Post your Ryzen memory speeds!

Nightfire

[H]ard|Gawd
Joined
Sep 7, 2017
Messages
2,045
Does ram speed or core count affect cache speed? your L1 and L2 cache speed is 200 higher then mine o_O. On my 2600X your clock speed during your test is only 100mhz higher so wouldnt be the reason for the big gap.
A low trfc has been the single highest increase in aida performance for me, other than memory Freq speed and primary timings. Other than that, it may only be CPU freq as Formula.350 mentioned.
 

Ruiner

2[H]4U
Joined
Aug 22, 2000
Messages
3,064
2 rigs, both asrock b350 (most recent bios), one with a r3 1200x, 2nd with r5 1600, both running ripjaw black 3200, 2x4GB at 3200 under xmp profile.
 

illusio13

n00b
Joined
Feb 20, 2019
Messages
49
Corsair Vengance LPX 2x4GB @ 3200 C-L 16 :

Been able to do XMP since bios ~ G/H ( MSI X370 GPC ); 2933-3000 before!.

On my R5 1400, would boot into and game @ 3333 auto-everything, but not system-wide stable!

Tightend down to 14-15-14-14-27-46-227-1T @ 1.45-1.506v bios with NB/Soc @ ~ 1.170v. When OC R5 1400 ~ 3.8-3.9Ghz getting ~47-48000MB/s @ 71-76ns.

Can lower clocks to 14-14-14-14 etc @ 3000-3066 with same voltages but again, not really stable, everywhere.

On R5 2600, 3400 is tunable@ stockV ( can boot @ 3466-3566 and get into windows......; before whea errors start:-( ), played around a little, been Cpu Oc'ing just lately; but more the mass of info relating to DDR OC'ing puts me off...:-(

A clear guide on both terms and flows between relationships would significantly; in my opinion, help many people with extracting max Performance/Efficiency from their DDR!!

Curently getting @ ~ 61-63ns @ ~ 50000MB/s with almost ( "The Stilt's - Fast DDR4@3200 CL-14 ) timings; that 15 value above will not drop...........................................; missing something fundamental here, since i get the feeling this Corsair E-Die Single-rank 2x4GB kit, is not that bad!:)

Edit : clarity.
 

Formula.350

[H]ard|Gawd
Joined
Sep 30, 2011
Messages
1,107
A clear guide on both terms and flows between relationships would significantly; in my opinion, help many people with extracting max Performance/Efficiency from their DDR!!
I fully agree. Problem is that most of the people, like extreme overclockers, who are the most suited for creating a guide like this.... aren't willing to share the finer details on this stuff *sigh* Reason they are vague or just flat out remain tight lipped (with the except of The Stilt), is because they participate in competitions where they can win quite a bit of money (either actual, or through winning high end hardware). Therefore, these secrets are what gets them those wins :(

So while I can appreciate the logic from that standpoint, I hate that we've reached that point as hobbyists to not want to share that info. It used to be that everyone was willing to share and help each other, which is how it should be. Alas, someone decided monetize it all through competitions and here we are.
 

illusio13

n00b
Joined
Feb 20, 2019
Messages
49
I fully agree. Problem is that most of the people, like extreme overclockers, who are the most suited for creating a guide like this.... aren't willing to share the finer details on this stuff ...................

Fully agree with the setiments expressed :-( Like you, and all of us on these type of forums ARE Hobbyists; and for me personally; take great, vicarious pleasure :) from helping others to also achieve extra performance/efficiency; or even just to have a better Gaming/ Media experience on their PC's consoles etc...

Its what eventually caused me to leave IT..., arseholes did not want share; ruined the whole experience for me...!!!

Have seen a couple of posts on various forums which have a half-decent stab @ collating a coherent DDR OC'ing guide..., but still......, not finding it easy to grasp those nuances.... which is F*****G me Off!! : Especially since comprehension happens... yes?

Hopefully somebody around these parts; who is much [H}arder than me can explain / chip in etc :)

Edit : Just like the post i made about my Tv firmware update; which has given me VRR-Like behaviour, as well as all metrics going up!. Since i have used this Tv for 2+ years before this particular update, the Night & Day difference is staggering ( brought unbidden tears to my eyes ! ).. So got straight on AMD reddit, posted here as well as Hexus.............; i'm NOT the only person to have a LG Tv........, wanted to share the love-in what i am experiencing......... ergo; all the above:)
 
Last edited:

Priller

Gawd
Joined
Sep 8, 2013
Messages
572
2700X on C7H @ 4.050Ghz with 1.232(1.256 VID) Volts Using P-State Overclocking. SOC 1.107V

Flare X 3200, 14-14-14-14-28-48-260-1T @ 3466 With 1.400V And Geardown enabled.

It is hours on end 100% load encoding stable and I'm still dialing in the all core overclock.
 
Last edited:

illusio13

n00b
Joined
Feb 20, 2019
Messages
49
2700X on C7H @ 4.050Ghz with 1.232(1.256 VID) Volts Using P-State Overclocking. SOC 1.107V

Flare X 3200, 14-14-14-14-28-48-260-1T @ 3466 With 1.400V And Geardown enabled.

It is hours on end 100% load encoding stable and I'm still dialing in the all core overclock.
Excellent subs @ those speeds! - nice underlying IF speedup, I can feel those! :); can not get my ( 14-15-14-14-27... ) CAS 15 down, no matter what i try........; all types of " Voodoo" nonsense... Ehhh!

Stopped there.
 

illusio13

n00b
Joined
Feb 20, 2019
Messages
49
Excellent subs @ those speeds! - nice underlying IF speedup, I can feel those! :); can not get my ( 14-15-14-14-27... ) CAS 15 down, no matter what i try........; all types of " Voodoo" nonsense... Ehhh!

Stopped there.
Why can i NOT have the same inherent functionality for DDR4 ( Go slower ! - go Lower :) ); less volts; tighter timings : amongst these slow DDR4 speeds are some senarios where ( Desktop ) feels ........., Just like @ 3200 or 3400..............., some proper nonsense on one forum somewhere; the dude was a wizzard, i'm sure :) @ ~ 98 - 112ns :-( Voodoo!

On topic : If my DDR4 sub- system is setup for GS-GL :), then IF..........; should GS-GL, on a MUCH larger multiplier............?

Why, .........
 

Priller

Gawd
Joined
Sep 8, 2013
Messages
572
Excellent subs @ those speeds! - nice underlying IF speedup, I can feel those! :); can not get my ( 14-15-14-14-27... ) CAS 15 down, no matter what i try........; all types of " Voodoo" nonsense... Ehhh!

Stopped there.
I've tried a bit faster on ram speed and even gotten into Windows as high as 3600 but any stress and it crashes instantly. I've found the limit of my ram at 1.40 volts to be around 3501. Im sure the board, CPU IMC, & the ram will go higher with more voltage but I keep my gear for a long time so I go towards longevity instead of getting every last drop out of it.
 

illusio13

n00b
Joined
Feb 20, 2019
Messages
49
I've tried a bit faster on ram speed and even gotten into Windows as high as 3600 but any stress and it crashes instantly. I've found the limit of my ram at 1.40 volts to be around 3501. Im sure the board, CPU IMC, & the ram will go higher with more voltage but I keep my gear for a long time so I go towards longevity instead of getting every last drop out of it.

Totally agree.
 

Below Ambient

[H]ard|Gawd
Joined
Dec 5, 2001
Messages
1,441
the only guides i used are the tutorials for Ryzen memory calculator (the youtube video in the FAQ section) and I watched the two hr video of buildzoids regarding memory overclocking on hte gigabyte K7, and lots of pain and suffering and windows OS installs... you cant really make a guide because of too many variables considered, mostly cooling, motherboard choice, and chosen memory kit. I wont even deal with those that chose to run hynix or micron memory because they simply arent friendly or tweakable in comparison to b die..Ryzen Timing checker is a great way to share your settings but every time I share my RTC with someone else they cant run the same settings... and Ryzen memory calculator takes it a step too far and wants you to change a bunch of settings that will simply brick your board, so it takes some hand on and some knowledge to know what to change and what not to change...

buildzoids ryzen memory overclocking video (CAUTION VERY LONG)

Ryzen memory calculator video (disregard any settings on the far right they will brick your board)
 

Nobu

2[H]4U
Joined
Jun 7, 2007
Messages
3,693
the only guides i used are the tutorials for Ryzen memory calculator (the youtube video in the FAQ section) and I watched the two hr video of buildzoids regarding memory overclocking on hte gigabyte K7, and lots of pain and suffering and windows OS installs... you cant really make a guide because of too many variables considered, mostly cooling, motherboard choice, and chosen memory kit. I wont even deal with those that chose to run hynix or micron memory because they simply arent friendly or tweakable in comparison to b die..Ryzen Timing checker is a great way to share your settings but every time I share my RTC with someone else they cant run the same settings... and Ryzen memory calculator takes it a step too far and wants you to change a bunch of settings that will simply brick your board, so it takes some hand on and some knowledge to know what to change and what not to change...

buildzoids ryzen memory overclocking video (CAUTION VERY LONG)

Ryzen memory calculator video (disregard any settings on the far right they will brick your board)
I didn't have any issues using ryzen memory calc with my hynix memory (other than the usual instability from my GigaB x370 Gaming K5). All (available) settings worked great, and my memory screams at 2933c14-15-14 (right up until it BSODs).
 

Formula.350

[H]ard|Gawd
Joined
Sep 30, 2011
Messages
1,107
Ryzen memory calculator video (disregard any settings on the far right they will brick your board)
You talking about these??

upload_2019-2-26_22-21-43.png



Cuz I think claiming they'll BRICK your motherboard (or RAM) is quite a stretch.

If you meant "Calculate EXTREME", that wouldn't really, either, but its RAM voltage it suggests definitely could fry your modules over time.
The DRAM Voltage and SOC Voltage suggestions are safe and quite helpful though.


That being said, personally, I've had zero luck trying the "Misc Items", "Termination Block Ohms" and "CAD_BUS Block Ohms" on my Titanium. Those for me have always resulted in a speed that would at the very least POST, now fail to POST, which is a bummer.
 

illusio13

n00b
Joined
Feb 20, 2019
Messages
49
the only guides i used are the tutorials for Ryzen memory................

buildzoids ryzen memory overclocking video
Ryzen memory calculator video
Dude :) Thanks for videos, WILL try to get something out of them.........:)

As an aside ........, do not know what i changed in bios ( ...cant be setting NB=1.2v, DDR@1.5v, autoSub@everything...., Maybe -12VOCExpander @10A, mins for overvolt protection, max for undervolt protection, enhanced setting on other two, LLC=1 for both....., think thats everthing..! ).... Voodoo, i say:)......................., so currently @ 3533!
upload_2019-2-27_9-20-33.png


Where before, would boot but like Nobu, BSOD, System-Halt error, Whea etc, etc......, at first application of load across system!

Pissed Off! dudes, DONT know whats going on......................:-(
 

Priller

Gawd
Joined
Sep 8, 2013
Messages
572
Looked at the Ryzen DRAM calculator and changed a couple of settings. It wasn't far off from what I had before.

It's down to 3466 14-14-14-14-28-42-256-1T. Changed procODT from 60 ohms to 53.3.

Strange enough most of my other secondary timings are lower then the FAST setting in the calculator.

I will say changing tRC and tRFC and the procODT was worth 9 points consistently in Cinebench R15.
 

Formula.350

[H]ard|Gawd
Joined
Sep 30, 2011
Messages
1,107
Dude :) Thanks for videos, WILL try to get something out of them.........:)

As an aside ........, do not know what i changed in bios ( ...cant be setting NB=1.2v, DDR@1.5v, autoSub@everything...., Maybe -12VOCExpander @10A, mins for overvolt protection, max for undervolt protection, enhanced setting on other two, LLC=1 for both....., think thats everthing..! ).... Voodoo, i say:)......................., so currently @ 3533! View attachment 144737

Where before, would boot but like Nobu, BSOD, System-Halt error, Whea etc, etc......, at first application of load across system!

Pissed Off! dudes, DONT know whats going on......................:-(
Be sure to run AIDA's Stress Test on "Cache" and "System Memory" to make sure it is actually stable.
 

Below Ambient

[H]ard|Gawd
Joined
Dec 5, 2001
Messages
1,441
I didn't have any issues using ryzen memory calc with my hynix memory (other than the usual instability from my GigaB x370 Gaming K5). All (available) settings worked great, and my memory screams at 2933c14-15-14 (right up until it BSODs).
2933 isnt actually pushing the envelope hence why i dont deal with hynix or micron... anything that doesnt do 3200 out of the box i would consider a disappointment

You talking about these??


Cuz I think claiming they'll BRICK your motherboard (or RAM) is quite a stretch.

If you meant "Calculate EXTREME", that wouldn't really, either, but its RAM voltage it suggests definitely could fry your modules over time.
The DRAM Voltage and SOC Voltage suggestions are safe and quite helpful though.


That being said, personally, I've had zero luck trying the "Misc Items", "Termination Block Ohms" and "CAD_BUS Block Ohms" on my Titanium. Those for me have always resulted in a speed that would at the very least POST, now fail to POST, which is a bummer.
I bricked a Gigabyte Gaming AX370 usingf hte far right column at 1.7vddr

Dude :) Thanks for videos, WILL try to get something out of them.........:)

As an aside ........, do not know what i changed in bios ( ...cant be setting NB=1.2v, DDR@1.5v, autoSub@everything...., Maybe -12VOCExpander @10A, mins for overvolt protection, max for undervolt protection, enhanced setting on other two, LLC=1 for both....., think thats everthing..! ).... Voodoo, i say:)......................., so currently @ 3533!

Where before, would boot but like Nobu, BSOD, System-Halt error, Whea etc, etc......, at first application of load across system!

Pissed Off! dudes, DONT know whats going on......................:-(
the AHOC helped most of all, its more about theory then walking someone through it...the same could be said of any kind of overclocking since you have the silicon lottery and a million other variables, Im glad you got something out of it... 3533 isnt a hard thing to do, just have to be daring with the volts
 

pitingres

Weaksauce
Joined
Jul 25, 2018
Messages
95
Thanks. Gah, why a video? I don't suppose there's a transcript? anyway I'll check it out, thanks.
 

Formula.350

[H]ard|Gawd
Joined
Sep 30, 2011
Messages
1,107
patience is a virtue
Especially with overclocking.
However, I think he meant in terms of being easier to both understand in text, as well as reference when applying the provided info.

I bricked a Gigabyte Gaming AX370 usingf hte far right column at 1.7vddr
I would've loved to hear the autopsy report on that board. I get the feeling this was a fringe situation and unlikely to manifest itself too often. I say this because these are, of course, settings that the manufacturers provide, so they'll have to have run tons of tests of their own.
For that matter, I can't help but think that it was more the 1.7V being fed to the DDR as the cause, given how extreme that is outside of sub-zero runs. In other words, a faulty component in the power delivery chain.

I presume everything else survived the board bricking; CPU and RAM still work fine?

Also, did the board power up but simply fail to POST, or did nothing happen when the power button was pushed?
 

pitingres

Weaksauce
Joined
Jul 25, 2018
Messages
95
patience is a virtue
LOL but using some of the rapidly diminishing store of life hours I have left to sit through a video of someone talking, when I could read the important bits in 10 minutes or less, isn't a virtue it's just a waste.
 

Below Ambient

[H]ard|Gawd
Joined
Dec 5, 2001
Messages
1,441
Especially with overclocking.
However, I think he meant in terms of being easier to both understand in text, as well as reference when applying the provided info.


I would've loved to hear the autopsy report on that board. I get the feeling this was a fringe situation and unlikely to manifest itself too often. I say this because these are, of course, settings that the manufacturers provide, so they'll have to have run tons of tests of their own.
For that matter, I can't help but think that it was more the 1.7V being fed to the DDR as the cause, given how extreme that is outside of sub-zero runs. In other words, a faulty component in the power delivery chain.

I presume everything else survived the board bricking; CPU and RAM still work fine?

Also, did the board power up but simply fail to POST, or did nothing happen when the power button was pushed?
The board was functional, but vecam
 

Below Ambient

[H]ard|Gawd
Joined
Dec 5, 2001
Messages
1,441
I bricked the board by trying to overclock...good learner's board too very basic buis with offset voltage and no limits, just bad vrms from a b swries
 

DogsofJune

2[H]4U
Joined
Nov 7, 2008
Messages
3,531
I got a silly question, whiskey induced blockage perhaps.

But, I used Thaiphoon and have my ram info, Manufacturer is Hynix. The selections in the Ryzen Calc ask if it's AFR, MFR, or CJR. Nothing in Typhoon says anything about those three selections.
Anyway to verify, or perhaps I simply am overlooking it.

snap32201960140PM.png
 
Last edited:

DogsofJune

2[H]4U
Joined
Nov 7, 2008
Messages
3,531
Well shit, That's what I went with anyways. Saw little gains over stock right now, but still tweaking.


I figured it was right there in front of me. In more of a hurry to fiddle with stuff than anything.

Ryzen Memory.png
 

Formula.350

[H]ard|Gawd
Joined
Sep 30, 2011
Messages
1,107
Well shit, That's what I went with anyways. Saw little gains over stock right now, but still tweaking.


I figured it was right there in front of me. In more of a hurry to fiddle with stuff than anything.

View attachment 145567
Do you have a kit of 4? Cuz you only have 2 selected in the program (DIMM Modules). Whether or not that'll change anything, I dunno.
 

Formula.350

[H]ard|Gawd
Joined
Sep 30, 2011
Messages
1,107
Given my lack of skills with overclocking (or maybe just my poor luck of the draw with my CPU's IMC), I've gone back to timings. Given I've worked on my Subs fairly hard in the past, not touching anything else really, I've opted to try whittling down the primary timings. On the latest BIOS (1.J0) I've been running those just at mostly stock 15-15-15, but 30 instead of 35. I've previously had it stable at 14-14-14-30, so I'm fairly sure that'd be fine on this new BIOS.

Long story short, I'm given 'er the beans right now at 13-13-13-30 1.40V :p (I typically like to stay 2-over on the tRAS, at least for starters).

I will note that I had gotten into one HELL of an issue with POSTing initially because I forgot to: #1 Disable GearDownMode to allow for Odd-CAS latency. #2 forgot to bump the voltage from the 1.36V I usually use.
So it would get just far enough into the POST that the BIOS wouldn't recognize a no-POST condition and auto-restart. Similarly, it wouldn't count it as a no-POST situation and so the "5 restarts" safety wasn't registering :| I don't advise my method of rectifying that, so I won't mention what I did lmao I was too lazy to Clr_CMOS, which is what you should do in this situation. As my dad would say: "Do as I say, not as I do." heh

As of posting, it's passed 15mins of AIDA Memory stress. Which, being it's 1:40am, I'm going to call that good for now (usually would've failed already). Tomorrow I'll do some proper HCI Memory Test and let it pass for 200%, that is if the computer wakes up from sleep at those timings. :p If it passes, I'll then try 12's

NOTE: AIDA's Cache-Mem is no better at 13's than at 15's, with the possible exception of Copy speed being higher. However, I'm hoping gaming performance may improve a bit, but we'll see.


UPDATE: Alright, today is the morrow (as AvE would say) and I've ran my Memtest as promised. I managed to get distracted (a good thing in this instance) which has allowed it to run way longer than I planned, which yielded more useful data.
Interestingly, not all 16 threads manage to run through the memory as fast as the next, which has me wondering if maybe there's some Page Filing going on. As such, it leaves me thinking perhaps the upcoming result may not be too big of a deal.

Completion range: 440% to 500% using 953MB allocation per thread (16 instances opened)
Errors: only 2, at roughly 275%, on different threads, but at the same time. (Solar flare? lol)

Configuration:
MSI X370 Titanium - BIOS 1.J0
Ry 7 1700X @ 3.85GHz - 1.35V
CPU-NB (aka CPU-SOC) - 1.1125V
RAM @ DDR4-3200 - 1.40V
Timings: 13-13-13-30 ProcODT 53.3Ω, GearDownMode Disabled, BankGroupSwap Disabled (Alt is enabled, MSI removed that option claiming it's APU-only)
Subtimings:
upload_2019-3-6_17-24-34.png


That's an older RyzenTimingsChecker but same Subs and I post from a laptop, so this is just easier. Also those three corrections on the right are more or less irrelevant since as far as I know, no motherboard provides those for adjustment (automagically set based on other timings).

All non-mentioned drive strengths and what not are set to Auto.

Anyways, I'm rather certain that even if those 2 errors are legitimately the RAM, that it's going to be perfectly fine for 99% of things, like gaming. EDIT: Ok maybe not, surprisingly. Fallout 76 has crashed twice in a rather short time, compared to its usual (which is not as often, go figure)
 
Last edited:

Below Ambient

[H]ard|Gawd
Joined
Dec 5, 2001
Messages
1,441
needs more soc voltage..trc is a bit high.. ive found cl13 to be waste of time... 14 is really the sweet spot then lower your trc as far as you can...

im enjoying this btw...
 
Last edited:

Formula.350

[H]ard|Gawd
Joined
Sep 30, 2011
Messages
1,107
needs more soc voltage..trc is a bit low... ive found cl13 to be waste of time... 14 is really the sweet spot then lower your trc as far as you can...

im enjoying this btw...
Are you suggesting it needs more SoC Voltage due to dropping the primary timings, or you just feel it's a little low in general?
Once upon a time I would disagree, as on much older BIOS versions I was perfectly stable at just 1.0V; however, these latest ones are a whole new beast. I'm suspect to think it's on account of the AGESA, but I really don't know.

How do you mean my tRC is low? It's actually higher than the age-old equation of tRP+tRAS=tRC. Once upon a time, I had huge issues getting it lower than that 54 with stability, despite it being higher than the math suggested. Though last night I applied what it 'should' be, 43, and things were otherwise fine. No change in stability for the good, but no change for the bad, either. On that note, I'm annoyed at MSI (or AMD?) for making the tCWL locked to evens, even with GearDownMode disabled. I'd like to have it at 13, but am forced to have it at 12 (I forgot to change that in my previous screenshot, which I'll address right now).
I'll have to try CL14 to compare, but one thing I did feel changed when it came to Fallout 76 was less microstutter and general frame drop. I'm not sure if it's my ancient Win10 version, changes to the game, or my GPU, but it's definitely not as smooth as I think it should be (or once was, which is why I question patches having screwed things up).

lol What aspect are you enjoying? My adventures in futility? :p The fact my tires are spinning but I seem to be going nowhere... haha
I know my lack of knowledge is holding me back somewhat, but I also feel like one of my hardware components may also be.


Honestly, I'd rather get 3333 stable, vs even something insane like CL10, because I'm not seeing any bandwidth change (or Infinity Fabric for that matter) via timings. Going to 3333 bumps Reads by 3000mb/s, Writes by 1500-2000mb/s and Copy by 1000-1500mb/s (with my very slightly relaxed 3200 sub-timings
 

Nobu

2[H]4U
Joined
Jun 7, 2007
Messages
3,693
Are you suggesting it needs more SoC Voltage due to dropping the primary timings, or you just feel it's a little low in general?
Once upon a time I would disagree, as on much older BIOS versions I was perfectly stable at just 1.0V; however, these latest ones are a whole new beast. I'm suspect to think it's on account of the AGESA, but I really don't know.

How do you mean my tRC is low? It's actually higher than the age-old equation of tRP+tRAS=tRC. Once upon a time, I had huge issues getting it lower than that 54 with stability, despite it being higher than the math suggested. Though last night I applied what it 'should' be, 43, and things were otherwise fine. No change in stability for the good, but no change for the bad, either. On that note, I'm annoyed at MSI (or AMD?) for making the tCWL locked to evens, even with GearDownMode disabled. I'd like to have it at 13, but am forced to have it at 12 (I forgot to change that in my previous screenshot, which I'll address right now).
I'll have to try CL14 to compare, but one thing I did feel changed when it came to Fallout 76 was less microstutter and general frame drop. I'm not sure if it's my ancient Win10 version, changes to the game, or my GPU, but it's definitely not as smooth as I think it should be (or once was, which is why I question patches having screwed things up).

lol What aspect are you enjoying? My adventures in futility? :p The fact my tires are spinning but I seem to be going nowhere... haha
I know my lack of knowledge is holding me back somewhat, but I also feel like one of my hardware components may also be.


Honestly, I'd rather get 3333 stable, vs even something insane like CL10, because I'm not seeing any bandwidth change (or Infinity Fabric for that matter) via timings. Going to 3333 bumps Reads by 3000mb/s, Writes by 1500-2000mb/s and Copy by 1000-1500mb/s (with my very slightly relaxed 3200 sub-timings
I would check windows' error log (forget which one) for memory errors. Sometimes, stuttering or general slowness can indicate memory instability, but errors in the log would be a dead giveaway.
 

Below Ambient

[H]ard|Gawd
Joined
Dec 5, 2001
Messages
1,441
Are you suggesting it needs more SoC Voltage due to dropping the primary timings, or you just feel it's a little low in general?
Once upon a time I would disagree, as on much older BIOS versions I was perfectly stable at just 1.0V; however, these latest ones are a whole new beast. I'm suspect to think it's on account of the AGESA, but I really don't know.

How do you mean my tRC is low? It's actually higher than the age-old equation of tRP+tRAS=tRC. Once upon a time, I had huge issues getting it lower than that 54 with stability, despite it being higher than the math suggested. Though last night I applied what it 'should' be, 43, and things were otherwise fine. No change in stability for the good, but no change for the bad, either. On that note, I'm annoyed at MSI (or AMD?) for making the tCWL locked to evens, even with GearDownMode disabled. I'd like to have it at 13, but am forced to have it at 12 (I forgot to change that in my previous screenshot, which I'll address right now).
I'll have to try CL14 to compare, but one thing I did feel changed when it came to Fallout 76 was less microstutter and general frame drop. I'm not sure if it's my ancient Win10 version, changes to the game, or my GPU, but it's definitely not as smooth as I think it should be (or once was, which is why I question patches having screwed things up).

lol What aspect are you enjoying? My adventures in futility? :p The fact my tires are spinning but I seem to be going nowhere... haha
I know my lack of knowledge is holding me back somewhat, but I also feel like one of my hardware components may also be.


Honestly, I'd rather get 3333 stable, vs even something insane like CL10, because I'm not seeing any bandwidth change (or Infinity Fabric for that matter) via timings. Going to 3333 bumps Reads by 3000mb/s, Writes by 1500-2000mb/s and Copy by 1000-1500mb/s (with my very slightly relaxed 3200 sub-timings
soc deals with the voltage required between the cpu and memory and once you hit 3200 i've found anything less than 1.15 crashes all teh time... heres the catch though, every ryzen has its sweet spot.. mine is 1.1875 anything more or less crashes.. i found mine just using aida and leaving my cpu and memory at what i know is a stable speed and then slowly brought it up until it started crashing and just being a little bitch lol... as far as tCWL all of the boards are like that, pretty sure Ryzen is hard coded to be like that, even setting mine to 15 results in 16, 13 being 14, 11 being 12, so on so forth... this was the same with the strix F I have now and the gigabyte I had before... so even numbers it is... ive found my TRC to be one significant factor in overall bandwidth... anything less than 48 and im just trying to get that CL down regardless of bandwidth...
 

Formula.350

[H]ard|Gawd
Joined
Sep 30, 2011
Messages
1,107
I would check windows' error log (forget which one) for memory errors. Sometimes, stuttering or general slowness can indicate memory instability, but errors in the log would be a dead giveaway.
You're not referring to WHEA are you? I've only experience one of those once, and that was a couple days ago (probably while trying 3466), which it produced a dialogue box.

When I get back into Windows I'll take a look.
I'm finding FO76 to be a far more enjoyable way to test stability since I get to actually do something while the "test" is running lol As such, I'm in the BIOS again to fiddle. :p

soc deals with the voltage required between the cpu and memory and once you hit 3200 i've found anything less than 1.15 crashes all teh time... heres the catch though, every ryzen has its sweet spot.. mine is 1.1875 anything more or less crashes.. i found mine just using aida and leaving my cpu and memory at what i know is a stable speed and then slowly brought it up until it started crashing and just being a little bitch lol... as far as tCWL all of the boards are like that, pretty sure Ryzen is hard coded to be like that, even setting mine to 15 results in 16, 13 being 14, 11 being 12, so on so forth... this was the same with the strix F I have now and the gigabyte I had before... so even numbers it is... ive found my TRC to be one significant factor in overall bandwidth... anything less than 48 and im just trying to get that CL down regardless of bandwidth...
Personally, I'm just amazed that I only now am needing this much SoC voltage to get the exact same speeds and timings stable, whereas the Summit Ridge focused BIOSes didn't need so much. I'm definitely not opposed to throwing more voltage at it though!

I can't help but think it could be that, but more-so the inconsistencies in what I've set in the BIOS and what HWiNFO reports. Which on that note, I'm also not totally sure which sensor is the "truth" heh
Example, CPU... There's the CPU VID the CPU Core Voltage, the Motherboard CPU VCore, and the MOSFET VOut.
IDLE:
CPU VID reports 1.35V (which I've set in BIOS),
CPU Core reports 1.337V,
Mobo CPU VCore reports 1.352V,
MOSFET reports 1.338V.

Cache Stress:
CPU VID reports 1.28V,
CPU Core reports 1.337V,
Mobo CPU VCore reports 1.394V,
MOSFET reports 1.338V.

As such, I've been inclinded to believe the MOSFET and CPU Core, since those remain static and I have a manual OC dialed in, so no CnQ or Boost occur.
Which brings me to the SoC, as... well I don't off hand remember what the CPU SoC voltage sensor reports, but the Motherboard SoC reports 1.12V, and MOSFET VOut is 1.109. Same cache load that increases to 1.125V for MoBo and decreases to 1.107V MOSFET (blips down to 1.105V). BIOS is set to 1.1125V

I had previously just bumped up the SoC LLC from Mode 3 to Mode 2, and that caused the MOSFET to only drop to 1.107V instead of 1.105V, which now I'm going to set it to Mode 1 and hope it doesn't droop.
For kicks I'm also going to try ProcODT of 48, but I won't hold my breath. (I know, it's a big no-no to change multiple things since you can't rule out what solved it, but I'm impatient lol)
 

Nobu

2[H]4U
Joined
Jun 7, 2007
Messages
3,693
You're not referring to WHEA are you? I've only experience one of those once, and that was a couple days ago (probably while trying 3466), which it produced a dialogue box.

When I get back into Windows I'll take a look.
I'm finding FO76 to be a far more enjoyable way to test stability since I get to actually do something while the "test" is running lol As such, I'm in the BIOS again to fiddle. :p
That sounds right. Yeah, I prefer gaming to looking through logs too, but I also don't want silent corruption to slowly kill my OS and personal files. ;)
 

Formula.350

[H]ard|Gawd
Joined
Sep 30, 2011
Messages
1,107
That sounds right. Yeah, I prefer gaming to looking through logs too, but I also don't want silent corruption to slowly kill my OS and personal files. ;)
I'm actually looking through Event Viewer quite extensively and besides maybe finding the issue behind why I can't browse the Laptop files over the network on the Desktop or some time (I attributed it to the laptop being vastly more up to date)... I'm not seeing anything that is showing continually logging of events.

So far pretty much everything is what I'd consider normal, where stuff runs and logs at boot, which the only errors being attributed to the P2P stuff above. In this context, that's a good thing (y)



EDIT: Setting the SoC's LLC to Mode 1 now indeed has it reporting as 1.113V from the MOSFET, with no droop under load. Motherboard reports it as 1.136V under load, and the CPU SoC sensor reads 1.12V. So it would seem they've changed around how the modes function as well, since previously Mode 3 kept it level all the time like it is now. Oh well. Maybe that'll indeed help with my issues? *crosses fingers*

UPDATE: So before I realized it...it's an hour and 45mins later with FO76 not even haven crashed. I'll continue crossing my fingers, but if I get to like 11pm without an issue then that's a very good sign!
UPDATE 2: :meh: Made it to 10:12pm and FO76 crashed just as I started fighting a Mythic Deathclaw. I guess that means it's time to throw a little bit more SoC voltage at it as Below Ambient suggested! :p
 
Last edited:
Top