Ryzen Owners Zen Garden

[Mussels]

You don't think it's due to bios updates? It could even be from Windows updates for all we know.

Since I upgraded my power supply I've actually found that my PBO undervolting stability has improved. Previously could only do negative 20 all cores stable, now I can do negative 30 all cores. So it could even be a power delivery degradation issue, nothing to do with the processor itself. It just seems like everyone is jumping to the conclusion that the silicon is to blame.

this is why i specifically said i dont know what the cause was, only that i didnt like the voltage creeping up as months went by

 

[tabascosauz]

I mean, my 3700x degraded at 1.2V
my brothers just had his 3700x start degrading at 1.3v

Maybe thats why they werent used as 3800X's, who knows. we dont even know if it was the boards VRMs, the chips, or changes in ambient weather (we had identical boards and CPUs)

What the UA batch numbers for these two 3700X's?

Let's face it, 3000 was basically a beta test for AMD's chiplet design (of which Ryzen 5000 is the finished product), and AMD treated its product and us buyers as such. It wouldn't be surprising to start seeing more stability issues with time, from some of the initial lower binned CPUs / harder-worked CPUs. 3600 and 3700X got the worse end of the stick - there were some really bad 3600s and 3700Xs in Q2 2019 and Q3 2019 production - like, 4.0 @ 1.3V bad.

AMD definitely pushed more Vcore through the 3000 chips than 5000. Probably in a bid to try to get the lower quality silicon as close to the "rated boost" as they could get, to avoid a class action lawsuit. I got used to seeing peak Vcore at up to 1.55V every other day on my 3700X for a meagre 4.28GHz effective, no PBO no offset no nothing just bone stock - and the same on 4 different motherboards, no less. The 5900X now peaks at about 1.49-1.5V @ 4.85-4.9GHz effective, stock.

Maybe I will just stick to IF and Memclk tuning..

The I/O die says hi and would like to remind that it too is a piece of silicon with finite lifespan in all seriousness, the entirety of my old chip was a [barely] living testament to "if it ain't broke, don't fix it" - read: don't overclock a turd .

So I learned my lesson and not fixing what's not broke is what I'm doing now

 

[freeagent]

I agree, I turned everything down.. its still pretty quick..

I am curious if the 3 series was a beta test, and maybe the XT is where they borrowed some of the refinement of the 5 series? I doubt it though..

 

[harm9963]

The results aren't very surprising anyway. It seems to be that six cores are usually Bronze, with the eight core chips as Silver. I'd imagine that Gold samples are reserved for Threadripper and Epyc.

Nope

 

[8BitZ80]

View attachment 190291


Nope

Given that it's a flagship processor it's not very surprising, brother.

 

[Gmr_Chick]

Hello, as long as you didn't apply to a profile or at startup, it should be fine. Either way thanks to @biffzinker for sharing the screenshots over here that I had sent in the discord, and to elaborate upon this that 1.5-1.55v Vcore is NOT using the protections from AMDs PBO system and manual removes these protections, 1.55v static through a single core is not the same as PBO and 1usmus removes the protections during this process and runs a Prime AVX1 test (more suitable towards RAM/IMC) for the testing process which drives the core to the ground and will cause degradation to the processor in most scenarios, Anyone using this with renoir avoid it period - unless you want your processor degraded
Some more fun of CTR2.0:
View attachment 190211
And a new article written on the matter from "Chip":

CTR Safety, Revisited

There are times when being a journalist is exciting.

chipsandcheese.com

Either way, do not trust this tool and run PBO instead or manual with safe vcore.
Edit: (another screenshot from a renoir sample)
View attachment 190213

@trickson Care to explain why you're bombarding posts with the 'haha' reaction on? Is there something we're doing wrong with CTR2.0 for it to degrade processors and consistently damage CPUs along with shoving down 1.5V into a 3970X? Or running completely unsafe all-core load vcore?????????? Care to elaborate?

Wondering this myself, as the Haha Wonder just reacted the same way to my post wondering why a Ryzen owner would ever use CTR given the problems mentioned.

 

[dyonoctis]

I'm a bit surprised that techpowerup haven't picked up on that new B550 board yet:
first thunderbolt 4 on AMD. I just wish that they would follow Gigabyte and make a ProArt ITX as well.

ASUS ProArt B550 Creator motherboard gets Thunderbolt 4 - VideoCardz.com

ASUS AMD motherboard with Thunderbolt 4 ASUS has the first Thunderbolt 4 motherboard. 12+2 Teamed Power Stage for stable and reliable power delivery under load Motherboards have not really been a hot topic in recent months due to the lack of new sockets. Only recently Intel has released its...

videocardz.com

 

[freeagent]

I have to say this 5600X is alotta fun. I have to question the contact of my Thermalright cooler now.. I can barely tame an all core of 4600 with max loads in the very low 80s, IF clocking is insane though.. I had a hard time at first and couldn't break 2000IF I haven't adjusted any other voltage outside of vcore and vdimm. Sub timings helped me out greatly, so its not just the CPU, I think the mobo has to want to play ball too, but hard to say because I am a noob.



I feel like I am tempting fate by trying for 2133.. not sure when I will run out of ram though.. never been this fast before with memtest

 

[PaulieG]

So, after getting a stable all core manual overclock of 4.7ghz @ 1.29v on my 5600x rock solid, I started playing with PBO and curve optimizer. Long story short, I seemed to be pretty stable on all cores at -20 CO getting boost speeds on all cores of 4.85ghz. Passed OCCT for 2 hours, ran all kinds of other benchmarks and left youtube on for 7 hours overnight without any reboots etc. No signs of any instability whatsoever.. but over the last two days I've had something odd happen. I can run the computer all day, doing all kinds of benchmarks and just day to day tasks without anything weird happening. However, if I shut the computer down overnight, when I turn it on in the morning the computer will try to post, but will behave in a way similar to a bad overclock. It will either just hang and not post at all or will come into the bios with an instability message, prompting me to press F1. I'll have to reset the bios to get it to post. Thing is, after this I can input the same overclock settings and it will post, boot into windows and run everything as perfectly stable. Anyone else experience this type of behavior? Any idea what's going on here?

 

[freeagent]

To me it sounds like a bad OC.

A weird thing with mine is that I can scale all the way up to 4900, but cant nail down more than anything over 4600 with really low voltage. That is 1 speed 1 voltage, no curves or anything like that.

 

[PaulieG]

To me it sounds like a bad OC.

A weird thing with mine is that I can scale all the way up to 4900, but cant nail down more than anything over 4600 with really low voltage. That is 1 speed 1 voltage, no curves or anything like that.

It just doesn't make sense to me. I pass everything, and I can go hours and hours of daily tasks, streaming, everything. No signs of instability whatsoever. It isn't until I turn the computer off and leave it off for several hours that this occurs.

As far as the manual OC goes, I posted a couple of weeks ago about the big jump in voltage needed for anything over 4.6ghz. Needed just 1.14v for 4.4,ghz 1.16v for 4.5ghz, 1.24v for 4.6ghz and 1.29v for 4.7ghz. My line in the sand on these chips is 1.3v, so I never really messed with 4.8ghz.

 

[freeagent]

Maybe she's not getting enough juice on startup? Like a GPU boosting too high? Sorry I am just guessing here. Thanks for some insight on voltage, I saw all kind of comments, but they seemed like guesses to me.

 

[dgianstefani]

To me it sounds like a bad OC.

A weird thing with mine is that I can scale all the way up to 4900, but cant nail down more than anything over 4600 with really low voltage. That is 1 speed 1 voltage, no curves or anything like that.

Ryzen as a platform not as rock solid as Intel.

Weird stuff like this happens. Try increasing a few other voltages other than just vcore.

 

[biffzinker]

It just doesn't make sense to me. I pass everything, and I can go hours and hours of daily tasks, streaming, everything. No signs of stability whatsoever. It isn't until I turn the computer off and leave it off for several hours that this occurs.

It could be SoC related instability. I just having a similar problem after a shutdown then posting the next power on. Bumped the OC on the memory, ram is stable seemed to be the IMC.

 

[PaulieG]

Ryzen as a platform not as rock solid as Intel.

Weird stuff like this happens. Try increasing a few other voltages other than just vcore.

It could be SoC related instability. I just having a similar problem after a shutdown then posting the next power on. Bumped the OC on the memory, ram is stable seemed to be the IMC.

The "weird stuff" makes me nuts. I've set the SOC to 1.1v because I have the memory overclocked. So maybe 1.15v? Anything on IOD? CCD?

 

[biffzinker]

I just bumped the SoC again last night to 1.21 from 1.2 and the weird behavior went away. Running with four ram sticks though.

 

[dgianstefani]

The "weird stuff" makes me nuts. I've set the SOC to 1.1v because I have the memory overclocked. So maybe 1.15v? Anything on IOD? CCD?

U can set both VDDG CCD to 1.05v, 1.075/09 max. As a general rule they should both be 0.05v minimum lower than SOC voltage.

VDDP Voltage should be 0.95 to 1v, max 1.1, general rule 0.025 minimum below VDDG.

 

[PaulieG]

I just bumped the SoC again last night to 1.21 from 1.2 and the weird behavior went away. Running with four ram sticks though.

I'm running 4 sticks also.

U can set both VDDG CCD to 1.05v, 1.075/09 max. As a general rule they should both be 0.05v minimum lower than SOC voltage.

VDDP Voltage should be 0.95 to 1v, max 1.1, general rule 0.025 minimum below VDDG.

This is what I needed. I'll give it a try tonight.

 

[dgianstefani]

I have my SOC offset at 0.0125+, SoC at 1.175v, CCD/IOD @ 1.09v, VDDP @ 1.05v.

But I have top tier cooling so bear that in mind.

 

[freeagent]

I'm running 4 sticks also.

This is what I needed. I'll give it a try tonight.

My system would do weird things with 4 sticks installed. Stable as a table except occasionally on a cold boot. I took the other pair out to let this thing spread its wings. I have had IF at 2166.. its psychotic.

 

[tabascosauz]

So, after getting a stable all core manual overclock of 4.7ghz @ 1.29v on my 5600x rock solid, I started playing with PBO and curve optimizer. Long story short, I seemed to be pretty stable on all cores at -20 CO getting boost speeds on all cores of 4.85ghz. Passed OCCT for 2 hours, ran all kinds of other benchmarks and left youtube on for 7 hours overnight without any reboots etc. No signs of any instability whatsoever.. but over the last two days I've had something odd happen. I can run the computer all day, doing all kinds of benchmarks and just day to day tasks without anything weird happening. However, if I shut the computer down overnight, when I turn it on in the morning the computer will try to post, but will behave in a way similar to a bad overclock. It will either just hang and not post at all or will come into the bios with an instability message, prompting me to press F1. I'll have to reset the bios to get it to post. Thing is, after this I can input the same overclock settings and it will post, boot into windows and run everything as perfectly stable. Anyone else experience this type of behavior? Any idea what's going on here?

I'm not sure exactly whether yours is cores- or SOC-related, but from my Matisse experience the failed cold starts until CMOS reset came as a result of pushing IF too hard (3733 for my old crappy 3700X). Sometimes I had issues with my SN750 disappearing from Windows with a stornvme error, as soon as I restarted upon seeing the drive disappear, it would fail to POST. I soon learned to brace for the worst upon the next restart as soon as I lost the drive.

I'm much more leery now about the CLDOs than I used to be. The conventional wisdom from the Memtesthelper guide is that the VDDGs should be about 0.04-0.05V lower than VSOC as they are derived from it (taking into account different SMU and SVI2 measurements of VSOC and droop), but I'm not so sure that is the right call. The recommended ranges for VDDG and VDDP never worked on my 3700X, either to enable stable 3733 operation, or to prevent the SN750 from dropping out, or the occasional cold start POST failures.

The other school of thought is that one of the VDDGs (sorry, I really can't remember if IOD or CCD should be the one set low) and VDDP should be set as low as they go - I think this came from users over on OCN. I think there is some merit to this claim as setting one of the VDDGs (again, sorry can't remember which) and VDDP low to about 0.95V and 0.9V respectively did make my stornvme errors and POST failures go away on my 3700X. The symptoms didn't come back, right up until I sold the 3700X.

I've not seen this behaviour on my 5900X yet, despite no other hardware having been changed aside from CPU.

 

[PaulieG]

I have my SOC offset at 0.0125+, SoC at 1.175v, CCD/IOD @ 1.09v, VDDP @ 1.05v.

But I have top tier cooling so bear that in mind.

I don't really worry about cooling. On my EK AIO and good case airflow, I never see temps go over 80c on OCCT or Realbench, regardless of all core overclock or PBO/CO.

I'm not sure exactly whether yours is cores- or SOC-related, but from my Matisse experience the failed cold starts until CMOS reset came as a result of pushing IF too hard (3733 for my old crappy 3700X). Sometimes I had issues with my SN750 disappearing from Windows with a stornvme error, as soon as I restarted upon seeing the drive disappear, it would fail to POST.

I'm much more leery now about the CLDOs than I used to be. The conventional wisdom from the Memtesthelper guide is that the VDDGs should be about 0.04-0.05V lower than VSOC as they are derived from it (taking into account different SMU and SVI2 measurements of VSOC and droop), but I'm not so sure that is the right call. The recommended ranges for VDDG and VDDP never worked on my 3700X, either to enable stable 3733 operation, or to prevent the SN750 from dropping out, or the occasional cold start POST failures.

The other school of thought is that one of the VDDGs (sorry, I really can't remember if IOD or CCD should be the one set low) and VDDP should be set as low as they go - I think this came from users over on OCN. I think there is some merit to this claim as setting one of the VDDGs (again, sorry can't remember which) and VDDP low to about 0.95V and 0.9V respectively did make my stornvme errors and POST failures go away on my 3700X. The symptoms didn't come back, right up until I sold the 3700X.

I've not seen this behaviour on my 5900X yet, despite no other hardware having been changed aside from CPU.

The IF is a good thing to explore. I have been pushing it pretty hard the last few days playing with some new ram.

 

[tabascosauz]

The IF is a good thing to explore. I have been pushing it pretty hard the last few days playing with some new ram.

One thing I know for sure: when faced with IF issues, higher VSOC up to 1.2V *can* solve your problems, but don't automatically jump to adding more VSOC - the CLDOs can play as big if not a bigger role. First knee-jerk reaction for early Ryzen 3000 adopters especially was to bump VSOC in the event that x RAM speed wasn't stable, which didn't make for a very consistent and effective solution even though it worked better for legacy Ryzen (Zen and Zen+).

It'll depend on your chip and its quality, but once the initial "moar VSOC" bandwagon was over, a lot of people were surprised about just how much they could achieve on just 1.0-1.1V VSOC even in the 3800MT/s ballpark.

If you're running on the edge of IF stability though (ie 2000MHz), I'd probably go close to 1.2V.

 

[biffzinker]

I'm not sure exactly whether yours is cores- or SOC-related, but from my Matisse experience the failed cold starts until CMOS reset came as a result of pushing IF too hard (3733 for my old crappy 3700X). Sometimes I had issues with my SN750 disappearing from Windows with a stornvme error, as soon as I restarted upon seeing the drive disappear, it would fail to POST.

So it’s IF related? I started having the missing boot drive every other post before nudging the SoC voltage.

 

[freeagent]

This is how much I am winging it..

I have not touched any of those voltages

Also never lost my drives, ever.. crazy. Asus must just make it easy? They took the adventure away