Trying to get my ram higher than 2666mhz and stable on new build

[kevintuna]

Soooooo, bad update =(

Every 2 hours of gaming, I'd get a blue screen of death. They were all "kernal 41", or memory dump related BSOD.

It appears that's usually the PSU or the ram. First thing I did was take the overclock off my CPU ( I did get some clock time out BSOD related to overclock, believe I just needed to increase voltage ).

Anyways, still happened with CPU at stock. So I thought i'd make sure everything with the PSU was connected well, so I re-connected everything. Still happening.

I then re-formatted and updated all drivers, still happened.

Decided to run a memtest with both sticks, on XMP at 3600. Memtest failed test 6 with over 400 errors.

memtest each stick individually on XMP at 3600, in different ram slots as well, and all those tests pass.

I now can't get XMP to boot while using the two sticks at all. I've tried increasing voltages a decent amount, no go. Kind of worried since it failed memtest anyways.

I've got both sticks running at 3333mhz ( as high as it will boot ), and it has passed memtest several times, and so far, no more BSOD ( only been a day though ).

Really unsure where to go at this point, I'm fairly happy with 3333mhz if it doesn't produce any errors, but now I'm just fairly confused to whats going on.

- - - Updated - - -

test*** I'm sorry if I have a bunch of spam or random posts, I was just trying to post my reply here once every few hours and see if it would work, and now they all seem to be showing up? I've created a mess!

 

[wingman99]

What does HWiNFO show for Vccsa and Vccio?
I had my memory overclocked to 3733 17-17-17-37 with passing memtest86 and after a while of gaming I had errors in memtest86 like you, so now it's 3600 15-15-15-35.

 

[kevintuna]

https://imgur.com/a/lvWFI

that's where I'm at right now, but I'm only at 3333mhz now, with 16-16-16-36, so you're doing better than mine.

Definitely lost the blue screens though.


5.0ghz on the master evo?! impressive man! I thought I was doing decent pushing it to 4.6, what voltage did you push that to?

Also, what kind of errors were you getting while gaming with the ram? Wondering if it's similar to mine.

 

[Johan45]

HWmonitor isn't showing the voltages we want to see. You said you tried adding voltage but what did you set? I would try IO at 1.27V and SA at 1.25V and see if it'll handle the 3600 then

 

[kevintuna]

HWmonitor isn't showing the voltages we want to see. You said you tried adding voltage but what did you set? I would try IO at 1.27V and SA at 1.25V and see if it'll handle the 3600 then

Just tried this, with DRAM to 1.385, no boot on XMP, or manually setting it to 3600. Tried 3400 as well, still no boot.




I previously was at 1.15V IO, 1.15V SA, 1.35V DRAM, 1.25V CPU core, on XMP, when it did boot, but i was getting BSOD / memtest fail.

I'm at 3333mhz, 1.15V IO, 1.15V SA, 1.35V DRAM, 1.25V CPU core, boot + pass memtest / now 20+ hours of gaming and no BSOD.

I find it weird I was able to boot on XMP before, and now no matter what voltage I raise I can't boot above 3333mhz.

Computers are crazy complicated things lol


edit *** the forum is working today, yay!

 

[Johan45]

VCCSA and VCCIO higher?

 

[kevintuna]

VCCSA and VCCIO higher?

how high do you think is safe? at 1.27 and 1.25 it's showing up as pink which I assume means danger in the bios

 

[Johan45]

That's just a mild warning. That should be fine even with air cooling
You can start lower (1.2 ish) and see how it goes

 

[wingman99]

https://imgur.com/a/lvWFI

that's where I'm at right now, but I'm only at 3333mhz now, with 16-16-16-36, so you're doing better than mine.

Definitely lost the blue screens though.


5.0ghz on the master evo?! impressive man! I thought I was doing decent pushing it to 4.6, what voltage did you push that to?

Also, what kind of errors were you getting while gaming with the ram? Wondering if it's similar to mine.

With my memory at 3733 speed I had one BSOD, then I did memtest86 again and this time it showed errors and my AUTO Vccsa is 1.28v, so I tried 3600 15-15-15-35 with memtest86 and it passed, since then my PC has been running trouble free. It is interesting for me also why it passed memtest86 at 3733 when my setup was new, after running stress testing for 24 hours then the memory system throws errors.

Here is my setup the first 4 hours and the last time I will run windows at 3733 speed because now I have errors in memtest86 and it would corrupt the windows install. When it comes to memory overclocking I always test memtest86 first before using windows.

 

[Woomack]

Memtest86 is not testing memory controller and cache stability, all is linked and we were talking about it many times. Simply, use multithreaded stability tests in Windows.

 

[wingman99]

The new Memtest86 works wonderfully checking for errors with the memory controller and memory, it is very consistent when retesting for the same errors with the same voltages and settings. I just spent the whole afternoon yesterday tweaking the memory controller voltage first to see if I could run 3733 17-17-17-37 speed memory and I was able to. Then I tweaked the memory controller voltages to find a safe margin, also not to overvolt the memory controller and memory. So now after testing with new memtest86 hopefully these voltages Vccio 1.150v Vccsa 1.285v Dram 1.40v are good, I just have one last 2 hour run with memtest86.

Testing memory in windows is not simple and takes weeks to achieve 100% stability in all running conditions without overvolting and I would have corrupted my windows install by now over and over.

 

[Johan45]

Is that right? 1.5V VCCIO, seems a tad high

 

[wingman99]

Is that right? 1.5V VCCIO, seems a tad high

Thanks it is a typo, I fixed it.

 

[Woomack]

The new Memtest86 works wonderfully checking for errors with the memory controller and memory, it is very consistent when retesting for the same errors with the same voltages and settings. I just spent the whole afternoon yesterday tweaking the memory controller voltage first to see if I could run 3733 17-17-17-37 speed memory and I was able to. Then I tweaked the memory controller voltages to find a safe margin, also not to overvolt the memory controller and memory. So now after testing with new memtest86 hopefully these voltages Vccio 1.150v Vccsa 1.285v Dram 1.40v are good, I just have one last 2 hour run with memtest86.

Testing memory in windows is not simple and takes weeks to achieve 100% stability in all running conditions without overvolting and I would have corrupted my windows install by now over and over.

Have you seen that info on wikipedia or something ? ... sorry but you read too much Internet and later repeat this gained knowledge in various posts. Big part of your posts are based on what you've seen on other forums and not much is your actual experience.

Even new memtest is not loading memory in 90%+ and is not testing memory controller and cache under full load. It's barely using multithreading and on some chipsets is not providing accurate results. You can use it to check memory cell by cell but even if everything passes then you may see instability in OS. It's because of how new chipsets work. You test it in OS also because there can be issues with high frequency memory allocation and stability. There is reason why memory vendors are not showing screenshots from memtest but they base on multithreaded Windows software while showing stability of higher frequency memory kits.

I had no corrupted Windows because of memory overclocking for a long time and I push memory up to 4500+. There are no issues like that if you have basic knowledge about memory settings and make all step by step.

 

[wingman99]

Have you seen that info on wikipedia or something ? ... sorry but you read too much Internet and later repeat this gained knowledge in various posts. Big part of your posts are based on what you've seen on other forums and not much is your actual experience.

Even new memtest is not loading memory in 90%+ and is not testing memory controller and cache under full load. It's barely using multithreading and on some chipsets is not providing accurate results. You can use it to check memory cell by cell but even if everything passes then you may see instability in OS. It's because of how new chipsets work. You test it in OS also because there can be issues with high frequency memory allocation and stability. There is reason why memory vendors are not showing screenshots from memtest but they base on multithreaded Windows software while showing stability of higher frequency memory kits.

I had no corrupted Windows because of memory overclocking for a long time and I push memory up to 4500+. There are no issues like that if you have basic knowledge about memory settings and make all step by step.

What I said above is what my experience is with memtest86 not what I read. I have proof my overclocked memory passed the software usage in the screenshot above in this post and failed memtest86. Also other folks in the forums use it for memory overclock testing. Memory errors corrupting windows has nothing to do with knowledge, it's memory communication errors that cause it.

If you don't use the new memtest86 where do you get your information? Memory data runs full speed with the memory controller all the time, If there is any errors my years of testing shows memtest86 finds it. I'm unable to run my memory at 3733 with decent timings and votages 17-17-17-37 without getting errors in memtest86. However memtest86 passes 3600 15-15-15-35 with no problem.

Have you tried the new memtest86? It has 13 different RAM testing algorithms, like hammer test. How do you know memtest86 is not utilizing the memory along with the controller 100%?

Memtest86 is for overclocking.

Why use MemTest86
Unreliable RAM can cause a multitude of problems. Corrupted data, crashes and unexplained behaviour. Bad RAM is one of the most frustrating computer problems to have as symptoms are often random and hard to pin down. MemTest86 can help diagnose faulty RAM (or rule it out as a cause of system instability). As such it is often used by system builders, PC repair stores, overclockers & PC manufacturers. https://www.memtest86.com/

 

[Woomack]

People who are testing higher frequency memory like memory vendors and many extreme overclockers are not even touching memtest86 for the reasons mentioned in my last posts.
I was testing memtest86 in the latest versions and it was showing me no errors many times while Windows tests like AIDA64 or even HyperPi 32M were crashing. As I already said, we were talking about it many times on the OCF. Memtest is not bad diagnostic tool but you have to know how to use it.
I would use memtest86 to check if memory has no errors at stock/declared speed but not at 3600+ using higher capacity modules.
Often issues with high memory frequency are not because memory is faulty or can't run at so high frequency but because signal quality is not good enough and Windows can't allocate high memory capacity without errors. This is also why for extreme overclocking at tight timings all are trying to limit memory capacity or use 32bit OS. This is also why memory PCB and traces design count.

Windows installation can't be corrupted if you overclock with small steps as I already said. Your answer only proves that you don't know how Windows works. Blue screen is only a warning before anything happens... something like protection so OS won't crash or so you won't lose data. Next step is data corruption. If you are pushing clearly unstable settings and multiple times try to run something then OS may crash. If you know basics of overclocking or used IC settings then usually there are no issues.

When they say that memtest is often used by overclockers then it doesn't mean it's for overclocking ... it's just used by overclockers. As I said, it's not a problem to use it but you can't base only on memtest86 while overclocking memory.
I said that in the past and will repeat again, I wish to see more of your original comments based on your actual tests and experience, not something based on other users comments or quotes from the Internet.

 

[wingman99]

I know how PCs work, you clearly don't understand from the things you say. It only takes one error to write to the wrong address on the HD or corrupt date that is being stored at a address. Memory with the IMC then CPU works with NTFS table stored in memory and if there is a error while writing to the drive there will be corruption. Machine-check exception is a hardware error that the processor detects and windows WHEA reports it. Not all parity errors are detected in transfer of data. If the memory is unstable do to overclocking the data can be corrupted. However, the parity bit for the bus transfer can be correct. Same with overclocking processor with corrupted calculations being undetectable by Machine-check. Machine-check exception detects system bus errors, memory parity errors with ECC, Cache parity errors. It does not detect miscalculations by the processor or corrupted data stored or retrieved in the overclocked memory sticks or processor Chache. That is why we use check sum software like memtest86, prime95 and Aida64.

Like I said above I have proof here in the thread with a screenshot of RealBench running 8GB memory at 3733 17-17-17-37 passing 4 hours. Then memtest86 failed constantly. And I've been in the forum every day for years and your the only one that has said not to use memtest86 for memory overclocking testing.

 

[Woomack]

This thread is already out of topic for longer and I see no point to continue this conversation as it clearly won't change anything.
Later I wonder why I get more pms with help requests than I see new threads on the forums.

 

[kevintuna]

I kind of just kept lurking and let it go off topic

You're both extremely knowledgeable, no need to have a pissing match with who knows more, when you both know more than 99.9 percent of computer users. I really appreciate both of your help and knowledge that has been posted in this thread.

I've still had my memory set to 3333mhz with zero crashes or BSOD... and I've really enjoyed gaming / surfing with zero computer issues.

Spent dozens of hours on this ram problem and it probably stresses me out more than it should.

I'd love to get my ram up to full speed, but I think I'll go back and visit it another time. Having way too much fun getting destroyed by teenagers in PUBG currently.

 

[wingman99]

Glad to hear you are enjoying the PC. 333Mhz is plenty fast for gaming and anything above that does not help much in gaming. I play battle field one whenever I can.