Ryzen 3600 Help?

[pio]

Hey guys,

So I'm not a complete noob, but with Ryzen 3000 I might as well be.  Any guides or tips?  I seem to be hitting a thermal wall and I can't for the life of me figure out why.  Using PBO and -100mV I can get 4.275GHz boost clocks, pretty consistantly.  Using manual vcore and multis I can push 4.3GHz, 1.27v but I hit 95*C near immediately on SmallFFT's on P95.  I'm sure this chip could do 4.3+ if I can figure out why its overheating so drastically.  I see people all the time on other forums saying 1.30v+ is safe.  PBO seems to be "safe" where its at since it throttles down to like 4.15GHz under smallfft's.

Any ideas?  Cooler is a Dark Rock Pro4, and the TIM application should be near perfect I've reseated like 20 times....pretty sure I'm a pro at that now.

EDIT:
I do now have a EKWB 240mm AIO I could replace the Dark Rock with.  However I don't think my cooler is currently saturated with a 65w TDP chip running at PBO.  It says 250w TDP right on the box.

Edited March 9, 2021 by pioneerisloud

[PCSarge]

3 hours ago, pioneerisloud said:

Hey guys,

So I'm not a complete noob, but with Ryzen 3000 I might as well be.  Any guides or tips?  I seem to be hitting a thermal wall and I can't for the life of me figure out why.  Using PBO and -100mV I can get 4.275GHz boost clocks, pretty consistantly.  Using manual vcore and multis I can push 4.3GHz, 1.27v but I hit 95*C near immediately on SmallFFT's on P95.  I'm sure this chip could do 4.3+ if I can figure out why its overheating so drastically.  I see people all the time on other forums saying 1.30v+ is safe.  PBO seems to be "safe" where its at since it throttles down to like 4.15GHz under smallfft's.

Any ideas?  Cooler is a Dark Rock Pro4, and the TIM application should be near perfect I've reseated like 20 times....pretty sure I'm a pro at that now.

EDIT:
I do now have a EKWB 240mm AIO I could replace the Dark Rock with.  However I don't think my cooler is currently saturated with a 65w TDP chip running at PBO.  It says 250w TDP right on the box.

ill stick my nose in here, as i started with a 3600 and moved to a 3900XT

 

that 3600, no matter how hard you slam it, will hit a clock wall at 4.4ghz unless you LN2 and voltage the snot out of it...... if you want a non high temp OC. PBO needs to be disabled, and that sounds like a possible bad cooler mount(pressure wise, not you wise). also, drop a straight edge on that IHS. some ryzen chips have been known to be a little convex on the IHS. my 3900XT running that speed with more heat output doesnt even come close to 80C. nvm 95C.

[pio]

23 minutes ago, PCSarge said:

ill stick my nose in here, as i started with a 3600 and moved to a 3900XT

 

that 3600, no matter how hard you slam it, will hit a clock wall at 4.4ghz unless you LN2 and voltage the snot out of it...... if you want a non high temp OC. PBO needs to be disabled, and that sounds like a possible bad cooler mount(pressure wise, not you wise). also, drop a straight edge on that IHS. some ryzen chips have been known to be a little convex on the IHS. my 3900XT running that speed with more heat output doesnt even come close to 80C. nvm 95C.

Really?
 

The only way I can get this thing to run WITHOUT hitting 85*C+ temps using SmallFFT's Prime95, is by using PBO. -_-  I can't even run it at 1.25v set in BIOS without it overheating in my opinion (hits like 87*C or so).  The only way it'd be possible for me to even get 4.1GHz manual would be under 1.20v.  There's absolutely no way I can even get 4.2 to run stable unless I'm doing something wrong.  PBO....its fine though.

Basically, mine will NOT overclock at all.  Since PBO is basically stock anyway.  Kind of disappointing really lol.

TIM application appears to be squishing out appropriately.  It does in fact look like there's not enough pressure, but there's no other way for me to mount the cooler.  I followed the directions on the Dark Rock Pro4, and its not like there's room for error in it lol.  Well.....it is me afterall, I'm sure I can find a way to catch it on fire.

So yeah, 1.25v ish and less, its fine manual.  1.25 itself is pushing it.  More than that, I hit over 90*C immediately.  Seems to almost act as if I'm hitting a voltage wall or something, like the CPU just can't dissipate the heat fast enough into the cooler.

Edited March 9, 2021 by pioneerisloud

[PCSarge]

30 minutes ago, pioneerisloud said:

Really?
 

The only way I can get this thing to run WITHOUT hitting 85*C+ temps using SmallFFT's Prime95, is by using PBO. -_-  I can't even run it at 1.25v set in BIOS without it overheating in my opinion (hits like 87*C or so).  The only way it'd be possible for me to even get 4.1GHz manual would be under 1.20v.  There's absolutely no way I can even get 4.2 to run stable unless I'm doing something wrong.  PBO....its fine though.

Basically, mine will NOT overclock at all.  Since PBO is basically stock anyway.  Kind of disappointing really lol.

TIM application appears to be squishing out appropriately.  It does in fact look like there's not enough pressure, but there's no other way for me to mount the cooler.  I followed the directions on the Dark Rock Pro4, and its not like there's room for error in it lol.  Well.....it is me afterall, I'm sure I can find a way to catch it on fire.

So yeah, 1.25v ish and less, its fine manual.  1.25 itself is pushing it.  More than that, I hit over 90*C immediately.  Seems to almost act as if I'm hitting a voltage wall or something, like the CPU just can't dissipate the heat fast enough into the cooler.

i havent air cooled since first gen i5, so air cooler wise i dont know how good the mounting pressure is on that thing. you may want to mount that AIO up. it will help you with more thermal mass for long load stability, and possibly more consistent mounting pressure. my 3600 only really wanted to stay completely under contol at 4250mhz. if i went to 4.3 or above the voltage jump was reduculous. i used to have this 3900XT on an all day 4.4, but i wasnt using it enough, so i threw PBO back on and undervolted to 1.15v and left it. Ryzen is much more finnicky overclocking and more time consuming. i suggest downloading ryzen master, leave the CPU at default with pbo on and everything. and run the ryzen master stress test, it wil ldetermine which cores are weaker, and you can overclock on a per core basis from within windows. beware as it does apply settings to bios if you hit apply.

 

Edited March 9, 2021 by PCSarge

[damric]

What I've noticed as the nanometers get smaller and smaller is that the heat problem isn't the cooler itself dissipating the heat, it's the chip getting the heat off the die itself and getting it through the IHS to heat sink. This has been a problem for intel as well for years. The heat flux coming from the CPU is just so dense in one tiny location that it's hard for even the best coolers to capture that heat.

 

One thing that helps is if your cooler is tight asf. My TPC-612 often cools as well or better than my TPC-812 because the mounting mechanism is different and have much more pressure despite having less heat pipes and half the fin stack size.

 

The other thing that helps is COLD. Heat naturally transfers to something colder. If you can keep your cooler plate cold, it will transfer that small dense heat flux much easier. Air coolers have a disadvantage here as they heat up to steady state quickly, meaning the base plate touching the IHS eventually gets quite warm while it's doing its job, by design. On the other hand a water block tends to stay cold much longer, especially if it is a custom loop with much reservoir which will take a long time to reach steady state.

 

I gave up trying to run PBO with my Ryzen 3600. It ran consistently near 1.5v and was always hot. I dialed in an all-core overclock of 4.2GHz and 1.25v with aggressive LLC and it generally hovers around 1.26v and stays much cooler that way. It ran very well like that with every air cooler I tested over the last few months, but kind of crappy on a H100i GTX (saturated too fast under load). Now I have a high volume custom loop so it's a non-issue. I'm about to start running some tests against the loop today in fact to see how well it stands up against the ProSiphon Elite I tested last month.

[UltraMega]

My only two cents is that overclocking on ryzen make such a small difference that it doesn't really matter anyway. 

[pio]

16 hours ago, damric said:

What I've noticed as the nanometers get smaller and smaller is that the heat problem isn't the cooler itself dissipating the heat, it's the chip getting the heat off the die itself and getting it through the IHS to heat sink. This has been a problem for intel as well for years. The heat flux coming from the CPU is just so dense in one tiny location that it's hard for even the best coolers to capture that heat.

 

One thing that helps is if your cooler is tight asf. My TPC-612 often cools as well or better than my TPC-812 because the mounting mechanism is different and have much more pressure despite having less heat pipes and half the fin stack size.

 

The other thing that helps is COLD. Heat naturally transfers to something colder. If you can keep your cooler plate cold, it will transfer that small dense heat flux much easier. Air coolers have a disadvantage here as they heat up to steady state quickly, meaning the base plate touching the IHS eventually gets quite warm while it's doing its job, by design. On the other hand a water block tends to stay cold much longer, especially if it is a custom loop with much reservoir which will take a long time to reach steady state.

 

I gave up trying to run PBO with my Ryzen 3600. It ran consistently near 1.5v and was always hot. I dialed in an all-core overclock of 4.2GHz and 1.25v with aggressive LLC and it generally hovers around 1.26v and stays much cooler that way. It ran very well like that with every air cooler I tested over the last few months, but kind of crappy on a H100i GTX (saturated too fast under load). Now I have a high volume custom loop so it's a non-issue. I'm about to start running some tests against the loop today in fact to see how well it stands up against the ProSiphon Elite I tested last month.

Okay....so maybe give this EKWB Basic 240mm AIO a go then huh?  I mean a Dark Rock Pro4 should be absolutely more than enough cooler for a 65w Ryzen 3600.  But, meh.  I'll try it if it means I can get 4.3.  I really do think 4.3 should be achievable all core on mine, it does it at 1.275v ish give or take, maybe 1.30v.  Problem is keeping it cool. -_-  Seems that voltage range is "safe", just dunno how if it won't stay below 100*C lol.
 

15 hours ago, UltraMega said:

My only two cents is that overclocking on ryzen make such a small difference that it doesn't really matter anyway. 

Oh no I'm fully aware of that, and IF clock / memory clock and timings make much more difference.  I already have that dialed in with my IF at 1866 and my RAM at 3733 CAS 16.  I can get 1900 IF clock but its not totally stable and requires BCLK adjustments to get there, so figured I'd go back to 1866.

It's more or less I want 4.3 because I know the chip SHOULD be capable of it safely.  I'm seeing 4.275GHz on all cores on PBO under "light" loads like gaming.  1.30v and under "should" be safe.  I don't get it lol.  That's more or less why I want it.  Because I can't get it and I should be able to.

[damric]

Keeping my fingers crossed for you but I suspect that your 24/7 clock may have to be less than 4.3GHz as mine is.

 

On a positive note, I was benching at 4.45GHz yesterday morning when my water loop was chilled from leaving the windows in that room open overnight. I could even run the AIDA64 FPU with AVX and only hit 76C with 1.456v applied.

 

https://docs.google.com/spreadsheets/u/0/d/15yM6y7f4AuRuj0qp0J4HX4Pyohc3-nGwIq9Kua9MQsA/htmlview#gid=0

[BWG]

I'm going to give you my perspective since zen3 has the same issues, if not worse. First, these chips can run at 90C+ and not thermal throttle. Not saying that's the best idea, but the spurts of usage PBO causes should not cause an issue. The chip is designed to run hott when needed, but we all hate seeing those high temps. I have 2x 360 RADS with RGB push fans that just look cool, and 6 ultra kaze pull fans (push fans are not sealed and have plenty of gaps and ultra kaze's are sealed to the back). I use an EK Supremacy Evo right now, but have a Heatkiller 4 Pro in the mail to me after ordering it like nearly 25 days ago!

 

Anyway, in your PBO OC Menu in the bios, you may be able to set the EDC, TDC, PPT limits. Do a p95 run and watch what hwinfo picks up as the max, and then set each to that. It should alone be a 10C or so drop. Let me know how that goes because PBO is way better than All Core Fixed Voltage in Single Threaded applications and low usage scenario's by about 5-10% per my testing.

 

CTR is another option you could look into. 1usmus made that program. Stick with the live builds. If you need help using that program, let me know. It will basically test for max all core oc and you can set it to do a short run at a higher clock, and then default to a lower OC for extended all core usage. When low usage, it defaults to PBO. They also have a beta I've been testing that does it's own PBO, but I haven't had good luck with it yet. It's still Beta.

[pio]

  

16 hours ago, damric said:

Keeping my fingers crossed for you but I suspect that your 24/7 clock may have to be less than 4.3GHz as mine is.

 

On a positive note, I was benching at 4.45GHz yesterday morning when my water loop was chilled from leaving the windows in that room open overnight. I could even run the AIDA64 FPU with AVX and only hit 76C with 1.456v applied.

 

https://docs.google.com/spreadsheets/u/0/d/15yM6y7f4AuRuj0qp0J4HX4Pyohc3-nGwIq9Kua9MQsA/htmlview#gid=0

Thanks.  I mean really PBO @ 4.275 mostly all the time on all cores (so gaming) is fine.  Folding I sit around 4.05-4.15GHz depending on the core (more at 4.15).  Overall it performs amazingly well and I'm super pleased with out of the box performance.  It's just....dang, there's NOTHING to tweaking these things at all.  Buy the right parts, plug in the stock / tuned numbers, and bam, you're good to go.  There's no fun in that.

Dude, I got in at 4.5 and primed till I saw 105*C.  Really wasn't very long, like 5 seconds, but still.  Didn't even think about it since voltages were all "safe" (I think that was 1.375v).  That's why I know this is a good chip, if I can figure out how to get over 1.275v to run without overheating lol.  I don't have any proof on that, and since I now KNOW I have a cooling problem somehow I'm not trying it again anytime soon lol.  So for now, we'll assume I get 4.275 off PBO lol.  Nothing else I've ran seems to work quite right just yet.

Isn't 1.275v kind of a really low voltage wall to be overheating at?  I mean really it should be closer to 1.25v if I want to keep under 90*C.  Everything I've been reading seems to say 1.35v is normal, even stock.  Hell, PBO gets to 1.45v at times.

 

13 hours ago, BWG said:

I'm going to give you my perspective since zen3 has the same issues, if not worse. First, these chips can run at 90C+ and not thermal throttle. Not saying that's the best idea, but the spurts of usage PBO causes should not cause an issue. The chip is designed to run hott when needed, but we all hate seeing those high temps. I have 2x 360 RADS with RGB push fans that just look cool, and 6 ultra kaze pull fans (push fans are not sealed and have plenty of gaps and ultra kaze's are sealed to the back). I use an EK Supremacy Evo right now, but have a Heatkiller 4 Pro in the mail to me after ordering it like nearly 25 days ago!

 

Anyway, in your PBO OC Menu in the bios, you may be able to set the EDC, TDC, PPT limits. Do a p95 run and watch what hwinfo picks up as the max, and then set each to that. It should alone be a 10C or so drop. Let me know how that goes because PBO is way better than All Core Fixed Voltage in Single Threaded applications and low usage scenario's by about 5-10% per my testing.

 

CTR is another option you could look into. 1usmus made that program. Stick with the live builds. If you need help using that program, let me know. It will basically test for max all core oc and you can set it to do a short run at a higher clock, and then default to a lower OC for extended all core usage. When low usage, it defaults to PBO. They also have a beta I've been testing that does it's own PBO, but I haven't had good luck with it yet. It's still Beta.

I have messed with my PBO settings as best I can per the Ryzen guides floating around other forums.  Never thought to check p95 like that and set the PBO accordingly.  I'll have to write that down and see if my BIOS has those options, or pull up my manual or something lol.

Does that temperature drop result in higher PBO clocks or something?  I'm boosting decently enough as is, and full load temps using PBO aren't horribad since it downclocks.  I'd just like to see all 6 cores a little higher under full loads since I know all 6 cores can clock higher and do.  I mean I could probably pull off a 4.25 all core no problem and keep temps under 90C SmallFFT's (folding about 75C).  I dunno, I'm still learning these things and so far not all that pleased with tweakability at all. -_-

Edited March 11, 2021 by pioneerisloud
added second quote

[BWG]

The reduced temps did allow me to do a higher PBO. Lower temps do help it do a little more clock speed.

[UltraMega]

I just went from Intel to amd and this is my first amd build. I can relate on feeling like the tweak ability is disappointing. Coming from Intel it feels like overclocking is essentially pointless, but I have to remember that it's also pretty pointless on Intel now too. CPUs are running at higher and higher clock speeds and basically taking over the overclocking themselves with better and better boost/turbo capabilities. 

 

We're never going to have gains even half as good as what sandy bride gave us again and overclocking today is way more hobby than practical. It used to be really practical. 

[PCSarge]

2 hours ago, UltraMega said:

I just went from Intel to amd and this is my first amd build. I can relate on feeling like the tweak ability is disappointing. Coming from Intel it feels like overclocking is essentially pointless, but I have to remember that it's also pretty pointless on Intel now too. CPUs are running at higher and higher clock speeds and basically taking over the overclocking themselves with better and better boost/turbo capabilities. 

 

We're never going to have gains even half as good as what sandy bride gave us again and overclocking today is way more hobby than practical. It used to be really practical. 

its gone from a practical "hobby" to just "a hobby" those of us who really like it wont let it die

[pio]

16 hours ago, UltraMega said:

I just went from Intel to amd and this is my first amd build. I can relate on feeling like the tweak ability is disappointing. Coming from Intel it feels like overclocking is essentially pointless, but I have to remember that it's also pretty pointless on Intel now too. CPUs are running at higher and higher clock speeds and basically taking over the overclocking themselves with better and better boost/turbo capabilities. 

 

We're never going to have gains even half as good as what sandy bride gave us again and overclocking today is way more hobby than practical. It used to be really practical. 

Oh by all means I agree completely with that.  I wasn't expecting even SB level overclocks, or even Nehalem level overclocks.  Hell, I wasn't expecting Core 2 Duo overclocks, or even FX level overclocks.  I just wanted to see PBO speeds as advertised on the box on all 6 cores instead of just 1.  I figured it could do at LEAST that much since it does at stock with one core at a time.  Guess not lol.  But I'll try BWG's advice next tweaking session, probably over this weekend while I'm building my Phenom build up (again).

[damric]

Have you tried raising VDDP and SoC voltages?

[pio]

13 hours ago, damric said:

Have you tried raising VDDP and SoC voltages?

VDDP, no.  Well, yes and no?  That's the one that I believe the guides said to set to 1.05v.  I had tried upwards of 1.10-1.15v yes.  Pretty sure its still sitting at 1.05v right now though, maybe 1.10v.  I noticed zero difference with this setting.

SoC, I have 2 settings for it and I could be using the wrong one maybe?  I'd have to reboot and double check for sure, but I'm about 99% sure I have it set "correctly", and yes I've raised it as high as 1.295v trying to get things tuned.  I believe I have it sitting at like 1.125v right now for 1866 IF clock.  Again, I have noticed absolutely zero difference changing this setting either.  I can get to 1866 IF using all auto settings and just adjusting IF and memory dividers and nothing else.