3950x Overclocking guide needed!!

it seems to be happening in the same consistent address range,

Lowest Error Address 0x1DD9CBF08 (7641MB)
Highest Error Address 0x1DF9CBFA8 (7673MB)

Bits in Error Mask 0000000010000000
Bits in Error 1
Max Contiguous Errors 1

upload_2019-11-29_15-57-49.png



thumb_show.jpg
 
it seems to be happening in the same consistent address range,

Lowest Error Address 0x1DD9CBF08 (7641MB)
Highest Error Address 0x1DF9CBFA8 (7673MB)

Bits in Error Mask 0000000010000000
Bits in Error 1
Max Contiguous Errors 1

View attachment 203340


View attachment 203346

You need to warranty swap that ram. Shouldn't be getting bit errors. My gskill did that and after warranty the new stuff was perfect.
 
disabled XMP, sitll getting the black screens with chrome and prime95:

upload_2019-11-29_16-11-28.png



upload_2019-11-29_16-12-21.png
 
EREK, that my exact ram, using 64gb's of this stuff. Use ryzen calculator to set "safe" timings at 3600MT and 1800 Fclk. Use "CJR" based ram setting (even tho you have DJR) set a 1.40vdimm and sett every single timing that program tells you, also use a 1.1SOC.. Report back
 
EREK, that my exact ram, using 64gb's of this stuff. Use ryzen calculator to set "safe" timings at 3600MT and 1800 Fclk. Use "CJR" based ram setting (even tho you have DJR) set a 1.40vdimm and sett every single timing that program tells you, also use a 1.1SOC.. Report back
did you have issues otherwise?
 
EREK, no sir, i have not had a single issue with this memory, but upon first bootup and windows install, the first thing i did was set my mem to those EXACT specs, then went to 3733 "fast" but used 16,19,19,36 for timing and not the suggested 21 and still not a single issue. DJR mem seems pretty good. Not saying this is the absolute fix for you, but if you still have errors and your appropriate voltages are set for mem, cpu, etc... and you have all those timings dialed in 100%, then without a doubt, you have a bad stick
 
The memory calculator is only as good as the memory controller on die. My CJR 2x16GB kit works perfectly fine at Cas 16 at 3600Mhz with a 3900x, but only at Cas 18 on a 3600 (and apparently the 3600x I have in there now). I have a 3800x coming to play with.
 
EREK, no sir, i have not had a single issue with this memory, but upon first bootup and windows install, the first thing i did was set my mem to those EXACT specs, then went to 3733 "fast" but used 16,19,19,36 for timing and not the suggested 21 and still not a single issue. DJR mem seems pretty good. Not saying this is the absolute fix for you, but if you still have errors and your appropriate voltages are set for mem, cpu, etc... and you have all those timings dialed in 100%, then without a doubt, you have a bad stick

can't get it to boot with DRAM Calculator recommended settings, so i bumped SoC to 1.1V and DRAM to 1.360V


2019-11-29 18:46:08 - [Data Error] Test: 4, CPU: 24, Address: 1DD9CBF08, Expected: F7F7F7F7, Actual: E7F7F7F7

it's always starting at this memory address:


Lowest Error Address 0x1DD9CBF08 (7641MB)
 
rma sir... no doubt about it. im sorry for your troubles

**EDIT** for what its worth, you have "rank" set to "1" and it should be set to 2 as those sticks are dual rank modules, see what settings change and try again
 
The memory calculator is only as good as the memory controller on die. My CJR 2x16GB kit works perfectly fine at Cas 16 at 3600Mhz with a 3900x, but only at Cas 18 on a 3600 (and apparently the 3600x I have in there now). I have a 3800x coming to play with.

Well the IO die is also subject to silicon lottery. Any impurities in that and it's less as good. I think thier 12 nm transistors.
 
Last edited:
Well the IO die is also subject to silicon lottery. Any impurities in that and it's less as good. I think thier 12 no transistors.

Considering how binned the 3950x, I'd assume that last of the weakest links. The ram is the weakest link by far here.
 
B7C9287C-7D7C-4986-981B-912D830C8973.jpeg
So even non XMP 2133MHz getting errors on the same CPU and sane address
 
Welp got errors on different cpu and different memory address with totally different ram 4x8GB not even XMP
9B3AAEBE-5666-4FFC-93D6-64A01F762D0C.jpeg
 
Maybe you can ship your chip to Smokingman to test?

I dont have an AM4 board anymore, or else I would gladly do it for you.

You could have a bad resistor or cap or copper trace on your mobo too?! Theres lots of smd filter caps and if one is bad it can cause dirty signals and subsequent bit errors.
 
Maybe you can ship your chip to Smokingman to test?

I dont have an AM4 board anymore, or else I would gladly do it for you.

You could have a bad resistor or cap or copper trace on your mobo too?! Theres lots of smd filter caps and if one is bad it can cause dirty signals and subsequent bit errors.

Put the G.Skill back in and it consisntently errors at the same memory address 7641MB
 
at this point i have absolutely no clue as to what to even consider RMAing... seems it could be anything

i even elicited help from TechPowerUp starting here @ https://www.techpowerup.com/forums/threads/ryzen-3950x-build-help.258856/post-4162130


(to be honest this has been the most difficult build i've ever gone through... my previous rig was stable for many many years... i'm about to just toss this all to the side just so i can have some peace of mind... i mean literally the CPU, Ram and motherboard could be at fault and basically the whole ordeal of having to wait weeks for the RMA process and start troubleshooting all over again one component at a time while being completely in a down state... i mean this build has taken up my entire week of vacation and ruined my thanksgiving, i've gotten basically no rest that i desperately need)
 
at this point i have absolutely no clue as to what to even consider RMAing... seems it could be anything

i even elicited help from TechPowerUp starting here @ https://www.techpowerup.com/forums/threads/ryzen-3950x-build-help.258856/post-4162130


(to be honest this has been the most difficult build i've ever gone through... my previous rig was stable for many many years... i'm about to just toss this all to the side just so i can have some peace of mind... i mean literally the CPU, Ram and motherboard could be at fault and basically the whole ordeal of having to wait weeks for the RMA process and start troubleshooting all over again one component at a time while being completely in a down state... i mean this build has taken up my entire week of vacation and ruined my thanksgiving, i've gotten basically no rest that i desperately need)

I went through this ordeal on a Ryzen 2700x build when the x470 platfrom still had some teething issues. Long story short, a bad board killed two sets of memory sticks before we figured out what was going on.
 
Erek make sure you click Import XMP on Ryzen Calculator.

Just something I noticed wasn't clicked. It has a factor in the ram timing calculation from the database in that software.

Lets revisit this one thing at a time.

If you try different ram on same motherboard and it errors out the board is suspect.

If you try different board and it errors out RAM or CPU is suspect.

If you try different board and different ram and errors out it is 100% cpu.

If you try another cpu in your board with your ram and it errors out its your board or your ram.

If you try your ram in another board and cpu and it errors out its your ram.

Or ... you can take it to a local PC repair shop, explain your level of expertise, and work a deal so that they can do all the testing across various boards etc.... and narrow it down for you. It might cost you 100 in labor but it will be much less than having to RMA the wrong parts over and over until you finally get it right.

Just an idea.
 
holy shit what a nightmare erek, sorry you are going thru this bs, as there is nothing worse then spending all this money on hardware and all you want is the enjoyment of building it and using it, NOT time spend repairing and troubleshooting. after reading thru the replies again, my best educated guess would be motherboard at this point. i mean, the cpu is a highly binned cpu, so these issues shouldnt happen, but anything is possible. hate thrwing more money, time, and frustration into it, but hopefully this adventure will be over soon.
 
One of the sticks individually gave an error, but i haven't been able to reproduce it yet after many retries:
upload_2019-11-30_5-54-43.png
 
With it still reporting CPU 24 basically all the time I have no clue if it’s the one ram stick actually or if that due to the lower load being single channel testing just makes it that much more difficult to produce an error now. Sigh. . I’ve only seen it produce an error testing one stick at a time just once so far
 
Erek make sure you click Import XMP on Ryzen Calculator.

Just something I noticed wasn't clicked. It has a factor in the ram timing calculation from the database in that software.

Lets revisit this one thing at a time.

If you try different ram on same motherboard and it errors out the board is suspect.

If you try different board and it errors out RAM or CPU is suspect.

If you try different board and different ram and errors out it is 100% cpu.

If you try another cpu in your board with your ram and it errors out its your board or your ram.

If you try your ram in another board and cpu and it errors out its your ram.

Or ... you can take it to a local PC repair shop, explain your level of expertise, and work a deal so that they can do all the testing across various boards etc.... and narrow it down for you. It might cost you 100 in labor but it will be much less than having to RMA the wrong parts over and over until you finally get it right.

Just an idea.

I did try a set of crucial 4x8 2400mhz and it gave a lot worse results. My 2x16 has been consistent on cpu 24 and the same starting memory address
 
You are overthinking this. RMA the board, RMA the CPU, RMA the memory. All of the vendors will test their stuff, and likely give it a clean bill of health and ship it back to you. Some part or parts may fail tests and get replaced. Once everything is back, reassemble and retest, all should be good. If not, *THEN* you have tested good gear with compatibility issues. *RIGHT NOW* you have *KNOWN BAD* gear and you're sitting and screwing with the settings instead of RMAing like you should. There is no setting tweak that will get around the *fact* that either your RAM, motherboard or CPU has issues. Get things fixed FIRST, then retest/tweak/adjust.
 
  • Like
Reactions: erek
like this
is it possible to configure iCUE but then disable the desktop app from starting up on boot and the profiles still work from the iCUE Service that runs? i don't want to keep a bunch of bloatware running all the time

At idle I never see icue even register in task manager when sorted by cpu utilization, so given we're talking massive cores/ threads these days, I no longer worry about it. I've been pretty pleased with it so far on my 1920x, we'll see how it does when the 3950x gets here.
 
  • Like
Reactions: erek
like this
I bought this kit for $57 at Best Buy as it appears on my mobo’s QVL: https://www.bestbuy.com/site/corsai...-desktop-memory-black/5822470.p?skuId=5822470



The Corsair passed a targeted run of memtest86 at 2133, and now I’m testing with XMP profile. Hoping all I’m dealing with is a bad G.Skill set. I started the RMA process. The gskill is also on the QVL

I was running Corsair Dominator Platinum 3600 c18 two sticks. Samsung B die.

Not a single inkling of an issue on my x570 board so you should be good. Now its on the trx40 board and same thing not an inkling of an issue. Now I need two more sticks. I had problem with Gskill but never corsair in the past. Gskill is still really good stuff. The Hynix C die is actually very very good silicon.
 
Back
Top