@@puffyips LOL... What guy? hope you don't mean buildzoid... Also I love his content, his cool af, but his CO videos are like the bare minimum, doing a global offset is just lazy and not the ideal thing to do, this hurts AT LEAST two cores, the highest performers, and this doesn't achieve full stability, doing per core CO does.
I've been useing curve optimizer on my 5800X3D for a while now (year 1?) on a Asus Dark Hero. Also tuned the max power draw and lowered that a lot. Since you can't really overclock the 5800X3D, lowering the offset to -30 on all cores, allows the cores to stay at 4.45GHz under almost all workloads. I think Cinebench stabilizes at 4.2 (was like 3.9 with stock settings) during all core, but all other situations, like gaming for me, it just runs at full speed and sips 65-75W. And when you have a custom loop with a 4090, it's nice to have a CPU that doesn't require you to get a second loop for the CPU. Also, everything is more quiet. I also run a power limit on the 4090 for the most part, to get a almost dead silent gaming rig.
This helps only if you use low TDP / Eco on thermally constrained system i.e. miniITX. Since in that case lowering the voltage removes some botleneck points for the CPU to give performance. I have the 5800X3D at 65W working in ~5% performance range from the default 105W in all core, and exactly the same in single core. You would expect 40% lower TDP to have more of a performance hit. And it does without the PBO2 Tuner.
The voltage/power required for the last 200-300 MHz is indeed insane, same with 6000/7000 GPU, they really do try to squeeze whatever possible with little regard to power draw
curve optimizer is good to shave off 10-15C and stop air cooler from screaming like a banshee. And it tecnically can give some performance if you're thermally limited
Any drop in temperature should give an ever so slightly higher performance on Zen, unless you somehow run at full boost in all core stress tests. With liquid cooling CO should be helpful as well, even if just a tiny bit.
I used -30 all-core on my 5800X3D - CB23 results: before: 14,900 @ 67 degrees after: 15,200 @ 60 degrees. Far more bothered about the temp losses than the MC gains.
i run -20 on my 5800x3d and this does make a difference. with it set to -20 the cpu will hold a boost of 4450 and 4250 at stock. also, i dont know if its my mb, but i can set this in the bios (asus x570 cf7).
Been running -30 on all cores of my 5800x3d for almost a year. Temps during gaming usually sit around 45-50c with my 360aio. I’m glad I did it too because when I’m playing BG3 in split screen that sucker can get up to 75c in some busy scenes.
Playing with PPT to find where it performs the same as stock is basically just monitoring how much power it uses for the workload by guessing and typing in the PPT number. Could just open HWiNFO64 without touching PPT and compare CO 0 vs -30.
The problem is that the 5700X3D has turds for cores that cannot go above 4.15GHz at stock BCLK, so CO just isn't going to do shit. CO on just about any other CPU does work to an extent. On a 7600/7600X/7700X you can get some good single and multi core performance increase. Now, is that performance noticeable in gaming? Mostly no, but sometimes yes. My 7800X3D saw a decent drop in temps and holds higher average frequencies and all I had to do was type -20. Just about every chip out there will do -20 and that has a very big decrease in temps and voltage. Also AM5 does support lower than -30 for CO values, but just about no CPU will be stable.
CO together with PBO took my 5950X from 24K in CBR23 to 29-30K in CBR23 whilst keeping temps under 80 on air. It’s a balancing act - if all the cores are running cooler they can all boost higher but they need the extra power headroom PBO allows.
Wait what? I do have access to curve optimizer in the bios on my 5800X3D o_O Are you looking in the right place? There are usually 2 places where you can config CO: AMD Overclocking panel and your motherboard standard panel, in the standard panel you can tune CO, but it is locked in the AMD overclocking.
I used PBO2 Tuner for my 5800X3D for quite a while, then MSI released a bios update for my B450 Gaming Pro Carbon AC that updated to AGESA 1.2.0.A which didn't just add KomboStrike but also gave me access to curve optimizer and PBO settings directly in the bios. There is definitely uplift. I'm running -30 across all cores, (edit)122,84,124 for PBO settings, and gave myself a 0.05 voltage offset with an LLC of 7 to help with temps. I post a 15300 in R23. I don't have patience to run single core though.
@@ActuallyHardcoreOverclocking most people don't keep spreadsheets of their benchmark results. I did this over a year ago. Why would I keep settings that didn't improve performance? However I just went back through and ran completely stock CPU settings with my current tightened DDR timings. Best of 3 runs stock on R23 was 14150. Best of 3 runs with PBO and curve optimizer was 15250. All the tests were off a fresh reboot with my default programs loaded.
I can get a 15600 with a 102 bclk, if i lower my DDR clocks to 3733, but I get random WHEA crashes, so I choose not to run those settings. With PBO and curve optimizer my clocks peak at 4450. With a 102 bclk I'll reach 4550. I've seen guys with external clock gens rocking 107 bclk with 4750 clocks, think their R23 scores are in the 16000's.
Curve optimizer helps with CPUs that have a higher boost table or higher boost limit, as well as a capped power limit where PBO cannot bypass. For example the Ryzen 5800x on my X570-E Strix motherboard will basically top out at about 160 watts (regardless of cooling), even if you set PBO to a higher setting, and thus PBO leads to a higher all-core clock speed, often you can get a good 50MHz+ additional clock speed compared to just the the PBO settings alone.
@@TimberWulfIsHere With you increase the limits with PBO, the all core power consumption, especially during workloads such as encoding with handbrake, will lead to power consumption in the 160 watt range on the 5800X (non 3D). The main issue is that the CPU needs even more power and while the motherboard can deliver far more, it seems that AMD has some other limitation imposed as the CPU level that prevents the boost function from pulling more than that. This is why curve optimizer ends up helping to further increase clock speeds.
Where do you find oit about these tools? Like for example, did everyone always know about ycruncher? Or is there a big forum post everyone goes to that i dont know about lol
5800x3d ppl usually just slap -20 on it and call it a day, way lower temps and less spikey fan noise. AND actually hte 5800x3d will happily use the 142 power limit or close to it but I think it just pumps the voltage for no reason. And seems all the limits set on the x3d chips by default are the maximum, so u can only lower them on PBO tuner
Maybe you are using old bios, I have access to curve optimizer on my X570 Elite for my 5800x3d for about a year...? I was using pbo tuner before though....
either you have a weird motherboard or this is 5700X3D specific but the 5800X3D has curve optimizer in the bios since many months. (i have three and a handful of B550 Boards)
Its MOBO specific. My Asus B450 doesnt have it in bios i have to use pbo tuner on my 5800x3d. I suspect only higher end boards B450 boards and B550s have it
Afaik PBO2 shifts the voltage-to-frequency curve (more from at the bottom, less at the higher to retain as much stability as possible), that shouldn't make the the CPU boost higher in any case, unless it was heavily thermally limited already and the reduction in voltage contributed to overcome that limitation to some degree, regardless, it will only allow it to sustain the same clocks for longer. It's just gonna take a breather basically. There's a setting for it to attempt to clock higher than the rater max boost clock by a certain amount (different motherboards call give it a slightly different name, but it's in the PBO section) but in my experience with my previous CPU, 5800X (which happened to be quite the sample) i had to use a tighter LLC to stabilize it though but it works for more than a year without ANY issues and since it's water cooled it can sustain 4.7GHz at all cores with temps being around the lower 70s iirc and without even caring it will boost up to 6 cores to 4.85GHz for a while if needed. 6/8 cores have set to be at -30, one is 28 and the other one is -26 because that's the 100% stable. I have of course tweaked the PTT, TDC and EDC for some headroom and some economy in between them but it can hold on without them, the tweaking is there more for precautionary reasons partly.
Asus has added their version of curve optimizer for the X3D chips on the main AI tweaker page instead of the AMD overclocking tab. I tried it with my 5800X3D and it works! A -30 all core curve seems to drop 50-100mv depending on the workload, and random temperature spikes are much lower during normal desktop use. Effective clocks haven't dropped and my cinebench/OCCT score is the exact same. Luckily my chip is good enough to handle the full -30 curve, but it sounds like most 5800X3D's can. Curious if the 5700X3D's lower bin will change the result.
Miss application of the tech . The amount of people running spires or stealths hitting max temps then employing curve optmiser has been a life saver for there boost clocks . You also mention you live in the uk so you are aware of extreme energy costs so its not like every parent wouldn't appreciate a yearly saving here with no performance loss or in some cases a gain? Also you state these OC tools do nothing on a CPU known to be locked down from OC tools especially after burning up those boards on the 7000 3d. The 3D on smaller coolers is surprisingly rough to cool as the vcache acts like a thermal layer . Theres plenty of reasoning why curve optimiser helps allot of people but not guys with large AIOs, then they wonder why there not getting benefits from additional thermal headroom. What next saying a diect to die kit does nothing because its running locked clocks on an already effective AIO ? This would be a weird application of the product also.
I had a 5800X3D on a asus B550-E and that got a bios update around the launch of am5 that let me set co in the bios.. a combination of -30 co and a power limit of 108w let me increase both single thread and multithread performance by 5%, with 6c lower max load temps
On air my 5800X3D will thermal throttle in 3s in Cinebench, -30 in the CO makes it takes a while more to happen and the scores goes from 12500ish to 14000ish.
Considering that those CPUs consume 20W at idle because of SOC and interconnects, I'd say 55W vs 66W is pretty good if you subtract the 20W it has to consume anyway to just work.
I think curve optimizer isn't really that much about performance per se, it's more about having lower temps which means lower noise. Many folks are very strict with noise so in that sense it's a good thing, especially for people like me who just slapped my good old wraith prism onto this thing, works like a charm i must say, no issue at all. Doesn't seem like it's any hotter than my old 3700x. I did notice more stable frequency while gaming when i adjusted the voltage, it stays pegged at 4.05 all cores at all times now, before i adjusted the voltage it fluctuated more between 3.95 - 4.05.
Hi guys! My understanding up to this point is, that boost clock is not related to core voltage but temperature. May it clock higher with better cooling? These boost algorithms are a miracle 😅
Curve optimizer will only provide a performance increase if the CPU isn't reaching the max boost clocks due to some other limitation. It helps if you're hitting a power, cooling, or FIT limit. It's a shame AMD doesn't allow increasing the max boost clock on X3D CPUs. Unlike running a static multiplier and voltage, increasing the max boost clock should be completely safe.
Yeah, when you have outrageous cooling already, PBO tuner probably just doesn't make sense when you need that extra voltage for higher clock speed stability. BUT!, because of how AMD put these CPU's together, at least with the 5800X3D as far as I am aware, due to the 90c throttle limit, it becomes harder to keep those peak speeds in hotter and hotter environments. So yes, especially in the case of lesser cooling solutions, the PBO tuner is really nice for dropping those temps to some extent per core, which is overall best for the CPU's construction with the extra cache sitting on top. I have a Dark Rock Pro 4 with two 3000rpm noctua industrials strapped into and onto it. Obviously I have no major issue keeping it cool at stock configurations. BUT! Due to those fans attached, I can cool the unit rather well at lower fan speed; but not quite well enough for the larger workloads. It works, but it will throttle. So essentially what I get is a quieter product that results in mostly the same performance at a slightly better power efficiency than the stock configuration. AND. If I want, I can now use the extra voltage headroom to up the BCLK by about I figure 5-10% eating into that thermal headroom. Main goal is to make it so the PC isn't heard by the microphone without having to do any major filter adjustments in OBS, etc; but still have the same or better performance if possible. Perhaps I should finally just move to liquid cooling, but... I kind of like the reliability of a non-liquid setup. And ultimately, I kind of need to upgrade my paste. I'm currently leaving about 5c of thermal headroom on the table by using some old Noctual H1 paste. I should probably be using thermal grizzly paste, or their solid pads for CPU and GPU respectively. Way better thermal conductivity comparatively on paper, and from the reviews I've seen so far; definitely better than the paste I am using right now anyways. (but I might as well use it all first.)
Very strange, Curve Optimizer definitely shows up when using CPU-Z for testing (I know its not great, but it loads and runs quickly so I have some idea of if I'm going in the right direction) I was applying my CO settings in BIOS, so having to re load cinebench every time would be annoying, I'm not trying that hard. My 5800X3D went from 602 single thread to 620, and from 6100 multi threaded to 6400.
Please educate me... My setup: ASRock X570 Taichi, 5800X3D. Curve optimizer is available in BIOS...so does this mean it just doesn't apply any settings? I'm confused
Ať least on my B550 Aorus board if I touch the BIOS setting it kills the boost all together (stuck on base frequency). Doing it via this tool keeps the boost on 4450Mhz
@@jakubknotek4891 BIOS PBO-> Curve optimizer set to -30, clocks at 4,35GHz on all cores. Changing PPT, TDC & EDC does nothing to performance, so i left them at default. Temps max 77 with small Arctic single fan tower. CinebenchR23=15,021 pts I'll try PBO2 Tuner next...
i have this chip my self and setting -30 in curve does makes mine to clock to 4050 mhz and reduced throttling a bit i would say its good to do that if ur in extremely hot country and dont have liquid cooling otherwise like he said it mostly worthless to do this unless in special circumtances like mine u can benfit a good amount
I've also noticed overclocking doesn't really do much anymore. I set my 5600X to 4.85 all core, and the difference wasn't noticeable at all. Same with my 3060. +150 on core gives me an ~8% uplift. Going from 50 fps to 54 fps isn't anything to write home about.
On my 7800x3d and X670e Hero curve optimizer works the way effective clocks get closer to the core clock. Since I’m running Bclk (synchronized mode) 101.15 for ram fine tuning at 6270MHz (m-die) curve optimizer per core varies-24 to -30 (not fully stable in Cyberpunk though it runs all the stress tests without an issue). Thus I get 5.1GHz all core (core clock monitoring) / 4.98-5.0GHz in R23 and around 4950MHz effective clock in R23. The max score was 19077. On the other hand if I switch to asynchronous mode and set eCLk to 105MHz (5.3GHz all core clock) I cannot apply negative curve optimizer values bc the system turns to pretty unstable. And with CO on Auto I get 5060-5080MHz in R23 core clock monitoring while effective clocks drop to 4890-4900MHz. And R23 runs slower at hypothetical 5.3 than 5.1 with negative curve optimizer. It’s about 18750 score. The question is how it scales in gaming workload when the load isn’t linear. Will try to run CP2077 2K resolution on my 4090 to check it out. Btw negative CO on 7800x3d in my case affects neither temperature nor voltage in Cinebench
Effective clock being lower than core clock means CPU doesn't get enough voltage supply during a given workload, it detects that and starts clock stretching. When you set CO to offset voltage curve, your CPU starts asking for less power and you get less load on the VRM, therefore you get less vdroop under load. You can defeat clock stretching by setting stronger LLC and/or by adding positive voltage offset until an effective clock starts to match the core clock under affected loads.
@@andriikruzhylo8996 that's not what I meant, I meant voltage offset, which is used by VRM, not CO. Keep CO negative, add a bit of voltage offset to the core voltage.
I've stopped using curve optimizer on my 7950X3D (thanks as this was causing all my stability issues and you were the one who told me to turn it off) but I did notice that having a low Curve Optimizer setting did help my score on longer tests as my CPU would sit around 79 to 80C instead of hitting 85C after a few mins which would give me a better score overall, for a quick run of Cinebench R23 I didn't see much of a improvement, maybe a few hundred points higher on a 36000 point score.
You probably should reconsider and try again. The instability from CO usually comes at idle or close to idle, when only the better/higher priority cores ask for voltage, because they ask for less voltage on the same CO value than their lower quality siblings (you can check per core VIDs under load in HWiNFO), and your CPU runs at the HIGHEST of the asked VIDs. Given that, if worst cores are loaded as well(all core loads), your better cores will run at the same voltage asked for by worst cores, which in turn will mask instability of your best cores. If you try to ease off on OC values for your better cores, you might just get back stability that you lost.
For example, on my 5700X, the best core asks for just 1.035 V with -30 CO, and worst asks for 1.115 V with -30 CO during Cinebench run. And during those all-core loads, it doesn't matter if my best is at -30 or -12 CO, in the end it gets the same voltage as my worst at -30. To be safe you can just match VIDs for each core using CO to your worst core's -30 CO and it should give the same improved Cinebench all-core performance/power use, just slightly worse power and much better stability under low core loads.
Can't speak for the 5700X3D but using a PBO per CCX undervolt my 5800X has a stable +300MHz boost in bios to 5.15GHz. 30,30,25,30,25,30,30,29. (6149 in Cinebench @90°C on 185w rated Thermaltake cooler, 4x8GB CL14@3600MHz, mesh case). It outperforms any 5800X3D in MSFS2020 when using above 1440p resolutions and ultra settings (rtx3060).
Hey Buildzoid. I've heard X570 Crosshair impact supports BCLK overclock in the range of 103-114 BCLK on some BIOS'es. Saw it on Asus and Overclockers forums, do not know if that's true. Do you think it might be the viable x570 option for 5700x 3D BCLK overclock? I remember you had the motherboard 3-4 years ago ... p.s. God bless you people ! Jesus Christ loves you 🙏
Avx is on a slider....at any given sse / core curve is also the avx curve which you can offset via voltage bandguard or ratio bin down 😅. So when you run an avx load you getting stock avx increase 100%value. This might be just Intel feature though. I run mine at 10%...avx is stable at Sse on 11th Gen
I'm running a -30 offset and a PPT of 75W on my 7950X3D because absolutely nothing I do with it seems to lose performance (Code compile and gaming. Often at the same time) but it makes my 360mm run much quieter... and I'm rather noise sensitive.
You can make your 360mm much quieter simply by making it spin slower under load, instead of limiting CPU power. Your CPU won't die from running slightly hotter.
@@igoresque I could. I could also just stop the fans completely and let it self regulate by thermal throttling. But as I wrote above I'm getting the same performance to within my measuring tolerance, and I'm getting better temps, lower noise and a lower power bill. And I'm obviously running a custom profile on the AIO anyway. The fans don't start until the liquid is 35C, so when I'm just reading documentation and reviewing patches only the 240mm front fans spin at 160RPM which is inaudible to me.
imo CO/undervolt won't give much more performance on capped/"bad" bin cores, usefullness is that it reduces temps and wattage.. which is really not needed on a 5700X3D
Idk why you had such bad experience with CO. On my 5700x OC'd with pbo to 4850mhz and has motherboard limits, so it pulls around 130W-140W so it's harder to keep it cool. So curve optimizer -30 all core can keep higher clocks for much longer because of lower voltages and less heat. Without CO i can get up to 100mhz less than with it. And that leads to a few hundred points(maybe more) less in cb23.
PBO2 tuner is best used when you utilize the CO modifier alongside increasing your PPT/EDC/TDC. You need to let the chip pull more current. The tool also bypasses the annoying EDC bug that nerfs your performance
Try changing your custom Pstate0, though that is also bclk dependent. If the math can be trusted, try 3900 MHz at pstate0 VID at 32 and then run the curve optimizer at -30 across the board again. I think you should be around 4.5GHz boost if there isn’t an issue running bclk that high. If so pstate0 at 4200 and VID 32 with bclk 107 shouldn’t be too high
if your cooling is on the edge you might not boost to that max you see and this might help. your cooling is just too good for this to be relevant. then again, this is just relevant for all-core workloads which are not really what X3D will be used for
I now have a video to link to people with unstable UV's, I know they wont watch it but ill feel better. BZ I hope you do a video on UV without changing the power limit one day, people dont understand that + to GPU core at same power limit is the same as UV. They always get confused when a GPU using 200W is still the same temp even with a UV!
i got just over 11k multicore score and 1570 single-core with same R5 5600 CPU on 550-plus ASUS mobo,32gb cl16 ram no custom timmings just preset, with undervolt -20 -20 other cores -18,although i did change mobo power limiter to extreme it goes over 100w in cinebench,artic frreezer 34 cheap 35€ air cooler i guess its nothing special either but i was able to manually oc to 4.85ghz all cores with only 1.325 volts stable with Cinebench temps under 90 for sustained load in Cinebench did not try Prime though for that long maybe it would crash but it was stable everywhere for my basic gaming and multimedia usage just power consumption without load was not worth it and performance increase in games were like none maybe here and there 3-5% max just not usable for gaming, but in rendering and such there were improvements up to 10% sometimes less or more depends on application
The X3D chips are great gaming chips, but the lack of overclocking is a drag, but that cache just can’t handle anything more. You’re right about the bclk on the X3D chips. It just trashes everything. Memory overclocking is a go on the 5800x3d. Not sure about the 5700x3d. I have a 5800x3d with 4x8 3600 Mhz dimms. You’d think the ram throughput wouldn’t be good with 4 sticks, but it gets about 49 gb/sec consistently.
yeah its only if you're really limited. Considering modern CPU's/GPU's is just a game of throttling limitations and overclocking is dead (cant run 5Ghz all core on a 5950x even though that is the 1 core firmware speed), the best we can do is reduce the headbutting of the power ceiling, like vega64 cards you can keep the same factory overvolt and try and go for more Mhz... but you stil run into the power limit, which sure if you raise the power limit 20% you can maybe get 5% extra on the core, however if you reduce the factory voltage it stops headbutting the power limit and clocks 5% higher as a result anyway. Modern overclockign software-throttle optimisation software you can reduce the throttling to gain more performance *sometimes* but the clock/power limits are not changable, IIRC AM4 is like 200 watts socket no matter what you do besides maybe going subambient (where a microcode switch gets flipped). No amount of voltage will allow any ambient cooled 5950x to do 5Ghz all core even if temps are okay because of the fun cap. I had a 5600 which at stock had 4.45Ghz 1c turbo and like 4.2Ghz all core With +200 frequency offset that was 4.65Ghz 1core turbo (so 5600x) but like 4.25Ghz all core. But then if you ran -30 CO across the board all 6 cores would just sit at 4.65Ghz which was a decent multicore improvement for now extra power and heat. Tried the same on a 5950x (no frequency offset) and all core R15 went from ~4Ghz to 4.25Ghz, Later combined with some Bclk tuning got to 4441mhz by the end of the run with a dark rock pro 4 air cooler getting over 30K in R23. I'd love to run 4.8Ghz all core (seeing as daily use has shown every core is perfectly fine running at that speed... on their own, the second more cores get used the power leash gets pulled and it goes back into gimp 4.2Ghz mode), this is all we have now. This is word soup because I'm in allot fo back pain.
The 142W CPU PPT value seems to be intended for dual CCD CPUs, because they can easily get to 35W - 37W SoC power under load, in case you run quad-rank DDR4 (e.g. 128GB) and connect all available PCIe lanes. The cores usually sit on a combined 105W - 107W then, depending on the AGESA version, since TDP for Ryzen 3000/5000 is only a fantasy number. My 5950Xs also draw 17W - 20W SoC power during idle compared to the ~7W SoC my 5800X draws idle. Before AMD fixed ErP Lot6 with one of the AGESAs last year, the idle power could occasionally even hit 30W SoC on the dual CCD CPUs. *edit: clarification
@@mparagamespackage power is always gonna be higher than SOC + core power, since your CPU has to have the interconnects powered at all times, maybe something else is going on as well, that isn't displayed. My 5700X idles at just over 19W with SOC at 8.5W when nothing is running except HWiNFO. SOC voltage set to 1V and 3733CL18 64GB RAM at 1.35.
5800x3d asus rog strix b550-f wifi gaming, -30 in bios same page top 3 , 100 , 70 100 61c single cores 1487 cpu multi cores 14475 . 4949 61c on 2 cores the rest 4649 61c. Im happy with my setup. Cinebench r23. Enjoy.
Curve optimiser is about to reduce the voltage of the CPU. Why is it there and not only Core voltage and/or load line calibration? I don't know. All I know is that thanks to this - X3D CPUs users can undervolt their CPUs even though CPU voltage is locked. :D And I am the person who uses PPT limits on my 5800X (NON 3D). With voltages and PBO tweaks I am able to achieve around stock performance with lower power consumption and much lower temps and fans speeds - noise. All those new fancy OC stuff is not to make your overclocking easier or better. It is all to push further the CPU on stock values, to overclock itself and beat competition. No one cares if it draws the amount of energy similar to small village. Do whatever it takes to go even 50Mhz higher! That is why many people nowadays are doing opposite things - undervolting and energy efficiency optimising. In many cases it also leads to better performance.
How does this software works? You gotta open it and dial the numbers every time you turn it on? If so not very convenient but it is a nice to have tool lowering power consumption is a big deal for many ppl.
u create a scheduled task in windows with a argument like this ( for example 12 Core 5900X ) -20 -10 -20 -20 -20 -10 -18 -20 -10 -20 -20 -10 395 120 120 first 12 values are CO , the rest is PPT EDC TDC
Not a amd thing but Intel could use the terminal velocity and vf more easily to use like PBO... I'm running a 12700k at 1.43v to 5.1/4.1ghz because i lost patience trying to do the fine tuning... Rams micron fdie 2x32 at 3900mhz cl18-21-21 1.28v
On my 5700X I initially just set all to -30, but within a week system crashed at idle. Set the best two cores to -25 instead and it was all good. Eventually came to -28 on a single best core and -30 on the rest, no issues whatsoever.
From what I’ve seen at least with the 5800x3d cpus the highest clock frequency I’ve seen from one was on an X570 unify (non x) using the external clock gen. I don’t disagree that you can run the base clock higher on b550 but the GPU won’t likely support it even at pcie gen 2 speeds.
in a perfect world, we should be able to clock this with a 200 mhz difference opposed to a 5800x 3d. let's say between 4300/4350 boost, and eventually all cores. A little frustating, for sure.
CO is complicated, if not done right there's just clock streaching and instability, like the 99% of the people posting -30 all core online. My 5700X is able to run +50Mhz with a mixture of -30 to -10 fully stable, more than that cause one core to no boost correctly, maybe is my bios bugged idk, but it's pain.
I've only ever seen instability as an issue with CO, never clock stretching. In my experience, clock stretching only happens when the supplied voltage is detected to be insufficient, which might happen due to manually set vcore/offset or a weak LLC setting. CO just straight up makes cores ask for less voltage for the given frequency than they should at stock, which is why it crashes when not enough.
That's not true when it comes to the X3D chips on AM4. They run at a low enough frequency that most can do -30 on all cores without negative effects. I've tested my 5800X3D in every benchmark going at -30 and performance has always been better than stock, complete with significantly lower temperatures. On the other hand, my 5800X non-3D most certainly couldn't do -30 on all cores, although the result was instability in light workloads (random BSoDs when watching RU-vid videos and such), not clock stretching.
@@igoresque and never seen clocks act crazy? My chip can pass +200Mhz just coz the effective clock never goes beyond 4.6Ghz, making stress testing useless
Probably cause you have used AIO 240 boost algorithm PB2 already maxout the speeds without need it use curve optimizer, that's why you don't see any true difference and also Boost Override didnt work , the combine curve optimizer (negative) with Boost override (max 200) give huge performance increase in 5000 series compare with the stock.
so I just tried this out at 0 I get 110 score in 7zip, if I set CO to -5 I get 112 if I set CO more aggressively I still get 112 i.e. -20 or -30 so 1.8% improvement... but a 20% power saving on a 5800x3d see my comment on the other 5700x3d video about why people are dumb talking about CO etc.
Price is a joke tbh for such downclock when u got 5800X3D for almost the same price. Was hoping for some clock gains here but naah, that thing more capped than 7800x3d
I use PBO2 for my Ryzen 7 5700X3d since July 2024. I set all cores to -25 and its rock solid playing all my games with Crysis Remastered being one. It reduced my CPU temp by about 8 degrees C. I used Beefy Tech's video to set it up.
Meanwhile my MSI MPG B550 Gaming Plus supports curve optimizer via AMD Overclocking menu. I can only go as low as -30. But that, paired with 95W limit, is enough to keep my 5800X3D below 70°C in Cinebench R20 with a 280mm aio while keeping the allcore boost clocks at 4,3 Gigahurtz and 4,45 GHz in games...
Ha this will be more like it... Next on shit you didnt know existed but normies run: MPT, ds state disabing and TempDependentVmin lol Will provide 69xt😂