Question DDR4 Chip and 10700K Overclocking Help.

Oct 24, 2019
79
12
45
1
Hi, I have two questions to ask. I am trying to find out what the chip of my memory is and if these voltages are good.

I have two TF3D48G2666HC15B01, which is 2666 CL15-17-17-35 1.2V with XMP enabled, i was able to do a quick stable OC to 2933 CL15-18-18-36 1.3V. I tried rising the voltage a bit to 1.35 and tightening the timings to 15-17-17-36, but seems unstable. I can't find anywhere the type of my memory chip, with these timings, do you guys have any clue?

About my voltages, i can't find my actual SA voltage readings in my BIOS, just the VCCIO, which i set to 1.1 (stock is 1.15). Is this too much or is it a reasonable value?
 
Last edited:

CompuTronix

Intel Master
Moderator
Thanks for the heads-up, DB.

Maikurosofuto,

I've reviewed each post in this thread, so to summarize, this appears to be where you're at:

i7- 10700K
5.1Ghz All Cores
79°C
LLC4
Adaptive Vcore, Offset - 0.105V (1.326V Peak running Prime95)

P95, AIDA64, CineBench R20, RealBench

Stable using IETU Settings
Unstable with IETU Settings in BIOS

Overclocking requires that we minimize as many variables as possible, of which there are dozens. So in order to compare apples to apples, we must always be very specific. Unless you're accustomed to closely observing the relationships between Core voltage, power consumption, V/F curves and thermal behavior, using both numerical values as well as graphical patterns, most users don't realize how various utilities and their assorted tests can differ drastically in the nature of the workloads they impose on a processor, which can dramatically affect stability.

“Stress” tests vary widely and can be characterized into two categories; stability tests which are fluctuating workloads, and thermal tests which are steady workloads. Prime95 v29.8 Small FFTs (AVX disabled) is ideally suited for testing power consumption and thermal performance, because it conforms to Intel's Datasheets as a steady-state 100% workload. You can also use v26.6 Small FFTs, which is an identical workload, but without any AVX code. (As per Intel's Datasheets, TDP and Thermal Specifications are validated without AVX.)

Utilities that don't overload or underload your processor will give you valid power and thermal baselines. Here’s a comparison of utilities grouped as thermal and stability tests according to % of TDP, averaged across six processor Generations at stock settings rounded to the nearest 5%:


Although these tests range from 70% to 130% TDP workload, Windows Task Manager interprets every test as 100% CPU Utilization, which is processor resource activity, NOT actual workload. Core temperatures respond directly to Power consumption (Watts), which is driven by Core voltage and workload. Prime95 v29.8 Small FFTs (AVX disabled) provides a steady 100% workload, even when TDP is exceeded by overclocking. (The topic of TDP is another involved discussion, which for the purposes of this thread, is unnecessary and will likely confuse the issue).

As you can see from the scale, Intel Extreme Tuning Utility is a fluctuating workload that's only about 80%. Conversely, AIDA64 has 4 CPU related stress test selections (CPU, FPU, Cache, Memory) which have 15 possible combinations that yield 15 different workloads and Core temperatures. That's a lot of variables and inconsistencies, which may explain when you plug IETU settings into BIOS, why utilities with heavier workloads crash. The vast majority of users don't specify exactly which test(s) they ran, nor do they typically mention ambient room temperature (normal is 22°C), which can as well be a HUGE variable.

Concerning LLC (Load Line Calibration); the purpose is to compensate for the difference between the "set" no load Core voltage in BIOS and the actual Core voltage when under a 100% workload. For example, if a wall socket in your house is probed with a voltmeter, then a very high load such as air conditioning switches on, you'll see the voltage "sag" by a volt or two, which is normal and expected. The same applies to processors, which is know as "Vdroop". Even the highest quality PSUs and efficient motherboard VRMs are affected.

When adjusting LLC, the goal is to find a setting which allows the Core voltage in Windows during a steady-state 100% workload to match the Core voltage set in BIOS. Intel intends that there should be at least a minimal sag or "undershoot", but any surge or "overshoot" is not recommended for CPU longevity. Motherboards with tight VRM regulation may vary by as little as 16 to 24mv, which, for example, means if BIOS is set for 1.260, then 100% workload in Windows should ideally vary from ~ 1.244 to 1.236. Keep in mind that since Vcore settings are in 5mv (.005) increments, but the resulting values are in 8mv (.008) increments, you will find that certain settings will cause the values in Windows to "toggle" between voltages more than others.

When workloads spike higher due to processing dense segments of code, power consumption also spikes higher which causes Core voltage to momentarily spike lower or "sag". It's during these moments of lower Vcore when the processor is most vulnerable to BSOD crashes. Therefore, when adjusting Core voltage and LLC in BIOS, it's critical to closely observe Core voltage behavior in Windows. A better method to successfully accomplish this task with the least amount of frustration is to run a steady-state 100% workload, which refers back to Prime95 v29.8 Small FFTs with all AVX test selections disabled.

When used in conjunction with HWiNFO, if you right-click on parameters such as Vcore, Package Power and Package Temperature (typically the hottest Core), it will open a graph for each, which will allow you to expand your view beyond the blinders of numerical values and take advantage of "the big picture". You can't expect to make these critical adjustments accurately when running fluctuating workloads that look like a bad day on the stock market. After you've found the most ideal LLC setting, you can then move on to stability testing with fluctuating workloads.

Here's how different workloads look on a graph:



Shown above from left to right: Small FFTs, Blend, Linpack and IntelBurn Test.​

Note the steady-state signature of Small FFTs, which allows accurate measurements of power consumption, Core voltage and Core temperatures. A steady 100% workload is key for testing so the CPU, cooler, socket, motherboard and voltage regulator modules can thermally stabilize.

Being fully aware of your test conditions and minimizing the variables involved will help you to achieve stability. As Phaaze88 has already been pointed out, although Silicon Lottery provides examples of various Core voltages used for different overclocking combinations, they're also specific regarding their QVL and test conditions. They also state that beyond Core voltage, LLC and AVX offsets, other BIOS adjustments such as PLL, SA and I/O voltages, Uncore or Ring ratio are typically unnecessary. To experienced overclockers, this suggests that for your overclock, which is not unreasonable, higher Vcore should be all that's needed.

As Unolocogringo and Karadjgne have corroborated, we know that since your RAM is somewhat less than a matched set, it introduces yet another set of variables. So to rule out RAM stability issues, while testing CPU overclock stability it's standard procedure to approach CPU overclocking and RAM overclocking separately; not simultaneously. I suggest that you run your RAM at stock settings, not XMP. After you achieve a stable CPU overclock, you can later tweak RAM settings, but beware that unstable RAM is the most expedient way to corrupt your software, which my esteemed colleague, Darkbreeze, has very clearly and adamantly emphasized. As such, since BSOD crashes are an inevitable part of the overclocking process, always perform a full system backup prior to conducting any overclocking endeavors.

Although you're running the latest version of IETU, here's a link to the latest version of its successor, Intel® Performance Maximizer for 10th Generation Intel® Core™ Processors. Not to introduce further variables, but you might want to give it a try instead of IETU.

You might also want to read my Intel Temperature Guide. Section 8 covers Overclocking and Voltage. Just click on the link in my signature.

CT :sol:
 
Reactions: Maikurosofuto
One thing I caught.
"I have two TF3D48G2666HC15B01 "
Part numbers are for a single stick of memory @ 8GB.
Mixing ram is not always stable.
For dual channel you need2 matched sticks from a set. not 2 single sticks.
Memory has a ton of secondary timings that must also match for stability.
That is why they are sold as Kits. The modules are tested to by the manufacturer to work together.
Buying 2 single sticks and putting them together is a crap shot if they work together.
 
Reactions: Maikurosofuto

Darkbreeze

Retired Mod
Yes, good catch. Something I usually latch onto right away, and I agree, if those sticks did not come in a kit TOGETHER, then you're fortunate they are even running together, much less attempting to overclock them. I'd recommend that you run them at the stock XMP profile speed rather than overclocking them, if they were not both a part of a matched set. And these days, to be honest, it doesn't make much sense not to simply just BUY DIMMs that are the speed you want to run.

You don't actually NEED "two matched DIMMs from a set" in order to run in dual channel. Being matched has nothing to do with running in dual channel, but what DOES matter, is that the stick are compatible ENOUGH to run together in dual channel, and by that we mean that the ENTIRE configuration of the DIMM (ICs, ranks, rows, primary, secondary and tertiary timings) are similar enough that the motherboard is able to make micro adjustments that are amenable to both sticks. Sometimes both sticks look at the motherboard and say "no, thanks, we're not playing that game", while other times they are more like "ok, I guess we'll play", and still others when the memory is very much the same or is matched simply works together with no micro adjustments to any of the timings being necessary by the motherboard.

Besides which, dual channel operation isn't the issue in question here really AND usually, if the DIMMs are specc'ed within the JEDEC standards for the platform (So, for Z490, 2666mhz or less) there is usually a LOT more forgiveness for differences between DIMMs and not using matched ones. Outside of the JEDEC standard configurations, and anytime the XMP "OC" or an actual manual OC is involved, you quickly begin to move further and further away from that ESPECIALLY when you start trying to tighten timings on top of it. My further advice on this would be, if you're going to do something other than simply run them at their XMP profile values, then either overclock them OR tighten the timings, but not both. Especially not when (Or IF) they are not from a matched set.
 

Darkbreeze

Retired Mod
Two questions before we get into anything "meaty".

One, why is your CPU fan showing only 555 RPM while your CPU_OPT is showing 1025 RPM? What CPU cooler do you have? What does the rest of your cooling configuration including your case model and case fans, look like?

Two, what BIOS version are you on currently?
 
Oct 24, 2019
79
12
45
1
Two questions before we get into anything "meaty".

One, why is your CPU fan showing only 555 RPM while your CPU_OPT is showing 1025 RPM? What CPU cooler do you have? What does the rest of your cooling configuration including your case model and case fans, look like?

Two, what BIOS version are you on currently?
1 - I'm using the Assassin III splitter on CPU_fan and a MasterFan MF140R ARGB on CPU_OPT, behind the cooler, on the case, as an auxiliary fan (Basically, Push/Pull/Pull). My case is the NZXT Noctis 450 ROG, with 1 x 140mm FN V2 Fan connected to the top (Exhaust) and 3 x 120mm FN V2 Fans connected to the front (Intake).
2 - Latest one, 1001 (July 1, 2020).
 
Hi, I have two questions to ask. I am trying to find out what the chip of my memory is and if these voltages are good.

I have two TF3D48G2666HC15B01, which is 2666 CL15-17-17-35 1.2V with XMP enabled, i was able to do a quick stable OC to 2933 CL15-18-18-36 1.3V. I tried rising the voltage a bit to 1.35 and tightening the timings to 15-17-17-36, but seems unstable. I can't find anywhere the type of my memory chip, with these timings, do you guys have any clue?

About my voltages, i can't find my actual SA voltage readings in my BIOS, just the VCCIO, which i set to 1.1 (stock is 1.15). Is this too much or is it a reasonable value? I tried to set to 1.09v, but my LAN controller seemed to be having problems, with sudden disconnections.
VCCIO is IMC controller voltage, should be fine if you leave it on auto. DDR4 can stand 1.6v and up, 1.35v is usual for RAM over 2400MHz and is in no way too high.
 
Oct 24, 2019
79
12
45
1
VCCIO is IMC controller voltage, should be fine if you leave it on auto. DDR4 can stand 1.6v and up, 1.35v is usual for RAM over 2400MHz and is in no way too high.
I set both VCCIO and "System Agent Voltage" to 1.09v, but as i said, my LAN controller was acting weird, then i set only the VCCIO to 1.1v and leave the "System Agent Voltage" to auto, but i can't find what values the "auto" is giving me, no readings are shown in the bios, i think this "VTT" value in HWInfo is the reading I am looking for, but I am not sure.
 

Darkbreeze

Retired Mod
System agent (VCCSA) and VCCIO wouldn't have anything to do with the LAN controller, UNLESS there was one or more bent pins and then anything could have something to do with anything. Otherwise, they are unrelated.

I personally wouldn't run ANY DDR4 memory kit at higher than 1.45v UNLESS the profile voltage is already higher than that, and then I wouldn't run it at anything higher than what the profile voltage is because if it's higher than that it's already a very high speed kit anyhow and probably doesn't need to have anybody overclocking it in the first place.

As far as overclocking your memory kit, sure, you can do that, but honestly you won't see enough benefit going from 2666mhz to 2933mhz while staying at CL15 to make it worth doing, and you're not going to get those sticks to run at CL14, guaranteed, with any increase in frequency along for the ride. Plus, if you're not aware, there is a very extensive stability verification process that is a MANDATORY REQUIREMENT if you don't want to end up with data corruption after making changes to the memory frequency or timings, and if you're not willing to do the stability testing as outlined, it's best to simply run them at the XMP profile and call it a day.

Testing your memory configuration to verify stability

Before you decide that this section is not worth your time or get lazy thinking you don't need to test because your system "seems" fine, with no obvious blue screens, freezing or restarting, let me make one thing VERY, VERY CLEAR.

ANY amount of instability in your memory configuration is enough to cause what are known as micro errors. This is a very miniscule error which, if it only happened one time might not ever be a factor but when it happens cumulatively in small increments over time, can result in complete and total corruption of your operating system, documents, game files, applications, music, movies, everything, to the point of being a complete and total loss with no chance of recovery.

Memory configurations that are not as close to 100% stable as possible are not a joke.
They WILL eventually cause widespread corruption of the entire file system.


Don't cut corners because it's simply not worth it. If you are unwilling to do the testing necessary to make sure the system is stable you should simply leave the memory at the default configuration and that includes NOT setting the memory to the XMP profile if the profile of the memory is beyond what the system automatically configures the memory speed and timings to by default. Do the testing. One day out of your life is not going to kill you but not doing it might make you wish you had died if you lose a lot of very important information and personal files that can't be replaced.



Memtest86


Go to the Passmark software website and download the USB Memtest86 free version. You can do the optical disk version too if for some reason you cannot use a bootable USB flash drive.

Create bootable media using the downloaded Memtest86 (NOT Memtest86+, that is a different, older version and is outdated). Once you have done that, go into your BIOS and configure the system to boot to the USB drive that contains the Memtest86 USB media or the optical drive if using that option.


Click here to download Memtest86 USB package

Create a bootable USB Flash drive:


1. Download the Windows MemTest86 USB image.

2. Right click on the downloaded file and select the "Extract to Here" option. This places the USB image and imaging tool into the current folder.

3. Run the included imageUSB tool, it should already have the image file selected and you just need to choose which connected USB drive to turn into a bootable drive. Note that this will erase all data on the drive.



No memory should ever fail to pass Memtest86 when it is at the default configuration that the system sets it at when you start out or do a clear CMOS by removing the CMOS battery for five minutes.

Best method for testing memory is to first run four passes of Memtest86, all 11 tests, WITH the memory at the default configuration. This should be done BEFORE setting the memory to the XMP profile settings. The paid version has 13 tests but the free version only has tests 1-10 and test 13. So run full passes of all 11 tests. Be sure to download the latest version of Memtest86. Memtest86+ has not been updated in MANY years. It is NO-WISE as good as regular Memtest86 from Passmark software.

If there are ANY errors, at all, then the memory configuration is not stable. Bumping the DRAM voltage up slightly may resolve that OR you may need to make adjustments to the primary timings. There are very few secondary or tertiary timings that should be altered. I can tell you about those if you are trying to tighten your memory timings.

If you cannot pass Memtest86 with the memory at the XMP configuration settings then I would recommend restoring the memory to the default JEDEC SPD of 1333/2133mhz (Depending on your platform and memory type) with everything left on the auto/default configuration and running Memtest86 over again. If it completes the four full passes without error you can try again with the XMP settings but first try bumping the DRAM voltage up once again by whatever small increment the motherboard will allow you to increase it by. If it passes, great, move on to the Prime95 testing.

If it still fails, try once again bumping the voltage if you are still within the maximum allowable voltage for your memory type and test again. If it still fails, you are likely going to need more advanced help with configuring your primary timings and should return the memory to the default configuration until you can sort it out.

If the memory will not pass Memtest86 for four passes when it IS at the stock default non-XMP configuration, even after a minor bump in voltage, then there is likely something physically wrong with one or more of the memory modules and I'd recommend running Memtest on each individual module, separately, to determine which module is causing the issue. If you find a single module that is faulty you should contact the seller or the memory manufacturer and have them replace the memory as a SET. Memory comes matched for a reason as I made clear earlier and if you let them replace only one module rather than the entire set you are back to using unmatched memory which is an open door for problems with incompatible memory.

Be aware that you SHOULD run Memtest86 to test the memory at the default, non-XMP or custom profile settings BEFORE ever making any changes to the memory configuration so that you will know if the problem is a setting or is a physical problem with the memory.

After your memory will pass Memtest for 4 full passes, it is still not necessarily stable, but it is a good start and you should move on the the last phase of testing using Prime95. See, there IS a light at the end of the tunnel.







Final testing with Prime95

It is highly advisable that you do a final test using Prime95 WITH AVX and AVX2 disabled, and run a custom configured Blend test. You can also use the Blend mode option as is, but after a fair amount of personal testing, asking questions from some long time members with engineering level degrees that have forgotten more about memory architectures than you or I will ever know, and gathering opinions from a wide array of memory enthusiasts around the web, I'm pretty confident that the custom option is a lot more likely to find errors with the memory configuration, and faster, if there are any to be found.

Please note as this is rather important, if you prefer, or have problems running version 26.6 because you have a newer platform that doesn't want to play nice with version 26.6, you can use the latest version of Prime95 with the Custom test selected but you will need to make the following change.

In the bottom of the Torture test selection popup menu there will be some options for disabling AVX. I recommend that you do so, not because we are doing thermal testing and require a steady state workload (Which AVX wouldn't affect anyhow, as Computronix explained to me), but because the last thing you need during memory testing is having to worry about CPU temperatures, and you will, with AVX enabled.

So, uncheck the option for AVX2. That will un-gray the option for AVX, and uncheck that box as well.

Now open Prime95.

Click on "Custom". Input a value of 512k in the minimum FFT size field. Leave the maximum FFT size field at 4096k. In the "Memory to use" field you should take a look at your current memory allocation in either HWinfo or system resource monitor. Whatever "free" memory is available, input approximately 75% of that amount. So if you currently have 16GB of installed memory, and approximately 3GB are in use or reserved leaving somewhere in the neighborhood of 13GB free, then enter something close to 75% of that amount.

So if you have 13GB free, or something reasonably close to that, then 75% of THAT would be 9.75GB, which, when multiplies times 1024 will roughly equal about 9984MB. You can average things out by simply selecting the closest multiple of 1024 to that amount just to keep it simple, so we'll say 10 x 1024= 10240mb and enter that amount in the field for "Memory to use (MB)". We are still well within the 13GB of unused memory BUT we have left enough memory unused so that if Windows decides to load some other process or background program, or an already loaded one suddenly needs more, we won't run into a situation where the system errors out due to lack of memory because we've dedicated it all to testing.

I've experienced false errors and system freezes during this test from over allocating memory, so stick to the method above and you should be ok.


Moving right along, do not change the time to run each FFT size. Leave that set to 15 minutes.

Click run and run the Custom test for 8 hours. If it passed Memtest86 and it passes 8 hours of the Custom test, the memory is 100% stable, or as close to it as you are ever likely to get but a lot of experts in the area of memory configuration suggest that running the extended Windows memory diagnostic test is also a pretty good idea too.

If you get errors, (and you will want to run HWinfo alongside Prime95 so you can periodically monitor each thread as Prime will not stop running just because one worker drops out, so you need to watch HWinfo to see if there are any threads not showing 100% usage which means one of the workers errored and was dropped) then you need to either change the timings, change the DRAM voltage or change the DRAM termination voltage, which should be approximately half of the full DRAM voltage.

There are also other bios settings that can affect the memory configuration AND stability, such as the SOC, VCCIO and system agent voltages, so if you have problems with stability at higher clock speeds you might want to look at increasing those slightly. Usually, for Intel at least, something in the neighborhood of 1.1v on both those is pretty safe. There are a substantial number of guides out there covering those two settings, but most of them are found within CPU overclocking guides so look there in guides relevant to your platform.

As a further measure of assurance that your WHOLE configuration is stable, you can download and run Realbench for 8 hours. If the system freezes or fails when running Realbench with your full memory amount set, try running it again but select only half your amount of installed memory.
 
Oct 24, 2019
79
12
45
1
System agent (VCCSA) and VCCIO wouldn't have anything to do with the LAN controller, UNLESS there was one or more bent pins and then anything could have something to do with anything. Otherwise, they are unrelated.

I personally wouldn't run ANY DDR4 memory kit at higher than 1.45v UNLESS the profile voltage is already higher than that, and then I wouldn't run it at anything higher than what the profile voltage is because if it's higher than that it's already a very high speed kit anyhow and probably doesn't need to have anybody overclocking it in the first place.

As far as overclocking your memory kit, sure, you can do that, but honestly you won't see enough benefit going from 2666mhz to 2933mhz while staying at CL15 to make it worth doing, and you're not going to get those sticks to run at CL14, guaranteed, with any increase in frequency along for the ride. Plus, if you're not aware, there is a very extensive stability verification process that is a MANDATORY REQUIREMENT if you don't want to end up with data corruption after making changes to the memory frequency or timings, and if you're not willing to do the stability testing as outlined, it's best to simply run them at the XMP profile and call it a day.

Testing your memory configuration to verify stability

Before you decide that this section is not worth your time or get lazy thinking you don't need to test because your system "seems" fine, with no obvious blue screens, freezing or restarting, let me make one thing VERY, VERY CLEAR.

ANY amount of instability in your memory configuration is enough to cause what are known as micro errors. This is a very miniscule error which, if it only happened one time might not ever be a factor but when it happens cumulatively in small increments over time, can result in complete and total corruption of your operating system, documents, game files, applications, music, movies, everything, to the point of being a complete and total loss with no chance of recovery.

Memory configurations that are not as close to 100% stable as possible are not a joke.
They WILL eventually cause widespread corruption of the entire file system.


Don't cut corners because it's simply not worth it. If you are unwilling to do the testing necessary to make sure the system is stable you should simply leave the memory at the default configuration and that includes NOT setting the memory to the XMP profile if the profile of the memory is beyond what the system automatically configures the memory speed and timings to by default. Do the testing. One day out of your life is not going to kill you but not doing it might make you wish you had died if you lose a lot of very important information and personal files that can't be replaced.



Memtest86


Go to the Passmark software website and download the USB Memtest86 free version. You can do the optical disk version too if for some reason you cannot use a bootable USB flash drive.

Create bootable media using the downloaded Memtest86 (NOT Memtest86+, that is a different, older version and is outdated). Once you have done that, go into your BIOS and configure the system to boot to the USB drive that contains the Memtest86 USB media or the optical drive if using that option.


Click here to download Memtest86 USB package

Create a bootable USB Flash drive:


1. Download the Windows MemTest86 USB image.

2. Right click on the downloaded file and select the "Extract to Here" option. This places the USB image and imaging tool into the current folder.

3. Run the included imageUSB tool, it should already have the image file selected and you just need to choose which connected USB drive to turn into a bootable drive. Note that this will erase all data on the drive.



No memory should ever fail to pass Memtest86 when it is at the default configuration that the system sets it at when you start out or do a clear CMOS by removing the CMOS battery for five minutes.

Best method for testing memory is to first run four passes of Memtest86, all 11 tests, WITH the memory at the default configuration. This should be done BEFORE setting the memory to the XMP profile settings. The paid version has 13 tests but the free version only has tests 1-10 and test 13. So run full passes of all 11 tests. Be sure to download the latest version of Memtest86. Memtest86+ has not been updated in MANY years. It is NO-WISE as good as regular Memtest86 from Passmark software.

If there are ANY errors, at all, then the memory configuration is not stable. Bumping the DRAM voltage up slightly may resolve that OR you may need to make adjustments to the primary timings. There are very few secondary or tertiary timings that should be altered. I can tell you about those if you are trying to tighten your memory timings.

If you cannot pass Memtest86 with the memory at the XMP configuration settings then I would recommend restoring the memory to the default JEDEC SPD of 1333/2133mhz (Depending on your platform and memory type) with everything left on the auto/default configuration and running Memtest86 over again. If it completes the four full passes without error you can try again with the XMP settings but first try bumping the DRAM voltage up once again by whatever small increment the motherboard will allow you to increase it by. If it passes, great, move on to the Prime95 testing.

If it still fails, try once again bumping the voltage if you are still within the maximum allowable voltage for your memory type and test again. If it still fails, you are likely going to need more advanced help with configuring your primary timings and should return the memory to the default configuration until you can sort it out.

If the memory will not pass Memtest86 for four passes when it IS at the stock default non-XMP configuration, even after a minor bump in voltage, then there is likely something physically wrong with one or more of the memory modules and I'd recommend running Memtest on each individual module, separately, to determine which module is causing the issue. If you find a single module that is faulty you should contact the seller or the memory manufacturer and have them replace the memory as a SET. Memory comes matched for a reason as I made clear earlier and if you let them replace only one module rather than the entire set you are back to using unmatched memory which is an open door for problems with incompatible memory.

Be aware that you SHOULD run Memtest86 to test the memory at the default, non-XMP or custom profile settings BEFORE ever making any changes to the memory configuration so that you will know if the problem is a setting or is a physical problem with the memory.

After your memory will pass Memtest for 4 full passes, it is still not necessarily stable, but it is a good start and you should move on the the last phase of testing using Prime95. See, there IS a light at the end of the tunnel.







Final testing with Prime95

It is highly advisable that you do a final test using Prime95 WITH AVX and AVX2 disabled, and run a custom configured Blend test. You can also use the Blend mode option as is, but after a fair amount of personal testing, asking questions from some long time members with engineering level degrees that have forgotten more about memory architectures than you or I will ever know, and gathering opinions from a wide array of memory enthusiasts around the web, I'm pretty confident that the custom option is a lot more likely to find errors with the memory configuration, and faster, if there are any to be found.

Please note as this is rather important, if you prefer, or have problems running version 26.6 because you have a newer platform that doesn't want to play nice with version 26.6, you can use the latest version of Prime95 with the Custom test selected but you will need to make the following change.

In the bottom of the Torture test selection popup menu there will be some options for disabling AVX. I recommend that you do so, not because we are doing thermal testing and require a steady state workload (Which AVX wouldn't affect anyhow, as Computronix explained to me), but because the last thing you need during memory testing is having to worry about CPU temperatures, and you will, with AVX enabled.

So, uncheck the option for AVX2. That will un-gray the option for AVX, and uncheck that box as well.

Now open Prime95.

Click on "Custom". Input a value of 512k in the minimum FFT size field. Leave the maximum FFT size field at 4096k. In the "Memory to use" field you should take a look at your current memory allocation in either HWinfo or system resource monitor. Whatever "free" memory is available, input approximately 75% of that amount. So if you currently have 16GB of installed memory, and approximately 3GB are in use or reserved leaving somewhere in the neighborhood of 13GB free, then enter something close to 75% of that amount.

So if you have 13GB free, or something reasonably close to that, then 75% of THAT would be 9.75GB, which, when multiplies times 1024 will roughly equal about 9984MB. You can average things out by simply selecting the closest multiple of 1024 to that amount just to keep it simple, so we'll say 10 x 1024= 10240mb and enter that amount in the field for "Memory to use (MB)". We are still well within the 13GB of unused memory BUT we have left enough memory unused so that if Windows decides to load some other process or background program, or an already loaded one suddenly needs more, we won't run into a situation where the system errors out due to lack of memory because we've dedicated it all to testing.

I've experienced false errors and system freezes during this test from over allocating memory, so stick to the method above and you should be ok.


Moving right along, do not change the time to run each FFT size. Leave that set to 15 minutes.

Click run and run the Custom test for 8 hours. If it passed Memtest86 and it passes 8 hours of the Custom test, the memory is 100% stable, or as close to it as you are ever likely to get but a lot of experts in the area of memory configuration suggest that running the extended Windows memory diagnostic test is also a pretty good idea too.

If you get errors, (and you will want to run HWinfo alongside Prime95 so you can periodically monitor each thread as Prime will not stop running just because one worker drops out, so you need to watch HWinfo to see if there are any threads not showing 100% usage which means one of the workers errored and was dropped) then you need to either change the timings, change the DRAM voltage or change the DRAM termination voltage, which should be approximately half of the full DRAM voltage.

There are also other bios settings that can affect the memory configuration AND stability, such as the SOC, VCCIO and system agent voltages, so if you have problems with stability at higher clock speeds you might want to look at increasing those slightly. Usually, for Intel at least, something in the neighborhood of 1.1v on both those is pretty safe. There are a substantial number of guides out there covering those two settings, but most of them are found within CPU overclocking guides so look there in guides relevant to your platform.

As a further measure of assurance that your WHOLE configuration is stable, you can download and run Realbench for 8 hours. If the system freezes or fails when running Realbench with your full memory amount set, try running it again but select only half your amount of installed memory.
I'm aware about those memory stability check to avoid data corruption, i'm just trying to find the sweetspot and then start the tests, if anything goes wrong, i'll just step down and start testing again. Taking advantage of the aforementioned programs, would MemTest64 be an effective form of benchmark?
I just got this mobo + cpu kit recently, so I'm still downloading certain programs and I wanted to make a list of some good and reliable ones for benchmarking. Right now, I'm only focusing on RAM, I will consider the CPU later.

[...] VCCIO wouldn't have anything to do with the LAN controller.
And about that...are you sure? I always found these terminologies a bit confusing, but according to what I've read, VCCIO is directly related to CPU I/O. I might be wrong, that's why i'm here asking for help.
According to this article: "VCCIO: Available starting with the second-generation Core i CPUs (“Sandy Bridge”), this voltage is used for feeding all input/output (I/O) pins of the CPU, except memory-related pins. On CPUs that have this voltage, it is also used to feed the thermal control bus (PECI, Platform Environmental Control Interface). "
 
Last edited:

Darkbreeze

Retired Mod
Well, yeah, you're right, it COULD have an effect on I/O like LAN, but it's extremely doubtful, especially since you're well within not only tolerance, but recommended voltage. I normally tend to run VCCIO and VCCSA for most Intel generations at about 1.1v but then again I'm normally running both fast RAM and an overclock of the CPU. Quite often on Intel systems you might have to use at least a mild overclock to overcome problems with fast memory kits, but that's probably not the issue here.

I'd try both settings at 1v and see how you do.
 
Oct 24, 2019
79
12
45
1
Well, yeah, you're right, it COULD have an effect on I/O like LAN, but it's extremely doubtful, especially since you're well within not only tolerance, but recommended voltage. I normally tend to run VCCIO and VCCSA for most Intel generations at about 1.1v but then again I'm normally running both fast RAM and an overclock of the CPU. Quite often on Intel systems you might have to use at least a mild overclock to overcome problems with fast memory kits, but that's probably not the issue here.

I'd try both settings at 1v and see how you do.
Set both settings to 1v, but still no VCCSA voltage readings, i think my mobo just don't monitor it. Sadly, the free MemTest86 have only 4 pass now, which i already tested and got 0 errors. Is the MemTest64 a realiable (free) alternative?
 

Darkbreeze

Retired Mod
Yeah, I don't know why yours isn't showing it, because most HWinfo screenshots I've seen do, and mine does. Really it doesn't matter though, if it's set in the BIOS, that's all that matters. You really don't need to monitor VCCIO and VCCSA in Windows. Just set it and forget it unless there is a problem, then adjust it.

You only need to do four passes of Memtest86. No, do not use any other Memtest unless you pay for the full paid version of Memtest86 which has three more tests not used in the free version and unlimited passes. I'd just move on the custom Prime95 test as outlined in my guide.

 
Oct 24, 2019
79
12
45
1
Yeah, I don't know why yours isn't showing it, because most HWinfo screenshots I've seen do, and mine does. Really it doesn't matter though, if it's set in the BIOS, that's all that matters. You really don't need to monitor VCCIO and VCCSA in Windows. Just set it and forget it unless there is a problem, then adjust it.

You only need to do four passes of Memtest86. No, do not use any other Memtest unless you pay for the full paid version of Memtest86 which has three more tests not used in the free version and unlimited passes. I'd just move on the custom Prime95 test as outlined in my guide.

Hey, just a quick update. I did all the RAM testing thing and was able to achieve 5.1Ghz all cores with an adaptive voltage offset -0.105V (1.326V Peak running Prime95), LLC4 and 79C, is this a good result? Still testing, but everything seems stable with P95, AIDA64, Cinebench20 and Asus Realbench so far.
 

Karadjgne

Titan
Ambassador
VCCIO affects the voltage for the Super I/O, part of which is the Lan controller on some mobo's, according to what I've pieced together.
But found this.

Yes,VCCSA and VCCIO are pre-haswell terms , now VCCIO is split into VCCIO-A and VCCIO-D. CPU Digital I/O voltage = VTT in HWinfo. CPU System Agent voltage = VCCSA in HWinfo.

VCCSA: System agent voltage. Helps with DRAM overclocking. 1.15~1.30 is more than enough for maximizing DRAM clocks. You can use 1.30v when pushing high DRAM frequencies to take the IMC out of the equation.

IO A: IO analog voltage. May help with DRAM overclocking. 1.15~1.30 is more than enough for maximizing DRAM Clocks most times.

IO D voltage: IO digital voltage. Critical rail for DRAM overclocking. 1.15~1.30 is more than enough for maximizing DRAM frequency. IO D can be sensitive on some processors when cold and too high may worsen cold bug behaviour. The right level of I/O A and I/O D will help maximize memory OC, too much or too little can cause code ‘55’ on the Hex display during POST or induce instability.
With VCCIO originally set lower than 1.0v, would explain a bunch of instability.

Intel limits on DDR4 under XMP 2.0 are 1.5v no matter what speed or timings and absolutely recommend not going higher. It's the only thing I've ever found for actual 'safe' limits on DDR4 voltages.
 

Darkbreeze

Retired Mod
My Skylake system, both boards (Hero VIII and Gaming 5), with clearly are POST-Haswell, are not like that. They are both VCCSA/System agent and VCCIO, only. There is no VCCIO or VTT at all on the Gaming 5 shown in HWinfo. On my Hero VIII though, it shows VCCIO and VCCSA, no VTT, so clearly there are some discrepancies from gen to gen and board to board, even within the same gen.

And, the RECOMMENDED VCCIO and VCCSA voltages for overclocked systems are pretty clearly lined out, and have worked well for me on every board I've configured using them, with no weird problems with any I/O or LAN specific hardware because of it.

Whether or not that holds true for Canyon lake, IDK, because I haven't worked with any Canyon lake systems yet, but from what I've read I don't see that there are any huge discrepancies between that architecture or any previous Skylake based architecture when it comes to these settings and configurations.

Even the Intel datasheet recommends up to 1.06v for stock operation, while up to 1.25v seems pretty common for overclocked systems and even higher if it's really necessary. So far as I know, we are STILL on Skylake architecture, even with Canyon lake process.

 

Karadjgne

Titan
Ambassador
Revamp of a revamp of a revamp. Intel could use a fresh coat of imagination, this repeated skylake architecture is getting long in the tooth.

But yes, totally agree, there's multiple naming schemes across vendors, even boards. I've never seen a listing for VCCIO A/D but I'm guessing it does exist, probably in the Extreme or Godlike boards etc. With lesser boards it's probably rolled into the same heading, change one/change both. The bios on my 3rd gen Asus had PLL and System Agent, don't remember even seeing VCCIO or VTT or any of the others listed. But it was an entry level Z77, so a simplified bios compared to even the lower ranked ROG boards.
 
Oct 24, 2019
79
12
45
1
Hey guys, sorry to bother you again. I removed the "solved" because i got into another trouble now, i found a way to check if my OC gonna crash my PC instantly, and it's checking Intel's XTU V/F curve. Even setting everything to auto and raising only the clock multiplier, just to check my max Vcore, my PC just reboots without a single error report (i.e BSOD), i'll attach a video as an examble of this occurrence . Any idea of why is this happening? It doesn't happen when I put everything in stock or just overclock the memory.
 

Darkbreeze

Retired Mod
I don't like XTU and I don't recommend using it.

If you've tested with Memtest86, Prime95 (Small FFT for thermals and either the Blend mode or the custom settings mode I outlined in my guide) and Realbench, then there is really no reason whatsoever to need to test with anything else but if you STILL want to, then I'd recommend using OCCT or Intel burn test, but really, if you've run these others and were stable, it's sort of putting a boot to the ashes of the body you already burned.
 
Oct 24, 2019
79
12
45
1
If you've tested with Memtest86, Prime95 (Small FFT for thermals and either the Blend mode or the custom settings mode I outlined in my guide) and Realbench, then there is really no reason whatsoever to need to test with anything else but if you STILL want to, then I'd recommend using OCCT or Intel burn test, but really, if you've run these others and were stable, it's sort of putting a boot to the ashes of the body you already burned.
I don't like XTU either, but i just wanted to get the default V/F curve and that's what happened. I can't consider my OC stable if i'm having trouble even with XTU. Anything i can mess in the BIOS to fix these crashes?
 

Darkbreeze

Retired Mod
Honestly, it's probably because your memory configuration just isn't stable enough. Obviously something needs either more voltage or a lower frequency setting if you see crashes that you don't see at the stock settings.
 

ASK THE COMMUNITY

TRENDING THREADS