What ram stability test are you running these days?

SpongeBob

The Contraceptive Under the Sea
Joined
Jan 15, 2011
Messages
798
What memory stability tests are you guys running these days? I don't really want to install anything to a flash drive to boot from and run it for like 8-24. Anything newer that does a great job that you guys are running? I'm not talking like AIDA64 Cache and memory benchmark or something. Great benchmark but I'm looking for something a bit more thorough than benchies and games.
 

Canon

2[H]4U
Joined
Aug 12, 2004
Messages
2,586
Go to your start menu and type "windows memory diagnostic". It will give you a prompt to restart and run.

My only other experience is Memtest86 on a CD/DVD or flash drive.
 
Joined
Feb 3, 2007
Messages
1,021
I graduated from Memtest86 about 10 years ago when I had trouble with memory in windows that had been test fine with memtest86. I started looking for an app that would test it from within windows to avoid this. I had used Memtest86 for 10 or more years prior spending hours upon hours with it as far back as Athlon XP with Corsair B5 memory on DFI LanParty Mobo

I use HCI Design Memtest Pro the paid version. At the time it was like 7 bucks and I still use the license today. What I like about HCI Design Pro is it automatically calculates how many open instances are required to saturate the available memory you have in your system for a proper test. You run it until all instances have completed all test 100% or more if you feel spunky. I have not had a failure due to faulty memory in windows ever since after timings were tested stable with the HCI memtest.

You can accomplish the same affect with the free version but you have to do the calculations yourself.

In all fairness Memtest86 still has it merits under the right circumstance. Like say before windows or other OS is installed or used. The worst part of memory failures is how badly windows behaves when OS corruption occurs causing a format and reload to fix. Those who want to overclock memory test it before using the timings in windows to avoid corruption and getting things close to stable before testing from within Windows.

Bottom line they both have a purpose but if you are not overclocking memory and just using XMP settings then HCI Design Memtest is all you need.
 

Nasgul

Limp Gawd
Joined
Jun 11, 2005
Messages
136
Mine is the easiest. I put it in, it works!!! You are good to go.

This method has served me well since 1998.

Although in 2005 I bought a Corsair kit from eBay and one module was bad, I proceeded with my "method" of stability and tested ONE module at a time, and voila! One was bad because it wasn't letting the system boot. So I RMA it, and long gone are the days when you could buy CORSAIR RAM w/o having to provide a POP.
 

drescherjm

[H]F Junkie
Joined
Nov 19, 2008
Messages
14,926
I have in the past used HCL memtest the free version to diagnose problems with RAM in windows. You end up launching a bunch of instances until you see almost all of your ram covered in task manager. When I have had problems HCL was the fastest to identify the problem. I have seen cases where it took memtest86+ days to detect bad ram while HCL took only hours. Although that was an earlier version.
 

dasa

Gawd
Joined
Feb 19, 2012
Messages
534
When overclocking RAM on Ryzen I use TestMem5.
https://www.overclock.net/attachments/tm5-zip.341454/
The errors can give some guidance to what is wrong.
tm5_1usmus_download-jpg.jpg
 

MaZa

2[H]4U
Joined
Sep 21, 2008
Messages
3,354
HCI Memtest. Ryzem memory timing tool offers a nice frontend for it (no need to launch multiple instances of it to fill your RAM). It is very effective and finds overclocking related instabilities within few hours if there is any. Memtest86 is also good but as far as I know it is better for detecting a faulty ram than OC unstable. It does find very bad cases of OC instability just fine but it does not push memory as hard as HCI Memtest does so finding more subtle errors can take much longer.

I have heard good things about Karhu memory test but it is an paid app and frankly, as a mere hobbyist I find getting a paid stress test silly when free (and very effective) ones have existed for so long.
 

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
58,939
Mine is the easiest. I put it in, it works!!! You are good to go.

This method has served me well since 1998.

Although in 2005 I bought a Corsair kit from eBay and one module was bad, I proceeded with my "method" of stability and tested ONE module at a time, and voila! One was bad because it wasn't letting the system boot. So I RMA it, and long gone are the days when you could buy CORSAIR RAM w/o having to provide a POP.

This. I don't bother with stability testing unless there is something wrong with the system.
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,308
Go to your start menu and type "windows memory diagnostic". It will give you a prompt to restart and run.

This is the most horrible weak memmory test and will have tons of false negatives. it was the absolut bottom of around ~20 tested. do NOT use this
It would pass memory for over 17 hours testing that the pc could not even boot windows on ( and had graphical glitches on the screen)

It is more garbage than nunchuks


Memtest86 (v7.x+) and Prime95 Blend mode had the best performance of memory testing. however I do not have the test at hand at work
 

pendragon1

Extremely [H]
Joined
Oct 7, 2000
Messages
32,840
It would pass memory for over 17 hours testing that the pc could not even boot windows on ( and had graphical glitches on the screen)
how? you have to launch it from within windows.



memtest or memtest+ are fine.
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,308
how? you have to launch it from within windows.



memtest or memtest+ are fine.

1: No you dont you can use a start key for it.
and even if you didnt knew this. it not hard to turn of your computer after setting it up from within windows and replacing test ram for then to boot into the windows memory diagnostics

2: memtest86+ is generally inferior to memtet86 v7.x or higher due to lower stress levels (multicore usage is not fully supported in memtest86+) at leat according to my lastet test i dont think metest86+ have had any improvements since then
 

pendragon1

Extremely [H]
Joined
Oct 7, 2000
Messages
32,840
1: No you dont you can use a start key for it.
and even if you didnt knew this. it not hard to turn of your computer after setting it up from within windows and replacing test ram for then to boot into the windows memory diagnostics

2: memtest86+ is generally inferior to memtet86 v7.x or higher due to lower stress levels (multicore usage is not fully supported in memtest86+) at leat according to my lastet test i dont think metest86+ have had any improvements since then
oh ok.
my memtest86+ stick is a couple weeks old and supports multicore and both work fine in my experience.
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,308
oh ok.
my memtest86+ stick is a couple weeks old and supports multicore and both work fine in my experience.

What version are you using some kind of beta version ?
The latest full release 5.01 is well known to have issues with parallel multi-core with crashes and false positive. that why it works in default in a round robin method aka single core way but change the cores for each pass. hence the lack of the stress level of the fully parrallel multicore support that is in passmarks memtest86 v7.x.
This well know limitation on memtest86+ var highly debated on the forum that sadly appears to be gone.

im also not sure what you define as fine. I never said it did not work "Fine" i siad it was enferio to memtest v7.x


Anyway I found one of the old test on one of my backups archieves






Code:
3770K 4.4GHz 32GB
2x Kingston KHX2400C11D3/8GX
2x Kingston KHX1866C10D3/8G
2133MHz Dual channel

11-13-13-30             10-12-12-28     10-11-11-28     10-11-11-26     10-11-11-25     10-11-11-24     9-11-11-28      9-11-11-26      9-11-11-24
CMD2T V1.65             CMD1T V1.65     CMD1T V1.65     CMD1T V1.65     CMD1T V1.65     CMD1T V1.65     CMD1T V1.65     CMD1T V1.65     CMD1T V1.65

Memtest86 V7.03         OK  6:48        OK 12:50        D   0:13        D   0:01        C   0:05        C   0:00        C   0:00        C   0:00
Memtest86 V4.37         OK  4:07        OK  4:15        OK  4:30        D   0:38        D   0:05        D   0:00        D   0:00        D   0:00
Memtest86+ V5.01        OK 11:30        OK  4:50        OK  4:42        D   0:04        D   0:05        D   0:01        D   0:00        D   0:00
Prime95 V28.10 Blend    OK  4:03        OK  4:01        D   0:34        D   0:00        N/B             N/B             N/B             N/B
Intel XTU               OK  4:05        OK  4:05        OK  4:05        C   0:04        N/B             N/B             N/B             N/B
HCI Memtest             OK  4:10        OK  9:07        D   0:05        D   0:00        N/B             N/B             N/B             N/B
7-zip V16.04 Benchmark  OK  8:55        OK  4:23        OK 10:04        D   0:03        N/B             N/B             N/B             N/B
Windows Standard        OK  4:00        OK 10:00        OK  4:00        OK 10:00        D   0:05        OK 13:10        OK  7:48        D   0:00
Windows Extended        OK  5:33        OK 11:07        OK 10:30        OK 10:14        D   0:05        OK  9:05        OK  4:33        D   0:00
AleGr Memtest           OK  4:00        OK 12:07        D   0:18        D   0:00        N/B             N/B             N/B             N/B
RightMark Memtest       OK  6:05        OK  8:16        OK  4:11        D   0:32        N/B             N/B             N/B             N/B
Goldmemory              OK  6:09        OK  9:55        OK  4:55        D   0:20        D   0:00        OK  6:35        OK  4:05        D   0:00
TechPowerUP             OK  4:00        OK  4:00        D   0:11        D   0:00

                        11-13-13-30     11-13-13-30     11-13-13-30     11-13-13-30
                        CMD1T V1.6      CMD1T V1.55     CMD1T V1.50     CMD1T V1.45

Memtest86 V7.03         OK  7:20        OK  9:28        D   2:26        D   1:22
Memtest86 V4.37         OK 15:22        OK  4:09        OK  4:06        OK  9:51
Memtest86+ V5.01        OK 10:51        OK 11:28        OK  7:56        OK  5:28
Prime95 V28.10 Blend    D   7:53        D   0:36        D   0:27        D   0:05
Intel XTU               OK  4:05        OK 10:49        OK 14:26        OK  5:30
HCI Memtest             OK  5:03        D   0:27        D   0:08        D   0:05
7-zip V16.04 Benchmark  OK  5:51        OK  4:00        OK  4:05        OK  4:57
Windows Standard        OK  5:12        OK 10:04        OK 17:00        OK  7:21
Windows Extended        OK  7:24        OK  7:12        OK 11:30        OK  6:46
AleGr Memtest           OK 11:09        OK  8:34        OK 11:17        OK  8:45
RightMark Memtest       OK  8:24        OK  6:55        OK  6:52        OK  4:55
Goldmemory              OK  7:05        OK 11:36        OK 10:03        OK 10:05
TechPowerUP             OK  4:00        OK  4:00        OK  4:00        OK  6:00


2600K 4.5GHz 16GB
2x Crucial BLT4G3D1608DT1TX0
2x Crucial BLT4G3D1608DT1TX0
1600MHz dual Channel

9-9-9-24                8-8-8-24        7-8-8-24        7-7-7-24        7-7-7-22        7-7-7-20        7-7-7-18        7-7-7-16        7-7-6-24        7-7-6-22
CMD2T 1.5V              CMD1T V1.5      CMD1T V1.5      CMD1T V1.5      CMD1T V1.5      CMD1T V1.5      CMD1T V1.5      CMD1T V1.5      CMD1T V1.5      CMD1T V1.5

Memtest86 V7.03         OK  4:16        OK  4:54*       OK  8:21*       OK 21:17        D   0:40        OK  4:01*       OK 23:42        D   0:00        D   0:00
Memtest86 V4.37         OK  4:05        OK 10:05        OK  6:00        OK  7:38        OK  5:49        OK 17:57        OK  9:13        D   0:17        D   0:10
Memtest86+ V5.01        OK  4:07        OK  7:29        OK  5:01        OK 20:00        OK 12:30        OK  9:16        OK  6:10        D   0:00        D   0:25
Prime95 V28.10 Blend    C   6:43        OK  4:15        C   0:38        C   0:44        C   0:48        C   0:42        C   0:06        C   0:59        D   7:02
Intel XTU               OK  4:05        OK  4:05        OK  4:05        OK  4:05        OK  4:05        OK  4:05        C   1:44        OK 17:33        OK 13:40
HCI Memtest             OK  4:15        OK 13:04        OK  4:35        OK  8:44        OK 16:47        OK 24:43        OK  9:07        D   0:00        D   0:00
7-zip V16.04 Benchmark  OK 11:56        OK  9:58        OK 23:58        OK  6:03        OK  6:01        OK  4:43        OK  4:43        D   2:46        D   5:05
Windows Standards       OK  9:56        OK 12:05        OK  6:00        OK 23:04        OK 10:07        OK 13:01        OK 20:03        OK  9:23        D   0:01
Windows Extended        OK 24:00        OK 20:32        OK 10:15        OK  7:20        OK 23:51        OK 11:27        OK 10:54        OK  8:00        D   0:03
       
                        9-9-9-24        9-9-9-24        9-9-9-24        9-9-9-24        9-9-9-24        9-9-9-24
                        CMD?T V1.45     CMD?T V1.4      CMD2T V1.35     CMD2T V1.3      CMD1T V1.3      CMD1T V1.3

Memtest86 V7.03         OK 15:33        OK  4:03        OK  4:19        OK 27:09        OK 16:59
Memtest86 V4.3.7        OK 10:21        OK  6:25        OK 25:20        OK 22:27        OK  7:24
Memtest86+ V5.01        OK 24:44        OK  9:59        OK 18:47        OK 14:48        OK 20:55        OK  8:38
Prime95 V28.10 Blend                                    OK 18:18        OK121:33
Intel XTU               OK 11:23        OK 11:36        OK 23:48        OK 22:00        OK 20:55        OK  4:00
HCI Memtest             OK 14:07        OK  6:46        OK 13:59        OK  5:24        OK 28:21        OK 23:22
7-zip V16.04 Benchmark  OK 22:25        OK 15:03        OK  4:21        OK  6:50        OK 11:16        OK 12:41
Windows Standard        OK 12:25        OK 14:04        OK  9:04        OK  5:11        OK 23:14        OK 18:07
Windows Extended        OK  9:59        OK 11:15        OK  8:15        ok 24:48        OK 22:13



Standby Prevention
HCI memtest     NO
Prime95         NO

Legend:
OK      Passed
D       Error Detected
C       System Crashed
N/B     Not Bootable

*note might have row hammer issues

Bassically if a memory test declared a ram modul OK after 4 hours i would declare it as ok (not found any errors) itensify the errors on the module and retest
The above is the testing time how aver only the 4 hours mark was use to determine the validation to keep it far for every tools

makes it very to see easy which program detects smaller errors and detects them the fastes
This was just one of the test's
 

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420

Attachments

  • memtest 5.PNG
    memtest 5.PNG
    16.8 KB · Views: 0
  • zen timing3.PNG
    zen timing3.PNG
    32 KB · Views: 0
Last edited:

dasa

Gawd
Joined
Feb 19, 2012
Messages
534
I just recently upgraded to 32GB from 16GB. 4x8 SR and I can run other memory testing programs fine however this app always gives me errors.
Few things wrong there for four sticks.
SCL probably need to go back to 4 and try RTT 6-3-3 as that seems to work well for 4 sticks.

4 sticks at 1T GDM disabled can be difficult for some sticks especially on MB with a thin PCB (yours has 6 layers I thick which is not bad). I had to set AddrCmdSetup 56 to get 1T stable with 4 sticks of 4400c19 on A2 PCB but two sticks was fine as was 4 sticks of 3200c14 A1 PCB. Changing to AddrCmdSetup 56 does come with a small performance hit so only use it if you need to.
Since 4 sticks at 1T is hard on the MB\IMC you may need ~100mv more VDDG IOD and maybe a bit more SOC.

This wont help stability but once you get it stable at those settings try these tRFC or relax timings a bit and try push those sticks to 3733-4000 with ~1.45v or ~1.5v with a fan aimed at them.
tRFC 252
tRFC2 187
tRFC4 115
 

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420
Few things wrong there for four sticks.
SCL probably need to go back to 4 and try RTT 6-3-3 as that seems to work well for 4 sticks.

4 sticks at 1T GDM disabled can be difficult for some sticks especially on MB with a thin PCB (yours has 6 layers I thick which is not bad). I had to set AddrCmdSetup 56 to get 1T stable with 4 sticks of 4400c19 on A2 PCB but two sticks was fine as was 4 sticks of 3200c14 A1 PCB. Changing to AddrCmdSetup 56 does come with a small performance hit so only use it if you need to.
Since 4 sticks at 1T is hard on the MB\IMC you may need ~100mv more VDDG IOD and maybe a bit more SOC.

This wont help stability but once you get it stable at those settings try these tRFC or relax timings a bit and try push those sticks to 3733-4000 with ~1.45v or ~1.5v with a fan aimed at them.
tRFC 252
tRFC2 187
tRFC4 115
Thanks for the suggestions.

Going to up the SOC and VDDG IOD first at current settings.

When you say change SCL back to 4 you are referring to both of these

zen timing31.PNG


And RTT 6-3-3?

I will try make the changes and test in abit. For now I will stay at 3200 speeds until I get 0 errors then i can go up in speed. However I don't want to install any fans so 1.5v is out. 1.45v should be doable but I will have to monitor ram temps at that speed.
 
Last edited:

dasa

Gawd
Joined
Feb 19, 2012
Messages
534
I will try make the changes and test in abit. For now I will stay at 3200 speeds until I get 0 errors then i can go up in speed. However I don't want to install any fans so 1.5v is out. 1.45v should be doable but I will have to monitor ram temps at that speed.
Stability can decrease at higher temps so try keep them under 45-50c if you can.
Even 1.45V may be to much with 4 sticks depending on case airflow.
With my 3200c14 sticks one experienced negative scaling of tRCD if it went over 1.44v even though other timings benefited from more V.
 

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420
Stability can decrease at higher temps so try keep them under 45-50c if you can.
Even 1.45V may be to much with 4 sticks depending on case airflow.
With my 3200c14 sticks one experienced negative scaling of tRCD if it went over 1.44v even though other timings benefited from more V.

At my current speed 3200 @ 1.35v i'm at 38-44c range for temps.

1631396472296.png


Increasing VSOC and VDDG IOD alone one less error.

1631396422063.png


Still a work in progress.

1631397866137.png
 
Last edited:
Joined
Apr 9, 2021
Messages
732
Timings look reasonable. Some of them are probably a little looser than they need to be for 3200/1.35V if they are a good bin. I would enable GDM and see if that clears up any errors. Then go from there.
 
Last edited:

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420
Timings look reasonable. Some of them are probably a little looser than they need to be for 3200/1.35V if they are a good bin. I would enable GDM and see if that clears up any errors. Then go from there.

I have decided to go back to Stock XMP to test then will start redoing all timings. Figured I might as well just nuke it from orbit I got time tonight.

1631400312182.png


no errors so I may now try going back to my original settings but to enable GDM.

1631400644784.png
 
Last edited:

dasa

Gawd
Joined
Feb 19, 2012
Messages
534
RttWr RZQ/3 is likely helping on auto vs off on your previous settings which is better suited to two sticks.
RttPark RZQ/1 can be a little hard on the RAM apparently if you start increasing V beyond 1.4v I hear it is best to run higher values.
 

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420
RttWr RZQ/3 is likely helping on auto vs off on your previous settings which is better suited to two sticks.
RttPark RZQ/1 can be a little hard on the RAM apparently if you start increasing V beyond 1.4v I hear it is best to run higher values.

I'm back to my original settings with these changes and no errors so far.

1631402177681.png



Latency looks like this I still have to renable PBO and CO

1631402425107.png
 

Furious Nerd

2[H]4U
Joined
Sep 14, 2006
Messages
3,415
-Run HWinfo and monitor for WHEA errors and whatever else is important to you. WHEA errors will come up fast in desktop and is my first sign before I ever need to start torture testing that whatever change to timings or voltages made things unstable already.

Also in my case it revealed that my motherboard dynamically adjusts RAM voltages depending on load. Which is retarded because say my timings can be stable at 1.5V load with my Samsung B-die, but I didn't realize before that during idle situations my motherboard decides to drop vram voltages to frigging 1.44V. So super low latencies are out of the question because they wouldn't be stable at 1.44V while being stable at their load voltage of 1.5V.. I still don't know if this isa bug or a feature. But it allowed me to better understand some things and compensate.

-TestMem5 w/ Extreme @anta777 preset as first line after timings and voltage adjustments. Finds errors early on quickly. If it errors - back to the drawing board.

-If that passes I run OCCT for 30-60 minutes... memory test and/or CPU test (extreme preset also tests memory. And the heat produced by the CPU test also heats up the case/RAM, so it starts to simulate the heat conditions produced by more demanding tasks such as gaming (still no GPU load/heat here) that will make your TM5 stable run suddenly unstable - the extra heat from other components affects RAM stability.

-If it passes I run Prime95 either/or Large FFTs / Blend (to also include CPU heat) overnight, or when I know I won't need to use the computer for several hours during the day. Usually if it doesn't find errors within the first 3 hours it'll probably pass an overnighter, but I've had a few rare instances where it failed 6-7 hours in. So once you get close to your "final"/'I'm pretty sure it's gonna be stable all-around' territory, that's when I pull out Prime95. I do this as the 3rd step after each timing-set adjustment, before moving on to the next batch of timings I'll work on.

-The real and final test is gaming, though. The heat produced by the RAM, CPU, AND by the GPU will do a number on your previously seemingly stable RAM OC. RAM is heat sensitive just like your CPU and GPU OC is. So gaming throws the entire package at it. Just play a demanding game an hour or two. If it doesn't crash, and the above tests pass, that's when I personally call it stable.

This is the best and only memory OC guide you should follow.

Yesterday I completed my RAM OC and this is my personal final result - except I haven't tried lowering vddp and vddg after upping them yet, and didn't test below 1V SOC voltages because it's low enough to me:
RAM voltage is 1.44v idle and 1.5V load in memory testing utilities, and in-between that during video games.
Patriot Vipers - Samsung B-die.
Capture.JPG
 
Last edited:

dasa

Gawd
Joined
Feb 19, 2012
Messages
534
Not much info on RTT other than to say it can affect stability significantly is one example.
There is settings that are much better for two ranks of RAM and others that are much better four 4 ranks +.
 

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420
-Run HWinfo and monitor for WHEA errors and whatever else is important to you. WHEA errors will come up fast in desktop and is my first sign before I ever need to start torture testing that whatever change to timings or voltages made things unstable already.

Also in my case it revealed that my motherboard dynamically adjusts RAM voltages depending on load. Which is retarded because say my timings can be stable at 1.5V load with my Samsung B-die, but I didn't realize before that during idle situations my motherboard decides to drop vram voltages to frigging 1.44V. So super low latencies are out of the question because they wouldn't be stable at 1.44V while being stable at their load voltage of 1.5V.. I still don't know if this isa bug or a feature. But it allowed me to better understand some things and compensate.

-TestMem5 w/ Extreme @anta777 preset as first line after timings and voltage adjustments. Finds errors early on quickly. If it errors - back to the drawing board.

-If that passes I run OCCT for 30-60 minutes... memory test and/or CPU test (extreme preset also tests memory. And the heat produced by the CPU test also heats up the case/RAM, so it starts to simulate the heat conditions produced by more demanding tasks such as gaming (still no GPU load/heat here) that will make your TM5 stable run suddenly unstable - the extra heat from other components affects RAM stability.

-If it passes I run Prime95 either/or Large FFTs / Blend (to also include CPU heat) overnight, or when I know I won't need to use the computer for several hours during the day. Usually if it doesn't find errors within the first 3 hours it'll probably pass an overnighter, but I've had a few rare instances where it failed 6-7 hours in. So once you get close to your "final"/'I'm pretty sure it's gonna be stable all-around' territory, that's when I pull out Prime95. I do this as the 3rd step after each timing-set adjustment, before moving on to the next batch of timings I'll work on.

-The real and final test is gaming, though. The heat produced by the RAM, CPU, AND by the GPU will do a number on your previously seemingly stable RAM OC. RAM is heat sensitive just like your CPU and GPU OC is. So gaming throws the entire package at it. Just play a demanding game an hour or two. If it doesn't crash, and the above tests pass, that's when I personally call it stable.

This is the best and only memory OC guide you should follow.

Yesterday I completed my RAM OC and this is my personal final result - except I haven't tried lowering vddp and vddg after upping them yet, and didn't test below 1V SOC voltages because it's low enough to me:
RAM voltage is 1.44v idle and 1.5V load in memory testing utilities, and in-between that during video games.
Patriot Vipers - Samsung B-die.
View attachment 394361

No Whea errors here.

I'm already gone thru CO and the per core settings. I will produce whea is if put too high a negative on Core 0

Right now i'm running -15 on all cores and Core 3 which is my best core is -20 core 0 is second best core but that stays at -15. I think the biggest difference for me was enabling gear down which is needed for 4x8SR. When I was only on 16GB of ram I was able to disable it, but the move to 32GB shows it was necessary to test and recheck everything.

Now that everything is passing that test going to try the tRFC I was running at 16GB which was 256

Red are my current timings black is the preset i was using at 16GB. Going to Ingore the RTT settings in there now and will leave what i'm currently using. And of course Geardown stays on.

1631461857362.png
 
Last edited:
Joined
Apr 9, 2021
Messages
732
My timings for 3200 / 1.35V on pretty good b-die are similar to yours. Differences are tRRDL at 4, tWTR_S at 3, tWTR_L at 8, tRTP at 4, tWR at 10, tRFC at 252. I can go even tighter on a couple (tWTR_L at 7, tWR at 8) but doing so seems to hurt performance. I'm on a different platform altogether so YMMV.
 

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420
I think i'm done testing at 32GB @ 3200 / 1.35v with my B-die. PBO is on and CO also so my boost clocks are back to 5.05Ghz.

Now I can move on to testing at 3600.

1631465105377.png
 
Last edited:

dasa

Gawd
Joined
Feb 19, 2012
Messages
534
You will probably find that you can run GDM disabled but that it requires getting some other settings right.
At the end of the day it is much easy er with it enabled but GDM disabled 2T can be faster than GDM enabled 1T.

Here are my 3200 settings with 4 sticks of A2 b die.
3200c12.jpg


The best I got out of my 3200c14 kits.
3771c15 low v a.jpg


And what I run my 4400c19 kits at when I want the extra speed.
I do get WHEA errors over 3733 which can affect minimum FPS but disabling the logging of those errors minimizes it.
4000c15 stable.jpg
 

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420
You will probably find that you can run GDM disabled but that it requires getting some other settings right.
At the end of the day it is much easy er with it enabled but GDM disabled 2T can be faster than GDM enabled 1T.

Here are my 3200 settings with 4 sticks of A2 b die.
View attachment 394555

The best I got out of my 3200c14 kits.
View attachment 394554

And what I run my 4400c19 kits at when I want the extra speed.
I do get WHEA errors over 3733 which can affect minimum FPS but disabling the logging of those errors minimizes it.
View attachment 394556
I will try to see if I can get it working with GDM off eventually but will save it next weekend. I'm going to put the machine at its current settings thru my normal workloads for the work week.
 

mgty23

[H]ard|Gawd
Joined
Mar 16, 2017
Messages
1,187
Karhu Ram Test ,here :)No errors on stock 10850K and Gskill 2x16GB XMP 3200
 

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420
You will probably find that you can run GDM disabled but that it requires getting some other settings right.
At the end of the day it is much easy er with it enabled but GDM disabled 2T can be faster than GDM enabled 1T.

Here are my 3200 settings with 4 sticks of A2 b die.
View attachment 394555

The best I got out of my 3200c14 kits.
View attachment 394554

And what I run my 4400c19 kits at when I want the extra speed.
I do get WHEA errors over 3733 which can affect minimum FPS but disabling the logging of those errors minimizes it.
View attachment 394556


Had some time to test tonight and it looks like these changes were enough for me to run with GDM disabled now with no errors.

1631580923543.png
 
Last edited:

Makaveli@BETA

2[H]4U
Joined
Mar 24, 2004
Messages
2,420
None. It just works.

When running default settings yes, however when overclocking and tuning memory these applications are needed for checking stability. I learned a fair abit about tweaking RTT settings from this thread alone so well worth it to me for the knowledge.
 

daglesj

Supreme [H]ardness
Joined
May 7, 2005
Messages
5,437
When running default settings yes, however when overclocking and tuning memory these applications are needed for checking stability. I learned a fair abit about tweaking RTT settings from this thread alone so well worth it to me for the knowledge.

Yeah if you want to waste your life to little benefit I guess. :LOL:

I set Command rate from 2T to 1T if available and thats it. I never bother to test after. Why does anyone need to spend hours and hours testing their PC to go from 188FPS to 189FPS?

Tweaking ram is mostly wasting your time now. This isn't 2003 anymore IMO.
 
Joined
Apr 9, 2021
Messages
732
Yeah if you want to waste your life to little benefit I guess. :LOL:

I set Command rate from 2T to 1T if available and thats it. I never bother to test after. Why does anyone need to spend hours and hours testing their PC to go from 188FPS to 189FPS?

Tweaking ram is mostly wasting your time now. This isn't 2003 anymore IMO.

It's only a waste of time if you don't enjoy doing it.
 
Top