F5F Stay Refreshed Hardware Desktop Are faster clocks slowing things down?

Are faster clocks slowing things down?

Are faster clocks slowing things down?

S
Simon_303
Member
169
06-09-2022, 03:04 PM
#1
My 5800x3D runs on a 140mm AIO because of ITX constraints, so the initial temperatures were right around 90°C, which isn’t ideal. That’s why I reduced the voltage to lower the temps and improve core clock stability. At 1.144V (-0.075v), my all-core clocks reached about 4.25GHz, and Cinebench R23 scored 14200 with temps in the 75-80°C range. When I went even lower to 1.08V (-0.15v), speeds jumped to ~4.35GHz but the scores dropped to 12225 and temps fell to 60-70°C. HWinfo also shows an "average effective clock" that’s quite low at those settings. Is it possible HWinfo is underreporting the speeds, or are they actually running at 3.7GHz at 1.08v? Also, BIOS negative offset was used for undervolting.
S
Simon_303
06-09-2022, 03:04 PM #1

My 5800x3D runs on a 140mm AIO because of ITX constraints, so the initial temperatures were right around 90°C, which isn’t ideal. That’s why I reduced the voltage to lower the temps and improve core clock stability. At 1.144V (-0.075v), my all-core clocks reached about 4.25GHz, and Cinebench R23 scored 14200 with temps in the 75-80°C range. When I went even lower to 1.08V (-0.15v), speeds jumped to ~4.35GHz but the scores dropped to 12225 and temps fell to 60-70°C. HWinfo also shows an "average effective clock" that’s quite low at those settings. Is it possible HWinfo is underreporting the speeds, or are they actually running at 3.7GHz at 1.08v? Also, BIOS negative offset was used for undervolting.

C
Create_Miner
Junior Member
24
06-10-2022, 07:23 PM
#2
It’s likely the performance will be a bit higher than expected but only for a short time due to power limits. The temperatures are lower, so it attempts to increase output, but it gets restricted by power or temperature again. A 140 AIO won’t suffice for this CPU type; I know you’re using an ITX setup, but you should have considered that more carefully if you want the best results. Instead of upgrading, why not retain the existing system and repurpose it for another purpose? You could build a larger case with a 280mm AIO for better performance.
C
Create_Miner
06-10-2022, 07:23 PM #2

It’s likely the performance will be a bit higher than expected but only for a short time due to power limits. The temperatures are lower, so it attempts to increase output, but it gets restricted by power or temperature again. A 140 AIO won’t suffice for this CPU type; I know you’re using an ITX setup, but you should have considered that more carefully if you want the best results. Instead of upgrading, why not retain the existing system and repurpose it for another purpose? You could build a larger case with a 280mm AIO for better performance.

D
dorklad
Member
68
06-13-2022, 06:10 AM
#3
Clock stretching is the toughest aspect of overclocking fresh components. Ryzen processors handle it using PBO and reduced voltage, Nvidia GPUs manage it, Intel chips occasionally do it, and it's fairly typical on new hardware—still pretty frustrating to work with.
D
dorklad
06-13-2022, 06:10 AM #3

Clock stretching is the toughest aspect of overclocking fresh components. Ryzen processors handle it using PBO and reduced voltage, Nvidia GPUs manage it, Intel chips occasionally do it, and it's fairly typical on new hardware—still pretty frustrating to work with.

E
eduardodd08
Posting Freak
852
06-13-2022, 08:59 AM
#4
Opposite of what many think, reducing voltage isn't harmful—it actually sacrifices performance.
E
eduardodd08
06-13-2022, 08:59 AM #4

Opposite of what many think, reducing voltage isn't harmful—it actually sacrifices performance.

L
ljhawk2000
Member
55
06-13-2022, 04:51 PM
#5
I shouldn't have heard any of this nonsense. It's a gaming CPU that runs smoothly at 50-60°C with a 0.075v undervolt, and boosts nicely to 4.5ghz. What's the issue? For work tasks, at 0.075v it's fine, and I'm seeing 4.25ghz—still safe even at stock 90c. Previously, I was getting 3.9-4.0ghz, which felt a bit low. So, 140mm is more than enough. I just wanted to know why my performance dropped when clocks were higher and temps lower. The other person explained it involves clock stretching, and I'm looking into it now. Against common belief, at -0.075v my speed jumps up by 200MHz and temperatures drop by 10°C. I was curious about what happens beyond that point, and a helpful Bullzoid video confirmed the same behavior. The 5800x3D has an overly aggressive voltage curve, so undervolting only offers benefits without downsides.
L
ljhawk2000
06-13-2022, 04:51 PM #5

I shouldn't have heard any of this nonsense. It's a gaming CPU that runs smoothly at 50-60°C with a 0.075v undervolt, and boosts nicely to 4.5ghz. What's the issue? For work tasks, at 0.075v it's fine, and I'm seeing 4.25ghz—still safe even at stock 90c. Previously, I was getting 3.9-4.0ghz, which felt a bit low. So, 140mm is more than enough. I just wanted to know why my performance dropped when clocks were higher and temps lower. The other person explained it involves clock stretching, and I'm looking into it now. Against common belief, at -0.075v my speed jumps up by 200MHz and temperatures drop by 10°C. I was curious about what happens beyond that point, and a helpful Bullzoid video confirmed the same behavior. The 5800x3D has an overly aggressive voltage curve, so undervolting only offers benefits without downsides.

D
DlNKLEBOY
Member
176
06-13-2022, 06:47 PM
#6
Execute a workload similar to Linpack Xtreme and observe your GFlop results. You'll notice what I mean, especially during the 10GB test. Clock speeds aren't very useful without sufficient power to run them.
D
DlNKLEBOY
06-13-2022, 06:47 PM #6

Execute a workload similar to Linpack Xtreme and observe your GFlop results. You'll notice what I mean, especially during the 10GB test. Clock speeds aren't very useful without sufficient power to run them.