CAD_BUS = All 24 ohms
The XMP values that refused to work were tCL at 16 and tRC at 54 and they needed adjustment to post.

I was able to calculate the tRFC values based on the SPD data. The BIOS/UEFI kept wanting to use 312 on AUTO but that simply would not work.

DRAM Calculator suggested 16 for tCL, tRCD, tRP, and 307-312 for tRFC however this just would not work.

According to SPD this kit is Samsung b-die however I suspect probably not the best quality b-die that does CL14 that you constantly hear about.

ASRock's ability to change the SOC voltage was a bit out of the way. You have to find an option called "SOC OVERCLOCK VID" and enter a hexadecimal value to change the voltage. I came across a German youtube video that provided the VID values. (probably due to BIOS/UEFI updates the VID values may be different) On my system the value "54"=1.016v, increase the value to lower the voltage, decrease the value to increase the voltage. There is a SOC offset option however don't use that. I found in using that the BIOS/UEFI pushed my SOC over 1.2v unexpectedly.

I hope this is helpful to those that want to use 32GB (2 sticks, dual rank) configuration at DDR4-3200 with Ryzen. (same settings work for DDR4-3266)

DDR4-2933 to DDR4-3200 makes a big difference. I couldn't get DDR4-3333 working without some awful timings, not worth it. (perhaps workable with better motherboard & cpu)

(edited post for more completeness)

Last edited by A Computer Guy; 07-01-2019 at 08:55 PM.
Reply With Quote


  #14  
Old 06-16-2019, 11:58 PM
Gronar Gronar is offline
Registered User
Gronar's PC Specs
 
Join Date: Mar 2019
Location: Round Rock, Tejas
Posts: 105
POST ID # = 1008251
Gronar Reputation: 10
Default

Awesome info dude! My crypto mining machine has an ASRock board in it and I'm seriously considering going back to one for this 2600X gaming rig when the X570's are released.
Reply With Quote


  #15  
Old 06-20-2019, 11:08 PM
A Computer Guy A Computer Guy is online now
Registered User
A Computer Guy's PC Specs
 
Join Date: Jan 2019
Posts: 130
POST ID # = 1008726
A Computer Guy Reputation: 10
Default

I had a breakthrough. Event though the SPD indicated Samsung b-die DRAM Calculator's "Samsung OEM" profile more closely matches the actual timings of this ram module.

After using those numbers I finally have nearly the original XMP timings locked in at DDR4-3200 and it's seems stable (4 passes of memtest86)

(correction: 7/1/2019 - I was mistaken I accidentally selected DDR4-3066 in the BIOS/UEFI instead of DDR4-3266 without realizing it initially so these settings are actually for DDR4-3066)

Code:
DDR4-3066
 16 - tCL   (from XMP)
 18 - tRCD  (from XMP)
 18 - tRP   (from XMP)
 36 - tRAS  (from XMP)
 56 - tRC   (from XMP 54+2)
  6 - tRRDS (from XMP)
  9 - tRRDL (from XMP)
 34 - tFAW  (from XMP 36-2)
  4 - tWTRS
 12 - tWTRL
 24 - tWR
  5 - tRDRDSCL
  5 - tWRWRSCL
480 - tRFC  (calculated from SPD data) 
 16 - tCWL
 12 - tRTP
  6 - tRDWR
  3 - tWRRD
  1 - tWRWRSC
  7 - tWRWRSD
  7 - tWRWRDD
  1 - tRDRDSC
  5 - tRDRDSD
  5 - tRDRDDD
  8 - tCKE
DRAM = 1.35v
SOC = 1.012v (SOC OVERCLOCK VID = 54)
Gear Down Mode = Enabled
Power Down Mode = Enabled
BGS & BGS-ALT = AUTO
ProcODT = 60 ohms
RTT_NOM = RZQ/7(34)
RTT_WR = RZQ/3(80)
RTT_PARK = RZQ/1(240)
CAD_BUS = All 24 ohms

Last edited by A Computer Guy; 07-01-2019 at 08:59 PM.
Reply With Quote


Reply

Tags
agesa, ryzen

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -4. The time now is 09:13 PM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2019, vBulletin Solutions, Inc.