Can a PSU upgrade give stability to an unstable overclock?

[void pointer]

Hello everyone!

 

My PC has a ryzen 5 3600 that I have tried to overclock following Buildzoid guides on YT, maxed out the "safe" voltage at around 1.33v,  the highest I got was 4.1 ghz all core but it doesn't last more than 30 minutes running prime 95 before failing. Also tried 4.2 ghz and was able to play some games but it crashes prime instantly so I discarded this one. In both overclocks I sometimes get the overclock related BSODs.

 

I'm trying all of this using a Corsar CX 650w grey label which is currently placed at the B tier in the PSU tierlist and I was wondering if its possible that upgrading to a Tier A Seasonic Focus gm 750w unit would help even a little bit on giving the overclock some stability, I know its not a matter of wattage and that probably is more related to silicon lottery or to the fact that my mobo is kinda bad (aorus b450 elite) but I wanted to know if the PSU that I plan to upgrade to has less voltage ripple and that could contribute to a stable overclock.

 

 

[Chris Pratt]

In some cases, yes. If the PSU has a high amount of ripple or poor transient response, it could make it extremely difficult to dial in a stable OC. However, Corsair's CX line is pretty good overall, so I wouldn't expect switching to something else to help in that regard.

[Founders]

Your PSU is fine, I would say your chip is just a little weak, but it's a 3600 so it's not going to have the best binning to begin with. How does your chip do with PBO enabled? I know it pushes your voltage pretty high but the 3600 is not a hot chip so you may be better off sticking with that.    

[void pointer]

12 minutes ago, Founders said:

Your PSU is fine, I would say your chip is just a little weak, but it's a 3600 so it's not going to have the best binning to begin with. How does your chip do with PBO enabled. I know it pushes your voltage pretty high but the 3600 is not a hot chip so you may be better off sticking with that.    

I noticed that two specific cores are the ones that struggle and that fail first at prime95 always so yes you are right my chip is really weak. With PBO I have mixed thoughts on it because I dont see as much of an improvement in cinebench scores compared to the manual OC but might as well get back to PBO

[void pointer]

8 minutes ago, Chris Pratt said:

In some cases, yes. If the PSU has a high amount of ripple or poor transient response, it could make it extremely difficult to dial in a stable OC. However, Corsair's CX line is pretty good overall, so I wouldn't expect switching to something else to help in that regard.

That's what I thought but wasn't quite sure if my psu was good enough to rule out an improvement after switching to something else. I really needed the second opinion so thanks!

[void pointer]

Also my PSU is like 4 years old and it's getting close to the 5 year warranty expiration date, should I worry about that or just let it run until it dies?

[Founders]

6 minutes ago, void pointer said:

I noticed that two specific cores are the ones that struggle and that fail first at prime95 always so yes you are right my chip is really week. With PBO I have mixed thoughts on it because I dont see as much of an improvement in cinebench scores compared to the manual OC but might as well get back to PBO

You can try running ClockTuner for Ryzen, just make sure to read up on it before you run it. It tunes per core to give you the best results it can, it should be able to help get you a stable overclock at lower voltages than PBO would require.  

[Somerandomtechyboi]

27 minutes ago, void pointer said:

Hello everyone!

 

My PC has a ryzen 5 3600 that I have tried to overclock following Buildzoid guides on YT, maxed out the "safe" voltage at around 1.33v,  the highest I got was 4.1 ghz all core but it doesn't last more than 30 minutes running prime 95 before failing. Also tried 4.2 ghz and was able to play some games but it crashes prime instantly so I discarded this one. In both overclocks I sometimes get the overclock related BSODs.

 

I'm trying all of this using a Corsar CX 650w grey label which is currently placed at the B tier in the PSU tierlist and I was wondering if its possible that upgrading to a Tier A Seasonic Focus gm 750w unit would help even a little bit on giving the overclock some stability, I know its not a matter of wattage and that probably is more related to silicon lottery or to the fact that my mobo is kinda bad (aorus b450 elite) but I wanted to know if the PSU that I plan to upgrade to has less voltage ripple and that could contribute to a stable overclock.

 

 

Ryzen is just garbage at oc since they already run close to diminishing returns zone from the factory, only way to get a meaningful increase is running 1.45v max where you can expect 4.4-4.6ghz if your chip isnt total garbage

 

1.35v is conservative, wayyy too conservative, its either stick to 1.25-1.3v for a slight overclock that prioritizes efficiency or balls to the walls overclock for max performance but slightly accelerated degradation (doesnt really matter for a year or 2), going anywhere inbetween is just stupid since you arent really gaining much performance yet your efficiency is ruined

 

I think der8auer tested ryzen degradation at 1.5v and seems like some dont and some do degrade but ofc its at max load all the time, just stick to 1.45v or 1.25v for an oc, anything in between makes no sense

[Mark Kaine]

12 hours ago, Somerandomtechyboi said:

1.35v is conservative, wayyy too conservative, its either stick to 1.25-1.3v for a slight overclock that prioritizes efficiency or balls to the walls overclock for max performance but slightly accelerated degradation (doesnt really matter for a year or 2), going anywhere inbetween is just stupid since you arent really gaining much performance yet your efficiency is ruined

ok, you might not know , but maybe you do, so at 1.25v my 3600 crashes immediately, so there has to be some tricks to get this working, i thought of using LLC for example?

 

thing is i tried to follow a GN "guide" specifically for the 3600 and i went all the way up to what he considered "safe" (i dont remember, maybe 1.36v or something) but it just would BSOD on windows log in every time.

And he also said he had to go 1.43 or something to get it stable, but wouldn't recommend doing so for 24/7 oc,  aka its a shit chip for overclocking lol, but afaik he didnt say anything about LLC, so any thoughts about that?

[Somerandomtechyboi]

23 minutes ago, Mark Kaine said:

ok, you might not know , but maybe you do, so at 1.25v my 3600 crashes immediately, so there has to be some tricks to get this working, i thought of using LLC for example?

 

thing is i tried to follow a GN "guide" specifically for the 3600 and i went all the way up to what he considered "safe" (i dont remember, maybe 1.36v or something) but it just would BSOD on windows log in every time.

And he also said he had to go 1.43 or something to get it stable, but wouldn't recommend doing so for 24/7 oc,  aka its a shit chip for overclocking lol, but afaik he didnt say anything about LLC, so any thoughts about that?

Idk, im on 775 where llc is either max or none, i choose no llc cause llc fks my overclocks, though board is garbage so that also fks my overclock

 

I do know llc can actually have benifits on newer platforms cause theres diff levels but idk about that since i havent f around with newer hardware. Though i think llc does kill cpu abit faster atleast on old platforms cause set cpu to deathzone with no llc then its fine vibing at 2.03v or something for 45nm but set llc and it just dies so thats another reason but sht doesnt degrade till like 1.65v on 45nm anyways so like i give a sht about that

[Mark Kaine]

19 minutes ago, Somerandomtechyboi said:

Idk, im on 775 where llc is either max or none, i choose no llc cause llc fks my overclocks, though board is garbage so that also fks my overclock

 

I do know llc can actually have benifits on newer platforms cause theres diff levels but idk about that since i havent f around with newer hardware. Though i think llc does kill cpu abit faster atleast on old platforms cause set cpu to deathzone with no llc then its fine vibing at 2.03v or something for 45nm but set llc and it just dies so thats another reason but sht doesnt degrade till like 1.65v on 45nm anyways so like i give a sht about that

oh ok i see, yeah they have these different levels precisely because of different degredation levels i think… generally up to llc3 is considered "safe" …i also had to set llc2 with my old 2200g, wouldn't boot when oc otherwise whatsoever…

im also just seeing  the ryzen clock tuner says to set llc4 (for MSI) so that is quite high, and probably also why i didnt get anywhere when i tried to oc my 3600 (i was mostly just curious) im fine with its performance as is, i dont even need pbo, 200 points more in cb is not something id notice in rl applications, i dont think…

 

Thing is most guides ive seen dont mention LLC at all, theyre just like "set this rather high voltage , set a really high oc, it just works…"  aka garbage lol.

 

(yeah, GN was different, he basically said more like "dont bother" lol)

[Somerandomtechyboi]

2 hours ago, Mark Kaine said:

oh ok i see, yeah they have these different levels precisely because of different degredation levels i think… generally up to llc3 is considered "safe" …i also had to set llc2 with my old 2200g, wouldn't boot when oc otherwise whatsoever…

im also just seeing  the ryzen clock tuner says to set llc4 (for MSI) so that is quite high, and probably also why i didnt get anywhere when i tried to oc my 3600 (i was mostly just curious) im fine with its performance as is, i dont even need pbo, 200 points more in cb is not something id notice in rl applications, i dont think…

 

Thing is most guides ive seen dont mention LLC at all, theyre just like "set this rather high voltage , set a really high oc, it just works…"  aka garbage lol.

 

(yeah, GN was different, he basically said more like "dont bother" lol)

Interesting that llc does help with newer hardware, well i think ill move to x79 and just skip x58 entirely since i can buy some dead x79 board for cheap, so if i wanna oc there im going completely blind with only intel max safe volt spec at 1.55v for 22nm and oc the balls off bclk as my only knowledge on newer platforms. If i cant find deathzone voltages then ill just take the board max volt and -0.4v and consider it safe i guess

 

 

Well most guides are aimed at beginners who dont know anything about oc and they are supposed to be universal so makes sense that some of the more complicated settings arent mentioned, ofc not good for any meaningful oc cause either they are stupid and being wayyyyy too conservative with voltage or they are just not giving enough information

 

For a proper oc youll have to learn yourself from scavanging around on the internet and finding bits that help with oc like raising bclk, i know high bclk on x58 means lower volt required for a given oc freq for whatever reason (something related to modulation or something), but i didnt know the other platforms actually benifitted from high bclk, thats why im gonna target proper x79 over chinese x79 due to more settings and being able to adjust bclk for higher freq, ive even seen someone do 5.1g on a sandy e with just 1.4v and 130bclk. Its all those little bits of info that really help with your oc and actually make it meaningful and an actual performance gain that isnt so small that might aswell just undervolt instead

[void pointer]

23 hours ago, Founders said:

You can try running ClockTuner for Ryzen, just make sure to read up on it before you run it. It tunes per core to give you the best results it can, it should be able to help get you a stable overclock at lower voltages than PBO would require.  

the thing with ClockTuner is that the current version works better for ryzen 5000 so I didnt have any luck with that one when I was following the tutorial got stuck in one step that was meant for ryzen 5000 and my motherboard gives like faulty voltage readings from software since works with offsets and there is no static voltage setting. I will rewatch the PBO guide that Anthony put out on ltt to see if I can get an improvement. The thing that I was aiming at initially is that my chip does like 1400 points on cinebench r20 and I noticed that most ryzen 3600 did like 1600 so with the oc I managed to get close to 1700 points but not stable enough 

[void pointer]

23 hours ago, Somerandomtechyboi said:

Ryzen is just garbage at oc since they already run close to diminishing returns zone from the factory, only way to get a meaningful increase is running 1.45v max where you can expect 4.4-4.6ghz if your chip isnt total garbage

 

1.35v is conservative, wayyy too conservative, its either stick to 1.25-1.3v for a slight overclock that prioritizes efficiency or balls to the walls overclock for max performance but slightly accelerated degradation (doesnt really matter for a year or 2), going anywhere inbetween is just stupid since you arent really gaining much performance yet your efficiency is ruined

 

I think der8auer tested ryzen degradation at 1.5v and seems like some dont and some do degrade but ofc its at max load all the time, just stick to 1.45v or 1.25v for an oc, anything in between makes no sense

that der8auer video was testing a 5600x and 5800x, would the same principle apply to a 3600?. I think I will give your suggestion a try and go for it with higher voltage but its kinda dangerous for me I think because my motherboard works with offsets and there is not static voltage setting and no LLC

[void pointer]

10 hours ago, Mark Kaine said:

ok, you might not know , but maybe you do, so at 1.25v my 3600 crashes immediately, so there has to be some tricks to get this working, i thought of using LLC for example?

 

thing is i tried to follow a GN "guide" specifically for the 3600 and i went all the way up to what he considered "safe" (i dont remember, maybe 1.36v or something) but it just would BSOD on windows log in every time.

And he also said he had to go 1.43 or something to get it stable, but wouldn't recommend doing so for 24/7 oc,  aka its a shit chip for overclocking lol, but afaik he didnt say anything about LLC, so any thoughts about that?

the sporadic BSODs that I get with 1.33v on my 3600 are usually at the windows log in as well. I will check that GN "guide" to see if I should give higher voltage a try