Just an fyi, but before touching load line calibration check for stability of the curve offset first. If you don’t need to touch the load line, don’t. It still increases voltage under load compared to auto which leads to higher temps. This is just to make certain offsets stable when needed.
Thank you for this. Informative without going too far down the rabbit hole and losing the point. I see lots of undervolt videos and nothing about the load line. It helped me get more boost and cooler cooler temps. Can't beat that.
Came from Intel 6700k. I got that to 5ghz with 1.45 volts. Amd 7900x3d now and it’s so different. Helpful guide! I’m close to world record on 3d mark for a 3080 and 7900x3d set up (says “record for this exact combo of 3080 and 7900x3d” so I’m chasing that haha.
5ghz on a 6700k vs a modern cpu isn’t even comparable lol. Also the 7900x3d is a weird cpu where you have the voltage limitations of the 3d cpus and reduced benefit of 3d vcache since it’s only on half the cpu
@@josephbargo5024 obviously not comparable, but considered that majority of people could not hit 5ghz was pretty crazy. The boost stuff with amd is weird but I found not worth it really to mess with. I didn’t get much gains from anything or any guide I’ve tried, and I’ve tried a lot.
Thank you for that. Without the extra step that you showed I could only cap out at -33. I could do higher for Cinebench but it failed on stability tests. Once I adjusted the LLC I was able to get it to pass OCCT running for an hour at -40 and capped out at 75C. it should be noted that the LLC does increase temps the higher you go but I was still able to keep everything at a reasonable temp with a 280mm AIO.
My maaan! I've got the gigabyte Aorus x670 Elite AX REV 1.2 board and the bios is the same as yours. Great explaining what this negative CO does even though I already knew from hours and hours of reddit threads =). I came here to see if I could learn something new that's all! I'm a happy subscriber and hope you can bring more nice content soon for my 7800X3D.
@@beardoguyyes of course it got the bclk multipliers and the voltage settings. I haven't made any successful boots with increasing the bclk even to 102.if I had the Asus main board there is modes like eclk 1 and 2 and also asosynchronos mode. My board doesn't have any of those. Those high end Asus board cost like 700 bucks. Mine was around 200, 250 bucks. U get what u pay for. I'm currently happy with the ram tunings I got it to run at CL 26 even though it's cl30 rated. Hynix A die 😅
Damn. I wish I would have seen this video when it dropped. I’ve been researching and tinkering on my 7800X3D and not a single person has mentioned anything about LLC. Here I was, happy with my -15 offset. I’ve just set my SOC to the standard 1.250 which is perfectly reasonable for my cpu. But now I’m gonna have to go and tinker a little more. Thanks for the awesome tips man! Never seen anything else from your channel, but I’ll sub!
It's likely its a core or two holding you back from going lower. My 7800X3D only does -10 all core, but I just put some time into doing a per core CO and I can do like -40 on 6/8 cores.
@@billedwardz Whelp, I upgraded my cooler to a new 280mm AIO. Currently running -30 all core. Passed every stress test and stability test I've thrown at it over the last 2 days. No crashes yet haha. It's just odd, because I literally haven't changed any other settings. No clue why, but I'll take it!
@@billedwardz R23 and the entire OCCT suite. Threw in a few games and of course some light load with chrome browsing etc. But! After a few days of running great, the Small/Extreme CPU test in OCCT caused a crash haha so I’m gonna sit at -25 offset for a bit. Tested everything there and it was fine. I’ll run more tests as the days go by. But all in all, I’m really happy with the Chip.
@@ChrisLaneification Yeah I have a similar story which is why I asked, I initially was using y-cruncher, which apparently is a more moderate AVX2 or AVX512 load, which got me to -40 on most cores. a few days afterwards I loaded up OCCT and used Small/Extreme/AVX2 and had to lower everything to -30. Honestly we both could probably stay with our initial COs and be fine because IMO OCCTs stress test is probably harder on this chip than anything it would encounter during regular use.
I was like you back in the mid 20's. I manually enter all my DRAM timings now. Notice how I called it DRAM instead of RAM? I also correct people when they count DRAM speed in Hertz instead of transfers per second. Wanna buy that awesome ROG whatever mobo? Too bad it has throughhole DIMM sockets. Wanna know why that is bad? Trust me you don't. The fact that I can't bring myself to call a DIMM socket a RAM socket like all the normal human beings is reason enough for any of you to steer clear of any overclocking whatsoever. It all started when I first overclocked my Pentium 3 733 50 800 MHz back in the year 2000. The fact that I kind of know physics and material science makes all this unimaginably worse. There is still hope for you. Cease all overclocking immediately. This includes using the curve optimizer because, well you really don't wanna know why.
Thank you for this video man! No nonsense talk. No bullsh**t, Simply good and nice explanation. Through this I finally found the LLC on my Aorus X670. I put my 7800X3D with -45 and Extreme LLC up to 4900Mhz. Lets see how far I can go. Thank you buddy.
You probably don't have to worry about it because you're undervolting, but the issue with setting a stiffer LLC is it can cause voltage spikes on transient loads. Youre fully loaded, the LLC is compensating, there's a cache or memory stall which unloads your CPU but because you're running extra vdroop compensation your voltage spikes until the power stages catch up or the CPU loads again. Finding the stiffest value that doesn't cause this requires hooking up an oscilloscope to the voltage probe points on your motherboard, as it varies by model/manufacturer/bios revision.
I've been building pcs for fun for over 20 years. I retired early ,age39 (hey Fu I'm still young at heart) lol but 1) I always buy nvidia gpus And 2) up until last year always purchased Intel cpus. Last year I built 3 pcs for myself. Total overkill, one was for my virtualpinball build I built,the other was for my 32inch rec room masters arcade cabinet build. Both pcs are identical, msi mpg x570 gaming edge wifi motherboards, Ryzen9 5900x cpus, Nzxt 360mm aio coolers , both in corsair 5000D airflow cases. The only difference was (remember this was during the gpu drought) my virtual pinball machine has a rtx3090 founders edition, and my arcade has a rtx3080ti. I fell in love with amd cpus specifically that ryzen 9 5900x 24threads @4.7ghz was no problem. Ok sorry for the backstop but if everyone here is into pcs and tuning them etc, I love seeing peoples builds and hearing them talk about their systems just as car enthusiasts have car shows. Anyway, I went with the gigabyte Aorus X670E Master motherboard (thank God so nothing exploded) ryzen 9 7950X3D, grabbed the only pcie5 m.2 drive I could find (microcenter Inland 2tb pcie5 m.2 drive for only $179!!, and yes it gets read and write speeds over 10,000gb/s) It took lots of updating drivers and bios's but finally the cpu is running as intended. I think I just set my negative to -20 and left it. Right now I'm using the motherboard above, so hoping some of you are familiar with the layout. 1) I'm unsure but are you saying the graph you show , if its still on a negative curve , even on extreme, it's safe to leave it like that? 2) under where we turn pbo on to advanced, then motherboard, there's a scalar option that's on auto but goes 1-10. Should we just leave that alone? Lastly, I want to apologize for all the questions, if you got to know me you would know I try and help every person I can for arcade builds, pcs etc. But I'm a grown man and can admit, all my years of experience isn't really helping with this ryzen9 7950x3d. It idles around 40°C and usually games or benchmarks around 60-65°c. Does that seem like normal temperatures? My cpu cooler is once again the NZXT 360MM kraken. I appreciate anyone whose read this, I have a disability after almost being killed at work so I'm still trying to fight my way back mentally. As I always do, I like and subscribe to any fellow tech youtuber, even though I'm new at it and onky have 1500members lol 😅
Thank you guys for taking the time to read my long ramblings lol I've loved building pcs since my first 386sx 25mhz pc build. I think it had windows 3.1? Or windows for workgroups whatever was before windows 95. Lol it basically could only run the paint app. Then I went to a computer show , some of you who are in your 40's like me remember having to go to computer shows to buy our hardware before the internet. Lol scored a intel 486DX 66MHZ CPU, WITH 4MB OF RAM. LOL God anyone remember having to run memmaker? Because of the way we had to run games in DOS? Lol
Coupled with the other replies, you must remember that the x3d is different than the x, which you normally wouldn't set to the highest setting. The x3d voltage won't go above its limit. LLC only helps the vdroop remain more stable.
After you pass a prime95 run of about 6-8h you can say the settings are good. I can run cinebench with even -45 on my 7800x3d but man...the second I press start in prime...boom, crash. After that I started to test each core in OCCT and found out that, 7 of the 8 cores can run at between -21 and -25 and one core is absolute "garbage" and can only run at -13. Yes, curve optimizer is the way to go to lower the temperatures but the architecture on the am5 platform, especially on the 7800x3d is a bit weird, you cannot boost that high the clocks and needs at least a few days of testing to say that your system is 100% stable. After doing all these tweaks I then tested in prime95 and OCCT for about 12h each (the average time my pc is open) and had 0 errors. I saw someone said that picking individual offsets for each core can lower your performance...that is 100% false. You gain performance by doing so, cause each core can boost higher depending on the offset. If you can make 7 cores to boost at a bit over 5ghz and one stays below 5, will always bee better than having all cores limited to 4.8.
The 7000 series x3D chips are technically locked. The 7800x3D will not boost past its top speed of 5050hz without some infinity fabric or multiplier tweaking as the boost overide settings do nothing. It's online but I'm not sure if it is prime stable.
The guy who said you lose performance by applying per core offsets was Michael, one of the first comments. He is greatly mistaken and is applying trying to set per core FREQUENCY in CCD/CCX (not per core CO), which in that case can't be done as it defaults to the lowest core's speed. However, (as you and others have mentioned) per core CO is completely different as it adjusts the voltage offset (not frequency) and it works great . Just a thought on stability testing: Some of the stability tests you mentioned are unrealistic compared to what most people will actually stress their systems in real-world cases. I tend to stress for my worst-case real usage. You could be losing a lot of performance if you stress test for something way outside of your worst-case real-world usage scenario... just something to consider.
@@jimmyhuang7481 You are correct, some of the higher-end mobos, such as the Asus x670e Hero, have a separate bClK adjustment. You set eCLK to asynchronous mode, bCLK1 stays at 100, and bCLK2 you would adjust for cpu frequency (to like 106 or 108.15, etc). Employing that function/strategy/technique along with others can achieve some great overclocks on a processor that is otherwise "locked". I've had a lot of fun with the 7800x3d :)
@@RKBenchmarker i always stress test for worst case scenario, but as you said that is unrealistic for most cases, as I usually game on my pc, for example forza5 or cod can run with no problems, cyberpunk crashes my system the second I play start. Sometimes even the basic youtube videos crashed my pc. Since then i try to find the perfect balance between 100% stability and usual workload/gaming stability. But you cannot go wrong with 100% stability, even at a loss of 2-3 fps.
@@Focaz With your knowledge, you and others could easily set different bios profiles for a set of different games/scenarios if you wanted to take the time. Just another option in the long line of options :) Game hard my friend!
Mine can only do -20 negative offset. Your one must be a super golden sample. But, I made sure mine is stable on every stress test and game test possible
@@AverageAsianJoe Those are the weakest cores, so offsetting those isn't going to reduce temps all that much. In fact, the gaming cores have less room for offset than the other cores. The main reason why the 7950X3D is lower power than a 7950X is because of those gaming cores being lower clocked and temperature limited by the 3DVCACHE, which heats things up. So the main problem is the weaker cores run hot, and can't be undervolted as much, while the stronger cores run cooler at the same clocks but also reach higher clocks. The problem with LLC though is while it stops Vdroop, the opposite side of that coin is you get voltage spikes when load drops, which is actually worse for your CPU in the long run, because you're getting a lot of voltage pumped through a core that is essentially idle, which is far worse, because you're not doing anything with that energy, it's just scouring through the circuits with nowhere to go but to dissipate into heat, which means you might not get higher overall temps, but you will get very high temps over a smaller area of the chip that the sensors won't pick up. So it's like if a small part of your cores suddenly went up to 100c, but in such a small area that as it dissipates through the chip, by the time it reaches the sensor it's only reading at 70c or less. This means very sharp spikes in voltage/energy, which is worse for electronics than a gradual shift. Temps aren't what kills hardware, it's voltage over time, and voltage spikes. The sudden shifts in voltage results in more dramatic energetic activities in the materials of the chip. This is one reason why the IHS is soo thick in the first place, it's to keep different areas of the chip from becoming too different in temperature, otherwise those temperature differences at such small scales results in more stress on the materials, including physical cracking and delamination from each other. That's the main reason why the VCACHE is limited in temp. Not because it can't run at those temps, but because it would crack/delaminate from the CPU, because it's not so much built into the CPU, but rather, pasted on top of it and connected by thin circuits. If it was actually manufactured with the chip itself, rather than manufactured separately, the temp limits wouldn't even exist.
just for information. running llc on extreme will generate extreme voltage spikes for a short time when the load drops off. this can lead to huge voltage numbers for a short period of time. i dont recommend running llc on extreme!
@@Trampus10-4 make a video how to make our 7950x3d perform the same as 7800x3d for gaming I mean how does a $360 cpu beat the mighty 7950x3d 650$ usd CPU in gaming Make a video for us 7950x3d users , i want the same performance as 7800x3d is giving
I didn't win the Silicon Lottery and left it on auto and got better performance. So placed my NVME Drive to the middle of the MB and got PCIE4 instead of PCIE3 and got better performance but ran hott. My graphics card slowed down so I placed it back and the NVME is just a little slower but the GPU is at full speed. Cheers!
What wasnt mentioned is that you can also get a mobo that has an external clock generator and overclock the cpu bus. Something to consider additionally is that extreme LLC will put strain on the VRM. Depending on your board and your cooling scenario this can affect the life of your motherboard. Air cooling is actually better for cooling the VRM.
Overclocking the FSB is just an all around bad idea because now you are talking about OC’ing multiple components within the system………thereby increasing points of failure.
@@arlowicks9359 Yes……you are correct, the fsb can be overclocked by itself without affecting other components……yet……… it is still a bad idea and probably not worth the effort for the minor gains and eventual instability. What’s stable today probably won’t be stable in a few days.
@@techluvin7691 Whether it is a good idea or not, I'd actually have to try it to know. It worked out really well on intel sky lake providing massive gains on locked processors. This is different as the I believe the x3d processors are running closer to their limits than the locked intel chips.. As for eventual instability, thats nonsense... Like I said, maybe good, maybe bad.
CPU soc voltage is very important, not just something people are talking about these days lolllllll and for x3d parts, i would seriously consider looking into the purpose of vdroop. Cool overview though. Also worth disabling integrated graphics, actually makes a diff believe it or not
This makes sense, but then Bullzoid suggests that LLC is generally a bad thing, because of the spikes it causes. Confusing... Couldn't you just reduce the CO offset and get the same VCORE result as raising the LLC, but with less transient spikes?
whoop whoop whoop 😏 thanks bro I follow your guide for my 7900X , I still stable at negative 45 with extreme level calibration , also I just add in my setting temps to reach until 85 in my Aorus Master X670E , got on cinebench23 a score of 29700 , and my volts do not pass the 1.20 level , I monitoring my voltage in multitask heavy application or sleep mode doing nothing and again my volts dont even reach to 1.20v
Yes, most CPUs won't be able to do -40 however, LLC prevents the significant vdroop under full load that would be incurred from such as high negative offset. At the same time, the negative CO offset prevents the higher voltage spike that would normally occur after transitioning from full load to low load or idle; due to using LLC.
Wait..... But if you're reducing the voltage per clock under a specific load..... Then raising the LLC..... That means that the voltage is higher at a specific load so you're undoing the benefits you just got from Curve optimiser. This does not make sense to me
CO undervolts the old 5Ghz. But now at same voltage you get 5.2Ghz. This is not stable at the same voltage so you need overvoltage, since the CPU is locked, LLC is overvoltage workaround.
Negative 15 BSOD's me. However, my 7800x3d did win on the FLCK lottery and im able to run a 6400mhz memmory kit with 2133mhz on the infinity fabric. So the lottery is not speficly bound to the curve optimiser.
I used 35 for the undervolt and scored 37103 on cinebench R23, however i hit a high temp of 85.9 degrees. I see you hit over 90 degrees. No thermal throttling for me, i changed my thermal throttle to 90 degrees anyway, but im just ssying. I could probably go negative 40 but its gonna put my temps over 90 degrees. Im using a 360mm nzxt kraken AIO cooler which is really good, 85.9 is the hottest ive ever seen it get
I just switched to these AMD CPUs and clearly I need to read up and learn. Because this video to me seems like there's a single attribute (core voltage) that is controlled by two settings, curve and LLC. And that we can lower core voltage more if at the same time we raise it using LLC. I must have misunderstood something.
I have 7950x3D and I am with MB X670E-PLUS with the "Negative 20" curve Optimizer my PC restarted 2 times in a matter of 3hrs ish maybe? I also Set the Thermal Throttling to be at 75* Cel. Would that mean that my CPU doesn't tolerate even -20 o the curve settings?
This was a great great video!!! Now Ryzen directly communicates withe memory controller for the ram; will setting the soc LLC make a difference in the ram or cpu?? Great fookin video mate!!
I have the Terra case (mini ITX) I was struggling with finding a fan that could cool my 7800x3d. I have set my offset to negative 35 to reduce temps which is great . Will changing the LCC off auto increase cpu temps the higher I increase it?
Not sure what I’m doing wrong. Using a 7800 x3d and a rog strix b650a. Set the negative on minus 35 on all cores. And the LLC on level 3 of possible 4. My voltage goes up a little from 0.993V to 1.065V during cinebench. Tried some other LLC levels but only the cpu temperature decreases a little bit. Voltage goes up. Did I miss something?
So I'm assuming from the comments that for a CPU like mine, 7900X3D, it would be better to offset specific cores instead of having all of them on an auto value. So like -30 on most but -15 on the few that tend to get weird right?? Would this be enough for better performance or would LLC help?
i have 7800x3d with EXPO, -30 Curve, PBO 85 and thermal throttle limit is to 85c. I have tested in Cinabench r23 for 30min, and i get 18143 with core max temp on 70.2c
Cinebench is a benchmark and does not test for complete accuracy. You can be Cinebench stable but still crash inexplicably every month or week when some error hits a critical part. It really depends on what you can deal with. If you don't mind some random crashes, use Cinebench as a quick and dirty stability test. Otherwise use Prime95 which loads the CPU and matches computed values with known accurate values. Use the blend test in Prime95 to stress both CPU and memory controller. You will find that you will have to dial back a lot to get it Prime95 stable. And do run it for at least an hour because 1 pass of small fft (everything laoded into cache) will take up to 10 min before it moves to large fft (stresses memory controller as some RAM is used). Run it over night without any errors and you are rock stable.
That was not the point of the video; however, your point stands and it is good info to share. The only adjustment I'd make to your comment is that some of the stability tests are unrealistic and most users will never run their system even close to that. It is better to test stability for one's specific use case. CB23 is a good place to start and if one experiences a crash while doing something else, WHEA errors can help to determine which core/s is the issue... then tweak its CO offset.
@@RKBenchmarker I thought that was what I mentioned. If you can deal with it, use Cinebench as a quick and dirty stability test. But I don't think loading all cores or having transient core spikes are that uncommon. It happens all the time when loading Windows.
Could you link the discord? I just got my 7800x3d I've never over or undervolted before. What happens if the system is unstable and shuts down, how do you get back to the bios to fix the number? Right now my cpu averages 4.4ghz with PBO turned on and no curve optimizer.
checking this video was nice info due im soon upgradeing to a 7800x3D but i did notice you hit 90c on both the core and die temps more or less noticed any thermal throttleing etc bcs of that since it seems high i also know after checking videos that the 7000 series seems to run abit higher temps then the 5000 series but 90c feels abit hot what cooler are you using?
i use a Be quiet Dark Rock Pro 4. Set your PBO Thernal Limit to 80°C and check your Vcore max in uefi. is it higher than 1.25 set it manualy to 1.2 or - 1.15 volt
Don’t think that an AIO is gonna give you -10 C. That ain’t happening. I cool an i9-12900k with a Noctua NH-D15S and it barely hits 82 C at 100% CPU load.
CO is undervolting for the old frequency but actually the voltage stays the same and now you get a higher Frequency at same voltage, as measured with a voltmeter...
i found that changing some values in curve optimizer influences how other cores react,so i can go -33 on some cores and max -11 on others and if i set the -33 to something like -15 than the ones that couldn't run lower than -11 now can run -15... is this a llc problem? it's on auto for me.
Asus LLC is opposite of others; Asus, highest number is most voltage applied to vdroop. (On other boards, the lowest number is most voltage applied to vdroop).
I need your advice. Two CPUs (7950x3d and 7800x3d) are available for sale: a new 7800x3d for €359 and a 2-month-old 7950x3d for €445. Which one should I go with? (Money is not an issue, and I have a 360 AIO cooler.) Which one should I choose for performance and future-proofing?
any idea why no matter what curves i set i get no instability also 7800xd im only scoring 15-16k cb23 temps maxing around 81 it seems like no matter what setting i do nothing changes. i feel like i should be benching 19k ddr56000, 360 aio volts also always look the same under all settings i try
I have the 7800x3d in the dan a4 h2O... would it be better to do settings like this or to just use the ECO mode? Not sure if it's more of a personal preference for tradeoffs in performance and temperature or if one is known to be overall better than the other
You can use ECO mode to lower the overall power consumption but that shouldn't be combined with what is shown here. This video focuses on leveraging Curve Optimizer along with LLC to get better performance than stock, or getting the same performance at lower wattage which results in lower temperatures.
Please help. Recently i swapped from intel to an 7950x3d cpu with an b560e motherboard, but after swapping my nvidia 4080 started to rise the temps in idle mode up to 60 degrees.. usually used to be around 35 degrees in idle , have any toughs?
Yes, the 4080 is 320w normal usage 280-300w thats a lot of heat. Install the exhausting fans. Also dont use negative curve -40 all cores like here. With 6000mhz ram and loose timings like 40 40 40 it might work. But if u will go 6400 cl28 or 7800-8000 cl36 100% wont work. Especially in the games. The programs might work but there will be a micro stuttering. With 6400 -30 with 8000 -25 or -20. If u dont believe try to set default 4800 cl 40,40 40 and -50 negative all cores 100% will work. The high temps are always bcos of heat accumulation. So i would say exhaust fans are the best. 🎉😊
The firsr option made my pc totall crash not boot anymore nothjng at 25. So thats one i dont gonna try anymore. Other steps have not tried it. I have now ryzen master used it puts it on 5825+- with 28degrees idle watercooled and max 75 full load stress test
Curve optimizer only works if you don't have adequate cooling, otherwise precision boost overdrive will push itself to the max under thermal limits 90 Celsius
@@MisakaMikotoDesu that's because the 3d versions can't be overclocked. It's literally hard locked and if it senses you are trying to force overclock it will go into a safe mode, and auto down clock
@@MisakaMikotoDesu all you do for the 3d versions is optimize under cooling conditions because the stacked 3d vcash make the chips run significantly hotter. That's why they don't allow overclocking and a lot of people without proper cooling experience thermal throttle
@@mattshaw4016 It works on my machine. Changing the curve optimization to negative values allows my CPU to run at higher frequencies. Yes I'm pushing at the thermal limit but the fact is that I'm able to do it even on a x3D chip. Motherboards have been allowing you to do things AMD doesn't want for decades now.
@@MisakaMikotoDesu that's what I was telling you. But it doesn't push higher frequency, it pushes the highest frequency on all cores that AMD allows the chip to run at. Like I said most people are thermal throttling, and don't even know that their chips are not running at default clocks lol. The higher benchmark scores are do to holding that default highest frequency for longer periods of time. I know this because I build PCs with 420mm water cooling and 140mm fans X6. The processor reaches 68 Celsius max. And there are no problems, I don't even have to mess with the curve optimizer, it literally changes nothing other than a few watts.
complete beginner here, if i used an air cooler and wanted to lower the average power usage during load and idle while achieveing close to the same performance, what can i do? ,
MSI does not have the graphical slopes for load line calibration and the only options are auto and modes 1 to 8. No useful guide in the bios. The explanation in this video is still too technical to translate in simple terms.
It is working on msi as well, just find the right option. And -40 might work like here with 5600mhz 30-36-36, but if your ram is 6400cl32 then u need to give more V, because at the same cpu speed let say 5ghz the ccd will have more data to work with from memory controler therefore it will need more V to work properly.
I have an ASUS motherboard, and I'm hesitant to try this because the LLC options are just levels numbered 1-8, and I don't know how that compares to your "extreme" setting, and my BIOS doesn't include a graph.
Same issue , can't find any info about Asus LLC , so max probably not a good idea .I'm using LLC 6 , Asus bios manual recommends 5 for overclocking so I bumped up by one
Asus LLC settings are opposite of other mobo as Asus' highest number (8 in some cases) is the max LLC setting; whereas 1 is the max llc setting and a larger number is the lower settings on other mobos. Setting max llc on an x3d chip is not the same as setting max llc on a non-x3d (which you normally wouldn't do). However, if you can set llc at a lower v-droop setting and still get your max performance, all the better.
I tried LLC 8 on my Asus mobo with co -37 , instantly almost hits 89 , artic freezer ii 240 open/bench build , at least with LLC6 hovers around 82 , might dial back more , no crashes with R23 or corecycler but instant crash with OCCT
@@RKBenchmarkerhey man you seem really knowledgeable about this. But I always like doing more research and getting more sources. Trouble is not many people are talking about this other than this video I'm especially interested in seeing more info/sources about 3d chips capping voltage making messing with LLC safer. My specific situation: Applying per core co offsets that seem stable with core cycler and all core stability tests and gaming. But are weirdly unstable at lighter loads (usually reopening a browser with existing tabs from idle causes bsod). I'm thinking LLC can help with this. Thoughts?
I follow all his settings for my 7950x3d and it crashed when running CPU benchmark , BLUE SCREEN so if you don't wanna kill your CPU don't follow this video or don't do EXTREME profile, don't put -negative 40 try other value -30, -20
Dude, I have a 7900X3D and also a Noctua NH-D15 and my Idle temps are mostly around 55C. I got a curve offset of -15 and a 85C temperature limit set in BIOS, but this only helps with temperatures and clock speeds in full load. What else can I do to improve idle temps? Or should I maybe not even be worried about those temps? Also the highest clock speed I can reach is 5.2GHz, never even close to 5.6GHz. I can not use a curve optimizer of -20, so I think I got bad silicon. What do you suggest? Thx
X3D isn't meant to actually get max clocks in benchmarks, even with good liquid coolers. It can clock that high, but it will always reach temp limits before it will clock that high. Those idle temps seem a bit high though, but maybe not. I'd say 45-50 might be better, but it's probably nothing to worry about. About all you can do is set a more aggressive fan curve for the CPU fan. So instead of 20% fan speed at idle, you'd start with 40% at idle and step up from there. i.e. 50% at 60c, 70% at 70c, 80% at 80c, 90% at 85c. Usually CPU fan curve settings have about 4-5 steps, so you'd just try to evenly distribute the fan curve so it ramps up smoothly. Buying a better fan would also help. Try the Noctua industrial fans. They go to 3000RPM, but they're basically identical to the normal noctua fans at 100%, when used at only 50%, but with twice the static pressure. So you don't have to run them at 3000RPM, but you also can if you don't mind the noise for heavy tasks, and they would definitely make a big difference in the amount of air being pushed through.
@@peoplez129 thanks for your reply. I installed some monitoring and benchmarking software in the meantime and did some testing. In single core benchmarks or benchmarks which only stress a few cores the boost reaches 5.5 or even 5.6 GHz, so I think everything's alright now. The rather high idle temps are probably also because of my case and the whole setup. I got a sound dampened case, so the side walls are closed. I run the front and back fans pretty slow via a fan controller, because I want a silent build. That's the reason why it runs pretty hot. Already got a silent CPU fan curve as well.
@@dominik2631 Since you want a quiet system but are choking your cpu due to lack of air flow, I would suggest upgrading your case to a Fractal Torrent as it is the best ranked air case out there, and can run those large front fans slow making it extremely quiet with great airflow! The other choice is get a good AIO, which will help significantly.
The load line calibration on asus much different then this, and majority of people commenting on here have asus and not Aorus, so sadly this should say FOR AORUS bios only
So I have a x670e-a and I’ve set the llc to maximum and I see no difference in performance. If I set it to standard or normal I get the same scores in cinebench. This normal?
Hey man, i stumbled across your video and i have a question. I recently bought and build my new system with a Ryzen 7 7800x3d, now im running into a problem with my idle temps, under full load my Peak core voltage is around 1.08-1.10, and while idling it sits around 1.12-1.1.13. do you have any idea how i can fix this? I've tried nearly everything from updating bios, reseating cpu cooling, re applying thermal paste, power plans, updating bios etc but i can't seem to get my idle voltage down. Right now the cpu idles around 53-55 celcius with a 360mm corsair h150i, in game it won't even go above 65 celcius and under full cinebench load this changes to around 77-80 celcius. which in my eyes, does seem like allright temperatures... I've got no clue why my idle voltage is higher than my full load voltage... if anyone got a clue please let me know
Hello guys, I have a problem with this settings. I have set a curve optimizer per core between negative 15 to negative 30 and loadline calibration to low, but when i run games, benchmark etc, my VID core have peak to 1.365Volt and when i simply wathc youtube o doing nothing the voltage VID stay from 1.2 to 1.3volt. Have you any advice? thanks :)
Hi im hafiz from malaysia , just bought new proc & mobo and i got some issue with my 7800x3d 🥲 when i run multi core at cinebench r23 temp go high 90* 😭 and idle 50* . if i run some apps going to 70* . can someone help me ? 😢 x670 aorus elite ax r7 7800x3d corsair 5600mhz 16x2 aio 360 aorus . help me please 🙏🏻
@@GameTechReviews i get 80 normally with an aio cooler without any power limits. I lowered the power limit to 75W. PPT Limit [mW] → 75000 (75W) and look at the Cinebench score and Temps its crazy give it a shot just change PPT dont mess with the other settings
@@unalert2928 make a video how to make our 7950x3d perform the same as 7800x3d for gaming I mean how does a $360 cpu beat the mighty 7950x3d 650$ usd CPU in gaming Make a video for us 7950x3d users , i want the same performance as 7800x3d is giving
