As someone who owns the 5800x3D and undervolted it, you've limited the PPT TDC EDC a little too much with 100 70 100. Try 122, 82 and 124 while keeping the -30 on all cores, you should see better results while keeping the temps down simultaneously.
You will loose Cinebench (and Gaming) score if you limit PPT, EDC and TDC to these common values . Just undervolt to -30 (or how deep you can go stable) and you get a massiv boost in games and CB. The CPU will stay cool and saving. Disable CPPC Preferred Cores (but keep CPPC enabled), since it can make problems in games and lower the FPS and is not needed. And if you check Ryzen Master under stress testing conditions than you will see that the 58X3D never reaches 142, 95, 140 stock limits. Max under workload is ca 115, 70, 115. So limiting is not needed at all. Just go for the Curve Optimizer. Classic UV & OC is not possible on a 3D (for good reasons). Easy going with the 58X3D: CO to -25 or -30 and Preferred Cores Disabled. That's all you need to do :)
@@jonathanparle8429 i can only do curve optimizer all core -25 rockstable. Not every CPU can do -30, only the golden ones. But -25 is still a huge difference.
@@__Pathfinder__ Thanks. I now have mine at minus 30 and fully stability tested. A fellow flight simmer mentioned to me that they had their at minus 50 (which seems astonishing) but even at minus 30 temps are easily within the safe envelope and the power draw is significantly below the AMD spec limits. That said, I might try lower just out of curiosity even though at minus 30, I get 100% speed from all cores 100% of the time (if the applications demand it of course). I assume it cannot hurt other than I might begin to lose stability.
@@jonathanparle8429 You need to play different games for hours. If you get reboots and WHEA 18 errors, than you are to low. It will not show directly... takes a week or two of intense playing different games. Not every mainboard can go lower than -30. I think the Asrock boards are not limited. In my case some games work with -30 and some only with -25. So it takes time to figure it out. But nevermind, not really much difference between -25 or -30 in terms of temp. Ingame performance stays the same...
I have my 5800X3D running just with the curve optimizer -30 on all cores. The difference to stock was +200Mhz, -1-3Watt and -1-3C°. It is running at 4,45Ghz maxed on all cores at all times with CO -30. With 105W max (usually 65-85W in games) and 22C° roomtemp. BUT it depends on your hit in the silicon lottery. I know many people that don't hit -30. It also depends on your board.
If you can not hit -30 than disable AMD CPPC Preferred Cores (since it makes problems in some games and lowers FPS). But keep CPPC enabled. I couldn't get stable -30 and than i enabled Global C Stats, CPPC and disabled Preferred Cores. Rockstable everywere after that with -30. PPT, EDC and TDC is not really needed since it will lower your workstation performance in cinebench etc.
I only just got mine yesterday and have been too busy installing Windows, games and other apps to do any proper undervolt testing yet. But after 4 hours of up time after the initial Windows install, I reduced the curve to minus 10 and after another 4 hours of perfection in game testing took it down to minus 20. Although it is perfectly happy there as well I'm not going lower without proper testing rather than just testing the game installations.
I UVed my 5600x3d with Kombo Strike 30 as well, didn't change anything else. It initially lowered temps significantly to about 72C all core from 90C. Seems to have popped up to about 80C last I checked, so either my fans are clogged or it got significantly warmer (it's probably the latter)
I am late to the "party" but I run PPT=120, TDC=85, EDC=120. This is a slightly modified 122/82/124 that many people recommend for gaming, both work well, it just seems like my system like that 120/85/120 SLIGHTLY better, its almost within variance, I am pretty much min/maxing at this point. This is all with NO LLC, and my phase control on EXTREME for a full 16 phase power delivery, and a -30 voltage offset. I may start playing around with manual undervolting, I did that with my 3800X where I slightly overclocked while undervolting, I got my 3800X to run 4.5Ghz all core at 1.25v, so its possible but silicon lottery plays a significant factor in capability and my 3800X seems to be a 1% binned chip. But right now with no LLC, extreme power phase profile, a -30 offset, and a 120/85/120 I am very happy with the results. My 5800X3D idles at 42C, and with Cyberpunk 2077 on max settings at 1440P (yes I know its not 1080 so its less CPU intensive) it sits at 55C-58C. Prime95 Small FFT 30 minute testing saw a temperature of 77C vs stock where it hit 90C, that's a 13C difference under extreme testing!
you may not want to use extreme llc depending of the board.. it may increase your temps quite alot. i always use llc at the second highest setting at the most.. you find the best setting when you lock your vcore and see in which setting it changes the least when idle to 100% usage.. less the better. / the more stable the vcore the better.
I do own a 5800X3D with a 4080 and gaming in 1440p and after many many tests my best results came with CO -25 all cores (-30 was not stable for me) and PBO limits on stock (disabled on my mobo). I played with these a lot but in the end stock values seemed to do the trick on most scenarios and it seemed snappier overall. And also something else that helped a bit was turning CPPC preferred cores option to Disabled. All the cores for this CPU can hit 4.45 GHz no swet so there is no need for having preferred cores enabled. Some people who play mostly competitive like to turn off C-states as well to improve 1% lows but that didn't have any impact for my case.
Preferred cores should be disabled since it makes problems in some games and lowers FPS. I got freezes at -28 and than i enabled Global C State and CPPC and disabled Pref. Cores. After that -30 is rockstable for me. But every system and CPU is different... PPT, EDC and TDC will lower your workstation performance in cinebench etc. Best compromise is just to leave it stock. PBO is disabled with the 58X3D. That's right light that. No OC on a 3D for reasons...
Install AMD Ryzen Master and run the cuve optimizer auto offset, choose per core and the let the app tell you how much values each core can handle. This will make your undervolting much easier. One or more cores might be experiencing instability at higher clocks with the offset you used and have to be clocked down, thus, the performance hit.
Yes, I used this and it worked well. I had some instability but I fixed that over time by going in and raising the lowest voltage by 1 mv each time I got bounced and running my normal routine again for a while. Gradually my system completely stabilized. Not very scientific, I know, but it worked. Running a 5800X and under the heaviest loads I can't make it hit thermal throttling (85*) and it runs at a constant 4.6-4.8 GHz (Fuma Scythe 2 air cooler in 23* ambient room).
But this does not help the instability at lower voltages which is - at least in my experience - where the pitfalls exist with the curve optimiser. You can easily select offsets and very simply test for stability at full load but there is no reliable and 100% repeatable test for stability at the lower voltages on the curve when the CPU might have next to nothing or little to do. And in those situations the relative undervolting is much higher than at the top of the voltage curve. You can really only test that using the computer "normally", as in using it with the applications you would normally use and over a period of some weeks. I wish we had the equivalent of the ASUS GPUTweak OC Scanner, but for the CPU. It would mean a stable voltage frequency curve at all voltage points rather than just at flat out.
You were getting much better 0.1% lows with the undervolt, this will make your long term experience far better compared to slightly higher or lower average FPS
I just done it in Asus MBO bios with PBO2 -30 all cores, didnt touched that oder parametres...and i got all cores to 4550mhz and lower temperature about 10 12c using DarkRock 4 cooler. Fps is higher across board.
I have a 5600x3d with the only setting in the BIOS changed being PBO set to -25 all cores and yesterday while playing rdr2 I was monitoring everything with HW monitor. And at some point I reach 4.55-4.67GHz on all cores. It's limited to 4.45 and I've never seen it above that ever. I'm going to recheck again soon and also try to see some graphs. Did it just peak there for a millisecond or did it actually hold it for a period of time? I do not know. I do have my fan curves pretty aggressive on CPU and GPU and neither went above 61c.
I've had the 5800x3d in 2 rigs for about a year now, and I loved the -30 PBO it crushed all benchmarks and stayed so cool and pinned to the max freq I thought the cpu was rock solid. but I've been dealing with micro stutters in BR type competitive games, and it was the PBO tune.
my issue seemed to be memory stability. upping DDR voltage a tad 1.36 and applying @Mohamed-zy1er ppt tdc and edc of 122 82 124 allowed my to run crash free and Studer free while gaming.
Interesting results David. I have two Ryzen 7's, An earlier build with a 5800x and one with the 5800X3D. My original 5800X I needed to under volt with PBO + Curve. (-20 offset, Wasn't stable any higher of an offset) This Setup performs excellent with Cool temps and higher sustained clock speeds. (CineBench R23: 16000/1600 Multi/Single scores) (32 gigs of Ripjaws 3600mhz CL 16. two sticks) As for the Ryzen 7 5800X3D.. I didn't bother to touch the under volting, Only because of the needed power for the extra cache supply. I'm just using a brute force air build with an incorporated Noctua D15 and Lian Li Lan Cool 216 case. Also my 3D chip has the RTX 4080. No heat issues to worry about. I did weigh in the idea of under volting but realized it ran so smooth and cool the way it was setup. Thanks for the enlightening video. Cheers!
I'm happy with -30 all core and 114, 75, 115, since it keeps me right under 80, where throttling starts. If you have a good cooler though, 122/82/124 are the gaming-biased settings floating around. 100/70/100 sounds very conservative.
If you keep stock and check Ryzen Master in stress testing than you will see that the 58X3D never reaches more than 115, 70, 115. It will never get to AMD stock max values of 142, 95, 140. So you can just keep it stock ;) It's not a 58X... And Zen 3 throttling starts at 90c not 80c!
@@__Pathfinder__Good to know about the AMD targets. I guess it's the -30 all core that's putting in the work for me. I should say "pulling boost" rather than throttling. My silicon lottery definitely gets tame over 80.
@@MrCthur ah ok since throttling is a automatic clocking down function from AMD to prevent overheating. This starts at 90c on every Zen 3 CPU. Temps over 80c are still perfectly fine.
I used amd ryzen master to find the optimal ppt tdc and edc which i adjusted while running cinebench in the background. on a ryzen 7 5700x my PPT =135w TDC =85A EDC 140A. What i did next was used the scalar x10 added 200mhz on to clock speed and used -26 on all cores, i ran into some instability when the cpu was doing very little work, so i futher tuned and realised that the prefered cores, the gold star and the silver dot and cores that worked with these on low workloads needed less negative on them, so my cores now look like this core 1= minus 24, core 2= minus 18, core3= minus 26, core 4= minus 26, core 5= minus 24, core 6= minus 24, core 7= minus 18, core 8= minus 26. on my cpu core 2 , core 7 , core 1 , core 5 , core 6 all have lower negative numbers as these cores are the most used for general useage the rest of the cores only come into use under heavy workloads and so have a higher negative number. when the popular most used cores are in full useage it doesnt really matter how much negative you put on the least used cores as most of the cpu is powered up by the time these cores are used, if you put too much negative on the most used cores these cores work on there own and so therefore need more voltage to handle smaller more used tasks.
Thank you. I have the same cpu, I tried your initial settings. Seems to be stable for me. Had to drop my first 2 cores to 25 and left the rest on 30. My temps are higher than before while running cinebench but I’ll keep adjusting until I get a nice balance.
You dont need to power limit (PPT/EDC), you can use your voltagelimit to do that if you want it to actually run cooler, more on that later. Because in reality, PBO tuning is technically not an undervolt. The voltage is determined by the power draw, temperatures and other factors, and the CPU will then pick the corresponding frequency associated with that voltage via its internal curve. PBO tuning or curve optimizing, shifts the frequencies associated by the voltage. So, at the given voltage based on power draw yada yada, it picks a higher frequency. Therefor, any claims of lowered temps and power draw is untrue, since it still uses the same voltage and package power, it just clocks higher. At -20 you already hit 4450 all-core in gaming workloads, so going from -20 to -30 does nothing. But, you can still utilize the extra range in another way. What you can do, is combine a NEGATIVE OFFSET on your vcore with PBO. Something like -50 to -100mvtt. That will drop the voltage, but also the frequency. You can use the additional range from -20 to -30 to curve optimize as far as you can go, to not have clocks drop below your target. I ended up with something like -75mvtt negative offset, and -25 on PBO tuner, to hit 4450 in gaming workloads. It still goes to 4425 in blender or cinebench, but i dont care, i dont do professional workloads with the CPU. That will, unlike only PBO tuning, drop the temperatures and power draw. Also, using the voltage limit like this means it'll not bounce off the power target and drop clocks. Which you're not verifying on-screen. You need to actually hit 4450, otherwise you're just throttling it. No wonder its worse.
PBO can't be used with a 3D and is therefore always disabled. No OC with a 3D for reaons! PPT, TDC EDC is not needed since the 3D goes only to 115, 70, 115 und stress testing (Ryzen Master). It never comes close to AMD stock limits of 142, 95, 140. So no tuning needed. Only Curve Optimizer to -25 or -30. 3D is not like a regular Zen 3. You can't do classic UV and OC. So just CO to -30 and CPPC preffered Cores off. That's all what's meaningful...
@@__Pathfinder__ Using HWInfo with default TDC, EDC and PPT settings when running CB R23 it shows max TDC ~71-72A, EDC ~119-120A, PPT ~105-106A. So you are technically correct. You can limit to 70,100,100 but gains/losses will be marginal at the cost of possible instability.
@@Mopantsu And no need for that. Curve Optimizer as low as possible/stable is more than enough for this CPU. These limiting settings are more a thing for the regular Zen 3.
@@__Pathfinder__ I did see additional cooling while maintaining performance when using PPT/TDC/EDC changes. I run 120/85/120, the EDC being at 120 for me matches my motherboards max capability of 120A to the CPU/SOC as its a dual 60A on the X570 ROG Crosshair VIII. I slightly dropped the PPT and increased the TDC to 85W which is 10 less than stock, and then the 120A as explained before. I feel like the 140A on the EDC at stock settings was just so the CPU could take advantage of the uber high end boards like the MEG Godlike, AORUS Xtreme, and the Dark Hero to name a few. I think there are a total of 5 X570 boards that have 70A - 90A mosfets that can actually pull the 140A EDC limitation, but its not needed unless they start adding in 5950X3D chips.
As someone who recently bought a entry level system with an 5600 cpu and rx6750xt 12gb gpu I'm curious about what your final take is on whether the 5800x3d was a worthwhile upgrade. I was thinking that maybe in a year or two it might make be a quick (and hopefully cheap by then) upgrade option for me without the need for a major system overhaul.
You have an MSI motherboard so try kombo strike 3 setting. That's how I run my 5800x3d. I've heard that's the same as curve optimizer -30 all core, but kombo strike 3 seems to run cooler than CO at the same performance. I think your PBO settings are too restrictive, hurting performance under a ray tracing load but that's a guess.
Huge improvements at -30 for my 5800x3d. I get on average 10+ degrees less and 10-20 less watts of power usage. Originally my 5800x3d was peaking at aound 80c in a game like cyberpunk 2077, but now it just hovers around 70c, watts peaked at 105, and now it peaks at 95. I'm seeing a lot less down clocking too, because of the temps.
Great video, good quality video and audio. Maybe try out different levels of negative values for curve optimizer. Also keep an eye on the clock speed, as certain games like csgo perform better with higher clocks. Good luck mate with your channel, just ignore people who indulge in pure hate.
Honestly, I'd rather look at those 1% and 0.1% lows for this undervolt, while undervolting didn't change much for the AVG FPS, it certainly helped keep the FPS much closer to the average, which says a lot for frame skipping, means we would see little to no frame skips, where keeping it stock meant you'll notice those frame drops
I must say I have a fractal torrent case and added a 3 pin 120mm exhaust fan behind the cpu too. The 5800x3D idles at around 45°C in this hot summer, window open. While heavy gaming it's around 70 degrees, heavy cpu benchmarks at 89 if I remember correctly, never throttles with a peerless Assassine 120 SE. So far I saw no reason to UV it in terms of temps or performance.
Guys, should I upgrade to 5800X3D from a 5800X on 49" 5120x1440p @ 165Hz monitor with 7900 XT ? 1440p SuperUltraWide is closer to 4K and it's a known fact that the "3D upgrade" is NOT worth at that resolution ? 🤔
si vous jouer uniquement en 5120x1440P avec une 7900XT cela aura pu d écart avec un 5800X seule la 1080P montrera une différence visible dans votre résolution pas d interet si ce n est avoir le meilleur CPU pour l AM4 pour étre paré pour le jeu si vous avez besoin polyvalence le 5800X est mieux
For my 5700X3D I didn't really see the point of undervolting. Sure, in Cinebench it clocked from the 3.9 - 3.95GHz range at the standard voltage up to 4.05GHz constant after undervolting. However, when gaming it would clock up to it's max 4.05GHz anyway at standard voltage. Because it's not as heavy CPU load as stuff like Cinebench. And I don't use my PC to run Cinebench all the time :P I use it for regular browsing, youtube, email and gaming. So basically my cooler is good enough to keep it at max boost at default voltage when gaming. It may make a bit more sense with the 5800X3D if you don't get it's full boost when gaming with it's default voltage. Like if it clocks down to 4.25 - 4.35GHz when gaming. Undervolting it will make it run it's full 4.45GHz boost when gaming.
Mine went straight to -30 all core - thank you, hard-capped boost😁 (I inched my thermal and power limits down until my CB23 scores started to drop, then left them there.)
I have an ASRock B450M Pro4 and plan to buy the R7 5800X3D. I would like to do what you suggest, but I don't know how to do it. Could you provide a video that explains the process in detail? Thank you.
My answer is to low ppt, edc and tdc + to aggressive undervolt This -30 shit is so awful After several hours of testing like 10 hours everything were nice And then mwz in the red zone, massive hoarders of zombies, bosses, the weather effect got me an bsod Cause ram issue After enabling xmp2 instead of 1 and changing ppt, edc and tdc its working perfectly fine
My 5800x3d runs 88 80 120 limit with all core -30 and seems to be completely stable for about a half a year. Before it had 88 65 90 limit + all core -30 and I was having stuttering a bit. Also I have cppc and svm mode disabled
As nice as undervolting the 5800X3D in a typical desktop scenario is, where it really shines is in SFF builds. The 5800X3D is an absolute beach to cool in SFF, not just because of the incredibly inefficient heat transfer from the CCD to the IHS due to the V-cache chip in-between, but also because you're limited by your cooling capacity. My case is a 7.5L i100 Pro/Hummingbird, so I don''t even have the luxury of an AIO - 60mm low-profile is the best I could get. So I got an ID-Cooling IS55, one of the best in the business, and put on some Conductonaut liquid metal to give me the best possible results. CPU fan speed was set to max at all times, though it is damn near silent. All that for 60+ on idle and a constant 91 degrees on load with inconsistent clocks. Just turning the PC on and loading into Windows caused the CPU to exceed 80 degrees. It would hit 4.55GHz, but rarely as spikes and mostly hover around 4.3GHz and often as low as 3.7GHz momentarily. All core clocks in Cinebench 2024 would settle at around 4.0GHz after 10 minutes, also fixed at 91 degrees. At any rate, the fact that I could still boost in this situation meant that my cooling was just barely not enough, especially because I noticed the clocks being a little more consistent just by having the A/C running. So I went into the BIOS (Asus X570i) and set a basic -0.2v offset, and left the previously set performance enhancer to "level 3 (OC)" as it was. Temps on load are now on average around mid 70s, with occasional peaks at 85-ish. Idle temps now sit below 50, sometimes below 40 with the A/C running. That alone would've been a great result already, but due to the massive improvements in temperature the clocks greatly improved too. Cinebench all core load now sits at 4.3GHz constant at 86 degrees max, and when gaming it hits 4.55GHz quite often, usually 4.45GHz and rarely dropping below 4.3GHz. I could probably do better with some more in-depth tuning, but I don't really need it. Most of my concern was regarding the high temperatures, so any performance improvement I got out of it was a bonus. I don't even have a sufficiently fast enough GPU to hit a CPU bottleneck, even though I play at 1080p on a RX 6800, so this'll do for now.
Benching a CPU without showing CPU clock is kinda ...wrong. We have no clue whats going on with the clock speed. Is it fixed or fluctuating? Does is drop at some scenarios or not?
can you use kombo strike with gameboost and pbo bios setting ?or can you use all 3 at once or is using only 1 or 2 options best? right now i think I have kombo on 3 ,and PBO auto, game boost off set.installed my chip last night? TY .on a MSI X570 gaming plus. I do know how to manually tune PBO
Just listen professional tip. Don't change PPT TDC or EDC, Because once you change one it will bottle neck the other 2. For the best result choose PBO optimizer -30! Because the curve optimizer in bios won't have same result as the pbo optimizer which has slightly better cooler temperatures and and clock high 4450mhz! You just need to set task scheduler that the pbo tuner will set with every reboot.
If you do like a 102-103 bclk overclock, it'll probably still work at stock voltage without breaking PCI-E or IO ports or the ram, that'd be probably a better choice than undervolting with a typical large tower cooler as these single ccd ryzen cpu's don't really pull much power therefore they're easy to cool. Also I think your cyberpunk experience is worse because of ryzen cpu's clock stretch a lot, I have a ryzen 7600, I got it overclocked at 5.45 ghz stable at stock voltage, however when I increase the clock speed to 5.5, y cruncher doesn't crash, it just prompts a lower score, the reason why is that when ryzen cpu's don't have enough voltage by like a tiny bit like probably 0.01 mv or something, it'll drop the frequency or do something to drop the benchmark score. It runs 2 hours stress test with Linpack with no memory errors, but it also drops the tflop numbers which I see in the linpack. So this cpu at this voltage is not stable at 5.5 ghz, It'll probably go away if I give it like 10-20 mv or something. So even if your Ryzen cpu doesn't crash, it can drop the clock speed down the base frequency for like 0.00001 sec quite frequently and your monitoring application may not even register it.
If you're getting more or less the same performance for significantly less power then I'd call that a win, no? Also as you say, the 0.1% lows were much better in a lot of your tests, which is arguably more important than getting an extra 10-20fps (especially when you're already well over 100fps!). That translates to fewer "stutters" and thus a smoother gameplay experience overall.
so what would be good for b550m ds3h mobo i just got the 5800x3d and need to undervolt what would be good im -30 on all cores what else should i do im running 120 75 110 i dont see any difference please does anyone have any suggestions?
Im getting restarts in games demanding games actually with 3600mhz ram i have the same as you 32 gb and i searched that ryzen 7 5800x3d max speed of ram is 3200 how make that work please help me thanks!
My PC wouldn't boot with RAM @3600 XMP profile, I fixed it by: 1. Manually setting DRAM voltage in bios @1.35V (like my XMP profile) and then in windows I opened Ryzen Master application, and manually put the timings of my XMP profile (you can skip this if your PC gets to Windows without crashing with your XMP/DOCP profile selected in BIOS) 2. In Windows Ryzen Master - Increase VDDCR SOC from 1.0 to 1.1V (+0.1) 3. In Windows Ryzen Master - Increase CLDO VDDP, CLDO VDDG CCD, CLDO VDDG IOD from 0.9V to 0.95V (+0.05) 4. Save and then apply profile You just have to test for stability, but it should do the trick.
odd... I use 3600mhz ram as well and I run it at 3600mhz using MSI's Memory Try IT set at 3600 18-20-20-42 no issues but disabled XMP.... I will try going 16-19-19-39 and see if I get even faster performance... my ram stock timing is 18-22-22-46 (I think).
I like tests like this which improve gaming but I don't think you need to undervolt, especially since the cpu is great out of the box or unexperienced. If you cap frames, your cpu will work less and never get hot. Im running my 5800X3D at stock settings so it can decide itself how much power to use or not. The hottest ive seen it get is 65 degrees in games. In Cinebench it was around 77 maximum. Air cooled with Dark Rock 4
@@michealmeyers9789 I agree but it is not always guaranteed for everyone and every game. You never know how a game will react to the cpu. some games will require the cpu to use all its strength to perform properly
Your results were lower than they should be because the power was set too low. Try the undervolt and DON'T touch the power limits leave them default then see the difference......
@@davidnott_ I tried the power limit thing with the negative offset curve on a Asus B450-f gaming motherboard and what i found it that for some reason my chip degrades in performance if i lower it on one of the 3 power settings even if it's just a small amount that i change it. I think it was the TDC. When i tried it again with just a negative offset of -30 my benchmark scores went up and the chip stays locked at 4450mhz all core and about 10c cooler. I'm sure reducing the power limits is just another method of trying to do what the undervolt is already doing so doing both together maybe overdoing it depending on the chip/motherboard. Some people seem to get away with doing both. Another thing i thought about is maybe it's a BIOS issue? If there is a newer version of the BIOS maybe it will work better.
Honestly, all your results are likely just a test to test variance. Reasons being: 5800X3D is a locked down chip with temperature limits. (not factor in your case) PPT, EDC, TDC on this chip are an issue because they can't be increased, lowering them just means you're creating even tighter power limit and reducing the max performance even more. In your case the limits you've set are not a factor because this CPU does not even reach the limits you've set in gaming loads. You would need to run something like CBR23 that maxes the CPU power to see the difference. For your testing the only benefits there are the lower temps from CO being at -30. That said CO of -30 is rarely fully stable especially with default LLC settings and while you may have not encountered crashes, it may have resulted into clock stretching and thus even higher test to test variance.
David Nott : AMD is known for their advanced capability to configure their CPU-s very very efficiently and very very optimal. But again, WTF are you talking about with the lowering of voltage and thus the power consumption. I have a Ryzen 7 5800X3D and @stock frequencies with PBO enabled and Turboing over all cores enabled and this CPU physically cannot get over 111 Watts of total power consumption as a socketed CPU alone !!!! Why is there any need to lower the power consumption since this is a VERY VERY LOW POWER CONSUMPTION for a 8c/16t 2.nd fastest gaming CPU on the planet ?!?!?!?! The temps also physically cannot get over 80°C in my usage when I have 150+ tabs opened in my browser + playing RDR 2 + decompressing 90 GB large pack of files at the same time. Or don't go over 84°C when stressing it for hours in Prime95 or FurMarks CPU burner test. I've also tested the Ryzen 7 7800X3D when building my friends RIG....and at the same condition, cooled by the same Thermalright Phantom Spirit 120 SE cooler, physically cannot get over 83 Watts of total power consumption as a socketed CPU alone. And the 7800X3D is the No.1 fastest gaming CPU....performance wise shitting all over the i9 14900K in every game, especially if coupled with the RTX 4080, RTX 4080 Super or RTX 4090.
@@Blaze1984HH I did exactly that because I know of the Small FFT-s in Prime95. Just as I know of the Small Data Set in OCCT and just as I know of FurMarks CPU Burner, Intel CPU Burner and the continuous run of CB20 and CB23. After around 15 hours of continuous benchmarking my Ryzen 7 5800X3D I recorded the highest peak PWR consumption of 111 Watts in HWInfo and via calculater current clamps with my osciloscope :). So buttercup, you don't know squat about any AMD cpu-s. In contrary to Shintel, AMD NEVER LIES about their PWR consumption in CPU and GPU department :). Those are provable facts by me or any unbiased normal user anytime and anywhere :).
If you're not an overclocker than just go back to stock. Chip has been out for quite a while. If viewers have questions, the answers are already out there on the internet....
Why stick with a 5800X3D when you can easily sell it for $250, the board for $100 and the ram for $50. That's what you can get a 7600x/7700x/b650/32gb CL30 6000MT ram for. The value still in these old parts can still pay for much of the upgrade. I don't know why people would hold onto them when it's dirt cheap to get basically the same performance sometimes more sometimes less with the 7600x/7700x. I had a used 5800X3D for a month that I paid $263 for. It was certainly better than my 5900X which I sold for $260. Then a few weeks later I realized the 7600X on AM5 was on average 3% faster. Well I sold the 5800X3D & Asus TUF X570 for $380 and the 64gb ram for about $100. I went to a new 7600X for $200, open box X670 Aorus for $130, $110 new Trident CL30 6000MT. Then that trickled down to every AM4 system inside the house. When below retail AM5 deals, boards and ram appear I would part out AM4 parts for usually not much more than what I sold the parts for. I have two 5800X, 1 X570S Tomahawk and 96gb of DDR4 left to sell off. AM4 selling is slowing down, 3 months ago this stuff sold fast. Now it's much slower.
@@Kapono5150 If the CPU is cooler, the ambient air temperature will be cooler so motherboard, ram, drives, PSU etc will be cooler and so will be the GPU too. When running cooler, system stability is improved and device life is maximized. CPUs and GPUs will achieve better boost clocks, higher and for longer. In the heat of the middle of summer for example is when the gains will be even more impressive and important. Hey even the room will even be cooler. It's basically a win, win, win, win, win, ..., ... for not much effort and zero money. Satisfying too... nerd juice 🧃
As someone who owns the 5800x3D and undervolted it, you've limited the PPT TDC EDC a little too much with 100 70 100. Try 122, 82 and 124 while keeping the -30 on all cores, you should see better results while keeping the temps down simultaneously.
Thanks, I’ll try that!
He is right particularly 125 85 90 are perfect with -25 negative all cores, but it make it 5 ° hoter, bye @@davidnott_
I ll try thx. In my case at -30 i have random reboots After BIOS update. So i m using -20 now, cuz i dont have time to crunch and test each core.
There is not too much..
My 5800x3d 88/60/90 and -30, it runs better than stock 😂
In games 4450mhz, Cinebench 4200mhz and it's just way cooler
thankyou!
You will loose Cinebench (and Gaming) score if you limit PPT, EDC and TDC to these common values . Just undervolt to -30 (or how deep you can go stable) and you get a massiv boost in games and CB. The CPU will stay cool and saving. Disable CPPC Preferred Cores (but keep CPPC enabled), since it can make problems in games and lower the FPS and is not needed. And if you check Ryzen Master under stress testing conditions than you will see that the 58X3D never reaches 142, 95, 140 stock limits. Max under workload is ca 115, 70, 115. So limiting is not needed at all. Just go for the Curve Optimizer. Classic UV & OC is not possible on a 3D (for good reasons). Easy going with the 58X3D: CO to -25 or -30 and Preferred Cores Disabled. That's all you need to do :)
Appreciate it, will try it!
Glad you guys have done the hard work. I only built my machine yesterday and appreciate the collective wisdom to make the setting up much easier.
@@jonathanparle8429 i can only do curve optimizer all core -25 rockstable. Not every CPU can do -30, only the golden ones. But -25 is still a huge difference.
@@__Pathfinder__ Thanks. I now have mine at minus 30 and fully stability tested. A fellow flight simmer mentioned to me that they had their at minus 50 (which seems astonishing) but even at minus 30 temps are easily within the safe envelope and the power draw is significantly below the AMD spec limits. That said, I might try lower just out of curiosity even though at minus 30, I get 100% speed from all cores 100% of the time (if the applications demand it of course). I assume it cannot hurt other than I might begin to lose stability.
@@jonathanparle8429 You need to play different games for hours. If you get reboots and WHEA 18 errors, than you are to low. It will not show directly... takes a week or two of intense playing different games. Not every mainboard can go lower than -30. I think the Asrock boards are not limited. In my case some games work with -30 and some only with -25. So it takes time to figure it out. But nevermind, not really much difference between -25 or -30 in terms of temp. Ingame performance stays the same...
I have my 5800X3D running just with the curve optimizer -30 on all cores. The difference to stock was +200Mhz, -1-3Watt and -1-3C°.
It is running at 4,45Ghz maxed on all cores at all times with CO -30. With 105W max (usually 65-85W in games) and 22C° roomtemp.
BUT it depends on your hit in the silicon lottery. I know many people that don't hit -30. It also depends on your board.
If you can not hit -30 than disable AMD CPPC Preferred Cores (since it makes problems in some games and lowers FPS). But keep CPPC enabled. I couldn't get stable -30 and than i enabled Global C Stats, CPPC and disabled Preferred Cores. Rockstable everywere after that with -30. PPT, EDC and TDC is not really needed since it will lower your workstation performance in cinebench etc.
I only just got mine yesterday and have been too busy installing Windows, games and other apps to do any proper undervolt testing yet. But after 4 hours of up time after the initial Windows install, I reduced the curve to minus 10 and after another 4 hours of perfection in game testing took it down to minus 20. Although it is perfectly happy there as well I'm not going lower without proper testing rather than just testing the game installations.
I UVed my 5600x3d with Kombo Strike 30 as well, didn't change anything else. It initially lowered temps significantly to about 72C all core from 90C. Seems to have popped up to about 80C last I checked, so either my fans are clogged or it got significantly warmer (it's probably the latter)
I am late to the "party" but I run PPT=120, TDC=85, EDC=120. This is a slightly modified 122/82/124 that many people recommend for gaming, both work well, it just seems like my system like that 120/85/120 SLIGHTLY better, its almost within variance, I am pretty much min/maxing at this point. This is all with NO LLC, and my phase control on EXTREME for a full 16 phase power delivery, and a -30 voltage offset.
I may start playing around with manual undervolting, I did that with my 3800X where I slightly overclocked while undervolting, I got my 3800X to run 4.5Ghz all core at 1.25v, so its possible but silicon lottery plays a significant factor in capability and my 3800X seems to be a 1% binned chip.
But right now with no LLC, extreme power phase profile, a -30 offset, and a 120/85/120 I am very happy with the results. My 5800X3D idles at 42C, and with Cyberpunk 2077 on max settings at 1440P (yes I know its not 1080 so its less CPU intensive) it sits at 55C-58C. Prime95 Small FFT 30 minute testing saw a temperature of 77C vs stock where it hit 90C, that's a 13C difference under extreme testing!
you may not want to use extreme llc depending of the board.. it may increase your temps quite alot.
i always use llc at the second highest setting at the most.. you find the best setting when you lock your vcore and see in which setting it changes the least when idle to 100% usage.. less the better. / the more stable the vcore the better.
I do own a 5800X3D with a 4080 and gaming in 1440p and after many many tests my best results came with CO -25 all cores (-30 was not stable for me) and PBO limits on stock (disabled on my mobo). I played with these a lot but in the end stock values seemed to do the trick on most scenarios and it seemed snappier overall. And also something else that helped a bit was turning CPPC preferred cores option to Disabled. All the cores for this CPU can hit 4.45 GHz no swet so there is no need for having preferred cores enabled. Some people who play mostly competitive like to turn off C-states as well to improve 1% lows but that didn't have any impact for my case.
Preferred cores should be disabled since it makes problems in some games and lowers FPS. I got freezes at -28 and than i enabled Global C State and CPPC and disabled Pref. Cores. After that -30 is rockstable for me. But every system and CPU is different... PPT, EDC and TDC will lower your workstation performance in cinebench etc. Best compromise is just to leave it stock. PBO is disabled with the 58X3D. That's right light that. No OC on a 3D for reasons...
how can i test if its stable?
Install AMD Ryzen Master and run the cuve optimizer auto offset, choose per core and the let the app tell you how much values each core can handle. This will make your undervolting much easier. One or more cores might be experiencing instability at higher clocks with the offset you used and have to be clocked down, thus, the performance hit.
Yes, I used this and it worked well. I had some instability but I fixed that over time by going in and raising the lowest voltage by 1 mv each time I got bounced and running my normal routine again for a while. Gradually my system completely stabilized. Not very scientific, I know, but it worked. Running a 5800X and under the heaviest loads I can't make it hit thermal throttling (85*) and it runs at a constant 4.6-4.8 GHz (Fuma Scythe 2 air cooler in 23* ambient room).
But this does not help the instability at lower voltages which is - at least in my experience - where the pitfalls exist with the curve optimiser. You can easily select offsets and very simply test for stability at full load but there is no reliable and 100% repeatable test for stability at the lower voltages on the curve when the CPU might have next to nothing or little to do. And in those situations the relative undervolting is much higher than at the top of the voltage curve. You can really only test that using the computer "normally", as in using it with the applications you would normally use and over a period of some weeks. I wish we had the equivalent of the ASUS GPUTweak OC Scanner, but for the CPU. It would mean a stable voltage frequency curve at all voltage points rather than just at flat out.
You were getting much better 0.1% lows with the undervolt, this will make your long term experience far better compared to slightly higher or lower average FPS
It's completely fine at stock settings
I just done it in Asus MBO bios with PBO2 -30 all cores, didnt touched that oder parametres...and i got all cores to 4550mhz and lower temperature about 10 12c using DarkRock 4 cooler. Fps is higher across board.
I have a 5600x3d with the only setting in the BIOS changed being PBO set to -25 all cores and yesterday while playing rdr2 I was monitoring everything with HW monitor. And at some point I reach 4.55-4.67GHz on all cores. It's limited to 4.45 and I've never seen it above that ever. I'm going to recheck again soon and also try to see some graphs. Did it just peak there for a millisecond or did it actually hold it for a period of time? I do not know. I do have my fan curves pretty aggressive on CPU and GPU and neither went above 61c.
I've had the 5800x3d in 2 rigs for about a year now, and I loved the -30 PBO it crushed all benchmarks and stayed so cool and pinned to the max freq I thought the cpu was rock solid. but I've been dealing with micro stutters in BR type competitive games, and it was the PBO tune.
my issue seemed to be memory stability. upping DDR voltage a tad 1.36 and applying @Mohamed-zy1er ppt tdc and edc of 122 82 124 allowed my to run crash free and Studer free while gaming.
Interesting results David. I have two Ryzen 7's, An earlier build with a 5800x and one with the 5800X3D. My original 5800X I needed to under volt with PBO + Curve. (-20 offset, Wasn't stable any higher of an offset) This Setup performs excellent with Cool temps and higher sustained clock speeds. (CineBench R23: 16000/1600 Multi/Single scores) (32 gigs of Ripjaws 3600mhz CL 16. two sticks) As for the Ryzen 7 5800X3D.. I didn't bother to touch the under volting, Only because of the needed power for the extra cache supply. I'm just using a brute force air build with an incorporated Noctua D15 and Lian Li Lan Cool 216 case. Also my 3D chip has the RTX 4080. No heat issues to worry about. I did weigh in the idea of under volting but realized it ran so smooth and cool the way it was setup. Thanks for the enlightening video. Cheers!
Actually now that I think of it... I did adjust a -35mv power offset for the CPU in my Asus bios. (Not PBO though) To take the heat edge off it.
The ones you didnt see an increase on is where the power limits got in the way
I'm happy with -30 all core and 114, 75, 115, since it keeps me right under 80, where throttling starts. If you have a good cooler though, 122/82/124 are the gaming-biased settings floating around. 100/70/100 sounds very conservative.
If you keep stock and check Ryzen Master in stress testing than you will see that the 58X3D never reaches more than 115, 70, 115. It will never get to AMD stock max values of 142, 95, 140. So you can just keep it stock ;) It's not a 58X... And Zen 3 throttling starts at 90c not 80c!
@@__Pathfinder__Good to know about the AMD targets. I guess it's the -30 all core that's putting in the work for me. I should say "pulling boost" rather than throttling. My silicon lottery definitely gets tame over 80.
@@MrCthur ah ok since throttling is a automatic clocking down function from AMD to prevent overheating. This starts at 90c on every Zen 3 CPU. Temps over 80c are still perfectly fine.
I used amd ryzen master to find the optimal ppt tdc and edc which i adjusted while running cinebench in the background. on a ryzen 7 5700x my PPT =135w TDC =85A EDC 140A.
What i did next was used the scalar x10 added 200mhz on to clock speed and used -26 on all cores, i ran into some instability when the cpu was doing very little work, so i futher tuned and realised that the prefered cores, the gold star and the silver dot and cores that worked with these on low workloads needed less negative on them, so my cores now look like this core 1= minus 24, core 2= minus 18, core3= minus 26, core 4= minus 26, core 5= minus 24, core 6= minus 24, core 7= minus 18, core 8= minus 26.
on my cpu core 2 , core 7 , core 1 , core 5 , core 6 all have lower negative numbers as these cores are the most used for general useage the rest of the cores only come into use under heavy workloads and so have a higher negative number. when the popular most used cores are in full useage it doesnt really matter how much negative you put on the least used cores as most of the cpu is powered up by the time these cores are used, if you put too much negative on the most used cores these cores work on there own and so therefore need more voltage to handle smaller more used tasks.
Thank you. I have the same cpu, I tried your initial settings. Seems to be stable for me. Had to drop my first 2 cores to 25 and left the rest on 30. My temps are higher than before while running cinebench but I’ll keep adjusting until I get a nice balance.
You dont need to power limit (PPT/EDC), you can use your voltagelimit to do that if you want it to actually run cooler, more on that later. Because in reality, PBO tuning is technically not an undervolt. The voltage is determined by the power draw, temperatures and other factors, and the CPU will then pick the corresponding frequency associated with that voltage via its internal curve. PBO tuning or curve optimizing, shifts the frequencies associated by the voltage. So, at the given voltage based on power draw yada yada, it picks a higher frequency. Therefor, any claims of lowered temps and power draw is untrue, since it still uses the same voltage and package power, it just clocks higher.
At -20 you already hit 4450 all-core in gaming workloads, so going from -20 to -30 does nothing. But, you can still utilize the extra range in another way. What you can do, is combine a NEGATIVE OFFSET on your vcore with PBO. Something like -50 to -100mvtt. That will drop the voltage, but also the frequency. You can use the additional range from -20 to -30 to curve optimize as far as you can go, to not have clocks drop below your target.
I ended up with something like -75mvtt negative offset, and -25 on PBO tuner, to hit 4450 in gaming workloads. It still goes to 4425 in blender or cinebench, but i dont care, i dont do professional workloads with the CPU. That will, unlike only PBO tuning, drop the temperatures and power draw. Also, using the voltage limit like this means it'll not bounce off the power target and drop clocks. Which you're not verifying on-screen. You need to actually hit 4450, otherwise you're just throttling it. No wonder its worse.
PBO can't be used with a 3D and is therefore always disabled. No OC with a 3D for reaons! PPT, TDC EDC is not needed since the 3D goes only to 115, 70, 115 und stress testing (Ryzen Master). It never comes close to AMD stock limits of 142, 95, 140. So no tuning needed. Only Curve Optimizer to -25 or -30. 3D is not like a regular Zen 3. You can't do classic UV and OC. So just CO to -30 and CPPC preffered Cores off. That's all what's meaningful...
@@__Pathfinder__ Using HWInfo with default TDC, EDC and PPT settings when running CB R23 it shows max TDC ~71-72A, EDC ~119-120A, PPT ~105-106A. So you are technically correct. You can limit to 70,100,100 but gains/losses will be marginal at the cost of possible instability.
@@Mopantsu And no need for that. Curve Optimizer as low as possible/stable is more than enough for this CPU. These limiting settings are more a thing for the regular Zen 3.
@@__Pathfinder__ I did see additional cooling while maintaining performance when using PPT/TDC/EDC changes. I run 120/85/120, the EDC being at 120 for me matches my motherboards max capability of 120A to the CPU/SOC as its a dual 60A on the X570 ROG Crosshair VIII. I slightly dropped the PPT and increased the TDC to 85W which is 10 less than stock, and then the 120A as explained before.
I feel like the 140A on the EDC at stock settings was just so the CPU could take advantage of the uber high end boards like the MEG Godlike, AORUS Xtreme, and the Dark Hero to name a few. I think there are a total of 5 X570 boards that have 70A - 90A mosfets that can actually pull the 140A EDC limitation, but its not needed unless they start adding in 5950X3D chips.
As someone who recently bought a entry level system with an 5600 cpu and rx6750xt 12gb gpu I'm curious about what your final take is on whether the 5800x3d was a worthwhile upgrade. I was thinking that maybe in a year or two it might make be a quick (and hopefully cheap by then) upgrade option for me without the need for a major system overhaul.
You have an MSI motherboard so try kombo strike 3 setting. That's how I run my 5800x3d. I've heard that's the same as curve optimizer -30 all core, but kombo strike 3 seems to run cooler than CO at the same performance. I think your PBO settings are too restrictive, hurting performance under a ray tracing load but that's a guess.
Yeah I think you’re right. I’m now running at just -30 and that’s it. I’ll retest but not sure it warrants a new video
Huge improvements at -30 for my 5800x3d. I get on average 10+ degrees less and 10-20 less watts of power usage. Originally my 5800x3d was peaking at aound 80c in a game like cyberpunk 2077, but now it just hovers around 70c, watts peaked at 105, and now it peaks at 95. I'm seeing a lot less down clocking too, because of the temps.
Great video, good quality video and audio. Maybe try out different levels of negative values for curve optimizer. Also keep an eye on the clock speed, as certain games like csgo perform better with higher clocks. Good luck mate with your channel, just ignore people who indulge in pure hate.
Thanks for the kind words! Don’t worry, I’ve got broad shoulders! 😅
Honestly, I'd rather look at those 1% and 0.1% lows for this undervolt, while undervolting didn't change much for the AVG FPS, it certainly helped keep the FPS much closer to the average, which says a lot for frame skipping, means we would see little to no frame skips, where keeping it stock meant you'll notice those frame drops
I must say I have a fractal torrent case and added a 3 pin 120mm exhaust fan behind the cpu too. The 5800x3D idles at around 45°C in this hot summer, window open. While heavy gaming it's around 70 degrees, heavy cpu benchmarks at 89 if I remember correctly, never throttles with a peerless Assassine 120 SE.
So far I saw no reason to UV it in terms of temps or performance.
I just leave mine on stock. Running with an AIO which keeps it 70 degrees or below on average.
Same
Guys, should I upgrade to 5800X3D from a 5800X on 49" 5120x1440p @ 165Hz monitor with 7900 XT ?
1440p SuperUltraWide is closer to 4K and it's a known fact that the "3D upgrade" is NOT worth at that resolution ? 🤔
si vous jouer uniquement en 5120x1440P avec une 7900XT cela aura pu d écart avec un 5800X seule la 1080P montrera une différence visible dans votre résolution pas d interet si ce n est avoir le meilleur CPU pour l AM4 pour étre paré pour le jeu si vous avez besoin polyvalence le 5800X est mieux
For my 5700X3D I didn't really see the point of undervolting. Sure, in Cinebench it clocked from the 3.9 - 3.95GHz range at the standard voltage up to 4.05GHz constant after undervolting. However, when gaming it would clock up to it's max 4.05GHz anyway at standard voltage. Because it's not as heavy CPU load as stuff like Cinebench. And I don't use my PC to run Cinebench all the time :P I use it for regular browsing, youtube, email and gaming. So basically my cooler is good enough to keep it at max boost at default voltage when gaming. It may make a bit more sense with the 5800X3D if you don't get it's full boost when gaming with it's default voltage. Like if it clocks down to 4.25 - 4.35GHz when gaming. Undervolting it will make it run it's full 4.45GHz boost when gaming.
Mine went straight to -30 all core - thank you, hard-capped boost😁
(I inched my thermal and power limits down until my CB23 scores started to drop, then left them there.)
I have an ASRock B450M Pro4 and plan to buy the R7 5800X3D. I would like to do what you suggest, but I don't know how to do it. Could you provide a video that explains the process in detail? Thank you.
Enabled PBO2 and CO -30 all cores with ALF III 360 Black using the AM4 offset brackets idles at 28c never goes over 64c in gaming stays at 4.554ghz.
I don't know why you don't know this but showing clock speed would be crucial to this kind of testing.
2:53 .. why is VRam considerably higher on the stock compared to the undervolted BM when its the same scene?
good question!
My answer is to low ppt, edc and tdc + to aggressive undervolt
This -30 shit is so awful
After several hours of testing like 10 hours everything were nice
And then mwz in the red zone, massive hoarders of zombies, bosses, the weather effect got me an bsod
Cause ram issue
After enabling xmp2 instead of 1 and changing ppt, edc and tdc its working perfectly fine
I switched from 3700x to 5600x just for heat now I want a 5800x3d but I don’t know if I want that heat 5600x runs cool
My 5800x3d runs 88 80 120 limit with all core -30 and seems to be completely stable for about a half a year. Before it had 88 65 90 limit + all core -30 and I was having stuttering a bit. Also I have cppc and svm mode disabled
As nice as undervolting the 5800X3D in a typical desktop scenario is, where it really shines is in SFF builds. The 5800X3D is an absolute beach to cool in SFF, not just because of the incredibly inefficient heat transfer from the CCD to the IHS due to the V-cache chip in-between, but also because you're limited by your cooling capacity.
My case is a 7.5L i100 Pro/Hummingbird, so I don''t even have the luxury of an AIO - 60mm low-profile is the best I could get. So I got an ID-Cooling IS55, one of the best in the business, and put on some Conductonaut liquid metal to give me the best possible results. CPU fan speed was set to max at all times, though it is damn near silent. All that for 60+ on idle and a constant 91 degrees on load with inconsistent clocks. Just turning the PC on and loading into Windows caused the CPU to exceed 80 degrees.
It would hit 4.55GHz, but rarely as spikes and mostly hover around 4.3GHz and often as low as 3.7GHz momentarily. All core clocks in Cinebench 2024 would settle at around 4.0GHz after 10 minutes, also fixed at 91 degrees. At any rate, the fact that I could still boost in this situation meant that my cooling was just barely not enough, especially because I noticed the clocks being a little more consistent just by having the A/C running.
So I went into the BIOS (Asus X570i) and set a basic -0.2v offset, and left the previously set performance enhancer to "level 3 (OC)" as it was. Temps on load are now on average around mid 70s, with occasional peaks at 85-ish. Idle temps now sit below 50, sometimes below 40 with the A/C running. That alone would've been a great result already, but due to the massive improvements in temperature the clocks greatly improved too. Cinebench all core load now sits at 4.3GHz constant at 86 degrees max, and when gaming it hits 4.55GHz quite often, usually 4.45GHz and rarely dropping below 4.3GHz.
I could probably do better with some more in-depth tuning, but I don't really need it. Most of my concern was regarding the high temperatures, so any performance improvement I got out of it was a bonus. I don't even have a sufficiently fast enough GPU to hit a CPU bottleneck, even though I play at 1080p on a RX 6800, so this'll do for now.
Did you do pbo tuning also? I have same cooler with this cpu in a very small pc case.
Benching a CPU without showing CPU clock is kinda ...wrong. We have no clue whats going on with the clock speed. Is it fixed or fluctuating? Does is drop at some scenarios or not?
For the performance it gives, it’s pretty efficient. Especially compared to the intel cpus that released around it.
can you use kombo strike with gameboost and pbo bios setting ?or can you use all 3 at once or is using only 1 or 2 options best? right now i think I have kombo on 3 ,and PBO auto, game boost off set.installed my chip last night? TY .on a MSI X570 gaming plus. I do know how to manually tune PBO
Just listen professional tip. Don't change PPT TDC or EDC, Because once you change one it will bottle neck the other 2. For the best result choose PBO optimizer -30! Because the curve optimizer in bios won't have same result as the pbo optimizer which has slightly better cooler temperatures and and clock high 4450mhz! You just need to set task scheduler that the pbo tuner will set with every reboot.
Proof?
If you do like a 102-103 bclk overclock, it'll probably still work at stock voltage without breaking PCI-E or IO ports or the ram, that'd be probably a better choice than undervolting with a typical large tower cooler as these single ccd ryzen cpu's don't really pull much power therefore they're easy to cool. Also I think your cyberpunk experience is worse because of ryzen cpu's clock stretch a lot, I have a ryzen 7600, I got it overclocked at 5.45 ghz stable at stock voltage, however when I increase the clock speed to 5.5, y cruncher doesn't crash, it just prompts a lower score, the reason why is that when ryzen cpu's don't have enough voltage by like a tiny bit like probably 0.01 mv or something, it'll drop the frequency or do something to drop the benchmark score. It runs 2 hours stress test with Linpack with no memory errors, but it also drops the tflop numbers which I see in the linpack. So this cpu at this voltage is not stable at 5.5 ghz, It'll probably go away if I give it like 10-20 mv or something. So even if your Ryzen cpu doesn't crash, it can drop the clock speed down the base frequency for like 0.00001 sec quite frequently and your monitoring application may not even register it.
If you're getting more or less the same performance for significantly less power then I'd call that a win, no? Also as you say, the 0.1% lows were much better in a lot of your tests, which is arguably more important than getting an extra 10-20fps (especially when you're already well over 100fps!). That translates to fewer "stutters" and thus a smoother gameplay experience overall.
I see people suggesting using Ryzen Master to help optimise settings. I thought Ryzen Master could not be used with the 5800X3D?
Yes you can’t use ryzen master for undervolting 5800x3D
@@RANMA212 Personally i prefer using the BIOS, not a fan of using software within Windows to make hardware changes.
@@Dazzxp yes yes me to i prefer undervolt by the bios
so what would be good for b550m ds3h mobo i just got the 5800x3d and need to undervolt what would be good im -30 on all cores what else should i do im running 120 75 110 i dont see any difference please does anyone have any suggestions?
I use -30 and 130 85 125 for limits and I hit 4,430 mhz in cinebench.
-25 all core 120 -75-110 (PPT TDC EDC) --- with this and Arctic freezer 3 360 i have max clock boost in cs2 all the time withhin 50-61* celsius ,
Clock stretching is a thing when undervolting. Never forget.
yo i tried undervolting and it's SOO good woow way less heat and more boost
Run stock disable PBO. set -26 to 30
I just uses a 3 Fan All in One which allows CPU to run cooler and don't void any warranty.
Hey if it works, great!
undervolting does not void warranty
If you slap af Artic freezer III on that there cpu. You would cool it even better,
Mine is rock solid at -30.
.01% really matters and that's huge gains from undervolting
No such options in my bios....
Im getting restarts in games demanding games actually with 3600mhz ram i have the same as you 32 gb and i searched that ryzen 7 5800x3d max speed of ram is 3200 how make that work please help me thanks!
My PC wouldn't boot with RAM @3600 XMP profile, I fixed it by:
1. Manually setting DRAM voltage in bios @1.35V (like my XMP profile) and then in windows I opened Ryzen Master application, and manually put the timings of my XMP profile (you can skip this if your PC gets to Windows without crashing with your XMP/DOCP profile selected in BIOS)
2. In Windows Ryzen Master - Increase VDDCR SOC from 1.0 to 1.1V (+0.1)
3. In Windows Ryzen Master - Increase CLDO VDDP, CLDO VDDG CCD, CLDO VDDG IOD from 0.9V to 0.95V (+0.05)
4. Save and then apply profile
You just have to test for stability, but it should do the trick.
odd... I use 3600mhz ram as well and I run it at 3600mhz using MSI's Memory Try IT set at 3600 18-20-20-42 no issues but disabled XMP.... I will try going 16-19-19-39 and see if I get even faster performance... my ram stock timing is 18-22-22-46 (I think).
One last question what setting in the bios did you use for CPU ratio and voltage? Auto, manual, what?
I like tests like this which improve gaming but I don't think you need to undervolt, especially since the cpu is great out of the box or unexperienced. If you cap frames, your cpu will work less and never get hot. Im running my 5800X3D at stock settings so it can decide itself how much power to use or not. The hottest ive seen it get is 65 degrees in games. In Cinebench it was around 77 maximum. Air cooled with Dark Rock 4
Actually undervolt this processor makes a lot of sense because the performance is increasing and the temperature are lower and fans are more quiet.
@@michealmeyers9789 I agree but it is not always guaranteed for everyone and every game. You never know how a game will react to the cpu. some games will require the cpu to use all its strength to perform properly
Your results were lower than they should be because the power was set too low. Try the undervolt and DON'T touch the power limits leave them default then see the difference......
Appreciate the tip, thanks
@@davidnott_ I tried the power limit thing with the negative offset curve on a Asus B450-f gaming motherboard and what i found it that for some reason my chip degrades in performance if i lower it on one of the 3 power settings even if it's just a small amount that i change it. I think it was the TDC. When i tried it again with just a negative offset of -30 my benchmark scores went up and the chip stays locked at 4450mhz all core and about 10c cooler. I'm sure reducing the power limits is just another method of trying to do what the undervolt is already doing so doing both together maybe overdoing it depending on the chip/motherboard. Some people seem to get away with doing both. Another thing i thought about is maybe it's a BIOS issue? If there is a newer version of the BIOS maybe it will work better.
Yrs use power limits only if you really require lower temperatures or lower consumption otherwise don't touch the power limits at all
And have fun with this beautiful peace of engineering!
Not worth undervolting in my case, just got worse performance overall. Just CO -30 and you're set.
I got better results with PBO neg-30, PPT-122, TDC-82, EDC-124
Honestly, all your results are likely just a test to test variance.
Reasons being:
5800X3D is a locked down chip with temperature limits. (not factor in your case)
PPT, EDC, TDC on this chip are an issue because they can't be increased, lowering them just means you're creating even tighter power limit and reducing the max performance even more.
In your case the limits you've set are not a factor because this CPU does not even reach the limits you've set in gaming loads.
You would need to run something like CBR23 that maxes the CPU power to see the difference.
For your testing the only benefits there are the lower temps from CO being at -30.
That said CO of -30 is rarely fully stable especially with default LLC settings and while you may have not encountered crashes, it may have resulted into clock stretching and thus even higher test to test variance.
Appreciate the comment, interesting, thanks!
Ok now how to under volt properly 😅
Open to any suggestions that may improve things
I will never understand those plants on PC videos xD
Haha 😜 gotta spruce up my shed somehow
kombo strike on msi platform
David Nott : AMD is known for their advanced capability to configure their CPU-s very very efficiently and very very optimal. But again, WTF are you talking about with the lowering of voltage and thus the power consumption. I have a Ryzen 7 5800X3D and @stock frequencies with PBO enabled and Turboing over all cores enabled and this CPU physically cannot get over 111 Watts of total power consumption as a socketed CPU alone !!!! Why is there any need to lower the power consumption since this is a VERY VERY LOW POWER CONSUMPTION for a 8c/16t 2.nd fastest gaming CPU on the planet ?!?!?!?!
The temps also physically cannot get over 80°C in my usage when I have 150+ tabs opened in my browser + playing RDR 2 + decompressing 90 GB large pack of files at the same time. Or don't go over 84°C when stressing it for hours in Prime95 or FurMarks CPU burner test.
I've also tested the Ryzen 7 7800X3D when building my friends RIG....and at the same condition, cooled by the same Thermalright Phantom Spirit 120 SE cooler, physically cannot get over 83 Watts of total power consumption as a socketed CPU alone. And the 7800X3D is the No.1 fastest gaming CPU....performance wise shitting all over the i9 14900K in every game, especially if coupled with the RTX 4080, RTX 4080 Super or RTX 4090.
k
That's bullshit. Run Prime 95 with small FFTs, the CPU will EASILY get to 130W. You don't seem to know wth you're talking about man...
@@Blaze1984HH I did exactly that because I know of the Small FFT-s in Prime95. Just as I know of the Small Data Set in OCCT and just as I know of FurMarks CPU Burner, Intel CPU Burner and the continuous run of CB20 and CB23. After around 15 hours of continuous benchmarking my Ryzen 7 5800X3D I recorded the highest peak PWR consumption of 111 Watts in HWInfo and via calculater current clamps with my osciloscope :).
So buttercup, you don't know squat about any AMD cpu-s. In contrary to Shintel, AMD NEVER LIES about their PWR consumption in CPU and GPU department :). Those are provable facts by me or any unbiased normal user anytime and anywhere :).
If you're not an overclocker than just go back to stock. Chip has been out for quite a while. If viewers have questions, the answers are already out there on the internet....
No
Only thing worth doing is CO, PPT/TDC/EDC crap is not worth it (it is only suitable for chips like 5800x that runs very hot)
Hie five to cat 🖐️😊
🫡
Why stick with a 5800X3D when you can easily sell it for $250, the board for $100 and the ram for $50. That's what you can get a 7600x/7700x/b650/32gb CL30 6000MT ram for. The value still in these old parts can still pay for much of the upgrade. I don't know why people would hold onto them when it's dirt cheap to get basically the same performance sometimes more sometimes less with the 7600x/7700x. I had a used 5800X3D for a month that I paid $263 for. It was certainly better than my 5900X which I sold for $260. Then a few weeks later I realized the 7600X on AM5 was on average 3% faster. Well I sold the 5800X3D & Asus TUF X570 for $380 and the 64gb ram for about $100. I went to a new 7600X for $200, open box X670 Aorus for $130, $110 new Trident CL30 6000MT. Then that trickled down to every AM4 system inside the house. When below retail AM5 deals, boards and ram appear I would part out AM4 parts for usually not much more than what I sold the parts for. I have two 5800X, 1 X570S Tomahawk and 96gb of DDR4 left to sell off. AM4 selling is slowing down, 3 months ago this stuff sold fast. Now it's much slower.
because AM4 Platform is still solid and you're stupid that's all folks 😂
All that for 3 more frames. No thanks 🙏
All that for better temperatures? Yes, thanks.
5-10 degree less temperature. Huge gain.
Yeah you're not aware of all the benefits so you shouldn't do it... but comment confidently about what you dont know about if you want to seem silly
@@DuXQaK with my 7800X3D, there’s really no difference, only a few degrees cooler but thanks 🙏 for reading my comment
@@Kapono5150 If the CPU is cooler, the ambient air temperature will be cooler so motherboard, ram, drives, PSU etc will be cooler and so will be the GPU too. When running cooler, system stability is improved and device life is maximized. CPUs and GPUs will achieve better boost clocks, higher and for longer. In the heat of the middle of summer for example is when the gains will be even more impressive and important. Hey even the room will even be cooler. It's basically a win, win, win, win, win, ..., ... for not much effort and zero money. Satisfying too... nerd juice 🧃