Hard crash

[Mussels]

SoC was in your screenshot as VSOC, as 1.08v
1.20v is the max you want to run for 24/7 safety, 1.15v should be fine for the 3800 range



VDIMM will help with BSOD's, app/windows crashes and regular restarts

VSOC helps with devices disconnecting and black screen freezes at lower loads, and WHEA errrors

 

[BlaezaLite]

Well it just did this with no CPU oc, so I'll leave it alone now. Although I am still getting a bsod after the bios screen, reset and then it's fine.

I'm probably going to end up divorced, but what ram should I be looking at to replace me Vengeance crap? Ideally, I'd like 32GB of B die, so I can have low timings. Is that a thing?

 

[Mussels]

To get full performance you'd need four ranks - so four 8GB dimms would do what you want, or two 16's
The best performing setups i've seen were CL14 3200 with four ranks of B-die, with minimal gains from that going upto 3600 or 3800 if the IMC and memory permitted it

You do get a good 10% boost in many tests with four ranks so it's worth it either way, but you'll still need to tweak voltages and timings since it's all overclocking anyway



Wanting more than 32GB of RAM or using an x3D chip B-die became irrelevant to me

 

[Waldorf]

@Blaeza
short of things like content creation/editing/encoding, going past 16gb wont do much,
maybe in some games like MS FS.

mine and a friends rig are almost the same, mine using 2 sticks 3600c14 and LC 2080S,
is almost as fast as his running 32gb, mainly because of the 4 sticks doing only 3200C18 and the ti being air cooled.

your board/ram is running below the cpus IF clocks, so i say upgrade that first and then go 3600 on the ram.
have no recent experience with amd gpus, so cant say how much it will gain running on a "500" series board.
lpx isnt really an issue as long as you know which kit to pick, as corsair tend to use different chips with identical specs,
so look for 3600C14 or kits above 4000, those are either samsung b-die or at least equivalent from other brands.
much easier to dial in from beginning (you can even just run xmp, but a little waste with b-die).

if you dont want to mess with ram settings much, the viper kits are decent, and usually some of the best value.
links are about the product, not really about where to buy..
4000 kit bdie
3600C14 bdie
lpx 3600C14 bdie

they usually can do better than those numbers (at identical V), just a matter of finding out what the kit likes,
one viper was doing good with lower timings, the other better with clocks, all LPX i had liked lower timings more.

 

[Manderion]

Hi @Blaeza , i had the same problem playing wz2.
Is it possible that u Set the Cmd2T in the DRAM advanced settings to 1T but also turned the GDM on in the AMD Overclocking settings? Then disable the GDM in the AMD Overclocking settings and use only the 1T setting in normal OC tab > advanced DRAM settings
Tell me if this helped you pls.

 

[3x0]

WHEA 18 in this case seems to be CPU related or CPU V-core.

Do you know if Bus/Interconnect error is only related to the memory controller or can it be a pure core instability? WHEA18 has Bus/Interconnect and Cache Hierarchy errors, so it would seem different parts of the CPU are to blame, no?

 

[Mussels]

Do you know if Bus/Interconnect error is only related to the memory controller or can it be a pure core instability? WHEA18 has Bus/Interconnect and Cache Hierarchy errors, so it would seem different parts of the CPU are to blame, no?

SoC is multiple parts in one, PCI-E, USB, DRAM controller, infinity fabric etc

It's pretty safe to assume they're related instead of focusing on which particular part errored

 

[BlaezaLite]

Well, since I got rid of PBO, my pc has been behaving itself with the only crash being in MW2 multiplayer, but that was just the game. So it appears the "safe" oc that PBO provided being the problem.

 

[Mussels]

Well, since I got rid of PBO, my pc has been behaving itself with the only crash being in MW2 multiplayer, but that was just the game. So it appears the "safe" oc that PBO provided being the problem.

PBO can only introduce instability in three ways

1: If you added to the boost MHz (max of +200) while undervolting
2: If you undervolted too far with curve optimiser (Same as #1, but people often test the undervolt and then add the boost, forgetting to retest it all)
3: If the CPU got far too hot, my memory overclock can go wonky if the CPU hits 75C - but it's also 64GB at 3800 so it's at the extreme end

 

[BlaezaLite]

PBO can only introduce instability in three ways

1: If you added to the boost MHz (max of +200) while undervolting
2: If you undervolted too far with curve optimiser (Same as #1, but people often test the undervolt and then add the boost, forgetting to retest it all)
3: If the CPU got far too hot, my memory overclock can go wonky if the CPU hits 75C - but it's also 64GB at 3800 so it's at the extreme end

I only had a +100mhz boost and I never undervolted. I had Ryzen Master auto oc for me. Maybe I'll re-paste as I thought a AK620 would keep a 5700X chilly, not get to 78-80C.

Edit... Just hit 82.4C as reported by HWinfo in R23. That shouldn't be THAT high, surely?

 

[hhb65]

I have a Ryzen system myself with a 5800x and a 6800xt paired with 32 GB ram 3600 overclocked to 3766. The setup is running with good cooling and a +100 mhz offset with individual curve settings for each core. I have been able to stresstest and run higher offset setting in other games, but MW2 is a very good indicator to test if your system is 100% stable. After i tweaked curve settings i have never had a crash in MW2 and the best coress boost to 4.950 mhz.

The auto overclock in PBO does not work for me since it will be a unstable setting is chooses, so I have tested a tweaked manually. It takes some time but is worth it. You can lower the by using pbo and negative core, but again you need to test each core for stability etc. and be aware of the thing Mussels wrote.

 

[A Computer Guy]

That's what I said

unflushed SSD caches = major data ouchy

 

[Mussels]

I have a Ryzen system myself with a 5800x and a 6800xt paired with 32 GB ram 3600 overclocked to 3766. The setup is running with good cooling and a +100 mhz offset with individual curve settings for each core. I have been able to stresstest and run higher offset setting in other games, but MW2 is a very good indicator to test if your system is 100% stable. After i tweaked curve settings i have never had a crash in MW2 and the best coress boost to 4.950 mhz.

The auto overclock in PBO does not work for me since it will be a unstable setting is chooses, so I have tested a tweaked manually. It takes some time but is worth it. You can lower the by using pbo and negative core, but again you need to test each core for stability etc. and be aware of the thing Mussels wrote.

sounds like you've got yours up to about the very best zen3 can achieve

 

[hhb65]

sounds like you've got yours up to about the very best zen3 can achieve

Yes. I love this setup. The 6800xt Asus is undervoltet to 1030 Mv with 15% plus powerlimit. SAM is enabled and the gaming experience so smooth on this AMD setup..very pleased i switched.

 

[RJARRRPCGP]

I have 2x16GB Corsair Vengeance. I thought they looked nice, I now know that they aren't very good, lol.

I've just found this, because I have a smooth brain...
View attachment 275146

That screams Fclk to me! Bus instability! (could very well mean more system bus termination voltage being required) (Possibly called ODT)

"Bus/Interconnect Error" ->Bus instability, most likely a bus termination issue with settings beyond default.

Looks like it has to be the Fclk and procODT-> What it's most likely called now. (or just "ODT"?) (on-die termination)
In the late-FSB days, it was "FSB Termination Voltage".

 

[A Computer Guy]

I have 2x16GB Corsair Vengeance. I thought they looked nice, I now know that they aren't very good, lol.

I've just found this, because I have a smooth brain...
View attachment 275146

I have an Corsair LPX kit that works really great with my Zen3 at stock settings for both RAM and CPU.
I'll give you the model number and version if you want to know.
It OC's to DDR4-3800 but I haven't bothered to dial it in.

 

[hhb65]

That screams Fclk to me! Bus instability! (could very well mean more system bus termination voltage being required) (Possibly called ODT)

"Bus/Interconnect Error" ->Bus instability, most likely a bus termination issue with settings beyond default.

Looks like it has to be the Fclk and procODT-> What it's most likely called now. (or just "ODT"?) (on-die termination)
In the late-FSB days, it was "FSB Termination Voltage".

I totally agree. Looks like FCLK to me also. If you have tested ram 100% and they are 100% error free, you still have to be sure that the Infinity Fabric (FCLK) speed choosen is stable. I tried getting WHEA error 18 when setting FCLK to high even though RAM was error free. Voltage lik V-SOC, VDDP and VDDG can sometimes be to high at default settings when using FCLK around 3733, 3800 area. Again only way is to test etc.

 

[BlaezaLite]

I have an Corsair LPX kit that works really great with my Zen3 at stock settings for both RAM and CPU.
I'll give you the model number and version if you want to know.
It OC's to DDR4-3800 but I haven't bothered to dial it in.

View attachment 278877
View attachment 278878

My Corsair Vengeance is at 3800 now and everything seems to be running fine. Thanks for the offer though!

I totally agree. Looks like FCLK to me also. If you have tested ram 100% and they are 100% error free, you still have to be sure that the Infinity Fabric (FCLK) speed choosen is stable. I tried getting WHEA error 18 when setting FCLK to high even though RAM was error free. Voltage lik V-SOC, VDDP and VDDG can sometimes be to high at default settings when using FCLK around 3733, 3800 area. Again only way is to test etc.

Ever since I turned off PBO, I've had no errors. Well, no system crashing errors anyway. So I blame PBO, lol.

 

[hhb65]

My Corsair Vengeance is at 3800 now and everything seems to be running fine. Thanks for the offer though!

Ever since I turned off PBO, I've had no errors. Well, no system crashing errors anyway. So I blame PBO, lol.

Hehe i understand you. People just need to accept that PBO takes time and alot of testing, but when you find the sweetspot, you can get a really nice combination of tempatures, power draw and efficiency. It is harder that old times overclocking tuning, but with OCCT, core cycler etc. there are some tools making it alot easier compared to without them.

 

[BlaezaLite]

Hehe i understand you. People just need to accept that PBO takes time and alot of testing, but when you find the sweetspot, you can get a really nice combination of tempatures, power draw and efficiency. It is harder that old times overclocking tuning, but with OCCT, core cycler etc. there are some tools making it alot easier compared to without them.

I will eventually get my head round PBO and probably call myself lots of things that aren't repeatable on here and curse that I didn't use it before. But I'll need to be having one of my really good days to attempt it again. Mental health is rather annoying sometimes.