Manual VID and multiplier instability comparison with CCD0-1 manual VID stability—this is why?
Manual VID and multiplier instability comparison with CCD0-1 manual VID stability—this is why?
first off heres the general specs: 5950x asus\rog crosshair viii hero (not dark hero) with latest 5002 firmware *has custom EKWB waterblock for cpu\motherboard\VRMs, its massive 64g ram, 4x corsair vengeance pro's, 16g 16cl@3600 (might be 18, whatever), docp enabled, otherwise the ram isnt overclocked past that (dont care to either, lets solve one issue at a time) strix 4090 on top pci with waterblock, not overclocked. this is my primary display card only using my samsung G9, only card\display that runs games. xfx amd 6700xt on second pci, not overclocked. this is just for my other 2 displays, msi/asus 32". only cause idk if adding additional monitors to my 4090 will impact gaming. plus in nvidia control panel i dont have the option for DSR or virtual resolution. the asus is 2560x1440 default, the msi is 1080 default, so for them to match my G9 i need virtual resolution to make it 2650x1440. yes it has a waterblock also quadro 4000rtx on bottom pci slot (disabled cause nvidia overlay displays utilization\temps\clocks\etc for the 4000 rather than the 4090), no monitors connected, no waterblock. 8x 1tb SSD's on various raid configs setup thru bios 2x 1tb nvme's non-raid, primary boot of course, plus stores some games seasonic 1000w psu corsair hx1500i psu (yes, two power supplies due to the gpu's and SATA drives) way more than enough radiators\fans for cooling. like... way too many corsair d5 pump res combo, 500ml (entire loop holds about a gallon of water, yeah...) --------------------------------------------------- software stuff --------------------------------------------------- all drivers updated\installed including latest chipset drivers and so forth windows 10 pro 22H2 (OS Build 19045.5440) according to "winver" Digi+ power control bios settings setup for OC'ing each time i do a cmos reset, enable docp, enable resizeable BAR i have the settings written down in the order of how its displayed in the menu, wont bother listing them since u wouldnt know what they are unless ur looking at the bios menu. its all stable and ok Important to note : when i stress test i do so in a real world type situation, not on super minimal services boot or safe mode type situation. i just figure if im gonna OC and be stable, best to do so with my trimmed windows services boot and applications loading. stuff like iCue, AI suite (do i even need it let alone Armory crate?), and whatever other software you'd normally have boot up on a daily driver machine. General questions to ask, just to start since i have all these GPU's, sata drives and nvme's, given my pci lanes are saturated could that affect overclocking\undervolting headroom? is there a specific order in which i should be overall, overclocking my cpu? for example, [reset cmos], enable docp, resize bar, setup digi+, save\restart, thinker with TSME and other stuff that has proven to increase OC headroom and reduce latency, do TPU II auto optimization then turn it off (some reason it reduces my temps drastically along with voltages - not sure what else its doing but hasnt been a problem). basically is there a general "step 1, 2, 3" method i should follow? cause if so that might be where im encountering issues would installing a fresh\trimmed windows install after OC'ing yield any benefits aside from the obvious clean install? having 4 dimms populated, would that effect my OC stability? i presume not only cause ive seen tons of YT videos of ppl with 2 or 4 dimms and still manage to hit 5.0 or higher since my ram is 3600 usually i set my fclk to 1800, sometimes auto on each cmos reset. although just for kicks one time i set it to 2000 and it booted up fine with no OC\PBO adjustments. is there any benefit to having a fclk higher than 1800? all that out the way i have noticed some weird irregularities in terms of voltage. if i set the cpu voltage to manual and do idk 1.2v then multiplier at 44x its either unstable or boot loops windows startup. yet on YT this guy has the same mobo\cpu as me but can do [email protected] all the way up to [email protected]. he dont even set the digi+ stuff, just clear cmos, enable docp\resize bar, save\restart, then only messes with the multiplier and core voltage manually, not with pos\neg offset or otherwise. then today i decided to try a per CCD manual frequency\VID, i forget what option it is but its under the advanced tab. anywho i just put VID as Auto (bios didnt complain), [email protected] and [email protected]. to my f*ing suprise it booted and is totally stable at 1.106v (according to HWinfo) even under load. plus it runs cooler than ive ever seen!!! prob cause the voltage is so low. i mean on a cmos reset, in bios im at ~59c @ 1.410v without tinkering with anything, these are all factory defaults. also im only at 43.25\41.00 just cause i wanted to see if it would work. wouldnt consider it an OC by any means, was just an experiment if you will. i do understand what vdroop is (thanks to a debaur, excuse my spelling if its wrong) which is why i compensate by tinkering with my digi+ stuff as thats where my LLC junk is. actually vdroop is, as i understand it, exactly why my system crashes when hitting "start" on R23. or is that a misunderstanding? i have had tons of stable overclocks even with PBO all core or per core just to be clear, even setting the multiplier and having a static cpu frequency. just never understood why on some configs it would crash when i hit start until recently. dumb as i may sound i always figured vdroop was a typo - wasnt until i seen it mentioned on many many forums did i realize i was just being naive. why is it that when i set my cpu multiplier to [email protected] and gradually increase voltage is it still unstable?? plus HWinfo does not reflect the voltage i set accurately. as an example if i set it to 1.4v HW info will say idk, 1.368v or something close. suppose that is to be expected and why my mobo has um, pinouts(?) for using a voltometer to get a more accurate power reading as it relates to the 24 pin mobo plug and bios settings why is setting the multiplier per CCD and VID STABLE versus setting the multiplier and voltage (literally on the advanced page, not within any tabs) completely unstable? i just find it weird that setting CCD0 to 43.25x, CCD1 to 41.00x and VID to auto (in same menu) gets me 1.106v versus [email protected] (or more on repeat tries) on advanced menu is completely unstable. whats going on there? sorry if all that looks like a hot mess, did my best to organize it so its easier on the eyes. id love to hear some feedback on all this stuff! ANY knowledge is greatly appreciated!! thanks a bunch peeps!
I don't understand why you attempted to locate this information independently rather than reviewing the numerous stories and analyses about Zen requiring high volt under idle, then returning to normal levels of 1-1.2 when under load? PBO with adjusted PPT offers the best solution instead of manually overclocking, as I assure you. Avoid wasting effort on overclocking RAM unless you're ready for most Zen 3 enhancements after enabling PBO with a stable negative CO value.
In this scenario, I believe the optimal approach for setting up PBO should align with the formula w=v*a or v=w/a, depending on how you interpret it. If PBO reduces voltage in stages across frequency and load, we can establish a fixed upper limit. For instance, targeting 1.5V (VID) for my CPU. Then apply maximum PBO, enable Ryzen support, perform stress testing, and compare the results against PPT (power per watt). A typical target might be around 400PPT, which translates to about 224W at 56% efficiency—meaning roughly 165W at full load. Keep in mind that voltage will drop under load, but we maintain a stable ceiling. This helps prevent thermal issues since the maximum voltage drop is controlled. Our PPT has its own limit, and I’ve observed it can get hot if it reaches its maximum. Given the MBO capabilities (around 190/200), we’ll adjust PPT until temperatures stay within safe ranges. Gradually tweak TDC/EDC and PPT to balance heat and performance during idle. Voltage and power are manageable with careful adjustments; just focus on keeping TDC/EDC close together, as this is crucial. Amps can be fine-tuned via PBO, which automatically lowers voltage per frequency demand. If we start at 1.5V, the required current will vary based on workload—single or multi-core tasks. Our calculations are sound; we must stay under 1.5V, respect PPT limits, and manage temperatures. The math is solid, but let’s keep it simple: control TDC/EDC, adjust PPT, and watch the balance.
Even at 65 watts, maintaining proper temperatures remains essential. Initially exploring PBO resources, I encountered many discussions and videos but none clearly explained the core calculations behind the three key figures. It wasn’t until recently that I found a visual aid to help understand how PBO functions. Based on my knowledge, PBO primarily relies on fine-tuning voltage (offset) according to frequency. If idle conditions have an offset that’s too high—say -30 versus -20—the processor won’t receive sufficient power in the low range, which could lead to instability like a 3.6 GHz clock speed drop. Under load, it might still operate at around 5 GHz. Stress testing isn’t necessary here. As I try to grasp these complex numbers, I notice others on forums show vastly different settings—some push maximum values while others use much smaller offsets. This inconsistency raises questions about why certain configurations use extreme offsets. With 16 cores, a uniform -30 across all seems counterintuitive. A mix of values near zero tends to perform better. If your magic numbers are set up correctly, adjusting them further could help. For example, using -7, -2, -8, or even -12 might be more effective than sticking with -30. The idea is to balance precision and stability. Many discussions highlight the importance of matching watts, amps, and voltage, so simplifying by focusing on one variable at a time makes sense. Understanding these relationships will clarify why some setups succeed while others don’t.
I tried to keep my explanation straight but it seems the details got mixed up. Let me simplify. There are different ways to interpret this, and each person might mean something slightly different. The example about cores and temperatures is confusing because smaller numbers can be bigger values. If another writer says the opposite, that’s possible too. I need more clarity to get it right.