Overclock.net banner

1 - 7 of 7 Posts

·
Registered
Joined
·
18 Posts
Discussion Starter #1 (Edited)
In one of his videos Der8auer said that NVME SSD caused problems during extreme OC. My OC isn't that XTREME but seems like I have one thing in common with great and powerful Der8auer now...

Well, here is the story how (as I currently think) overclock may disrupt your SSD work:
I have a 960 EVO. It is NVME and is plugged in lower M2 slot (one under the Z370-A PRIME heatsink) in PCIE x4 mode. A few weeks ago I tweaked my system (in signature) VCCIO/VCCSA and made them lower, which allowed me to set even lower Vcore. PC was tested with an 1 hour of Prime 95 ver. 26.6 1344k (non-AVX) and was stable. Also i played BF1 alot (for two weeks or so) and system behaved well too.
Settings were the following:
Vcore 1.355 (5.2 GHz) LLC6, CPU-Z vcore 1.34-1.36 or so.
VCCSA 1.15
VCCIO 1.15
0430 bios (a bit dated).
Memory stock XMP 3000 MHz 15-17-17-35

And everything was OK for about 2 wheeks until I installed Lineage 2 (game which is known for severe hard disk/SSD abuse due to poor and obsolete design). I launched it and started to have random hardlocks (everything doesn't move, sound buzzes). Hard locks happened when the game loaded large pieces of terrain.
I tried to test my SSD but I have no idea how to do it better so I just launched Samsung Magician Benchmark and ASS Benchmark. The results are below:
VCORE 1.355
Samsung Magician stops at 35%, 1 min no progress, and then bsod "dpc watchdog violation".
ASS Benchmark - hardlock at 4k-64rd
VCORE 1.365
Samsung Magician stops at 39%, 1 min no progress, and then bsod "dpc watchdog violation".
ASS Benchmark - passes, works good
VCORE 1.37
Samsung Magician - pass
ASS Benchmark - pass

RAM in stock (2133) doesn't help. Bumping IO/SA voltage to default 1.32/1.22 don't help either.

What I'm going to do:
1) Take old SATA SSD and test wether it will be stable. Old SSD is Crucial M4 so I won't be able to use Samsung Magician bench
2) Really don't know.

I want to ask you guys if you have encountered something similar before, maybe you will advice something...
a) What voltage apart VCORE I can bump to stabilize it? Is adding VCORE is right at all (really fixes problem not walkarounds it)?
b) What are good things to check my SSD for good condition and stability? If it has stability at all or it is thing between it and CPU which causes the problem?
c) Current SSD is too small (250 GB) and I wanted to change it for 500GB, what can you recommend? I don't know much about this TLC/MLC stuff... I don't mind SSD resource being lower but when it causes hardlocks it pisses me off (know probably it's my faulty settings somewhere but still)...
d) Read somehwhere (don't laugh) that in one SSD slot SSD is conneced top CPU "directly" while another uses "southbridge" to connect it to CPU. Is that true?
 

·
Premium Member
Joined
·
16,114 Posts
Why the resistance to update the BIOS?

Guessing your comment of it being 'dated', there's a newer version. That and you have a relatively new NVMe SSD that a BIOS update could potentially add support for or better compatibility (whether stated outright or not).
 

·
Registered
Joined
·
18 Posts
Discussion Starter #3
960 evo was released 1 year before coffee cpu's and mb's.

I stick to older bios coz I don't like changing things just because they are newer - if it works why shall I change it? It is tested and all issues are known, while new bios can bring god knows what. Sure i'll try it if nothing helps.
 

·
Mr.4way SLI
Joined
·
5,522 Posts
Changing the bios is the first thing I would try. before I went on to other steps.
 

·
Premium Member
Joined
·
4,981 Posts
About your question d), isn't everything connected to the PCH (the chipset)? The CPU has 16 PCIe lanes and those are all used for the graphics card slots, I would think. Only those two (or three) slots of the board are connected to the CPU's PCIe lanes, and the other PCIe slots and M.2 slots are all connected to the PCH's PCIe lanes.

This then makes me confused about your overclock causing instability for the NVMe drive's connection. Perhaps something else is happening, and the overclock is simply not 100% stable? There might be rare data corruption happening, and that's what broke contents on the drive or in RAM, and the hard-locks you are seeing.
 

·
Premium Member
Joined
·
16,114 Posts
960 evo was released 1 year before coffee cpu's and mb's.

I stick to older bios coz I don't like changing things just because they are newer - if it works why shall I change it? It is tested and all issues are known, while new bios can bring god knows what. Sure i'll try it if nothing helps.
Your first question I would agree with on some things, but considering all of the hardware security breaches as of this year...and the fact that your problem indicates something actually isn't working...*shrug*.
 

·
Iconoclast
Joined
·
30,430 Posts
NVMe drives tend to allow for much more I/O performance, which is why we buy them. More I/O performance means more CPU/IMC/memory/PCI-E controller load (especially if you have the spectre and/or meltdown mitgations enabled). It's not particularly surprising that a system that was tuned to the edge of stability with a slower drive needs a bit of a bump somewhere to regain full stability again.

Restoring VCCSA (which is used by the IMC and PCI-E/DMI controller) and/or VCCIO to their previously levels may also help, or may also be necessary. You could have introduced subtle instability by lowering them that wasn't revealed until you tested it with the faster SSD.

An hour of Prime95 isn't much of a test and won't touch anything I/O related at all.
 
1 - 7 of 7 Posts
Top