Advice on Max safe voltage on Mem

EL_JOCK said:
Glad your almost getting it, gives me hope one day i will get it over the line on 3200. Each time i read this thread i spend a few hours trying to get it to work, but no luck so far have tried everything people have said but its getting so far away from normal operation that i think there is more to it than a setting i can bump up or change, have been spending time getting 2933 as super tight as i can, so its not been a total waste of time
Click to expand...

Yeah really strange although 3466 I tried on and off for over a month and then something just clicked.

If you go over to the ryzen DDR4 thread on OCN the OP has a table of memory results with links to screenshots of timings etc. which is a good resource. It's where I found out I should try playing with DRAM termination voltage.

ProcODT has not shifted my F9 codes, I suspect a slow and painful seek through the range of CLDO VDDP values is my only hope :(
 
Hi Ste, 8Pack or anyone having issues with getting 3200 on a Gigabyte X370 Gaming 5 F6 bios. This might work on other system "Hell I'm not sure why it works at all" I did say it was more magic!

Now, please read some of the posts above for more context and good info from 8Pack and Ste, but in short i have a 1600, with 3600 C16 Samsung b die memory, could do 2944 no issues super tight speeds, 3200 was a no go with any volts from 1.35v up to 1.5v, soc 1v upto 1.2v. There just seemed to be a block did not matter how loose the timings went or how much voltage i applied to it.

So today i just went all out turned off my pc pulled the power cord, then the 24 pin, popped out the cmos battery then put a screw driver between the cmos pins held for 5 sec to totally clear the cmos, also pushed the power button a few times to make sure all the charge was out the system. (i think there is a name for this type of reset but not sure)

Any roads following that, plugged it all back together, booted into bios, (says it has been reset), set mem clock to 3200 (by typing it in, not using page up or down). typed in cas 16-16-16-16-36 (just wanted a basic easy timings to start with) all other settings on auto.

Go into voltage type'd in 1.45 volts for Dram & SOC 1.1

Save and exit, No F9 errors "Great", back into bios 1.4v Dram Save and exit, No F9 errors "super", back into bios 1.35v Dram save exit no F9 errors "What its working"

Boot into windows load up Aida64 run stability test as i type this its just passed the 30 minute mark no crashes no issues, this is the best place i have gotten to ever when going for 3200, will update as i do more testing but for what ever reason a full cmos reset seems to have cleared what ever issue the system was having.

take it easy
 
Update to the update,

So now running Aidia64 stabilty test on the following timings ( I know i still have more testing to do before its 100%)

14-14-14-14-32 @ 1.35v Soc 1.1

Trc 54
Tfaw 36
Twr 10
trdrdscl 2
twrwrscl 2
Trfc 312
Cmd 1T

Bank Group Swap Disabled
GearDownmode Disabled

Just passed 30 minutes with no issues this is a ground breaking moment for me as it has never come close to this even on high volts and slack timings.

It would appear that the full cmos reset has done something to clear a blockage. I hope this helps/works for others as it has for me, its been so vexing to have this gear and no get the best out of it, 3466 maybe the next mile stone, but need to get this 100% stable then get the CPU up to 3.8 and i will be super happy.

This is where i wanted to be 2 weeks ago, but so glad i can see the light at the end of the tunnel. Super thanks to Ste and 8Pack for the help.

UPDATE

Same memory settings with the CPU @ 3.8 1.32v passed an hour of Aida64, going to start the mem test thing see how it goes on but fingers crossed I'm all sorted.
 
Last edited:
Wow that's interesting. Good work! Glad it's going well, those are good timings too. I suspect you might get 3333 at those timings too, possibly as easy as the next click up... give it a go :)

My 3466 is super stable (1000% HCI with CPU at 3.9) but the cold boot F9 codes are not going anywhere. This morning it F9 coded 3x in a row on startup and reverted to 2133 which sucks. Might stick with 3333 for a bit but I reckon a slow and painful trawl through CLDOVDDP settings might bring success eventually. Just not sure I have the patience :)

No problem RE help let us know if you crack 3466 as then you'll be helping me :D
 
After some more testing got it to this stage for 3200

14-14-14-14-26 @1.35v SOC 1.1

TRC 42
Tfaw 36
TWR 10
TRFC 256
trdrdscl 2
twrwrscl 2

That is running really nice, quick sneak peak on the next line of testing 3333 booted no issue @ 1.35v soc 1.1 & 3466 booted no issues @1.40v (timings about the same as the 3200) but both of them need testing.

Had a little look at 3600 16-16-16-16-36. It did need 1.5v but had some F9 errors , however had loads left i could look at to get that working. If i get time in the morning i will start with 3466 as that might help more

Nice one
 
Wow that's sweet. I can't get near tRFC at that level. I think your c16 kit must be a much better bin than my c17 one. Shame as there is now only a tenner between them but when I bought it was 50!

Be interested to know whether your F9 free 3466 continues as mine was ok and then returned after a 12 hour cold boot.

Finally your 14-14-14-14-26 is interesting as the 26 should be at least the sum of the prior two for stability so 28 as a minimum. Once you are happy try some proper stress testing to see if it's HCI stable at 1000% which is the golden standard. Even if you have to slacken tRFC slightly and up the 26 to 28 you are laughing there serious result! well done :D
 
right o, took on board what you said and altered the settings to 14-13-13-13-26, (have been using "The Stilt's" settings off OCN), i really like that Ryzen timing checker program easy to see in windows what the numbers are.

Just running the memtest thing (12 instances running 1133 megabytes each), will leave that running this morning, as i'm baking a lemon tart in a bit, lets see if it can get to 1000%, just went over 100% so a few hours left to run.

Had a little play with 3466 this morning it boots ok, but its not stable, need to look at timings etc, what volts did you have for Dram and SOC?

Will keep ya posted!
 
Well, whats the thoughts on this, ran the mem test as outlined above, on the settings (14-13-13-13-26) (1133 megabytes per) 11 of the 12 went over 1000% as i was waiting for the last thread to hit 1000% before stopping all threads, i had 1 error at 995% on the last thread,

worth trying 14-14-14-14-28, or at that level is 1 error ok at that % of pass?
 
Personally I have not had a system as picky as ryzen on memory so I would slacken slightly for stability.

For 3466 I am 1.45 and 1.175. The latter can almost certainly be lowered but until I solve the F9 issues I can't see the point.

Your RAM is better than mine but I run 14-14-14-14-38-54 with 356 tRFC so perhaps concentrate on the 38, 54 and 356 but you shouldn't need to go so loose. Perhaps 36 50 and 312/333.
 
Ok, made a bit of time today in-between reading my Psychology book to do some testing at 3333 mem speeds.

Have put my CPU back to 3.8 as the jump to 3.9 put my Aidia full load temp up to 64c, which is ok, I just like my fan profile on the Kraken x62 as its low noise and keeps 3.8 just under 60c, don't want to make more noise with an increased fan profile.

So with that in mind,

Mem 3333@ 14-14-14-14-30-52 @1.35v SOC 1.1 (1T Gear down disabled / bank group disabled)

Tfaw 36
twr 12
Twtrl 12
trdrdscl 2
twrwrscl 2
trfc 267 (which give trfc (ns) 160.2)

12 instances of Mem test (1100 megabytes) running, it hit 1000% no errors

CinebenchR15 multi cpu test gives me 1296

Think i will leave it there for this week, need to start enjoying the PC a bit, however i do need a new GPU the 290 is a bit long in the tooth now.
 
I settled on 3333 c14 as it is completely stable and I was spending too much time testing higher speeds to get them completely stable.

Strangely they would pass bench marks but still cause bf1 to crash. So if bf1 was fine as well I took it as stable.
 
EL_JOCK said:
Ok, made a bit of time today in-between reading my Psychology book to do some testing at 3333 mem speeds.

Have put my CPU back to 3.8 as the jump to 3.9 put my Aidia full load temp up to 64c, which is ok, I just like my fan profile on the Kraken x62 as its low noise and keeps 3.8 just under 60c, don't want to make more noise with an increased fan profile.

So with that in mind,

Mem 3333@ 14-14-14-14-30-52 @1.35v SOC 1.1 (1T Gear down disabled / bank group disabled)

Tfaw 36
twr 12
Twtrl 12
trdrdscl 2
twrwrscl 2
trfc 267 (which give trfc (ns) 160.2)

12 instances of Mem test (1100 megabytes) running, it hit 1000% no errors

CinebenchR15 multi cpu test gives me 1296

Think i will leave it there for this week, need to start enjoying the PC a bit, however i do need a new GPU the 290 is a bit long in the tooth now.
Click to expand...

That's great and with tRFC at that level a great result. I can't even boot at that!

I have given up on the F9 situation at 3466 for now. Not least because the constant testing and failed boots are probs killing my hard drives off with the constant spin up and down! When I have a good chunk of time I'll try again but I think 3333 will do me for now. Might try for a tighter tRFC sometime though.
 
Ste said:
That's great and with tRFC at that level a great result. I can't even boot at that!

I have given up on the F9 situation at 3466 for now. Not least because the constant testing and failed boots are probs killing my hard drives off with the constant spin up and down! When I have a good chunk of time I'll try again but I think 3333 will do me for now. Might try for a tighter tRFC sometime though.
Click to expand...


Well bit of an update, it passed all the above then i was watching a video and my PC hung blue screen reboot F9 errors, now it could be my windows 10 install is messed up after all the fails and i was going to reinstall it on Tuesday. However i did notice today after testing was done i played a few rounds of PUBG and at random times my fps just tanked, i thought it was just the game (early access and all) however nobody else i was playing with had an issue, it only happened for a few seconds but i had never had it happen like that before/

So right now back on 3200, will reinstall windows 10 on Tuesday to rule that out then back at it i guess
 
I've had that. A couple of others have too. Unless your aggressive tRFC has broken something (wouldn't cause lock up mind as it'd be corruption) it's just unstable memory.

Your timings are really aggressive though. I'd stick with 3333 but start to loosen slightly.

Also your CPU overclock may be pushed with the higher memory. I put vcore up a notch for 3333 just to be safe.
 
Could be however i went back to 2933 and it just kept happening, so did the full cmos reset again, put my settings back in with a bit more slack on things and its been OK, but need to run tests again!!!
 
cool, will set some testing aside tomorrow, have put those settings in and saved a profile. It booted first go so that's a good start
 
EL_JOCK said:
cool, will set some testing aside tomorrow, have put those settings in and saved a profile. It booted first go so that's a good start
Click to expand...

Yeah just done the same and set a HCI run off before work. Also booted straight in. Will check in at lunchtime...
 
I booted into 3466 fine this morning but it fell over on Aida64 within minutes. I used Dram at 1.45v as i did not know what they had used and thought well i can go down from that.

Currently sat with this and its working well for now, might look at 3333 again but think wait for new bios as on the gigabyte forum it appears Bios code 1.007 is on its way, so it might smooth things out

QWWGxBS.jpg.png
 
You must log in or register to reply here.
Back
Top Bottom