[SOLVED] Confusion regarding amd overclocking and undervolting with clockturner 2 Revision 5

Herr B

Commendable
May 29, 2020
179
36
1,690
So originally, I did not have plans to overclock my cpu but now I got quite intrigued by seeing clockturner which I got recommended here.

I know very little about overclocking so yeah. Perfect recipy to screw up a 5000$ computer as a first time overclocking project :).

Luckily I'm a graduated it technician so I know a little bit about computers and their internal workings. AdditionallyI did a lot of smaller projects involving hardware hacking so I am quite confident I can do it given some recherche. :)

First things first, let's clarify the basics of my system and my goals.

Hardware:
CPU: Ryzen 9 5950x
CPU Cooler: Noctua NHD-14
GPU: Asus rog RTX 3090 oc
Mainboard: Asus Rog Strix x570-I
Ram: 2 * Crucical ballistics 32gb 3600
PSU: Corsair SFX 750 platinum

Goal:
As you might have noticed, the parts are for a mini itx build. Right now the parts are still in full tower and air cooled since my case is not finished yet. The parts will be watercooled by 2 * 240mm radiators.

As this is a mini itx case I have restrictions with the psu which is sketchy on the rtx 3090 but the ryzen requires around 100 watts under full load so its sketchy but its fine.
Additionally, the Computer already acts as a heater when under full load and I have plenty of performance already. Im mainly after some minor performance optimisations / improvements and if possible to decrease the temperature a little so I don't sit in the sauna in summer.
Primairly its a workstation but I also play games on the beast. So a nice mix between efficiency, all core- and single core performance is the goal. Right now however, I feel there is always gonna be a tradeoff but we will see.
My cpu is now running at 1 volt. That seems very low to me but I wanted to find out the results. Base clock apparently stayed at 3400, boost clock went down from ~4890 to ~4070. This is where my confusion starts.

Questions:
As written before, I got clockturner recommended and this is how I got into this.
The CPU is undervolted considerably, the all core performance did a massive jump up (eventhough the max boost is considerably lower). My best guess is: It's because it's not running as fast but more steady and has more thermal headroom (due to the undervoltage). Additionally the heat now constantly stays at 50°c under load as opposed to 60-70°c. Those are major improvements for a workstation. The Productivity score for most professional programs roughly stays the same but with higher efficiency. At the same time all core perfomance made a big jump. I'm heavily relying on high parallelisation workloads which is why I bouht that beast in the first place. In that regard, I'm quite impressed with the result of that tiny program.
But here comes the hickup:
This might not apply to all scenarios. Single core performance got a huge drawdown due to the max boost increase. Eventhough the timespy benchmark scores went up by 200 points, I would see a considerable performance decrease in my current favourite game, eft which is incredibly cpu bound and not well optimized yet. The game was running smoothly before and now I encountered small hickups in menues which were annoying.
The fps would drop considerably. Basically from beeing stuck to the 144hz fps cap to around 125-146, depending on the complexity of the scene.
I hope, so far I understood everything correctly.
Looking at my beautifuy childish view of overclocking, It appears to me as if I have mainly achieved a great reduction in heat generation, a mediocre reduction of for the sacrifice of single core performance.

Is there such a thing as reducing power without impact in performance or is it always a tradeoff?

Next steps
To gain some more knowledge about the behavior of the system, I want to try to increase voltage ever so slightly step by step. Until I reach the Temperature of 60°c under full load which is the temperature where the cpu starts to throttle

At this point, I would expect this to be the "sweet spot" because the maximum performance is gonna be taken while my system can handle the heat. I don't know yet what to do about the single thread performance. My guess is now that I have to specify p1 profile and p2 profile separately. Unfortunately the profile has not saved on last run.

Question regarding the forum:

I'm not sure if I should have written that under the other thread. Or if it is running more like stack overflow where one separate question is one new thread.
 
Last edited:
Solution
So originally, I did not have plans to overclock my cpu but now I got quite intrigued by seeing clockturner which I got recommended here.

I know very little about overclocking so yeah. Perfect recipy to screw up a 5000$ computer as a first time overclocking project :).

Luckily I'm a graduated it technician so I know a little bit about computers and their internal workings. AdditionallyI did a lot of smaller projects involving hardware hacking so I am quite confident I can do it given some recherche. :)

First things first, let's clarify the basics of my system and my goals.

Hardware:
CPU: Ryzen 9 5950x
CPU Cooler: Noctua NHD-14
GPU: Asus rog RTX 3090 oc
Mainboard: Asus Rog Strix x570-I
Ram: 2 * Crucical ballistics 32gb 3600...
So originally, I did not have plans to overclock my cpu but now I got quite intrigued by seeing clockturner which I got recommended here.

I know very little about overclocking so yeah. Perfect recipy to screw up a 5000$ computer as a first time overclocking project :).

Luckily I'm a graduated it technician so I know a little bit about computers and their internal workings. AdditionallyI did a lot of smaller projects involving hardware hacking so I am quite confident I can do it given some recherche. :)

First things first, let's clarify the basics of my system and my goals.

Hardware:
CPU: Ryzen 9 5950x
CPU Cooler: Noctua NHD-14
GPU: Asus rog RTX 3090 oc
Mainboard: Asus Rog Strix x570-I
Ram: 2 * Crucical ballistics 32gb 3600
PSU: Corsair SFX 750 platinum

Goal:
As you might have noticed, the parts are for a mini itx build. Right now the parts are still in full tower and air cooled since my case is not finished yet. The parts will be watercooled by 2 * 240mm radiators.

As this is a mini itx case I have restrictions with the psu which is sketchy on the rtx 3090 but the ryzen requires around 100 watts under full load so its sketchy but its fine.
Additionally, the Computer already acts as a heater when under full load and I have plenty of performance already. Im mainly after some minor performance optimisations / improvements and if possible to decrease the temperature a little so I don't sit in the sauna in summer.
Primairly its a workstation but I also play games on the beast. So a nice mix between efficiency, all core- and single core performance is the goal. Right now however, I feel there is always gonna be a tradeoff but we will see.
My cpu is now running at 1 volt. That seems very low to me but I wanted to find out the results. Base clock apparently stayed at 3400, boost clock went down from ~4890 to ~4070. This is where my confusion starts.

Questions:
As written before, I got clockturner recommended and this is how I got into this.
The CPU is undervolted considerably, the all core performance did a massive jump up (eventhough the max boost is considerably lower). My best guess is: It's because it's not running as fast but more steady and has more thermal headroom (due to the undervoltage). Additionally the heat now constantly stays at 50°c under load as opposed to 60-70°c. Those are major improvements for a workstation. The Productivity score for most professional programs roughly stays the same but with higher efficiency. At the same time all core perfomance made a big jump. I'm heavily relying on high parallelisation workloads which is why I bouht that beast in the first place. In that regard, I'm quite impressed with the result of that tiny program.
But here comes the hickup:
This might not apply to all scenarios. Single core performance got a huge drawdown due to the max boost increase. Eventhough the timespy benchmark scores went up by 200 points, I would see a considerable performance decrease in my current favourite game, eft which is incredibly cpu bound and not well optimized yet. The game was running smoothly before and now I encountered small hickups in menues which were annoying.
The fps would drop considerably. Basically from beeing stuck to the 144hz fps cap to around 125-146, depending on the complexity of the scene.
I hope, so far I understood everything correctly.
Looking at my beautifuy childish view of overclocking, It appears to me as if I have mainly achieved a great reduction in heat generation, a mediocre reduction of for the sacrifice of single core performance.

Is there such a thing as reducing power without impact in performance or is it always a tradeoff?

Next steps
To gain some more knowledge about the behavior of the system, I want to try to increase voltage ever so slightly step by step. Until I reach the Temperature of 60°c under full load which is the temperature where the cpu starts to throttle

At this point, I would expect this to be the "sweet spot" because the maximum performance is gonna be taken while my system can handle the heat. I don't know yet what to do about the single thread performance. My guess is now that I have to specify p1 profile and p2 profile separately. Unfortunately the profile has not saved on last run.

Question regarding the forum:

I'm not sure if I should have written that under the other thread. Or if it is running more like stack overflow where one separate question is one new thread.
CTR 2.0 by 1usmus is not exactly an overclocking tool, it's more about finding best performance while using as low voltage as possible. As such it's "overclock" is not best tool to get most (all core) overclocked performance.
Ryzen Master in it's latest version is pretty good way to start overclock and then build on it by manual settings in BIOS. It could be a good base to start with. Although it has own stability checking, other checks by OCCT, Prime95 etc. goes further but CB r23 should suffice for preliminary testing.
 
  • Like
Reactions: Herr B
Solution