Overclock.net banner

6681 - 6700 of 9590 Posts

·
Overclock the World
Joined
·
2,638 Posts
tRC, tFAW, tRRDS and tRRDL are all not matching the XMP profile, is this normal?
GDM on does autocorrect on the first stage tCK, tWR, tRTP
on the 2nd stage it's the first 5 primaries that are rounded up ~ if it decides that it missmatches
GDM is 2.5T not 1.5T

Try to be GDM off 2T stable at first , till you have your CAD_BUS and RTT settings correct to power your dimms well
so, thankyou for pointing out my timing errors, care to tell if these are correct? ill stress test later.

(edit #3) im not looking to change rtt, or cad_bus strengths. if i set to
what you mention to the others with single rank 4x8 dimms then i have way WAY to many issues
trying to counter. but, seeing how i have no signs of instability issues, id have to assume there
what my board/sticks wishes to have, i probably could adjust but thats a headache trying to balance
Hmm, you really can start to focus on going CL15-15-15
Maybe similar to my timings with high tRRD & tWTR :D
don't forget 4x8 or 2x16 need tRDC_RD / 2 + 2 = tRDWR
~ with added tWRRD delay up to SCL state
* optionally you can run +1 instead of +2 but this needs every other timing to be aligned


Else you can test if RTT_WR /2 & maybe that + RTT_NOM /5 , breaks everything for you
(instability by them will be shown instantly on Error 1,11,10,6,15 ~ or full refusal to post at all)
And you can check if CsOdtDrvStr, 30ohm ~ causes instability on your current set & maybe if you get RTT_WR /2 to work, haha

It looks fine tbh, but it makes more sense to lower primaries and increase the delay between Read and between Write
tRRD & tWTR
instead of slowing down the main operation on the primaries :)
 

·
Registered
Joined
·
347 Posts
Gotcha, thanks! I'm having such a hard time finding stability, I'm actually currently running TM5 Anta trying 3600 CL16 XMP settings for the G Skill B Die kit (I have the 3200 CL14 G Skill B Die kit) and I entered in the XMP 16-16-16-16-36 1.35V, plus the tRC, tFAW, tRRDS and tRRDL and everything else on auto.. hopefully this will work
I have the same kit as yours, but I am running 4x16GB. The B-die kits sometimes need a bit more voltage to get better stability as it also depends on your motherboard layout (even though they are the same daisy chain topology).

Don't be afraid to bump up your DIMM voltage to 1.4V. This is what I can achieve with 4x16GB. So if you are running 2x16GB or 4x8GB, you can go way faster and tighter than my timings.

My DIMM voltage is 1.46V and VSOC is 1.09V.

2482785
 

·
Premium Member
Joined
·
6,167 Posts
Adding my 2 cents too. I noticed all these "proof" shots on the spreadsheet have these tests I pass but yet still have WHEA in the "Event Viewer". No one posts the pic of the Windows "Event Viewer"/ "System" to prove there was actually no WHEA. Even OCCT misses them unless it's truly a crappy RAM timing. Like HWiNFO showed me having WHEA's when I tried some ridiculous RAM timings that also failed DRAM Calc. For instance, I plugged in my pre-set for "stable" 4000/2000 that has @craxton 's settings in it (running 2 X 8GB , not 4 like him but it works). I have no errors, hangs, games run good, but it tosses WHEA 19's like no tomorrow in "Event Viewer" when a RAM test is run where it shows no error and passing.

I'm really interested to see people's "Event Viewer", like I posted one with redacted system name, showing my WHEA history. That would be true proof that these 3800/1900's and 4000/2000 runners are really error free. Even @craxton said he turns off every non-essential service to benchmark. While that's nice so one can pass a test, I think most of us are looking for daily 24/7 type stability.

So if I'm correct and no one can post a clean "Event Viewer" log, running any one of these RAM tests, one can't be positive that the PC is indeed WHEA free. The only no clean "Event Viewer" screen shots I can post happen with 3733/1866 after the 1.2.0.1 update or 3600/1800. Any other higher setting is a WHEA fest with any timing, any voltage but appears stable and passes all these tests that are being shown as "proof" of stability.

I'm WHEA free now but load this up with a game or RAM test and the WHEA 19's flow. The game will run fine for hours, no hitch but I noticed post gamer it will not open the window for AMD Radeon from the task bar to full screen. Everything else is fine. Run another game, surf the web, etc. No issue. check in "Event Viewer" and anything other than surfing, watching a movie, will result in WHEA 19 using 4000/2000.

I noticed some people posting "proof" where they must be using only one core active or serious LN2 to hit 5Ghz on a 5600X consistently. One ran in safe mode with 1 CCX disabled to achieve the score. Neither is proving a 24/7 stable system. None have a screen shot of "Event Viewer" where hidden WHEA's go. Some are using higher BLCK to overcome FCLK limits, would be nice to hear how to do it without making it so your drives still work and GPU. How far can one go? I'm thinking possibly to 102 before a serious issue.

Researching it just seems some are stretching more than clocks and don't want to look in "Event Viewer", call it clean after many RAM tests pass and no BSOD's. Because mine will pass all these tests fine but the Windows log doesn't lie. If error 19 doesn't matter there, then I've been stable at several speeds. All this is regarding results from the 5600X and running higher than 1866 IF without a WHEA 19. Sure, 5900X or 5950X have a much better chance at the very elusive 1900 or 2000 IF. The 5600X and 5800X are not so lucky, IMHO.

Prove me wrong and post some Windows logs that show no WHEA 19's. I'm not even saying people are purposely evading the truth, they assume by these test results there is no WHEA. I bet there's a very small handful of people that have a 5600X or 5800X that have clean logs running normal with a 1900 or 2000 IF in 1:1:1 and a medium sized with a 5900x or higher.

In any case, if you're claiming no WHEA 19 and you have them in the Windows Event Viewer, you're fooling yourself and some people hoping to find the "unicorn". Many will keep chasing after 3800/1900 or 4000/2000 because it seems like nearly all on here can achieve that. However, if the passing standard is providing the results from TM5, HCi, DRAM Calc, y cruncher, etc., showing zero errors found, then I should be posting my results as well. The real proof would look like this pic but error free, so the time/date match the rest. You can see my DRAM Calc shows no errors. View attachment 2482741 View attachment 2482743
I ran OCCT for an hour, my Windows log. It had zero WHEAs before I cleared it for OCCT as well.

2482790
 

·
Registered
Joined
·
21 Posts
GDM on does autocorrect on the first stage tCK, tWR, tRTP
on the 2nd stage it's the first 5 primaries that are rounded up ~ if it decides that it missmatches
GDM is 2.5T not 1.5T

Try to be GDM off 2T stable at first , till you have your CAD_BUS and RTT settings correct to power your dimms well
I have the same kit as yours, but I am running 4x16GB. The B-die kits sometimes need a bit more voltage to get better stability as it also depends on your motherboard layout (even though they are the same daisy chain topology).

Don't be afraid to bump up your DIMM voltage to 1.4V. This is what I can achieve with 4x16GB. So if you are running 2x16GB or 4x8GB, you can go way faster and tighter than my timings.

My DIMM voltage is 1.46V and VSOC is 1.09V.

View attachment 2482785
Thanks guys, I might give it another shot this weekend. Already spent 3 days trying to figure this out, this is the last result before I "gave up".

View attachment 2482646

Pretty loose timings overall, but I was still getting errors out the wazoo..

Currently using the primaries from the 3600CL16-16-16-36 kit with everything else on auto and I finally got some stability. I've lost a lot of hair over this for sure.. I'm not too sad about the small performance gains I've lost out on but I do get envious when I see people with same hardware (5900x, B550 Unify X, 2x16 B die) get to 3800/4000 MHz with 2:1 FLCK, and when I try their settings, I can't even post :confused:
 

·
Registered
Joined
·
150 Posts
@Veii
Ok I tried running off DF C-states and global C-states, but if it makes any difference at all, the results are maybe a bit worse.
Especially on the write speeds, though globally the results I got are about the same.

VDIMM 1,5
CPU VDDP 0,855
CLDO VDDP 0,850
CCD 0,850
IOD 1,050
SOC 1,100

Regular formulas
tRAS= tCL + tRP
tRC= tRAS+tRC
tWR= tRAS - tRCD
tRTP= tWR/2
tRRDS as low as possible
tRRDL > tRRDS, set to 6 to stay even
tFAW= 4 x tRRDS
tWTRS as low as possible
tWTRL = tRRDL+2
tRFC = (5 * tRC) + 8
t..SCLI just left at 5
tCWL I left at 14
tRDWR= (tRCD / 2) +2
tWRRD= X * SCL so I set it the same as SCL
couldn't find any formula on the other RDRD/WRWR settings so I went with Veiis 1/5/5 .. 1/7/7
tcke= 9 with setup 3/3/15
proc ODT I just left as is
RTT set to 7/0/5
CADBUS changed to 40/20/20/20

1T is better, but only runs in GDM on with these settings.

2482806
2482804
 

Attachments

·
Registered
Joined
·
140 Posts
Adding my 2 cents too. I noticed all these "proof" shots on the spreadsheet have these tests I pass but yet still have WHEA in the "Event Viewer". No one posts the pic of the Windows "Event Viewer"/ "System" to prove there was actually no WHEA. Even OCCT misses them unless it's truly a crappy RAM timing. Like HWiNFO showed me having WHEA's when I tried some ridiculous RAM timings that also failed DRAM Calc. For instance, I plugged in my pre-set for "stable" 4000/2000 that has @craxton 's settings in it (running 2 X 8GB , not 4 like him but it works). I have no errors, hangs, games run good, but it tosses WHEA 19's like no tomorrow in "Event Viewer" when a RAM test is run where it shows no error and passing.

I'm really interested to see people's "Event Viewer", like I posted one with redacted system name, showing my WHEA history. That would be true proof that these 3800/1900's and 4000/2000 runners are really error free. Even @craxton said he turns off every non-essential service to benchmark. While that's nice so one can pass a test, I think most of us are looking for daily 24/7 type stability.

So if I'm correct and no one can post a clean "Event Viewer" log, running any one of these RAM tests, one can't be positive that the PC is indeed WHEA free. The only no clean "Event Viewer" screen shots I can post happen with 3733/1866 after the 1.2.0.1 update or 3600/1800. Any other higher setting is a WHEA fest with any timing, any voltage but appears stable and passes all these tests that are being shown as "proof" of stability.

I'm WHEA free now but load this up with a game or RAM test and the WHEA 19's flow. The game will run fine for hours, no hitch but I noticed post gamer it will not open the window for AMD Radeon from the task bar to full screen. Everything else is fine. Run another game, surf the web, etc. No issue. check in "Event Viewer" and anything other than surfing, watching a movie, will result in WHEA 19 using 4000/2000.

I noticed some people posting "proof" where they must be using only one core active or serious LN2 to hit 5Ghz on a 5600X consistently. One ran in safe mode with 1 CCX disabled to achieve the score. Neither is proving a 24/7 stable system. None have a screen shot of "Event Viewer" where hidden WHEA's go. Some are using higher BLCK to overcome FCLK limits, would be nice to hear how to do it without making it so your drives still work and GPU. How far can one go? I'm thinking possibly to 102 before a serious issue.

Researching it just seems some are stretching more than clocks and don't want to look in "Event Viewer", call it clean after many RAM tests pass and no BSOD's. Because mine will pass all these tests fine but the Windows log doesn't lie. If error 19 doesn't matter there, then I've been stable at several speeds. All this is regarding results from the 5600X and running higher than 1866 IF without a WHEA 19. Sure, 5900X or 5950X have a much better chance at the very elusive 1900 or 2000 IF. The 5600X and 5800X are not so lucky, IMHO.

Prove me wrong and post some Windows logs that show no WHEA 19's. I'm not even saying people are purposely evading the truth, they assume by these test results there is no WHEA. I bet there's a very small handful of people that have a 5600X or 5800X that have clean logs running normal with a 1900 or 2000 IF in 1:1:1 and a medium sized with a 5900x or higher.

In any case, if you're claiming no WHEA 19 and you have them in the Windows Event Viewer, you're fooling yourself and some people hoping to find the "unicorn". Many will keep chasing after 3800/1900 or 4000/2000 because it seems like nearly all on here can achieve that. However, if the passing standard is providing the results from TM5, HCi, DRAM Calc, y cruncher, etc., showing zero errors found, then I should be posting my results as well. The real proof would look like this pic but error free, so the time/date match the rest. You can see my DRAM Calc shows no errors. View attachment 2482741 View attachment 2482743
This was my case and that's why I settled for 3733CL14 rather than 3800CLanything. IF 1900, eventually, sometimes after few days it'd register a WHEA 19 warning, single one, at random occasions, so it was not like "I game or stress and I get errors" but they were there. I gave up and tuned my IF1866 settings, it has been months since I did that and 0 whea errors.
 

·
Overclock the World
Joined
·
2,638 Posts
Thanks guys, I might give it another shot this weekend. Already spent 3 days trying to figure this out, this is the last result before I "gave up".

View attachment 2482646

Pretty loose timings overall, but I was still getting errors out the wazoo..

Currently using the primaries from the 3600CL16-16-16-36 kit with everything else on auto and I finally got some stability. I've lost a lot of hair over this for sure.. I'm not too sad about the small performance gains I've lost out on but I do get envious when I see people with same hardware (5900x, B550 Unify X, 2x16 B die) get to 3800/4000 MHz with 2:1 FLCK, and when I try their settings, I can't even post :confused:
It's A LOT , of work :)
It looks easy when the result is "just stable"
Check the DATA page for information on what each error means
it's average, but i think quite useful
@Veii
Ok I tried running off DF C-states and global C-states, but if it makes any difference at all, the results are maybe a bit worse.
Especially on the write speeds, though globally the results I got are about the same.

VDIMM 1,5
CPU VDDP 0,855
CLDO VDDP 0,850
CCD 0,850
IOD 1,050
SOC 1,100

Regular formulas
tRAS= tCL + tRP
tRC= tRAS+tRC
tWR= tRAS - tRCD
tRTP= tWR/2
tRRDS as low as possible
tRRDL > tRRDS, set to 6 to stay even
tFAW= 4 x tRRDS
tWTRS as low as possible
tWTRL = tRRDL+2
tRFC = (5 * tRC) + 8
t..SCLI just left at 5
tCWL I left at 14
tRDWR= (tRCD / 2) +2
tWRRD= X * SCL so I set it the same as SCL
couldn't find any formula on the other RDRD/WRWR settings so I went with Veiis 1/5/5 .. 1/7/7
tcke= 9 with setup 3/3/15
proc ODT I just left as is
RTT set to 7/0/5
CADBUS changed to 40/20/20/20

1T is better, but only runs in GDM on with these settings.

View attachment 2482806 View attachment 2482804
I have something interesting to share soon - something to abuse with DF states
for now DF states should be disabled :)
You can use overboost to your favor
With DF States utilizing all boost possible
Without DF states, if it's too unstable for some
More information in couple of days 🚀
Currently it's a bit too explosive and causes issues for CTR users :censored:
 

·
Overclock the World
Joined
·
2,638 Posts
Thanks guys, I might give it another shot this weekend. Already spent 3 days trying to figure this out, this is the last result before I "gave up".

View attachment 2482646

Pretty loose timings overall, but I was still getting errors out the wazoo..

Currently using the primaries from the 3600CL16-16-16-36 kit with everything else on auto and I finally got some stability. I've lost a lot of hair over this for sure.. I'm not too sad about the small performance gains I've lost out on but I do get envious when I see people with same hardware (5900x, B550 Unify X, 2x16 B die) get to 3800/4000 MHz with 2:1 FLCK, and when I try their settings, I can't even post :confused:
Balancing issues

But both of your issues say "timeout issue, lack of voltage or resistance"
This should be better for the start :)
 

·
Registered
Joined
·
23 Posts
After (a lot) of trial and error I got GDM 2T stable. tRFC seems to cap out around 285-290ns and tRC I can boot 54 with errors, but any lower and gotta reset CMOS. Any other suggestions?
2482830
2482831
 
  • Rep+
Reactions: Veii

·
Overclock the World
Joined
·
2,638 Posts
After (a lot) of trial and error I got GDM 2T stable. tRFC seems to cap out around 285-290ns and tRC I can boot 54 with errors, but any lower and gotta reset CMOS. Any other suggestions?
View attachment 2482830 View attachment 2482831
These timings by cm87 where good for Rev.E :)
2482842


Please compare how they perform
i think your RTTs are good , way better than RTT_PARK 1

Also good job !
Finally you can start to work with awkward experimental timings :giggle: (y)
 

·
Registered
Joined
·
21 Posts
It's A LOT , of work :)
It looks easy when the result is "just stable"
Check the DATA page for information on what each error means
it's average, but i think quite useful
I've been using this, this is a great resource, thanks for working on it :)

Balancing issues

But both of your issues say "timeout issue, lack of voltage or resistance"
This should be better for the start :)
I'll be sure to try those, thanks!
 

·
Overclock the World
Joined
·
2,638 Posts
I've been using this, this is a great resource, thanks for working on it :)



I'll be sure to try those, thanks!
If tWRRD 5 fails, 4 will work
If CAD_BUS 40-20-30-20 fails, 60-20-20-24 will work :)
If you still have random issues , for example error 11 or 1 - then give tRTP 14 a shot (half of tWR)
and if you still won't get error 1 or then maybe 10 away,you have to increase tRRD & tWTR (no way around it)
or just increase VDIMM +1

If you Error 0,10,6,1 (#1 together with the rest) then you need to change RTT or CAD_BUS (both likely)
 

·
Registered
Joined
·
21 Posts
If tWRRD 5 fails, 4 will work
If CAD_BUS 40-20-30-20 fails, 60-20-20-24 will work :)
If you still have random issues , for example error 11 or 1 - then give tRTP 14 a shot (half of tWR)
and if you still won't get error 1 or then maybe 10 away,you have to increase tRRD & tWTR (no way around it)
or just increase VDIMM +1

If you Error 0,10,6,1 (#1 together with the rest) then you need to change RTT or CAD_BUS (both likely)
2482847


Something is strange though, I even checked twice and tWR is for sure 28 in the BIOS..

And according to 13 it's most likely voltage related, I'll bump up VDIMM slightly and change my tCWL to 14 like @domdtxdissar mentioned
 

·
Overclock the World
Joined
·
2,638 Posts
View attachment 2482847

Something is strange though, I even checked twice and tWR is for sure 28 in the BIOS..

And according to 13 it's most likely voltage related, I'll bump up VDIMM slightly and change my tCWL to 14 like @domdtxdissarn mentioned
ZenTimings reads what the bios sets
it's strange
Will tWR 14 post at all ?
also try tWRRD 4 afterwards
 

·
Registered
Joined
·
315 Posts
These timings by cm87 where good for Rev.E :)
View attachment 2482842

Please compare how they perform
i think your RTTs are good , way better than RTT_PARK 1

Also good job !
Finally you can start to work with awkward experimental timings :giggle: (y)
Isn't tRAS = tCL + tRCDRD + x? That tRAS is too low even without the +x?

I've read +2 and +4 for tRAS. For my Rev.E 4x8 config on a 5900x, +2 gives worse scores than +4 in latency
 

·
Overclock the World
Joined
·
2,638 Posts

·
Registered
CPU: Ryzen 5 5600X, MB: X470 AORUS GAMING 7 GB, GPU: GTX 1080
Joined
·
37 Posts
Hey, @Veii got viper SR 2x8 16 but i get such " high " latency results, i couldnt get trcdrd to 15 no matter what i did

Thank you!!

2482851
 
6681 - 6700 of 9590 Posts
Top