11:39 Actually it is very easy to pinpoint the core that crashed. Just open your event viewer and create a custom view for the WHEA-logger. As soon as you crash, go to the viewer and check the APIC number that corresponds and divide it by 2 and there you have your crashed core. If you open CPUZ in the tools section, export an html and search for the APIC section and see which number corresponds to which core on your system. It might sound complex but it really is fast and accurate
You don't actually have to use CPUID to figure it out. If you have a 4 core processor with hyperthreading, it will be like this: Core 1 Thread 1 = APCID 0 Thread 2 = APCID 1 Core 2 Thread 1 = APCID 2 Thread 2 = APCID 3 Core 3 Thread 1 = APCID 4 Thread 2 = APCID 5 Core 4 Thread 1 = APCID 6 Thread 2 = APCID 7 For example, if you get an error with APCID 26, then it is the 12th core in the processor causing it (particularly the first thread on the core). If you don't have hyperthreading, then the APCID will simply be identical to the core number. With hyperthreading, each core has 2 APCIDs. The APCID always starts at 0, so your core 1 will have APCID 0 and APCID 1 with hyperthreading, or just APCID 0 without hyperthreading. Perhaps I'm over explaining, but you can see the pattern in my example above.
It was simple at beggining because the error in Event viewer was telling me the APIC ID... But after that my pc was crashing randomly wihout telling pe the APIC nr, just that the pc shutdown.... I tried to lower them by turn and it didnt work... I had to lower all of them by 5 and now is ok..
Stability testing can be done with OCCT which allows to cycle single thread variable loads across all cores and tells you which core was unstable. Very straight forward even with 16 core CPUs
Came back for a second comment. I watched so many guides on PBO and CO. While I kept tweaking and guessing it was when I watched THIS video that it clicked in my brain. Thank you for the best explanation of all time for this tech because this should be the official guide. I stopped at -15 on my best cores and -20 on the rest as this seems extremely stable and my benchmarks are the best theyve ever been. Keep up the amazing work and thanks so much as you've turned my gaming rig into a tuned machine with spreading your knowledge! 140PPT 90TCD 130EDC was my sweet spot for keeping this machine in check. :D 5800x + 6600 XT build.
While I was working on overclocking today I was watching the speed graph in Ryzen master and I was able to pick out the core causing the crashes pretty easily. Almost all of my cores on my 5950 were following nearly the same path on the graph when they ramped up and down except cores 1 and 3. I was able to pick out those 2 cores as the ones with instability and increased (decreased) my negative offset on the other cores 5 at a time until some of those cores also started to wander a bit from the others and started showing instability. I was able to pick through my cores and maximize the negative offset of each core in a pretty timely manner with minimal crashes this way. I recommend watching that speed graph for anyone that wants to save themself a bunch of time. It helped me immensely.
Thank you so much for taking the time to make this. Seriously. I was tearing my hair out trying to work out why my 5950x kept crashing with PBO on. Turns out the default PPT, TDC and EDC on my Aorus X570i pro wifi were set way too high. You rock, dude!
From my past year experience that seems to be more related to your temperature than the actual settings of these limits. If your cooling solution is not capable of handling a 250 watt cpu, you shouldn't play with overclocking. Also, undervolting with the curve optimizer can substantially decrease your temperatures so you should be able to push these limits higher.
of all the vids pertaining about this new function. You have really made it easy to understand properly. i have a 5800x and i have been trying to get it to run consistently. I tried CTR from 1smuz. its good but it gives me black screen lock ups whenever i run 3dmark. thanks for taking the time out to explain it =) my max on single core boost was 4.8 without crashing. now im at 4.95-5.04..
@@ryanlaseter7626 his new project hydra is good tho. really great on temps and the hybrid oc is great. the adjustable CO's across the frequencies has helped me stably run a 4133mhz ram kit on the 5800x. (at 1:1:1) prior to that i had to drop the kit down to 3800 or i'd lock up running r23 and even r20.. on the balls out profile i manage 17,900ish in r20. previous id hit like 16200 then lock up on the last section. oh and roughly 5.0-5.1ghz boost clocks @ 1.32mv with soc at 1.135.. barely goes over 80c on an EKWB AIO super the 360rad with 6 fans for push pull. and like 85c with just the fans on pull.
I’ve been looking for this concise of a video for in depth explanation of amd overclocking and you delivered very nicely, thank you. Deserves way more attention.
The most common crash after overvolting is changes between high load and all core idle. This is because those change points cause high ripple currents which may trigger the crash.
THANK YOU, I'm new to Ryzen and it is a whole new ball game. Your video really helped me understand how it all works. I'm on old school overclocker so this really helped me :)
Much prefer testing with occt. Detects errors which don't crash the pc/app and tells you which core os responsible for the error so you know what to focus on. I also tested single core by pinning the task to specific cores in task manager rather than always let it run on the same prefered core resulting in a more stable overclock (it helped me to detect the crashes which wouldn't occur if it ran on the prefered single core) my results for 5600x -12, -12, -5, -2, -15, -5.
Great video - explained properly and exact!!! Just tuned my 5900x and 3-4 cores are easily sticking to 5,04Ghz now, most of them up to 4,8-4,95 at single core tests. All core goes up to 4,6 - 4.67Ghz. Just made CCD1 -17 and CCD2 - 30 with EDC of 125, and PPT 250, TDC 170 and I almost never go over 70C with 360mm AIO :) Now it's almost perfect, but as I'm on linux and no windows, per cor settings won't be easy at all :D Still way better than before as I would go up to 4925 on the best cores and barely keeping 4,5GHz all core with over 80C temps and the same cooler :) If with the new settings I keep the EDC at 120, I barely go over 65C :D AMAZING video!!! Keep it up !!!
5:44 I have also noticed mentioned tendency myself: higher EDC gives higher temperature but clock speed is lowered, while lower EDC gives lower temperature but clock speed is higher. Assuming that the EDC limits are reached in both cases, it means that the computer draws more current and works slower, yet generates higher temperatures . This doesn't make sense. Unless we assume that at lower clock speeds and higher EDC, the processor is more heavily loaded but operates slower (or wastes power). I think that changing all these PBO settings (EDC, PPT, TDC) is merely an optimization for specific applications, rather than overclocking per se. A more loaded processor operates at a lower clock speed, while a less loaded processor operates at a higher clock speed. Which is better depends on the application.
Testing shows that EDC impacts performances differently at different values. To high impacts single core, to low impacts multi core. The sweat spot seems to be around 120-130 for optimal results on both on 2 ccd cpus.
I tried all of the overdrive settings on a 5800x with an Evo 212 v2 cooler im building and found that leaving the overdrive settings to auto and setting curve optimizer low as I could which was -30 (later set to -25 to ensure stability for my friend ) along with max cpu temp set to 80 c was the easiest option that had almost the same results as tweaking the PPT/EDC, i get 4600-4650mhz multi thread 4850 mhz single. If its not stable at the above settings change curve optimizer by 5 until you get a few stable runs of cine. If you so desire with the above settings you then should be able to run cinebench & then watch the PPT/EDC/TDC % usage values in ryzenmaster and then have a good idea of manually adjusting the values to match the desired max temps/mhz your looking for without setting a max temp. Example: the PPT showed 83% of 142w on cine with auto settings -30 co which is approx 118w @ the 80c limit i set so adjusting the PPT to 118w along with changing the other values the same way should result in the same temps without the lower temp limit . Before i started tweaking the settings my cpu Shot to 90 C (max) temps in a jiffy and although everything i have read says these temps are normal i wanted to reduce them to have a safer margin without much loss of performance.
Hey O!Technology, if you increase the 3 values of the PBO limit you will push more voltage in your CPU and you will have high temps. You should disable the PBO limit so the algorithm will find the lower voltage that the CPU can run the high frequencies. Also if you choose the lower power saving plan you will see a drop in temps without loosing performance. About 95% of the time your CPU will be at max 85% in usage so what you see in the stress test it will never happen (not even if you render a video). Aside from this the power saving is made to better adapt the consumption to the load and if the load needs all the cores at max, it will allow the voltage without loosing performance.
I didn't get the "setting PBO Limits" part. What is the meaning of "something you probably will never reach" value? If we have to set a value that something we will never reach, why don't we just type 9999? If that's the point.
What if I increase my negative offset to the maximum my bios will let me (30) for every core and it is still stable? Did I not utilize something to the most that I could have? I have an all core clock of 46.50 at 3.65v right now and it was pretty stable but I might have been able to increase it some.
dude you just saved me so much headache, I had all my pbo settings maxed I dropped to 120a on edc got 4.675ghz all core on cb20 but my overall was only 4.55, upped to 139a which is my sweetspot for 4.610 all core 😅 my man 💪 👏
Damn didn’t think about the 5900x having two different sets. I was putting all core to -20 and still getting bad temps! I suppose it was boosting way too much for it. Gonna try and put the first 6 cores to -15 and last 6 cores -20 or -25 and see what happens.
When using OCCT cpu test it tells you what core starta to do miscalculations if it's not stable that way you can fine tune curve optimizer super fast. I can do a negative offset of 30 on all cores but core 2 and 4 on my 5600x. Those I have at 15 and 5 respectively and that is rock stable. And even though it looks to boost higher than just PBO +200Mhz the performance is actually slightly worse in cinebench multicore than with just PBO +200Mhz. I havent set other limits than the motherboard though. Maybe thats why?
There is a way to find which cores crashed during a test. By using the WHEA log report in windows after a crash, you can see a message of a CPU ID like; "Processor APIC ID: 8 - A fatal hardware error has occurred". To see which core this belongs to you can open CPU-Z, click Tools and then "Save report as .txt". In this report the threads, Core IDs, CCXes and Nodes for your CPU is shown. For example, I have an AMD 5600X. An error message of "Processor APIC ID: 8" tells me that Thread 8 caused a fatal hardware error. Thread 8 is the first Thread on Core Nr 4 (Between cores 0, 1, 2, 3, 4 and 5 on my 6-core CPU).
When I was testing an all core OC on my 5900X, I found Cinebench, Blender OpenData, and an x264 encoding test to be most useful to test my CPU's stability. One particular scene in blender would push the CPU so hard that at anything higher than 4.4Ghz, it'd go past 90C with my cooler which was not so good 😂 but yeah, a plethora of benchmarking tools would be best to validate different workloads.
I had a stable overclocked 3950x at 4.3Ghz for over 9 months. It gave me a 5 to 15% performance boost over my stock settings 🤔. It also meant I had more heat and more voltage. TO ME the gains weren't enough for me to warrant keeping my 3950x overclocked. Today I just cut on my XMP profile and keep my 3090 overclocked with MSI AFTERBURNER. Maybe if I saw a 20% or higher performance boost I might still keep my 3950x overclocked. But I see no reason why I should since the stock value is closer to what I was able to get with an overclock MINUS the heat and voltage 🤔.
That's because no one should use a fixed clock overclock on Ryzen you're actually losing performance in most applications because the boost behavior is not working anymore.
Hey overclockers, ive never done this before. I have a 5900x and a 4080 super playing on 3440x1440 with some annoying bottlenecks on the factory CPU clocks. Should i try this on my tomahawk B550 motherboard? Cpu gets cooled with AIO Arctic Liquid Freezer II 240.
this is really cool and i was an allcore overclocker till this day, will give this a try, ryzens are really difficult to understand and to squeeze performance without sacrificing single or multi core scores.. is it possible to overclock per core?
Yep give this a shot! You kinda have to let it do its thing more than on intel cpus. Edit: Overclocking per core is possible but you will not be able to hit the same boost clocks as using curve optimizer still.
I think I found out why lowering the EDC amperage allows for higher clocks it seems after a point when it hits too much amperage the vcore will go down. So lowering the amperage allows for more vcore headroom.
Hello bro. I have r9 5900x cpu asus rog stix 3070ti 8gb oc gpu asus tuf gaming x570 plus wifi COUGAR MX660 MESH RGB 4x120mm RGB MESH GEX 850W 80PLUS GOLD MidT ATX Lian Li Galahad 240mm ARGB All-in-One CPU Liquid Cooler How do you think I can overclock the processor, graphics card and ram in this system? Not for rendering, just to improve game performance. because the monitor I use has a refresh rate of 280 hz and is in 1080p resolution. I need to get better performance to feel 280hz in fps games like pubg, so I want to overclock. may I help you?
This could be easily applied to undervolting scenarios too. Thanks for this video. I'm gonna put up my big boy pants now and undervolt PER CORE instead of all core, seems that is the intermediate-advanced way to apply better tuning values other than simply putting in all core due to the nature of CCD arrangement in Ryzen CPUs. This video onwards, I will need to read up on per core optimization.
I turned off all limits in MSI bios after worrying warning. my 5950x ascended to 5.2ghz and then tripped something and system froze. i guess I am doing it wrong. is 1.45v on cpu safe or shhould I stay under 1.3?
Great video ! So being close or around 1.5v in hwinfo is all normal? Should I use LLC with pbo? When using LLC3 it's stable but on auto, my per ccx at 1.3 won't be stable either
Almost all CPUs today have thermal throttling, which dials down the boost or core frequency when a top temperature limit is reached (to protect the CPU from damage). Is there motherboard functionality now where the performance/frequency/voltage of the CPU is regulated based on a thermal limit as the target ie overclock esp in low or low core loads if temperature allow it (beyond boost specifications).
Did you compare this vs the performacne of an all core overclock? In most scenarios an all core overclock gets better framerates and performance than curve optimizer. Curve optimizer makes it look like its hitting higher clocks but the performance doesnt scale with the clockspeeds.
"In most scenarios an all core overclock gets better framerates and performance than curve optimizer" Which scenarios would these be? I call BS. Show benchmarks!
Thank you for this Video. My cores dont want to boost Higher than 4.65 GHz. Any advice? I am Not running Into any Limit in Terms of EDC, PPT or TDC. Trying to get to 5 GHz Single Core btw
but now is my issue of CO I have all -30 apart from best cores -25 is that hurting them or? it's been fine running all the single cores for an hour with no occt errors
i cant make my pbo limits higher in ryzen master, only lower. do i have to uncap the pbo limits from bios first? how do i do that can someone explain i dont wanna mess anything up rn
I put the second highest setting on my Gigabyte motherboard called high as that seems to work well now that I tested more. Don't know what is higher on your motherboard 1 or 5 but if its an Asrock motherboard chances are 1 is a higher setting.
Some comments: 1) Not all Ryzen 5s and 7ss are 1-CCD. Some are 2-CCDs, which can be seen from e.g. ZenTimings debug report. 2) Clock Tuner for Ryzen can give comparable performance but much better volts/temps. But only on Windows. 3) OCCT or CoreCycler can be used to ease testing curve offsets per core. 4) There are faster ways of coming up with curve offsets, e.g. ruclips.net/video/4gDfMwOBnoE/видео.html
All true and good advices. Thanks for sharing. I know about all those but thought they were not really necessary especially clock tuner where you're sacrificing single core speeds.
@@OTechnology can some cores be enabled even though they might not meet grade quality in the ryzen 5600x, due to 1 or 2 ccd having 8 cores each, but 6 only are needed for 5600x.
I think there were some controversies (source from Reddit, thus anecdotal) regarding CTR 2.0 for Ryzen involving security flaws on AMD subreddit. Is there any reason to not simply use Ryzen Master to tune the CPU? Thanks for sharing the other additional tips. Going to do some due diligence after this.
its not normal that your cpu hit 90 C I have 5900x with noctua nh-d15 with stock fans and after copy-paste your settings in ryzen master, my max temp is 71C.
If I don't want manual PBO settings what should I use like Auto, Motherboard etc. Also what happens if you set these in the bios then change it in Ryzen Master does it override the bios?
Please, pleaseeee help...Im using 5800x on asus tuf gaming x570 plus...my temp's are more than 90* while gaming...Im not an expert on overvloacking, I tried some things in bios but not sure I did it right...can you tell me some values for my cpu for tdc, edc and ppt or I have to find my own values...I just want to lower my temperatures on cpu...I dont want them to go over 80*...can you help me please.??
5900X running -30 All Core and no crashing. Multi core clocks boost to 4.545GHz and single core to 4.95GHz. The BIOS won't allow anything less than -30, is there any way to squeeze more out of it?
I'm in the same scenario as you, ryzen 9 5900x with x570 asus dark hero. Latest bios and latest amd chipset driver. I have everything at -30 on all cores and ran OCCT for a few hours at stock power limits ppt=142,tdc=95,edc=140. Not a single crash with maximum cpu boost override at +100 to try to hit atleast 5ghz. Was peaking between a little over 4.9's ghz to 4.6 ghz on single core and multi cores were running at 4.5 ghz stable. Cinebench 23 scored single core at 1620, and multi core was at 21205. Which I read was as good as stock. My thermals did get a decent drop in cooler temperature by 5c. But no matter how low I go on my negative offset was for each core my cb23 scores didn't go up, until I bumped ppt to 165, tdc to 115, and edc to 160. My multi score went to 22450. But my single score always stayed the same. Thinking about removing some negative offset to my two best cores on my ccd0 and ccd1 probably at -15 or -20 to see if it'll at least peak at 5 ghz. But this stuff gives me a head ache for minimal gains.
I've been using Cinebench, AIDA, Geekbench, 3dmark and Prime95. My cpu is stable on everything i do, except on prime95 torture test. Would it be safe to assume that the torture test in prime95 is meant to test base clock speeds for fabric failure? I dont see any other software being able to push the CPU that high.
I personally think passing prime95 torture test with small ffts is sort of pointless if it passes all the other usual tests. You will never run into a scenario where you will be under anywhere near that level of stress. Unless you do machine learning or something crazy with your computer. In that case maybe just turn on pbo and leave the rest stock.
5600x here. Ended up with Negative 0 on best core (yes, ZERO) and around 20-25 on the rest. Does it make any sense that the difference is that much? oO Also, the crushed never happend at high load for me :) only drops to idle
That can happen yes. Its really a silicon lottery and I think if only one core is not doing well its totally fine. The multi core clocks will still be substantially improved.
@@OTechnology Thanks for reply. Well, my CPU overall seems very good. So the lottery is nice but odd :). I was testing it with usmus1 tool and there was info it's gold sample. Great content btw.
I was running an all core set to 4.75ghz at 1.2875v and was performing well but i knew that having a good pbo profile is much better but i just stopped caring as much when i used easy tune from gigabyte and had a perfect oc and the virus program it is corrupted and i couldn't get it to work again so i just used ryzen master, atm this is lookng good but haven't tried any games or other loads on it yet other than r23, however on a 5800x i am at 250,170,100 for power and at negative 30 for all cores but still stable is this right?
It used to be 500 MHz but a while ago manufacturers changed it via BIOS update to limit it to 200 MHz. Probably too many CPU's died and caused too much false warranty cases, which caused too much cost. But that's just a guess why they changed it. 200 MHz is the new maximum now.
Hey! I’m new to all this! I followed the first part of the video, modifying my PPT, TDC, and EDC. Here is the weird part. On my first test on cinebench (my baseline test) I already had the native “game mode” on ryzen master turned on (PPT: 720, TDC: 420, EDC: 480, very weird figures I know). I then modified all three values, and found that the sweetspot for my 5800X is an EDC of 80. That’s the highest clock speeds I was getting on HWmonitor. After running the test on cinebench I actually got a lower score, although as I said I was getting higher clock speeds. Any explanations. This stuff is proving too complicating for a noob like myself.
I have 5600x with Arctic Freezer 34 Esports DUO. I check the stats with HWinfo64. Testing with Cinebench R23. When my "PBO limit" is on "Motherboard" the temps go to 82, when "Disabled" temps are 66. With same other settings. Can you tell me why is that? I saw that when I set Curve Optimizer to all cores "-30" and PBO boost "+200", the Single core clock is higher "4,841" but the Multi-core goes down around "4,260 - 4,313"... Cinebench Multi-core test ~11000 or little lower. Like I said, if I change "PBO limit" to "Motherboard" the temps go 82 to get higher Multi-core clock, but it is not worth it. If I set PBO to just "Enable", without any other setting, the Single core is ~4,65 but the Multi-core goes up to 4,55, at the cost of 82 degrees, which is not good again... I want to achieve both high Single core and Multi-core with around 66-70 temps! Any suggestions? P.S. I didn't understand what is the manual method, calculation to determine the EDC, TDC, PPT values, this is why I tested with "PBO limit" either on "Motherboard" or "Disabled".
@@OTechnology Even enabling PBO may damage your CPU and the official party line of AMD is that CPU is no longer covered by warranty if you enable PBO. Of course, they would have hard time enforcing that limitation here in EU: because they obviously advertise PBO as a *feature* for both the motherboard and CPU, it must be supported feature and as such cannot be off-the-limits for warranty. For countries with less-strict consumer protection laws, I would consider twice before activating PBO.
As first, this is a very good video! Thank you But I have a question, does it have the same effect if I set all cores negative 5 instead of per core and then set all to negative? Cause a friend of mine can set negative 30 at all cores and still doesn't get a crash.
Nice and inspiring vid! One thing I found really annoying on my ASUS board is that after updating the BIOS which essentially updates the AGESA version. The off-set unit was re-shuffled. This resulted in a crush that damaged my windows once after loading the exact same BIOS profile after an upgrade. Don't know if that happens on other brands.
@@CidoraX The BIOS always resets itself to default after updating. I read from other sources that AMD changed the settings in certain AGESA version to have a more conservative strategy. But now it seems they changed it back a bit that allows me to go for a higher negative off-set value (My current BIOS version on my ASUS mobo is 2423)
Def not any more work to dial in CO on a 12/16 core Ryzen 9 unless you are actually spending hours increasing CO units on a per core basis to pinpoint which core crashes at a given offset. The only way to efficiently set your CO is to create an Event Viewer profile in windows to log when you get a crash. Even prime95, Cin. R23 can be stable and certain games/apps will crash after you get it dialed in for stress testing. The only way to accurately know which core is crashing is to log it. After a crash you go check your event log to see which core crashed and reduce the neg offset on that core. Also LLC can play a big role in CO stability. Increasing LLC is a must for stability with high offsets in my experience. I've only had 1 Ryzen 5000 not be able to hit over 5ghz light core workloads but it was also the only China batched chip 5800X I have ever tested with.
Do you have discord my game stutters trying to figure it out nothing is over heating I believe it’ll running the same motherboard you have in this video I’m Rather new to Ryzen I have the Ryzen 7 5800x
Your scores are way higher than mine with similar settings on my 5900x + aorus master combo. About 22,000 multi, 1600 single. 4.55GHZ all core and up to 5GHZ single. Any ideas?
Austin, you sound like you know what your doing. I could really use help on getting my system to run at its best. Honestly feel like it should be much better than it is. I’ll $$ someone to help me!! I’m running ekgaming vanquish 295 I believe is the name. It’s got fully liquid cooled even the crazy RTX 3090, ryzen 5900x, asus Prime pro motherboard, 32GB gskill 3200 ram, 1Tb Samsung 980 pro, another Samsung 1Tb 970 evo pro and 2 other drives I don’t use lol. I have been trying Ryzen master abs so on and I obviously don’t know a lot. Any interest in helping me? Would appreciate it thanks.
@@kaubut0145 I might be able to help you. I'll admit I'm no pro at overclocks, but I could at least try to help you out. Just give me a way to contact you and we can try to figure something out together.
it could be background tasks and general bloat. with my 5900x i can see a 3-400 point difference (in r20, so not the same as r23) depending on what other apps are running in the background. When I do my tests I make sure everything else is disabled first.
my 5900x can go negative 30 on all cores, pbo motherboard, +200mhz cbco, scalar 10x and still i am 5% in 3dmark cpu. dunno what i can do more.(edit:-from 4.95ghz i went 5.150ghz but still low scores in 3d mark 5-7%) 75C with nzxt x63
@@OTechnology i tried cb23(22800score)/3dmark cpu(11400), didn't have any issues (edit: also playing any games without issues). i ll try that software but it kinda beats me.
@@alexmckoy8898 yea those don't show instability because the cpu boosts differently per core so you need software that can do that which is corecycler.
Hi can you help? Everytime I try to use PBO (default values) or Auto OC (default values) I keep getting 96+ degrees and throttling down the Mhz to 560 or lower. It never crashes but it keeps overheating. So I can run only on default and that gives me around 60 - 65C. I'm using Prosiphon Elite Giant which should be one of the best coolers for AMD cpus according to benchmarks
Its only good for threadripper. On regular Ryzen CPUs a Noctua NH-12S beats it by a good amount and a NH-D15 is even better. Prosiphon only works for threadripper with a big IHS to spread the heat and the high TDP that can actually boil off the liquid. This has been tested by manu reviewers and people on reddit r/overclocking.
@@SilverIlly3 Oh then maybe you didn't put enough thermal paste? The X with 4 dots is usually the preferred method for threadripper. Only enabling PBO shouldn't be a problem for the prosiphon. Also better to have too much thermal paste than too little which will hurt performance while having too much you'll just make a mess is the worst that can happen.
@@OTechnology I had a PC professional apply the paste and he followed the Icegiant video tutorial and made 8 dots around it with enough thermal paste so that shouldn't be a problem.. any other idea of what could be an issue?
since my case scenario probably not the same than you been lower core count which often run in low usage scenario wouldn't going whit agressive negative offset on vcore voltage whit loadline calibration that overshoot a little for those rare time power is requiered would not end be best ways to achieve a théorical perfect balance of efficient as well as been powerful when requiered. -thx for sharing whit us? i would encourage you to try to avoid abreviation terme that can be unlcear for some user by giving more detail to there nature whit perharps some visual addtion text like you did in some occasion avoid to lose the viewer whit terme there not familiar since not everyonne first language is english tend to get lost with those i would lead to believe could yield nice result in the long terme
I would love to see some 5900x cinabench runs at 4.9 all core. Or close to it. My 5900x hangs out around 4.8-4.9 with only pbo2 and a 240mm aio and some mx5 and it NEVER goes about 68-69c I’m pretty impressed with its performance. I might go for 5.0 I believe the world record atm is 5.9 for a 5900x and 6.0 for a 5950x but that’s with liquid nitrogen. I’ll stick with water personally haha. Cheers. Next time plz show more over clicking and benchmarks, less just talking. Other than that, great video..👌🏼✌🏼
You never go above 68-69c while running cinebench with pbo2 on a 5900x? Thats crazy. What is your ambient temp, and what cooler are you running? My 5900x with an evga clc240 aio hits well into the 86-88c range in r20 with pbo2 and dialed in CO. It scores well though, around 9050-9100, just gets hot. Gaming it stays around 65-70 though. Ambient is around 18-19c (~65f)
@@Alex-wb2yw I’m using an arctic liquid freezer II 240aio and mx5 thermal paste. my ambient is around 20c. It helps that my aio is an intake. They run a bit hotter when you use the aio as exhaust. I bet that’s the main difference. This cooler is the best I’ve ever had. It’s better than a Corsair 360mm I had before, but that was exhaust also. Im sold on front aio Intake. Tubes down if you can. Lots of exhaust fans on top and rear also. Maybe I also got a golden sample silicone, not sure. My temps are legit low though. I think I’ve maybe seen 79c-81c once or twice while benchmarking. I’ll do a benchmark video here soon and show some scores. Subscribe so you don’t miss them. Cheers man
Did anyone succeed with this method? Don’t get me wrong, the video is awesome. But after getting Cinebench to pass the undervolting with flying colors, your PC will just crash when idle browsing the web or watching youtube. I pulled back on the failing cores just to end up with 3 back to 5 or even 0. Yet, constant crashes. I decided to take the slow path. Start all over slowly and let the PC prove stable before pushing further. Cinebench and corecycler (prime95) are simply not enough as the biggest voltage/frequency adjustemnt happens when running idle!!!
Maybe your particular CPU and motherboard needs to use a higher LLC setting. I'd set it to second highest as a start and go to max if that doesn't help.
No, I didn't succeed based on the method in the video. The All Core Curve Optimiser is just too rough. I prefer to go with the CoreCycler route and now have been using my PC for 2 days now, seems quite stable. The CoreCycler is a much better way IMO, as it identifies per core stability. From there, once you have come to the optimised curve, I moved everything up by 1 tick to ensure no cross instability caused by borderline cases.
Doing this does work, but be prepared to spend many, many hours working on it if you don't want to leave any performance on the table. on my 5900x I was able to get it to run 4.65 all core in r20, and score around 9050-9100. Single core routinely goes up to 5.05-5.1ghz under normal usage, and gaming i usually see 4.8-4.9ghz.
11:39 Actually it is very easy to pinpoint the core that crashed. Just open your event viewer and create a custom view for the WHEA-logger. As soon as you crash, go to the viewer and check the APIC number that corresponds and divide it by 2 and there you have your crashed core. If you open CPUZ in the tools section, export an html and search for the APIC section and see which number corresponds to which core on your system. It might sound complex but it really is fast and accurate
You don't actually have to use CPUID to figure it out.
If you have a 4 core processor with hyperthreading, it will be like this:
Core 1
Thread 1 = APCID 0
Thread 2 = APCID 1
Core 2
Thread 1 = APCID 2
Thread 2 = APCID 3
Core 3
Thread 1 = APCID 4
Thread 2 = APCID 5
Core 4
Thread 1 = APCID 6
Thread 2 = APCID 7
For example, if you get an error with APCID 26, then it is the 12th core in the processor causing it (particularly the first thread on the core).
If you don't have hyperthreading, then the APCID will simply be identical to the core number. With hyperthreading, each core has 2 APCIDs. The APCID always starts at 0, so your core 1 will have APCID 0 and APCID 1 with hyperthreading, or just APCID 0 without hyperthreading. Perhaps I'm over explaining, but you can see the pattern in my example above.
@@jonny6702 that was cool, thanks for the tip!
@@jonny6702 If I have a 4 core processor with hyperthreading - how can I get an error on the 12th core??
It was simple at beggining because the error in Event viewer was telling me the APIC ID... But after that my pc was crashing randomly wihout telling pe the APIC nr, just that the pc shutdown.... I tried to lower them by turn and it didnt work... I had to lower all of them by 5 and now is ok..
Stability testing can be done with OCCT which allows to cycle single thread variable loads across all cores and tells you which core was unstable. Very straight forward even with 16 core CPUs
Came back for a second comment. I watched so many guides on PBO and CO. While I kept tweaking and guessing it was when I watched THIS video that it clicked in my brain. Thank you for the best explanation of all time for this tech because this should be the official guide. I stopped at -15 on my best cores and -20 on the rest as this seems extremely stable and my benchmarks are the best theyve ever been. Keep up the amazing work and thanks so much as you've turned my gaming rig into a tuned machine with spreading your knowledge! 140PPT 90TCD 130EDC was my sweet spot for keeping this machine in check. :D 5800x + 6600 XT build.
Awesome! So happy to hear when my guide helped haha!
While I was working on overclocking today I was watching the speed graph in Ryzen master and I was able to pick out the core causing the crashes pretty easily. Almost all of my cores on my 5950 were following nearly the same path on the graph when they ramped up and down except cores 1 and 3. I was able to pick out those 2 cores as the ones with instability and increased (decreased) my negative offset on the other cores 5 at a time until some of those cores also started to wander a bit from the others and started showing instability. I was able to pick through my cores and maximize the negative offset of each core in a pretty timely manner with minimal crashes this way. I recommend watching that speed graph for anyone that wants to save themself a bunch of time. It helped me immensely.
Best video on this so far, and i watched tons on this topic. Well done!
Thank you!
Thank you so much for taking the time to make this. Seriously. I was tearing my hair out trying to work out why my 5950x kept crashing with PBO on. Turns out the default PPT, TDC and EDC on my Aorus X570i pro wifi were set way too high. You rock, dude!
From my past year experience that seems to be more related to your temperature than the actual settings of these limits. If your cooling solution is not capable of handling a 250 watt cpu, you shouldn't play with overclocking. Also, undervolting with the curve optimizer can substantially decrease your temperatures so you should be able to push these limits higher.
of all the vids pertaining about this new function. You have really made it easy to understand properly. i have a 5800x and i have been trying to get it to run consistently. I tried CTR from 1smuz. its good but it gives me black screen lock ups whenever i run 3dmark. thanks for taking the time out to explain it =) my max on single core boost was 4.8 without crashing. now im at 4.95-5.04..
Agreed, I like the idea of testing with Ryzen Master instead of doing the tweaking in your BIOS over and over 😬
@@ryanlaseter7626 his new project hydra is good tho. really great on temps and the hybrid oc is great. the adjustable CO's across the frequencies has helped me stably run a 4133mhz ram kit on the 5800x. (at 1:1:1) prior to that i had to drop the kit down to 3800 or i'd lock up running r23 and even r20.. on the balls out profile i manage 17,900ish in r20. previous id hit like 16200 then lock up on the last section. oh and roughly 5.0-5.1ghz boost clocks @ 1.32mv with soc at 1.135.. barely goes over 80c on an EKWB AIO super the 360rad with 6 fans for push pull. and like 85c with just the fans on pull.
You are the best! "Talking in a straight line" seems to be something special.
Precise, to the point, clear messaging. This deserves 10x more views.
I’ve been looking for this concise of a video for in depth explanation of amd overclocking and you delivered very nicely, thank you. Deserves way more attention.
The most common crash after overvolting is changes between high load and all core idle. This is because those change points cause high ripple currents which may trigger the crash.
THANK YOU, I'm new to Ryzen and it is a whole new ball game. Your video really helped me understand how it all works. I'm on old school overclocker so this really helped me :)
Great guide, just what I was looking for :)
Glad I could help!
@@OTechnology looking foward to the 4090 tier video
Much prefer testing with occt. Detects errors which don't crash the pc/app and tells you which core os responsible for the error so you know what to focus on. I also tested single core by pinning the task to specific cores in task manager rather than always let it run on the same prefered core resulting in a more stable overclock (it helped me to detect the crashes which wouldn't occur if it ran on the prefered single core) my results for 5600x -12, -12, -5, -2, -15, -5.
It would be so kind of you if u can share the Bios settings buddy
i do have the same CPU but not fully optimized
your help is appreciated.
Great video - explained properly and exact!!! Just tuned my 5900x and 3-4 cores are easily sticking to 5,04Ghz now, most of them up to 4,8-4,95 at single core tests. All core goes up to 4,6 - 4.67Ghz. Just made CCD1 -17 and CCD2 - 30 with EDC of 125, and PPT 250, TDC 170 and I almost never go over 70C with 360mm AIO :) Now it's almost perfect, but as I'm on linux and no windows, per cor settings won't be easy at all :D Still way better than before as I would go up to 4925 on the best cores and barely keeping 4,5GHz all core with over 80C temps and the same cooler :) If with the new settings I keep the EDC at 120, I barely go over 65C :D AMAZING video!!! Keep it up !!!
I just got a 5800x this video is very helpful thank you
This made it so clear to me. Thank you so much! Now I know where to start.
5:44 I have also noticed mentioned tendency myself: higher EDC gives higher temperature but clock speed is lowered, while lower EDC gives lower temperature but clock speed is higher. Assuming that the EDC limits are reached in both cases, it means that the computer draws more current and works slower, yet generates higher temperatures . This doesn't make sense. Unless we assume that at lower clock speeds and higher EDC, the processor is more heavily loaded but operates slower (or wastes power). I think that changing all these PBO settings (EDC, PPT, TDC) is merely an optimization for specific applications, rather than overclocking per se. A more loaded processor operates at a lower clock speed, while a less loaded processor operates at a higher clock speed. Which is better depends on the application.
Best video out there explaining this! Thank you!
Testing shows that EDC impacts performances differently at different values. To high impacts single core, to low impacts multi core. The sweat spot seems to be around 120-130 for optimal results on both on 2 ccd cpus.
I tried all of the overdrive settings on a 5800x with an Evo 212 v2 cooler im building and found that leaving the overdrive settings to auto and setting curve optimizer low as I could which was -30 (later set to -25 to ensure stability for my friend ) along with max cpu temp set to 80 c was the easiest option that had almost the same results as tweaking the PPT/EDC, i get 4600-4650mhz multi thread 4850 mhz single. If its not stable at the above settings change curve optimizer by 5 until you get a few stable runs of cine. If you so desire with the above settings you then should be able to run cinebench & then watch the PPT/EDC/TDC % usage values in ryzenmaster and then have a good idea of manually adjusting the values to match the desired max temps/mhz your looking for without setting a max temp. Example: the PPT showed 83% of 142w on cine with auto settings -30 co which is approx 118w @ the 80c limit i set so adjusting the PPT to 118w along with changing the other values the same way should result in the same temps without the lower temp limit . Before i started tweaking the settings my cpu Shot to 90 C (max) temps in a jiffy and although everything i have read says these temps are normal i wanted to reduce them to have a safer margin without much loss of performance.
Thanks for the amazing video, taught me a lot!!!
Hey
O!Technology, if you increase the 3 values of the PBO limit you will push more voltage in your CPU and you will have high temps. You should disable the PBO limit so the algorithm will find the lower voltage that the CPU can run the high frequencies. Also if you choose the lower power saving plan you will see a drop in temps without loosing performance. About 95% of the time your CPU will be at max 85% in usage so what you see in the stress test it will never happen (not even if you render a video). Aside from this the power saving is made to better adapt the consumption to the load and if the load needs all the cores at max, it will allow the voltage without loosing performance.
Very educational video bro, greatly appreciate it .
I didn't get the "setting PBO Limits" part. What is the meaning of "something you probably will never reach" value? If we have to set a value that something we will never reach, why don't we just type 9999? If that's the point.
What if I increase my negative offset to the maximum my bios will let me (30) for every core and it is still stable? Did I not utilize something to the most that I could have? I have an all core clock of 46.50 at 3.65v right now and it was pretty stable but I might have been able to increase it some.
Simple and good video for first timer oc'er like me 👍🏻. Other pbo2 guides i saw weren't this clear! Great job.
dude you just saved me so much headache, I had all my pbo settings maxed I dropped to 120a on edc got 4.675ghz all core on cb20 but my overall was only 4.55, upped to 139a which is my sweetspot for 4.610 all core 😅 my man 💪 👏
Damn didn’t think about the 5900x having two different sets. I was putting all core to -20 and still getting bad temps! I suppose it was boosting way too much for it. Gonna try and put the first 6 cores to -15 and last 6 cores -20 or -25 and see what happens.
damn i got the same issue lol
what gpu r u running
When using OCCT cpu test it tells you what core starta to do miscalculations if it's not stable that way you can fine tune curve optimizer super fast. I can do a negative offset of 30 on all cores but core 2 and 4 on my 5600x. Those I have at 15 and 5 respectively and that is rock stable. And even though it looks to boost higher than just PBO +200Mhz the performance is actually slightly worse in cinebench multicore than with just PBO +200Mhz. I havent set other limits than the motherboard though. Maybe thats why?
There is a way to find which cores crashed during a test. By using the WHEA log report in windows after a crash, you can see a message of a CPU ID like; "Processor APIC ID: 8 - A fatal hardware error has occurred". To see which core this belongs to you can open CPU-Z, click Tools and then "Save report as .txt". In this report the threads, Core IDs, CCXes and Nodes for your CPU is shown. For example, I have an AMD 5600X. An error message of "Processor APIC ID: 8" tells me that Thread 8 caused a fatal hardware error. Thread 8 is the first Thread on Core Nr 4 (Between cores 0, 1, 2, 3, 4 and 5 on my 6-core CPU).
When I was testing an all core OC on my 5900X, I found Cinebench, Blender OpenData, and an x264 encoding test to be most useful to test my CPU's stability. One particular scene in blender would push the CPU so hard that at anything higher than 4.4Ghz, it'd go past 90C with my cooler which was not so good 😂 but yeah, a plethora of benchmarking tools would be best to validate different workloads.
what mic you are using? :P
I had a stable overclocked 3950x at 4.3Ghz for over 9 months. It gave me a 5 to 15% performance boost over my stock settings 🤔. It also meant I had more heat and more voltage. TO ME the gains weren't enough for me to warrant keeping my 3950x overclocked. Today I just cut on my XMP profile and keep my 3090 overclocked with MSI AFTERBURNER. Maybe if I saw a 20% or higher performance boost I might still keep my 3950x overclocked. But I see no reason why I should since the stock value is closer to what I was able to get with an overclock MINUS the heat and voltage 🤔.
That's because no one should use a fixed clock overclock on Ryzen you're actually losing performance in most applications because the boost behavior is not working anymore.
Hey overclockers, ive never done this before. I have a 5900x and a 4080 super playing on 3440x1440 with some annoying bottlenecks on the factory CPU clocks.
Should i try this on my tomahawk B550 motherboard?
Cpu gets cooled with AIO Arctic Liquid Freezer II 240.
3:38 best tounge twister: "single CCD CPU"
this is really cool and i was an allcore overclocker till this day, will give this a try, ryzens are really difficult to understand and to squeeze performance without sacrificing single or multi core scores.. is it possible to overclock per core?
Yep give this a shot! You kinda have to let it do its thing more than on intel cpus.
Edit: Overclocking per core is possible but you will not be able to hit the same boost clocks as using curve optimizer still.
curve optimizer stability with negative offset need test only in idle
Nicely done, thanks
I was watching a video and noticed curve optimizer snd I don’t have that. We have the exact same setup. Why don’t I have it?
I disabled xfr and pbo in bios and my 5900x became stable , in games max 68c 4950mhz , cinebench r23 single:1638 multi:22339pts .
I think I found out why lowering the EDC amperage allows for higher clocks it seems after a point when it hits too much amperage the vcore will go down. So lowering the amperage allows for more vcore headroom.
Yep that seems to be what it is doing. Interesting mechanism to exploit.
Can you do a "which rtx 3070 ti to buy" video?
Second this!!
Hello bro.
I have r9 5900x cpu
asus rog stix 3070ti 8gb oc gpu
asus tuf gaming x570 plus wifi
COUGAR MX660 MESH RGB 4x120mm RGB MESH GEX 850W 80PLUS GOLD MidT ATX
Lian Li Galahad 240mm ARGB All-in-One CPU Liquid Cooler
How do you think I can overclock the processor, graphics card and ram in this system? Not for rendering, just to improve game performance. because the monitor I use has a refresh rate of 280 hz and is in 1080p resolution. I need to get better performance to feel 280hz in fps games like pubg, so I want to overclock. may I help you?
This could be easily applied to undervolting scenarios too.
Thanks for this video. I'm gonna put up my big boy pants now and undervolt PER CORE instead of all core, seems that is the intermediate-advanced way to apply better tuning values other than simply putting in all core due to the nature of CCD arrangement in Ryzen CPUs.
This video onwards, I will need to read up on per core optimization.
I turned off all limits in MSI bios after worrying warning. my 5950x ascended to 5.2ghz and then tripped something and system froze. i guess I am doing it wrong. is 1.45v on cpu safe or shhould I stay under 1.3?
Really good info!!
Great video ! So being close or around 1.5v in hwinfo is all normal?
Should I use LLC with pbo? When using LLC3 it's stable but on auto, my per ccx at 1.3 won't be stable either
normal, totally normal
www.reddit.com/r/Amd/comments/cbls9g/the_final_word_on_idle_voltages_for_3rd_gen_ryzen/
Almost all CPUs today have thermal throttling, which dials down the boost or core frequency when a top temperature limit is reached (to protect the CPU from damage). Is there motherboard functionality now where the performance/frequency/voltage of the CPU is regulated based on a thermal limit as the target ie overclock esp in low or low core loads if temperature allow it (beyond boost specifications).
When doing per core on a dual CCD, would it be smart to just leave all values at the first CCD at 0?
It's probably a little late for this, but wouldn't negative offset cause crashing when the CPU is idling as well?
No it does not apply an offset when idle
So I have a 5800x/3080 my frame rate is maxed out on my monitor at 165hz 1440p should I bother over clocking my cpu?
What about boost overdrive ???
Did you compare this vs the performacne of an all core overclock? In most scenarios an all core overclock gets better framerates and performance than curve optimizer. Curve optimizer makes it look like its hitting higher clocks but the performance doesnt scale with the clockspeeds.
"In most scenarios an all core overclock gets better framerates and performance than curve optimizer" Which scenarios would these be? I call BS. Show benchmarks!
Thank you for this Video.
My cores dont want to boost Higher than 4.65 GHz. Any advice?
I am Not running Into any Limit in Terms of EDC, PPT or TDC.
Trying to get to 5 GHz Single Core btw
but now is my issue of CO I have all -30 apart from best cores -25 is that hurting them or? it's been fine running all the single cores for an hour with no occt errors
i cant make my pbo limits higher in ryzen master, only lower. do i have to uncap the pbo limits from bios first? how do i do that can someone explain i dont wanna mess anything up rn
event viewer can help you locate the crashing core
what do you put on the Load Line Calibration when using curve opmitizer and pbo2?
when should i use Level 1 and when Level 5?
I put the second highest setting on my Gigabyte motherboard called high as that seems to work well now that I tested more. Don't know what is higher on your motherboard 1 or 5 but if its an Asrock motherboard chances are 1 is a higher setting.
How much Points you get in cinebench r20 ?
my 5900x did 9100 points
Some comments:
1) Not all Ryzen 5s and 7ss are 1-CCD. Some are 2-CCDs, which can be seen from e.g. ZenTimings debug report.
2) Clock Tuner for Ryzen can give comparable performance but much better volts/temps. But only on Windows.
3) OCCT or CoreCycler can be used to ease testing curve offsets per core.
4) There are faster ways of coming up with curve offsets, e.g. ruclips.net/video/4gDfMwOBnoE/видео.html
All true and good advices. Thanks for sharing. I know about all those but thought they were not really necessary especially clock tuner where you're sacrificing single core speeds.
@@OTechnology can some cores be enabled even though they might not meet grade quality in the ryzen 5600x, due to 1 or 2 ccd having 8 cores each, but 6 only are needed for 5600x.
I think there were some controversies (source from Reddit, thus anecdotal) regarding CTR 2.0 for Ryzen involving security flaws on AMD subreddit.
Is there any reason to not simply use Ryzen Master to tune the CPU?
Thanks for sharing the other additional tips. Going to do some due diligence after this.
@@OTechnology in CTR one can actually set single-core speeds higher than anywhere (incl. CO) manually, and even be stable
@@RomanOrekhov isn't CTR a fixed overclock?
Will you be doing any hardcore overclocking, zen 3 will benefit greatly from memory tuning in most case uses.
Haven't had time to do memory tuning on Zen 3 yet.
Thank you!
its not normal that your cpu hit 90 C I have 5900x with noctua nh-d15 with stock fans and after copy-paste your settings in ryzen master, my max temp is 71C.
Don't know, maybe my ambient temps were higher than yours and this 5900X is just particularly hot.
If I don't want manual PBO settings what should I use like Auto, Motherboard etc. Also what happens if you set these in the bios then change it in Ryzen Master does it override the bios?
Please, pleaseeee help...Im using 5800x on asus tuf gaming x570 plus...my temp's are more than 90* while gaming...Im not an expert on overvloacking, I tried some things in bios but not sure I did it right...can you tell me some values for my cpu for tdc, edc and ppt or I have to find my own values...I just want to lower my temperatures on cpu...I dont want them to go over 80*...can you help me please.??
5900X running -30 All Core and no crashing. Multi core clocks boost to 4.545GHz and single core to 4.95GHz. The BIOS won't allow anything less than -30, is there any way to squeeze more out of it?
I'm in the same scenario as you, ryzen 9 5900x with x570 asus dark hero. Latest bios and latest amd chipset driver. I have everything at -30 on all cores and ran OCCT for a few hours at stock power limits ppt=142,tdc=95,edc=140. Not a single crash with maximum cpu boost override at +100 to try to hit atleast 5ghz. Was peaking between a little over 4.9's ghz to 4.6 ghz on single core and multi cores were running at 4.5 ghz stable. Cinebench 23 scored single core at 1620, and multi core was at 21205. Which I read was as good as stock. My thermals did get a decent drop in cooler temperature by 5c. But no matter how low I go on my negative offset was for each core my cb23 scores didn't go up, until I bumped ppt to 165, tdc to 115, and edc to 160. My multi score went to 22450. But my single score always stayed the same. Thinking about removing some negative offset to my two best cores on my ccd0 and ccd1 probably at -15 or -20 to see if it'll at least peak at 5 ghz. But this stuff gives me a head ache for minimal gains.
The best way to overclock Ryzen 5000 is to live in a cold place
Too difficult to pinpoint the specific core that crashed??? Whea error anyone?
I've been using Cinebench, AIDA, Geekbench, 3dmark and Prime95. My cpu is stable on everything i do, except on prime95 torture test. Would it be safe to assume that the torture test in prime95 is meant to test base clock speeds for fabric failure? I dont see any other software being able to push the CPU that high.
I personally think passing prime95 torture test with small ffts is sort of pointless if it passes all the other usual tests. You will never run into a scenario where you will be under anywhere near that level of stress. Unless you do machine learning or something crazy with your computer. In that case maybe just turn on pbo and leave the rest stock.
5600x here. Ended up with Negative 0 on best core (yes, ZERO) and around 20-25 on the rest.
Does it make any sense that the difference is that much? oO
Also, the crushed never happend at high load for me :) only drops to idle
That can happen yes. Its really a silicon lottery and I think if only one core is not doing well its totally fine. The multi core clocks will still be substantially improved.
@@OTechnology Thanks for reply. Well, my CPU overall seems very good. So the lottery is nice but odd :). I was testing it with usmus1 tool and there was info it's gold sample. Great content btw.
I was running an all core set to 4.75ghz at 1.2875v and was performing well but i knew that having a good pbo profile is much better but i just stopped caring as much when i used easy tune from gigabyte and had a perfect oc and the virus program it is corrupted and i couldn't get it to work again so i just used ryzen master, atm this is lookng good but haven't tried any games or other loads on it yet other than r23, however on a 5800x i am at 250,170,100 for power and at negative 30 for all cores but still stable is this right?
why is max boost overclock override max 200mhz ?
It used to be 500 MHz but a while ago manufacturers changed it via BIOS update to limit it to 200 MHz. Probably too many CPU's died and caused too much false warranty cases, which caused too much cost. But that's just a guess why they changed it. 200 MHz is the new maximum now.
Hey! I’m new to all this! I followed the first part of the video, modifying my PPT, TDC, and EDC. Here is the weird part. On my first test on cinebench (my baseline test) I already had the native “game mode” on ryzen master turned on (PPT: 720, TDC: 420, EDC: 480, very weird figures I know). I then modified all three values, and found that the sweetspot for my 5800X is an EDC of 80. That’s the highest clock speeds I was getting on HWmonitor. After running the test on cinebench I actually got a lower score, although as I said I was getting higher clock speeds. Any explanations. This stuff is proving too complicating for a noob like myself.
Run cinebench more than once. If there is any activity on the CPU in the background (like windows update, defender etc) it will impact the score.
look at the effective clocks (shown in hwinfo), maybe your having "core stretching"
I have 5600x with Arctic Freezer 34 Esports DUO.
I check the stats with HWinfo64.
Testing with Cinebench R23.
When my "PBO limit" is on "Motherboard" the temps go to 82, when "Disabled" temps are 66. With same other settings. Can you tell me why is that?
I saw that when I set Curve Optimizer to all cores "-30" and PBO boost "+200", the Single core clock is higher "4,841" but the Multi-core goes down around "4,260 - 4,313"... Cinebench Multi-core test ~11000 or little lower.
Like I said, if I change "PBO limit" to "Motherboard" the temps go 82 to get higher Multi-core clock, but it is not worth it.
If I set PBO to just "Enable", without any other setting, the Single core is ~4,65 but the Multi-core goes up to 4,55, at the cost of 82 degrees, which is not good again...
I want to achieve both high Single core and Multi-core with around 66-70 temps!
Any suggestions?
P.S. I didn't understand what is the manual method, calculation to determine the EDC, TDC, PPT values, this is why I tested with "PBO limit" either on "Motherboard" or "Disabled".
6:56 "This is completely safe" after using PBO features that explicitly say "Warranty void" 🤔
Other than that, great video!
I didn't say anything wrong though? Thanks!
@@OTechnology Even enabling PBO may damage your CPU and the official party line of AMD is that CPU is no longer covered by warranty if you enable PBO. Of course, they would have hard time enforcing that limitation here in EU: because they obviously advertise PBO as a *feature* for both the motherboard and CPU, it must be supported feature and as such cannot be off-the-limits for warranty.
For countries with less-strict consumer protection laws, I would consider twice before activating PBO.
I also have 5900x cpu and I was wondering what the are the best settings for gaming?
This is. PBO with curve optimizer
Would’ve have been nice with some results. You seem to have done a bunch of tests already..
As first, this is a very good video! Thank you
But I have a question, does it have the same effect if I set all cores negative 5 instead of per core and then set all to negative?
Cause a friend of mine can set negative 30 at all cores and still doesn't get a crash.
other reviewers state to decrease more volts in the weaker cores and let the stronger cores with more volts
Nice and inspiring vid! One thing I found really annoying on my ASUS board is that after updating the BIOS which essentially updates the AGESA version. The off-set unit was re-shuffled. This resulted in a crush that damaged my windows once after loading the exact same BIOS profile after an upgrade. Don't know if that happens on other brands.
Always restore bios to default before updating
@@CidoraX The BIOS always resets itself to default after updating. I read from other sources that AMD changed the settings in certain AGESA version to have a more conservative strategy. But now it seems they changed it back a bit that allows me to go for a higher negative off-set value (My current BIOS version on my ASUS mobo is 2423)
Yeah reset bios by removing cmos battery. Had to do this when swapping 2700x to 5600x even with latest bios already updated
@@kleinwood9297 no they don't always automatically adjust mine obviously didn't
I got -30 on all cores in my 5800x with 4.7GHz and it seems stable. Is this normal?
Either you are one in a million ultra lucky or that is not actually stable.
@@OTechnology Yeah I had to move it down to -20 lol
@@sam_9228 haha that's normal.
Is it possible to decrease CPU temperature on ryzen 5800U?
yes, but use aatu/ryzen adj software.
just find ryzen controller then join their discord channel "Isle of zen"
What power plan are you using?
Just the default windows balanced. Ryzen 5000 doesn't need any special power plan.
@@OTechnology can you tell me what the min CPU state is set at in your balanced plan? Mine is set to 0% but others on Reddit said it should be at 5%
Def not any more work to dial in CO on a 12/16 core Ryzen 9 unless you are actually spending hours increasing CO units on a per core basis to pinpoint which core crashes at a given offset. The only way to efficiently set your CO is to create an Event Viewer profile in windows to log when you get a crash. Even prime95, Cin. R23 can be stable and certain games/apps will crash after you get it dialed in for stress testing. The only way to accurately know which core is crashing is to log it. After a crash you go check your event log to see which core crashed and reduce the neg offset on that core. Also LLC can play a big role in CO stability. Increasing LLC is a must for stability with high offsets in my experience. I've only had 1 Ryzen 5000 not be able to hit over 5ghz light core workloads but it was also the only China batched chip 5800X I have ever tested with.
Do you have discord my game stutters trying to figure it out nothing is over heating I believe it’ll running the same motherboard you have in this video I’m
Rather new to Ryzen I have the Ryzen 7 5800x
Ive tried all combinations of changing my curve values. No effect on my cinebench scores. Been messing with this all day. Same proc. Same bios.
No expert but have you tried adding the +200mhz in over drive .
Non on of this will work if your temps aren't low
@@wrusst yes. Full custom ek loop with 3 x 360mm rads.
@@teemohnay you tried upping the scalar to 5x or more which is supposed to push more volts and fake the chip is better than it is ?
@@wrusst tried it at all the Xs. All the way to 10x. No affect.
@@teemohnay my guess is one of your weak cores are undervolted I'd look at the skatterbencher guide as it's got a few more tricks
Your scores are way higher than mine with similar settings on my 5900x + aorus master combo. About 22,000 multi, 1600 single. 4.55GHZ all core and up to 5GHZ single. Any ideas?
Austin, you sound like you know what your doing. I could really use help on getting my system to run at its best. Honestly feel like it should be much better than it is. I’ll $$ someone to help me!! I’m running ekgaming vanquish 295 I believe is the name. It’s got fully liquid cooled even the crazy RTX 3090, ryzen 5900x, asus Prime pro motherboard, 32GB gskill 3200 ram, 1Tb Samsung 980 pro, another Samsung 1Tb 970 evo pro and 2 other drives I don’t use lol. I have been trying Ryzen master abs so on and I obviously don’t know a lot. Any interest in helping me? Would appreciate it thanks.
@@kaubut0145 I might be able to help you. I'll admit I'm no pro at overclocks, but I could at least try to help you out. Just give me a way to contact you and we can try to figure something out together.
it could be background tasks and general bloat. with my 5900x i can see a 3-400 point difference (in r20, so not the same as r23) depending on what other apps are running in the background. When I do my tests I make sure everything else is disabled first.
my 5900x can go negative 30 on all cores, pbo motherboard, +200mhz cbco, scalar 10x and still i am 5% in 3dmark cpu. dunno what i can do more.(edit:-from 4.95ghz i went 5.150ghz but still low scores in 3d mark 5-7%) 75C with nzxt x63
I doubt that's actually stable. Try running corecycler to test stability. github.com/sp00n/corecycler
@@OTechnology i tried cb23(22800score)/3dmark cpu(11400), didn't have any issues (edit: also playing any games without issues). i ll try that software but it kinda beats me.
@@alexmckoy8898 yea those don't show instability because the cpu boosts differently per core so you need software that can do that which is corecycler.
@@OTechnology i still use it and seems it's stable. never got a bsod/rr. idle, mid browsing, gaming, benchmarks.. nothing.
@@alexmckoy8898 Probably got a golden chip then!
You have to look at the effective clock. Not perf.
Hi can you help? Everytime I try to use PBO (default values) or Auto OC (default values) I keep getting 96+ degrees and throttling down the Mhz to 560 or lower. It never crashes but it keeps overheating. So I can run only on default and that gives me around 60 - 65C. I'm using Prosiphon Elite Giant which should be one of the best coolers for AMD cpus according to benchmarks
Its only good for threadripper. On regular Ryzen CPUs a Noctua NH-12S beats it by a good amount and a NH-D15 is even better. Prosiphon only works for threadripper with a big IHS to spread the heat and the high TDP that can actually boil off the liquid. This has been tested by manu reviewers and people on reddit r/overclocking.
@@OTechnology Yeah I'm using 3990x threadripper
@@SilverIlly3 Oh then maybe you didn't put enough thermal paste? The X with 4 dots is usually the preferred method for threadripper. Only enabling PBO shouldn't be a problem for the prosiphon. Also better to have too much thermal paste than too little which will hurt performance while having too much you'll just make a mess is the worst that can happen.
@@OTechnology I had a PC professional apply the paste and he followed the Icegiant video tutorial and made 8 dots around it with enough thermal paste so that shouldn't be a problem.. any other idea of what could be an issue?
@@SilverIlly3 fans too slow or bad case airflow would be my guess. Just seems like a cooling issue.
Finally someone with a correct graph for curve optimizer
Oh yeeee looking for that 1 core that crashes daily driving, yikes 😂 good idea about downing all the cores in the CCD by 1 💪
since my case scenario probably not the same than you been lower core count which often run in low usage scenario wouldn't going whit agressive negative offset on vcore voltage whit loadline calibration that overshoot a little for those rare time power is requiered would not end be best ways to achieve a théorical perfect balance of efficient as well as been powerful when requiered.
-thx for sharing whit us?
i would encourage you to try to avoid abreviation terme that can be unlcear for some user by giving more detail to there nature whit perharps some visual addtion text like you did in some occasion avoid to lose the viewer whit terme there not familiar since not everyonne first language is english tend to get lost with those i would lead to believe could yield nice result in the long terme
Micnya apa bang
Itu mic Blue Yeti
I was able to set my 5900x to negative 30 all core. Is this normal? Is it good / bad?
Im in the same boat. Motherboard limits and boost clock at 50. I havent seen any core crashes yet
Good stuff
I would love to see some 5900x cinabench runs at 4.9 all core. Or close to it. My 5900x hangs out around 4.8-4.9 with only pbo2 and a 240mm aio and some mx5 and it NEVER goes about 68-69c I’m pretty impressed with its performance. I might go for 5.0 I believe the world record atm is 5.9 for a 5900x and 6.0 for a 5950x but that’s with liquid nitrogen. I’ll stick with water personally haha. Cheers. Next time plz show more over clicking and benchmarks, less just talking. Other than that, great video..👌🏼✌🏼
You never go above 68-69c while running cinebench with pbo2 on a 5900x? Thats crazy. What is your ambient temp, and what cooler are you running?
My 5900x with an evga clc240 aio hits well into the 86-88c range in r20 with pbo2 and dialed in CO. It scores well though, around 9050-9100, just gets hot. Gaming it stays around 65-70 though. Ambient is around 18-19c (~65f)
@@Alex-wb2yw I’m using an arctic liquid freezer II 240aio and mx5 thermal paste. my ambient is around 20c. It helps that my aio is an intake. They run a bit hotter when you use the aio as exhaust. I bet that’s the main difference. This cooler is the best I’ve ever had. It’s better than a Corsair 360mm I had before, but that was exhaust also. Im sold on front aio Intake. Tubes down if you can. Lots of exhaust fans on top and rear also. Maybe I also got a golden sample silicone, not sure. My temps are legit low though. I think I’ve maybe seen 79c-81c once or twice while benchmarking. I’ll do a benchmark video here soon and show some scores. Subscribe so you don’t miss them. Cheers man
Did anyone succeed with this method? Don’t get me wrong, the video is awesome. But after getting Cinebench to pass the undervolting with flying colors, your PC will just crash when idle browsing the web or watching youtube. I pulled back on the failing cores just to end up with 3 back to 5 or even 0. Yet, constant crashes. I decided to take the slow path. Start all over slowly and let the PC prove stable before pushing further. Cinebench and corecycler (prime95) are simply not enough as the biggest voltage/frequency adjustemnt happens when running idle!!!
Maybe your particular CPU and motherboard needs to use a higher LLC setting. I'd set it to second highest as a start and go to max if that doesn't help.
No, I didn't succeed based on the method in the video. The All Core Curve Optimiser is just too rough. I prefer to go with the CoreCycler route and now have been using my PC for 2 days now, seems quite stable. The CoreCycler is a much better way IMO, as it identifies per core stability. From there, once you have come to the optimised curve, I moved everything up by 1 tick to ensure no cross instability caused by borderline cases.
Doing this does work, but be prepared to spend many, many hours working on it if you don't want to leave any performance on the table.
on my 5900x I was able to get it to run 4.65 all core in r20, and score around 9050-9100. Single core routinely goes up to 5.05-5.1ghz under normal usage, and gaming i usually see 4.8-4.9ghz.