Mobo Auto Settings for Ram Frequency overclock?

[richintheveins]

So for the past two weeks I have been figuratively and at times literally banging my head up against the wall attempting to stabilize HyperX Predator 4000 MHZ ram. I have used the XMP profiles (1 and 2) only to have instability issues (Freezing, Game Crashing, BSOD). I thought maybe it was defective ram but after tweaking the DRAM, Viccio and SA plus timing adjustments (16,16,16, 36 & 17, 17, 17, 38) I was able to run Memtest86 with no errors for 8 passes. Due to it being 4 sticks at a total of 32 gb of ram this process took around 24 hours. This left me even more perplexed to still have BSOD and crashes and even boot issues causing the system to automatically go back to default settings. 

 

I have verified that the instability comes from the ram because if I run it at 3200mhz or less all the instability goes away. 

 

I was at a loss so I back tracked my BIOS back to 1302, went into command prompt and ran the SFC Scan and the DISM Tool to restore Health. No help....

 

Last ditch effort I went into the Bios set everything to Auto and increased my frequency to 4000 mhz saved it and restarted in hopes my damn Mobo would tell me what it wanted to run it at this speed or just crash and I rip it out and move on.....

 

but strangely it booted.....

Strangely it chose pretty good settings on its own 1.35 DRAM, 1.25 IO, 1.25 SA. 

DRAM Timings 17,17,17,39 (the XMP profile is way looser)

 

When i got in I checked for any more updates on my windows (updated it to 1909) 

checked CPU-Z to verify that I wasn't dreaming annnnnd bam restarted it and it booted in really quick and I ran Aida64 for a memory stability test for over 4 hours and then played Division 2 and RE2 with no crashes and no BSOD for another 4 hours. 

Just to make sure I even did a cold boot by unplugging the power letting it sit for 5 min and then plugged it back in and still no shutting on and off just straight into windows. 

 

I'm not trying to say manual Overclocking is bad or anything but if anyone wants to try it out (if all else fails) it worked for me. Putting the tweaker on Manual and only changing the frequency allowed my Mobo to tell me what it needed to be stable. I'll continue to monitor for errors but so far so good. 

 

Still confused though! lol 

[LIGISTX]

4 minutes ago, richintheveins said:

So for the past two weeks I have been figuratively and at times literally banging my head up against the wall attempting to stabilize HyperX Predator 4000 MHZ ram. I have used the XMP profiles (1 and 2) only to have instability issues (Freezing, Game Crashing, BSOD). I thought maybe it was defective ram but after tweaking the DRAM, Viccio and SA plus timing adjustments (16,16,16, 36 & 17, 17, 17, 38) I was able to run Memtest86 with no errors for 8 passes. Due to it being 4 sticks at a total of 32 gb of ram this process took around 24 hours. This left me even more perplexed to still have BSOD and crashes and even boot issues causing the system to automatically go back to default settings. 

 

I have verified that the instability comes from the ram because if I run it at 3200mhz or less all the instability goes away. 

 

I was at a loss so I back tracked my BIOS back to 1302, went into command prompt and ran the SFC Scan and the DISM Tool to restore Health. No help....

 

Last ditch effort I went into the Bios set everything to Auto and increased my frequency to 4000 mhz saved it and restarted in hopes my damn Mobo would tell me what it wanted to run it at this speed or just crash and I rip it out and move on.....

 

but strangely it booted.....

Strangely it chose pretty good settings on its own 1.35 DRAM, 1.25 IO, 1.25 SA. 

DRAM Timings 17,17,17,39 (the XMP profile is way looser)

 

When i got in I checked for any more updates on my windows (updated it to 1909) 

checked CPU-Z to verify that I wasn't dreaming annnnnd bam restarted it and it booted in really quick and I ran Aida64 for a memory stability test for over 4 hours and then played Division 2 and RE2 with no crashes and no BSOD for another 4 hours. 

Just to make sure I even did a cold boot by unplugging the power letting it sit for 5 min and then plugged it back in and still no shutting on and off just straight into windows. 

 

I'm not trying to say manual Overclocking is bad or anything but if anyone wants to try it out (if all else fails) it worked for me. Putting the tweaker on Manual and only changing the frequency allowed my Mobo to tell me what it needed to be stable. I'll continue to monitor for errors but so far so good. 

 

Still confused though! lol 

It could be some weird memory controller thing... Remember, you are overclocking your memory controller via pumping the RAM speed, and 4000 is pretty high. When you run your AIDA64 run, I would stress not just the memory, but the entire suite of CPU tests as well. Load the CPU up with as much as you can, RAM as well, basically all the default checked boxes in AIDA's torture test to make sure the CPU can actually handle that RAM speed when the entire chip is being slammed to the max.

 

This is akin to a few years ago people with 6700k's had issues running "fast RAM" like 3200 MHz in a 4 DIMM config AND overclocking the core maybe say 4-4.2 GHz all core boost. Its just a crap shoot with silicon lottery... Some chips don't have an issue, my 6700k did decently well, but others were not as lucky. This is just a single example, but all CPU's are similar. I am not sure what the probability of issue is on a 9900ks when running 4 dimms of fast RAM is, but I am sure its not a 100% guarantee to work, and its very possible slight timing, voltage, or some tertiary timing/voltage/clock speed can result in instability.

 

Another good but not exactly "applicable" example is my current 8700k. Silicon lottery is really all there is to it, but my 8700k is freakishly good with voltage. 1.300v for a 5 GHz all core OC is a solid .05v bellow what most folks need... I just got a really lucky chip. My OG i7 920 did 3.86 GHz up from 2.66 on LESS THAN STOCK votls, talk about a golden chip. It wouldn't go a MHz higher, but it ran that speed with sub stock votls for YEARS without a single BSOD... Weird things can happen, and its really just statics/luck.

[richintheveins]

25 minutes ago, LIGISTX said:

It could be some weird memory controller thing... Remember, you are overclocking your memory controller via pumping the RAM speed, and 4000 is pretty high. When you run your AIDA64 run, I would stress not just the memory, but the entire suite of CPU tests as well. Load the CPU up with as much as you can, RAM as well, basically all the default checked boxes in AIDA's torture test to make sure the CPU can actually handle that RAM speed when the entire chip is being slammed to the max.

 

This is akin to a few years ago people with 6700k's had issues running "fast RAM" like 3200 MHz in a 4 DIMM config AND overclocking the core maybe say 4-4.2 GHz all core boost. Its just a crap shoot with silicon lottery... Some chips don't have an issue, my 6700k did decently well, but others were not as lucky. This is just a single example, but all CPU's are similar. I am not sure what the probability of issue is on a 9900ks when running 4 dimms of fast RAM is, but I am sure its not a 100% guarantee to work, and its very possible slight timing, voltage, or some tertiary timing/voltage/clock speed can result in instability.

 

Another good but not exactly "applicable" example is my current 8700k. Silicon lottery is really all there is to it, but my 8700k is freakishly good with voltage. 1.300v for a 5 GHz all core OC is a solid .05v bellow what most folks need... I just got a really lucky chip. My OG i7 920 did 3.86 GHz up from 2.66 on LESS THAN STOCK votls, talk about a golden chip. It wouldn't go a MHz higher, but it ran that speed with sub stock votls for YEARS without a single BSOD... Weird things can happen, and its really just statics/luck.

Oh I know its an IMC thing, but I've really worked hard to obtain stability. I will run a more broad stress test to slam it to the max. I typically avoid test that crank up the CPU, because its not real world experiences. I've never seen my CPU eclipse 55 degrees while playing AAA titles on max settings so slamming it to crank it up beyond 70+ degrees for 12+ hours is perplexing to validate daily usage. Def not saying there's no value just something I've always questioned. 

[LIGISTX]

28 minutes ago, richintheveins said:

Oh I know its an IMC thing, but I've really worked hard to obtain stability. I will run a more broad stress test to slam it to the max. I typically avoid test that crank up the CPU, because its not real world experiences. I've never seen my CPU eclipse 55 degrees while playing AAA titles on max settings so slamming it to crank it up beyond 70+ degrees for 12+ hours is perplexing to validate daily usage. Def not saying there's no value just something I've always questioned. 

Stability can be defined in many was, one is by the things you use the system for, but I chose to define it as achieving the same stability as the CPU ships from Intel with. Can a system be overall unstable while being stable for you use case? Yes, it can, but to me and most folks, that is not "stability". Intel and all chip manufacturers engineer and ship products that will run at 100% load within acceptable thermal limits for the life of the chip which is why I test my systems as such. Is a AIDA64 test indicative of a normal home user use case, no, absolutely not. But is it indicative of an enterprise use case or a render farm use case, likely yes, it in fact it may not even be true real world use case as the PCIe lanes are not being stressed via massive NVMe IOPS or other PCIe device loads places on the PCIe controller. We do testing in these worst case scenarios because that is how Intel engineers their chips to work... testing in any less demanding way is just accepting that you have in fact created an unstable system. This may be acceptable to you, but that is a judgement call.

 

A good example is a buddies 4U homelab with all 24 hhd sleds populated, dual socket x58 xeon and 128 GB of RAM. Very much accidentally, mostly at my own fault since I am his "hardware guy" I had the rear exhaust fans off which are responsible for a large chunk of the CPU cooling in a 4U case, especially considering the fan wall behind the HDD's had its fans running very slowly for noise constraint issues at the cost of HDD and CPU heat. Server ended up kicking out a over temp warning after a few months when the drives were doing ZFS scrubs which lead us to investigate and figure out the fan RPM issue stated above. I couldn't physically put my hands on the CPU heatsinks they were so hot, and the system ran like this for upwards of 6 months without a single perceivable issue. I 100% do not advice such a test, this could have been catastrophic, but, it goes to show that a system working as designed does ultimately retain 100% uptime stability even in a horrible situation that the safe guards of the chassis had no power in overcoming as I had intentionally added resistors to the fan cables for the noise reduction necessary. Granted, yes, this isn't a "home PC" used for gaming, its a homelab running ESXi with multile VM's and FreeNAS running under it, but to him, this system also means a lot more then his 8700k "home PC". This is once again an assumption of acceptability vs risk. He knew the risks of intentionally killing fan RPM for noise reasons, but I did make a mistake in verification of RPM after this choice was made. Parallels can be drawn to overclocking. Overclocking is inherently introducing risk to stability and uptime, but if you test in the worst case situation and the system can handle it, then you can have ultimate confidence it won't give you trouble in a "normal every day" use case.

 

Really long explanation to simply say: test in the worst way possible to ensure the tasks you do day in and day out will not have stability issues. Hope that makes sense...

[richintheveins]

Wow! Great explanation. Thank you for taking the time to provide examples as well. I can't say I'm not afraid to slam it because boy will it be disappointing if it freezes or gives me a BSOD, but I realize it's necessary. 

[richintheveins]

I ran Aida64 and it crashed :-/

 

I'm at a loss I've tried everything to get this to work even purchased two other 4000 mhz kits prior to the predator ram that crashed my PC as well. I've received many different errors since this venture of 4 8 gb sticks @ 4000 mhz. Would my PSU potentially be the issue (its 6 years old). 

 

Any ideas to make this work or am I stuck at a lower frequency ram. I attempted to use XMP profile 1 (3600 mhz) last night and run a stability test. I ran Aida64 for 10 hrs with everything checked with no errors. 

 

Am I stuck with this? I really want it to run at 4000. Performance isn't even the issue its more principle and the time invested to make this happen. 

 

Any ideas would help. 

 

Thanks!