Overclock.net - An Overclocking Community - Reply to Topic
Thread: *Official* Intel DDR4 24/7 Memory Stability Thread Reply to Thread
Title:
Message:

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in


  Additional Options
Miscellaneous Options

  Topic Review (Newest First)
01-21-2020 01:39 AM
satinghostrider Anyone able to better the timings as attached? Got that from @KedarWolf . Running the f4-3600c16d 16gb kit. (2x8gb) Getting in another pair next week. Board is Aorus Master on F11c official Gigabyte BIOS.

Thanks in advance Guys!
01-20-2020 03:30 PM
Falkentyne
Quote: Originally Posted by Jpmboy View Post
RMA? Based on bios-set voltages? Have you measured the actual voltage with a DMM? Remember, an OC on ram is an OC on the CPU too - as you discovered. It is very "normal" for XMP to require voltage tweaks for real stability.
I'm curious - how do you know the BSODs were memory related after HCi or RT? .. and not CPU IO ?
He needs to run GSAT, not Karhu. This sounds alot like the IMC/L3 cache is not stable.
If I remember your posts about Karhu, Karhu doesn't test IMC/L3 cache stability does it?

Better tests for this is Prime95 29.8 b6, 112k-112k in place FFT with AVX disabled (crashed threads or CPU Cache L0 errors, this test will test hyperthreaded IMC L3 stability quite well) or 256k-512k AVX disabled (with or without in-place, if in-place is disabled, a small part of RAM allocated, like 10%).

And errors with virtualized instruction registers (L0 cache) are related to hyperthreaded cores, which is heavily controlled by the IMC/L3 cache. Unless the BSOD is "memory_management", it could very well be related to that.
What BSOD's were showing up? If it's system service exception, IRQL_NOT_LESS_OR_EQUAL or something, that possibly isn't caused by the DRAM itself.
01-20-2020 02:38 PM
Jpmboy
Quote: Originally Posted by reachthesky View Post
I figured out the issue. It was lack of vcore.

At 2133 memory, cpu is stable 5ghz no avx offset with standard llc + 120mv offset + acdc 1/1.

When enabling xmp, the cpu requires an extra 60mv to be occt large avx stable. XMP also kept erroring unless I increased dram voltage past the xmp profile's 1.35v rating, no errors with 1.4v dram voltage. Should I RMA the kit since it errors unless voltage is above what the kit is specced for?

I appreciate your follow up btw.

I rebooted cl15 4133 with the additional 60mv vcore for stability. non avx occt large temps peaked at 94c lol. It's crazy how a ram overclock can add up to 30c in heat to non-avx loads at the same core frequency. non avx temps without the ram overclock(2133) were in the 60s lol. Just crazy.

The reason why I tried OCCT is because while I was able to pass 3200% karhu and 600% HCI without any errors with my cl15 4133 profile, I was still getting memory related bsods after passing those tests. Turns out karhu was a waste of money.
RMA? Based on bios-set voltages? Have you measured the actual voltage with a DMM? Remember, an OC on ram is an OC on the CPU too - as you discovered. It is very "normal" for XMP to require voltage tweaks for real stability.
I'm curious - how do you know the BSODs were memory related after HCi or RT? .. and not CPU IO ?
01-20-2020 09:30 AM
Gen.
Quote: Originally Posted by SuperMumrik View Post
Any idea where to improve? The chip gets really wonky above 4266 :/
Royal 4400 cl18 and 9900 "es" chip
4266 MHz

tCL=16
tRCD=16
tRP=16
tRAS=36
tRC=52
tCR=1 или 2

tWR=16
tRFC=288 (+-8; должно быть кратно 8, затем опустите или поднимите)
tRRD_L=6
tRRD_S=4
tWTR_L=8
tWTR_S=4
tRTP=8
tFAW=16
tCWL=16

tREFI=65534
tCKE=8
Все dr (если 2 плашки всего) & dd (если 1Rx8GB)=0
tRDRD_sg=6
tRDRD_dg=4
tRDWR_sg=10
tRDWR_dg=10
tWRRD_sg=30
tWRRD_dg=26
tWRWR_sg=6
tWRWR_dg=4

Quote: Originally Posted by Lalka228 View Post
Can you tell me.... I won in IMC silicon lottery?
so shame its not stable, i think because of motherboard limit of Dram 1.55 volt, i cant push it higher

Does IMC really matter in ram perfomance in your opinion?
У вас плохо прошла тренировка памяти!

4200MHz

tCL=16
tRCD=16(17)
tRP=16(17)
tRAS=36(37)
tRC=52(54)
tCR=1 или 2

tWR=16
tRFC=320 (+-8; должно быть кратно 8, затем опустите или поднимите)
tRRD_L=6
tRRD_S=4
tWTR_L=8
tWTR_S=4
tRTP=8
tFAW=16
tCWL=15 (Для ASRock)

tREFI=65534
tCKE=8
Все dr (если 2 плашки всего) & dd (если 1Rx8GB)=1
tRDRD_sg=6
tRDRD_dg=4
tRDWR_sg=10
tRDWR_dg=10
tWRRD_sg=29
tWRRD_dg=25
tWRWR_sg=6
tWRWR_dg=4
01-20-2020 09:00 AM
zGunBLADEz
Quote: Originally Posted by reachthesky View Post
Quote: Originally Posted by zGunBLADEz View Post
3200% on ram test not enough even hci theres not a % required the longer the test runs the better. If you failing ramtest at just %3200 your not stable at all cpu/ram or cache
I did 3200% because that what it says on the front page, same with 500% hci. Why is this stuff on the front page if it is wrong? Why are people's OCs posted if they are only tested to 200%-500%. Why am I given incorrect information as soon as I visit the front page of this thread. This **** doesn't help me.
Bcuz is there doesnt mean is 100% its just what's required to validate in this particulars topic.. You keep finding errors right? That means theres a problem somewhere you need to narrow to where exactly it is.

You can start validating the ram over clock with the cpu at stock and whats required to overclock the ram for example that includes io/sa and cache settings.

Also make sure the ram warmup pretty good, turn any rainbows leds off. You can even active cooling them for the test if you want. Trust me after those tests u ain't seeing those temps ever again if all you do is only game xD
01-20-2020 08:58 AM
reachthesky
Quote: Originally Posted by zGunBLADEz View Post
3200% on ram test not enough even hci theres not a % required the longer the test runs the better. If you failing ramtest at just %3200 your not stable at all cpu/ram or cache
I did 3200% because that what it says on the front page, same with 500% hci. Why is this stuff on the front page if it is wrong? Why are people's OCs posted if they are only tested to 200%-500%. Why am I given incorrect information as soon as I visit the front page of this thread. This **** doesn't help me.
01-20-2020 08:52 AM
zGunBLADEz 3200% on ram test not enough even hci theres not a % required the longer the test runs the better. If you failing ramtest at just %3200 your not stable at all cpu/ram or cache
01-20-2020 06:46 AM
reachthesky
Quote: Originally Posted by Jpmboy View Post
OCCT is not really testing the ram as rigorously as HCi or Ramtest, that's why the OP did not include it in the acceptable list for this 24/7 thread. You are absolutely right... you do not need OCCT for a gaming rig.
You should run a high-current stressor though. iBT serves this purpose wel. 5-10 loops. And use an AVX offset. It is there so that you can run higher clocks for non-AVX loads.
Server chips always had hard coded AVX offsets, they made their way to consumer desktops with X-class HCC chip families, and to desktop i7's and down... because AVX2 was showing up in more code and chips were gonna toast.
I figured out the issue. It was lack of vcore.

At 2133 memory, cpu is stable 5ghz no avx offset with standard llc + 120mv offset + acdc 1/1.

When enabling xmp, the cpu requires an extra 60mv to be occt large avx stable. XMP also kept erroring unless I increased dram voltage past the xmp profile's 1.35v rating, no errors with 1.4v dram voltage. Should I RMA the kit since it errors unless voltage is above what the kit is specced for?

I appreciate your follow up btw.

I rebooted cl15 4133 with the additional 60mv vcore for stability. non avx occt large temps peaked at 94c lol. It's crazy how a ram overclock can add up to 30c in heat to non-avx loads at the same core frequency. non avx temps without the ram overclock(2133) were in the 60s lol. Just crazy.

The reason why I tried OCCT is because while I was able to pass 3200% karhu and 600% HCI without any errors with my cl15 4133 profile, I was still getting memory related bsods after passing those tests. Turns out karhu was a waste of money.
01-20-2020 06:37 AM
Jpmboy
Quote: Originally Posted by reachthesky View Post
I ran hci and karhu for the memory. Occt brought out some instability as soon as I started the test, watchdog bsods. If I want to pass these stress tests with avx enabled, I have to use an offset otherwise temps are too much(with the ram oc enabled). But then its like i'm over tuning my system for workloads i'm never going to encounter since all I do is use this pc for gaming. How do I find the right balance? Is occt ideal here? Currently testing no memory OC and 5g with avx -2(acdc-1/1, standard llc, +120mv offset).

How long should I be testing in OCCT large for?
OCCT is not really testing the ram as rigorously as HCi or Ramtest, that's why the OP did not include it in the acceptable list for this 24/7 thread. You are absolutely right... you do not need OCCT for a gaming rig.
You should run a high-current stressor though. iBT serves this purpose wel. 5-10 loops. And use an AVX offset. It is there so that you can run higher clocks for non-AVX loads.
Server chips always had hard coded AVX offsets, they made their way to consumer desktops with X-class HCC chip families, and to desktop i7's and down... because AVX2 was showing up in more code and chips were gonna toast.
01-20-2020 12:38 AM
reachthesky I passed 1 hour of OCCT large with avx enabled at 5ghz with avx-2 with acdc-1/1 + standard llc with + 120mv offset. I did occt large avx on and off. Passed. This was with no memory OC.

Now I enabled XMP and i'm getting hundreds of errors in OCCT after a couple minutes, even after increasing vcore by 60mv. sa/io are set manually at 1.25v, dram at 1.35v. xmp 4000 cl17-17-17-37-2T

EDIT:I enabled enhanced multicore performance and i'm 12 minutes into occt large avx enabled and not getting any errors with 5GHz no avx offset. I think enhanced multicore performance needs to be enabled when using xmp or something or maybe it needs to be enabled when using dvid mode or something i dunno. This isn't the first time where enabling enhanced multicore performance allowed me to not fail tests right away.

EDIT: it ended up failing again but much later this time instead of 10 minutes in. Figured out the issue, My chip needs an extra 60mv vcore to be oct large avx stable when clocking ram to 4000. Yuck this chip is crap.

This thread has more than 10 replies. Click here to review the whole thread.

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off