s about overclocking and related topics
s about overclocking and related topics
Here’s a rephrased version of your text with varied wording:
Originally, I wanted to share my current setup details. The specs are just in case something relevant comes up:
-5950x, Asus/Rog Dark Hero Viii
I keep checking the latest BIOS updates regularly—usually just refreshing chipset drivers for AMD/X570. Everything seems fine; GPU drivers are up-to-date, and all graphics cards have the newest firmware.
Corsair Vengeance Pro RGB, cmw32gx4m2d3600c18—definitely the latest firmware too. The GPUs are running their latest drivers.
I have 4x dimms, each 16GB, totaling 18GB at 3600 MHz. I think the timing is around 18-15-15-22. Probably just a guess, but after checking the BIOS enough, it feels right.
The RAM was purchased in a set, so I ended up with two 2-packs. I double-checked the serial numbers—roughly, 90 is in 1A, 91 in 2A, 62 in 1B, and 63 in 2B (just random guesses, but they match their paired serial numbers). The custom Ekwb block was made for the Dark Hero, covering both CPU and VRMs.
Cryonaut thermal paste—Strix 4090 OC, custom block also. It fits my Samsung G9 only. The remaining GPUs are in the next slot down to the last one.
I used "speedster" AMD 6700X, with a custom waterblock (also from AMD). It handles my VG32 display well. Both are identical except brand and specs, but the bezel matches. I chose the 6700 because it’s easier to identify via Super Virtual Resolution.
I tried NVIDIA Control Panel for settings, but it was frustrating—setting a custom resolution was a hassle. I reached 60Hz from 165Hz and never hit 2560x1440. So I decided to let the AMD card handle everything. This card isn’t meant for gaming anymore; it’s just for general use.
I bought a NVIDIA Quadro 4000 RTX, set to GPU and compute tasks. No monitors connected. It’s mostly for mining. My PC rarely shuts down, so I’m planning to keep it running all night.
No waterblock found—I can’t locate one anywhere. I’m debating removing the card, swapping it out for an NVMe RAID card with bitrification. With four Kingston Fury drives (fastest and IOP), I’d add another 2TB on RAID 1.
"Gamemax" 80 Bronze 1050W PSU is handling the GPU. Since I’m considering swapping the Quadro for an NVMe card, I’ll be able to remove this card from the pile.
Corsair HX1500i (estimated wattage), 80 Platinum PSU—currently unused, but will replace both Seasonic and Gamemax when pulling out the Quadro. It’s a great deal! Around $250 for a PSU that costs over $500. The brand is new.
Watercooling gear: 500ml Corsair RES/Pump, DC5 naturally. I have a 1L "fish tank" for extra water. The UV LEDs under it look stylish on top of my case.
I installed two 360mm radiators—one slim and one 60mm. The larger one is mounted front-to-front pulling air in, the smaller one at the back pulling air out. There’s a separate pump to circulate water from the tank into the radiator and back again. I think with all radiators having a cyclone in the largest one, it’d help.
I have a 140mm (slim) on the rear exhaust, no fan—just passive cooling. The other radiators are 420mm by 60mm, also pulling air out. There’s a single 420mm radiator with a separate pump for the tank. I figured connecting it to another pump would work.
I have a custom block on the MMO/4090/6700, which includes Cryonaut and thermal pads. I followed instructions carefully, adding more of the same size across components that wouldn’t fit otherwise. I think if I broke down the GPU, it would make sense.
I verified proper CPU-CPU contact thanks to extra pads. I’m trying to spread heat evenly across this big block for better cooling. It looks like a challenge, but I’m determined.
I’ve been experimenting with "stock overclock" lately. Recently, I tried the PBO feature in BIOS. The latest version is installed, along with my chipset drivers. Fans are set via AI Suite app (Asus/ROG Dark Hero Viii).
First, I reset the BIOS, applied resizable BAR, set RAM via DOCP settings, and did a TPU II auto-tune. Cinebench R23 gave me 11 passes at 26,488, and a single-core 1,340 with only one pass. I ran a five-minute test—feeling like I should have seen better results.
During stress tests, clocks dropped, and temps stabilized around 75°C. CPU throttled but ran smoothly. Multi-core score was 26,488 at 11 passes, temps in the 75°C range, 4.1GHz. Single-core: 1,340 with a 19.77x MP ratio.
I tried PBO last night after saving bios settings to USB. After restarting and stress testing, I got lower scores than expected. This is my best setup so far. At this point, I’m hitting around 5.1GHz—though not all cores. The multi-core score is 24,288 with a 90°C temp.
I’m confused by the fluctuating scores. Why does changing PBO settings affect results so much? The temps are high but stable. I’m still trying to figure out how to get Ryzen Master working. It’s not working as expected—no dynamic OC option shown in many videos.
I read somewhere that disabling hyper-v/SMT/virtualization lowers scores. I disabled it and tried again, but nothing changed much. I’m stuck with a 4.0 or 3.8 boost while doing lighter tasks, then jumping to 4.9GHz when needed.
Single-core: 1535 MP ratio of 15.82x. The temps were around 74°C with 5.0–5.2GHz steady.
I saved PBO settings to USB and set defaults afterward. I configured DOCP, set RAM speed to 3600, and adjusted memory clock. At 1800 since it’s half my RAM. The BCLK is 100.00—so not that. I used stock RAM voltage (as on the box) at 1.35V and went into Digi+ tuning.
I’m unsure how Cinebench single-core stressing works, but Task Manager usually shows 0–3 doing work while others idle. If anyone can explain R23’s single-core performance, that would help.
Multi-core: 24,288 score, 90°C temps, 3.3–3.4GHz holding steady. Single-core: 1535 MP ratio of 15.82x. With 12+ hours of corecycler stress testing, I got a 4.2GHz score but throttled to 90°C.
I’m not sure why boosting to 4.0 or 3.8 while doing lighter tasks gives higher scores than boosting to 4.9GHz.
I noticed my CPU’s max clock drops from 5.0 to 3.6 to 4.8 under PBO, but it still stays around 90°C. I think the multiplier fixes clocks but doesn’t help much.
I’m trying to stress-test Ryzen Master settings. It’s not working—performance boost is off, and multiplier is at auto. I’m guessing core isolation isn’t enabled. I’ll deal with that later.
Still, Ryzen Master seems stuck. If I could run at 4.0 or 3.8 while doing less work, boosting to 4.9GHz would be ideal. With cooling time, that might be possible.
I’m not sure how Cinebench single-core stressing works—Task Manager usually shows 0–3 doing work. If you can clarify how R23 handles single-core stressing, it’d help.
I’m curious about PBO settings and their impact. Should I enable positive/negative values? Why does a lower negative number (like -30) mean better cooling? Should higher numbers be better? What’s the MP ratio for different core settings?
I’m trying to write down the best-performing cores in order, based on HWinfo data. The numbers are confusing, but I need this to make sense.
I’d appreciate any tips on getting Ryzen Master working. I think it’s not about drivers but a service I disabled at startup. I have questions like: What is the multiplier in Cinebench? How does single-core stressing work in R23? Why does auto-overclocking without PBO give more points than with it?
for clarity i've recorded the prefCore / Hardware Core numbers manually. Below you'll find the details. It looks like these values are very close, almost identical. 1 pref corresponds to core 1/2, which suggests the hardware ID is shifted by +1 from the pref core. Should I adjust my pbo settings depending on temperature or the pref/hardware cores? and how? what does it do? essentially, i need to understand whether i should base my configurations on these numbers or not. it seems like they start at zero, not one, so i'm entering values correctly after double-checking.
The numbers you mentioned refer to performance metrics like PPT, TDC, and EDC. These values indicate the power consumption and thermal characteristics of your processor. A lower PPT means better efficiency, which helps reduce temperatures on the TDC and EDC, especially during single-threaded tasks. Higher values suggest poorer efficiency and higher heat output. For single-thread loads, you want lower PPT to keep things cooler. The negative numbers (neg values) show how close a core is to optimal performance—closer to zero means better handling of those loads. With more cores, tuning each individually can help balance performance and cooling. It seems you're aiming for a mix where overheating cores get adjusted toward lower temps, which should boost their speed while keeping heat in check. Regarding your question about boosting frequency at a certain temperature, it depends on the specific model and settings—your experience with the 5900x to 5950 suggests you're trying to optimize for stability and performance. As for your boost frequency at a given temperature, that would depend on your BIOS/UEFI settings and the current thermal state of each core. If you're unsure, testing in real conditions is key. Your approach sounds logical, but always verify with actual measurements or manufacturer guidelines.
My CPU features several cores with clock speeds ranging from 4950 to 5150MHz, while others sit between 4850 and 4850MHz. Performance varies based on workload; Linpack Xtreme at 10GB loads around 4550MHz, and lighter tasks like TM5 run near 4750MHz across all cores. The older design handles stress well, though it struggles under extreme conditions. Edit: Ryzen Master isn't reliable—I stick to the BIOS instead.
hm well that's a solid point, not every processor can reach its full potential. in my setup I’m lowering the boost for the slowest unit and matching them with that speed, then increasing later. i wasn’t expecting it! based on what I’ve learned so far about these three factors, it feels more like a trial-and-error process. i’ve seen discussions where people with similar gear swap numbers—like 260/170/200, 225/325/245, and even edc185-190, tdc165-170, ppt270-300 (out of order). i tried the last one and found that pushing higher PPT values like 250-300 usually leads to lower performance. here’s a suggestion i found online—it outlines those three parameters clearly. it might help others who are reading. ppt mainly affects the maximum TDC, but it still plays a role in setting your peak limit. you can adjust ppt to 300w, but if your tdc is capped at 50a, you won’t reach that target. essentially, tdc sets the upper speed limit for your core. ppt is what dictates your ultimate clock rate. in short, higher tdc generally means faster clocks, though it eventually hits a ceiling. for me, it’s about balancing how fast the core spins versus how much power it uses. setting ppt too high can act like a limiter—tdc won’t go beyond that number. it starts to slow down and matches the workload’s demands. i think you’re onto something here. the tdc/edc values matter more than you think; they shape your performance curve. if you want to push further, you’ll need to tweak the limits carefully. this process feels a bit like tuning—slow and steady, but sometimes you have to guess.
You installed Ryzen Master and had to clear AMD GPU components beforehand. Restarting and installing it brought everything back to normal. Odd, but now the BIOS is working properly again.