Need help with OC - single cores on 5950x not holding boost, crazy fluctuation

[Heyblinken]

I have a 5950x that i am trying to tune on ryzen master with the help of some youtube videos. Things were going along nicely manipulating PPT, TDC, & EDC and was getting all cores to a stable ~4.39GHz with a temp of about 77C while running multi core Cinebench R23.... When i got to single core OC i fired up the cinebench single core and noticed my single core boosting to 4.7 then going back down to 4.1 to 3.5 back to 4.1 back to 4.7 and kind of rinses and repeats, spending seemingly more time below 4.0 GHz than above it and only briefly touching it's highest speeds before dropping. I noticed the guy in the video i was watching, with the same CPU and settings, had about 4.9-5.0 and it wasn't moving at all (he was close to 90C, i am way lower as stated before due to a thick 360 &480 radiator custom loop). I have been playing on this system since jan/feb and have definitely noticed the jumping/unstable boost while gaming but thought it was normal.... Is there an easy fix to this unstable boost issue? I would love to maintain a boosted speed with all the money i sunk into the chip itself and the custom cooling. Also the chip surface was far from flat, had real unsatisfactory temps early on with the custom loop so i lapped it with decent results. I only mention that because if this is an RMA issue, i am screwed. Thank you ahead of time for your assistance on this one and please let me know if there is more info i am leaving out that would help you understand my situation better.

 

MSI Gaming X Trio 3090 with EK Vector Trio water block|
AMD 5950x With EK Velocity water block
x570 Asus crosshair hero wifi

32GB of 4000MHz Trident z with DOCP set to 3933 due to stability issues at 4000

[Bombastinator]

22 minutes ago, Heyblinken said:

I have a 5950x that i am trying to tune on ryzen master with the help of some youtube videos. Things were going along nicely manipulating PPT, TDC, & EDC and was getting all cores to a stable ~4.39GHz with a temp of about 77C while running multi core Cinebench R23.... When i got to single core OC i fired up the cinebench single core and noticed my single core boosting to 4.7 then going back down to 4.1 to 3.5 back to 4.1 back to 4.7 and kind of rinses and repeats, spending seemingly more time below 4.0 GHz than above it and only briefly touching it's highest speeds before dropping. I noticed the guy in the video i was watching, with the same CPU and settings, had about 4.9-5.0 and it wasn't moving at all (he was close to 90C, i am way lower as stated before due to a thick 360 &480 radiator custom loop). I have been playing on this system since jan/feb and have definitely noticed the jumping/unstable boost while gaming but thought it was normal.... Is there an easy fix to this unstable boost issue? I would love to maintain a boosted speed with all the money i sunk into the chip itself and the custom cooling. Also the chip surface was far from flat, had real unsatisfactory temps early on with the custom loop so i lapped it with decent results. I only mention that because if this is an RMA issue, i am screwed. Thank you ahead of time for your assistance on this one and please let me know if there is more info i am leaving out that would help you understand my situation better.

 

MSI Gaming X Trio 3090 with EK Vector Trio water block|
AMD 5950x With EK Velocity water block
x570 Asus crosshair hero wifi

32GB of 4000MHz Trident z with DOCP set to 3933 due to stability issues at 4000

Don’t see a whole lot of 5950 OCing.  ryzen2 Ryzen2 was a pooorer overclocker than ryzen+ which was a poorer overclocker than ryzen1  so it’s not super common. Ryzen3 could break that trend I suppose.   Haven’t heard it did though.  That more or less encompasses my knowledge of ryzen3 overclocking. 
 

I’m trying to think of additional information that might be useful, One thing that might be useful to someone (though likely not me as it has already gone beyond my Ken) is the specific video you are talking about because someone might notice something done in the video that creates the situation you are trying to get at that got missed.
 

[Heyblinken]

5 part series but this one is where he starts manipulating values. Like i said before though, prior to any manipulation of the precision boost overdrive i gamed for months and definitely noticed via fpscounter that the boost clocks hitting 4.7-5.0 but holding for a second or less before dropping. So it's definitely been an issue since before I knew it was an issue. I am not familiar with the CPU boost stuff so had no idea that boost clocks were supposed to hold for longer than a split second lol.

[freeagent]

From what I saw..

 

Manipulating 12+ cores is much different than manipulating 6 cores. With my 5600X I set a max all core OC to see what the CPU would pull. I noted those values and entered them into advanced PBO. And I was able to tune it so it boosted to its top clock for just about everything. My 5900X is not like that. More power did not help. But less did. The CPU will take what it wants regardless of what you set. Example. I  could set a "limit" of 130a EDC. That's fine, but if the program calls for heavy crunching, it will take the full 180+ EDC. More PPT, less everything else. Don't forget Curve Optimizer. That is a big one.

[Heyblinken]

yea i played with the values for PBO as described in that video and was happy with the tuned all core clocks. I then moved to manipulate those values again for a single core an that's when i witnessed the crazy fluctuation. I was going to get to the Curve Optimizer at some point but didn't get that far due to the single core issue.

[Heyblinken]

this is what my fastest 2 cores are doing during a single core R23 Cinebench run. They seem to boost, drop and then the other one boosts and drops. Back and forth.... Why are 2 cores double teaming a single core benchmark? 

 

(4.625 was not the highest. i think i screenshotted as it was rising but the max single core boost i was getting was 4.85 in that test)

[Bombastinator]

4 minutes ago, Heyblinken said:

this is what my fastest 2 cores are doing during a single core R23 Cinebench run. They seem to boost, drop and then the other one boosts and drops. Back and forth.... Why are 2 cores double teaming a single core benchmark? 

 

(4.625 was not the highest. i think i screenshotted as it was rising but the max single core boost i was getting was 4.85 in that test)

Heat and luck? A random guess though.

Edited July 27, 2021 by Bombastinator

[freeagent]

36 minutes ago, Bombastinator said:

Heat and luck? A random guess though.

No it’s a delicate balancing act. I don’t know how it works. I just know how to exploit mine..