So hopefully this video settles things - undervolting 3rd gen Ryzen in the bios seems to result in truly strange performance, despite the CPU appearing functionally stable. With Ryzen Master Software, we can still tweak the 3900X to 1.1V and below, but at a cost to single threaded performance. Thanks to those who pointed out the errors, I don’t take this personally, it’s all about making better content in the end. Also thanks to Steve over at Hardware Unboxed who was able to test his 3900X on the Aorus X570 Xtreme this morning to confirm what we’re seeing here.
Amen. I blacklisted their network on my router because of their stunt. It was disgusting, cowardly, and bordered on criminal negligence as the video was so wrong, some kid saving for months building his first PC would have simply been out of all of that. Bad form.
@@GLXY23 The Verge posted a "how to" video guide for building your own PC. It was littered with false information and bad advice. They were ridiculed for this and they never corrected the video, acknowledged their mistake or apologised for it. Search UA-cam for "Verge PC build" and you'll see what I mean.
Thank you for being so passionate about your job, this was extremely appreciated by me and everyone is this community. And Yes i would love you doing a cryorig build with this chip running at 4.0 -.1 ghz, with maybe some emphasis on single-thread tasks, like games. Packing this thing in a S1 and being able to surpass/close the gap with the 9900k would be amazing
Why can't more people be like you and admit mistakes were made and verify/correct any erroneous statements? This is refreshing compared to big name journos that can't admit they were wrong ever.
I subscribed for no other reason than you were willing to make this correction. I love following people that can see their mistakes and work to ensure the correct information is put out there. Thank you
Admirable to be honest and all this work to maintain integrity. I really wanted to believe your last video was right, but wishful thinking is as wishful thinking does. Keep up the good work. Subbed
Looking forward to some updates from AMD and board makers on this! I saw the video and got super hyped. Did testing and got a little sad. Would you mind doing testing between 3600+ ram speed and 3533 and below and how it effects the boosting of the chips? My experience anything over 3533 causes the chip to max at at 4.250ghz but lower than 3600 it boosts naturally. Thanks for the updated video!
Looking up undervolting came across this video after hearing about your misinformation mishap, anyway owning up and making a follow up video about being wrong gets a follow, good job correcting
You are my hero. My temperatures dropped from around 50C to 60C at idle to 38C to 45C at idle. It runs much cooler, more silent, and my Cinebench R20 score went from ~7050 to ~7300. I used 1.17V and 4.2GHz from your chart with my 3900X and TUF Gaming X570 MOBO. Thanks so much!
Respect for correcting yourself. Also just a heads up, you really don't want to use that 1.4V OC 24/7, as maximum safe voltage on high-current loads seems to be around 1.325V.
Oh cmon, you think will blow at 1.4V??? Those cpu's are made to sustain even 1.5V, same as ryzen 2000. May suffer in 4-5 years from degradation if you stay at 1.5V but who cares. You will change your cpu anyway after so much time, stop being a bitch, as long as your temps are not in the high 80's you are completely fine.
fair play, refreshing to see an update, anyone who does these kind of tests knows that there is always glitches especially with new products. These things happen but again good to see a correction which was in itself interesting
I don't think it's a bug in the BIOS, it's just the way Ryzen 3000 seems to work with voltages. Basically it never crashes with low voltages but instead adjusting everything else in the background to keep running. The clockspeeds you see with monitoring tools are not the real clockspeeds, as no tool is quick enough to get the exact frequency when the CPU is changing it every 1-2ms.
This is great content! I upgraded from a 1700 and have been struggling with high voltage and temperatures. To avoid issues until it is fixed I am running at 3800mhz and 1.1volts. It's still a great upgrade until the voltage issues are fixed in the coming weeks.
Props for your dilligence and thanks for exploring the topic deeper. I am enjoying very much, how the community is exploring the nooks and crannies of Zen2. And an undervolted R9 3900 still seems like a legit idea for some SFF-edge-cases. Fun times right now. I am curious how Zen 2 settles in over the next monrths, when BIOSes, drivers and software matured.
Whats up Ali! I've been watching your videos for a little over a year now. I've been wanting so badly to jump into the SFF scene for a while now, and I think we're currently in a buyers market, so now is the time I'm going to do so. There's a case that has very little reviews (due to it's failed kickstarter campaign I imagine) called the Nouvolo Steck. Considering you're in my opinion the most knowledgeable youtuber in regards to SFF PC building, I was wondering if you have considered offering your thoughts on it. Regardless if it's video worthy, I think It's about time I commented to let you know how I feel about your content. - Hands down, this channel offers me the most pure and comprehensive information in regards to modern PC technology. It is presented in the most visually stunning, and professional quality available. It is evident how much time and effort is put into your research, presentation, and production. And your opinions are always offered, but never overshadow the facts. I have learned so much from your channel and can't express how much I appreciate your service. All that being said, following up your previous video with this one shows me how important your community is to you. Removing a video that you labored over shows such a dedication to accuracy and honesty. That action speaks highly of your character. I sincerely hope you get as much joy and reward creating this content, as you provide me to consume it. You rock! Thank you.
Thanks for the correction, I did wonder... Would love to see power consumption (idle & full load vs stock). Thanks for sharing, I always enjoy your videos.
I legit watched this video twice, very interesting! I hope that my theory of a 3900X essentially being 2 3600Xs is correct. Of course they are likely binned as better chips, but if I can get one that performs similarly to this, I will be able to keep my chip and VRM cooler than my 1600X is right now.
Respect for admitting that mistake and doing a new video to clarify what was going on! I am looking forward to the Ryzen 3900X streamign video. I get heavy stuttering and framesdrops @1080p 60fps or 900p 60fps medium with the 3900X. Far away from that magical 1080p 60fps slow 6k bitrate. Not sure how they achieved that - maybe it was a stream only rig and gaming was done on a second one?
I think the precision boost algorithm actually lowered the clock speed when voltage is not enough, just that he monitoring softwares are not reporting the correct numbers. Undervolting is still possible while using precision boot. I set a -0.0125V offset to Vcore and used a higher LLC setting. Result was about 5 degrees temperature reduction under full load. The multi core and single core performance was not affected. It’s not much but it works.
I'd particularly like to see it tested with a Noctua NH-L12 with the included 92mm bottom fan and a NF-A12x15 top fan. That combination fits a number of SFF cases and is likely to have enough cooling power for the 3900X, though undervolting may be helpful.
thanks so much bro I did a complete new build but from the first day I was getting high temps on R5 3600 like when gaming it got above 95 C, when I saw the Ryzen master utility it was running 4.2ghz on 1.3 v from default , and I am using stock cooler, watched your video , then set it to 3900mhz on 1 v for normal temps and you would not believe it am get not more than 75 C even when stress testing on AIDA 64, TYSM BRO
its known that the lower vcore increases clocks BUT it reduces the performance, makes no sense, the Ryzen chips wont crash, the cores just perform poorly at a reported "higher" clock.
I'm late to the party, but since no one else mentioned it: If you want to undervolt 3rd gen Ryzen CPUs, don't bother setting manual voltages or frequencies, lower the package power target (PPT) instead. The CPU will then figure out the best combination of voltage and frequency on its own and you'll still get single core boost. For example, my Ryzen 5 3600 has a default PPT of 88W. When I lower this to 50W, I only lose about 10% multi-threaded performance, single threaded performance stays the same and my fans can run at roughly half the speed. See also www.reddit.com/r/Amd/comments/ceakbs/if_you_want_to_save_powerreduce_thermals_reduce/
Good findings. I found this out on my 3700x about a year or so ago. Ryzen doesn't actually crash or blue screen due to low voltage it will run and manage like nothing is wrong but still trying to get the voltage it needs. I would laugh when ppl would it was stable at some low voltage and scores come back horrible.
stability isn't exactly black and white, you can be stable at higher clocks and lower voltages for everything except AVX workloads which is ok for people not doing video editing and just gaming.
What I mean is that you'll either find workloads that will crash it / generate errors or you won't. Here we are seeing what appears to be internal throttling because the CPU isn't happy with the Vcore - instead of just crashing. In a way this is potentially good as a failsafe, but in its current state I think we can all agree this is unexpected behaviour - especially at a moderate vcore of 1.25V.
Then they try and make it look minimal and 'innovative' and end up making the device overheat and kill itself in the long run all in the sake of business, ffs.
When i put 1.25V and 4.2Ghz. I noticed that now every single monitoring programme thinks that cpu always runs on 4.2Ghz and 1.25V Always. Any comments on this ? P.S. had Stock settings. 86deg in aids 64 stress. 3587 in C20 test. At 1.25V My ryzen 3600 shows 79deg in aida 64 with constant 4.2Ghz and 3748 in C20 test. At 1.2375V I got 76deg and 4.2Ghz and 3763 in C20. Single core 483 - same as stock 481
I have done quite a bit of research and testing and I want to share my settings with you to get your take on Vcore safety as I have been reading A LOT about static Vcore voltages. I currently have the following settings resulting in stable bench marking across R20, R15 and various others, as well as with my gaming sessions and studio work for design. I am using the Kraken Z63 and I idle around 45C to 55C and hit 79C in R20 with a score of 10100. While gaming I am usually in the range of 50C to 60C. Below are my OC settings: PBO disabled Global C state disabled LLC 3 Vcore 1.36 RAM 3200 Fabric 1600 Stock AMD Ryzen High Performance Power Plan CCD0 CCX 0 4.425 CCX 1 4.425 CCD 1 CCX 0 4.325 CCX 1 4.335 My main concern here is that many forums suggest that the static Vcore should never be set higher than 1.32 with many suggesting that 1.3 should really be the highest or else I will degrade my chip very quickly. In your video I believe you set the Vcore to 1.375 on the 3600. Has this impacted the CPU over time? In R20 my Vcore sits at around 1.250. While gaming it is always at the max setting of 1.36. When Idle, it's also at 1.36 all the time. I really want to stick to the manual OC because I do not like the heat production when using stock with PBO because the voltage jumps so much as does the heat. I really don't want to run OC settings that will degrade my CPU in less than a year so any feedback on the topic would be greatly appreciated.
The best stability test for me is playing "The Division 2". The game is so bloody sensitive and kicks out while all other tests run stable for some time. That's matching for CPU and RAM overclocking.
One thing you could add was how much voltage you can reduce from stock in BIOS without reducing performance. Seems to be somewhere between .05-.1v reduction.
1.3V or maybe even 1.2V seems to be a reasonable voltage that doesn't really lower performance, while still being very stable and lower than stock voltage.
Also just a note that 4.3GHz is a no go for games like CSGO. This game has a bug where things in the game will begin to look weird and "damaged" UI will appear and the console windows will display continuous errors.
Dude you are a legend, a little late to the party but thank to this i solved my temp issues and dropped 20 degrees of my max temp! THATS NOT A TYPO!!!!!!
Since CPU temperature matters a lot ... what were the CPU temps during your testing? Each 20C changes the frequency margin by 100MHz. It's a big deal. More importantly, depending on the load, when you use fixed settings like that and you happen to have a load which heats the CPU more than usual... now suddenly your stability margins disappear because your temps are 10C higher and you wind up with an unexpected crash. This is the down-side of using fully-manual OC settings (in addition to the single-thread performance downside). Second item... these BIOSes are very broken, for sure, but another way to approach this, and I know I'm probably beginning to sound like a broken record... but another way to approach this is to set the wattage for the socket in the BIOS to a lower-than-stock value and then maybe try adding in a negative VCORE offset. Assuming the BIOS implements the parameters properly. When you do this the system will run relative to the temperature you manage to get the CPU running at, will continue to boost single-core workloads effectively, and most importantly the system should (I'd like to say 'will' but I can only say 'should' at the moment) compensate for temperature. All automatically. Lastly, in terms of stability... that's a difficult question to address. Benchmark stability is very, very different from day-to-day stability. I get benchmark stability all the time but if I throw a real 24-hour workload at the system, such as bulk application compiles, settings which are stable for the benchmark wind up not being sufficient for the real workload. The system on a cool day verses the system during a heat wave... there are a lot of variables that manual settings cannot compensate for. -Matt
You need to measure the SVI2 TFN sensor to get an accurate voltage reading. Fact of the matter is, not one of those sensors measured by software is accurate, at least not compared to actually sticking the probe of a volt meter into the socket. But AMD engineers have confirmed that the SVI2 TFN is the closest approximation to what voltage the CPU is actually running at.
Same thing happened to me with my 3600. I was running a golden R5 1600 @ 3.9 GHz 1.21v (sff case, didn't want to go higher with the voltage for thermals). I thought I could quickly undervolt the 3600 and be on my way. Spent that whole night fiddling and trying to figure out why my performance numbers were absolute garbage. Set everything back to auto and it was fixed. Hopefully some improved BIOS versions come out soon!
So hopefully this video settles things - undervolting 3rd gen Ryzen in the bios seems to result in truly strange performance, despite the CPU appearing functionally stable. With Ryzen Master Software, we can still tweak the 3900X to 1.1V and below, but at a cost to single threaded performance.
Thanks to those who pointed out the errors, I don’t take this personally, it’s all about making better content in the end. Also thanks to Steve over at Hardware Unboxed who was able to test his 3900X on the Aorus X570 Xtreme this morning to confirm what we’re seeing here.
Can you monitor the clock frequency of infinity fabric as well? Maybe that gets automatically reduced when undervolting. Just an idea.
Thank you for admitting your mistakes, you have my respect.
good work dude!
Good job clearing this up. Now you just need to play with the temp cap in the manual PBO settings to see how that does vs a fixed clock and voltage.
So.. I read one thread not the long time ago and some guy said that FX-8350 has better results with higher voltage but the same frequency
Great work Ali, huge respect for pulling the video after all that hard work and correcting things so quickly. Great stuff mate 👍
Thanks man, really appreciate your time yesterday.
@@optimumtech I was more than happy to take a look, it was an interesting test.
Is this the internet version of a circle jerk?
@@happy0ne No it's two people who respect each other having a conversation, It must look strange to you as i am certain no one respects you.
@@admin8244 - well said.
Mad respect for posting corrections and pulling the video!
Love the integrity!
The integrity gesture is good, but if a channel like this loses its integrity it's game over.
I doubt this dude has any integrity. He’s probably faking it just for the views and white knights like you are falling for it.
PlasmaPower says the pony
@@Safetytrousers if one doest admit mistake he didn't have any in the first place.
Good job on the update. You did awesome!
See Verge, when you made a mistake and admit it, you would earn others respect!.
Amen. I blacklisted their network on my router because of their stunt. It was disgusting, cowardly, and bordered on criminal negligence as the video was so wrong, some kid saving for months building his first PC would have simply been out of all of that. Bad form.
I was subscribed to the Verge Science channel until this little incident. No respect for anything with a "Verge" label on it.
@@grizzly6699 what was the incident?
@@GLXY23 The Verge posted a "how to" video guide for building your own PC. It was littered with false information and bad advice. They were ridiculed for this and they never corrected the video, acknowledged their mistake or apologised for it. Search UA-cam for "Verge PC build" and you'll see what I mean.
@@grizzly6699 just saw it and im just baffled at them. They did alot of wrong in their build then i thought they did.
Least you can admit when you make a mistake, I respect that.
"If you lose credibility just by admitting fault, then you didn't have any in the first place" -Fujitora, One Piece
I'M LOOKING AT YOU, THE VERGE.
Wow a One Piece fan, and quoting Fujitora no less. You sir, have excellent taste.
11/10 best apology video on UA-cam
1.1V/10 😂
@@Gandi2000 well played hahaha.
UA-camrs admitting their mistakes always have my utmost respect. Thanks for clarifying.
much rather have experimentation and very quick correction, than the same old stuff from the other sites. Keep up the great work!
Thanks Reddit. Good job Optimum Tech!
Mad respect to you on issuing a correction!
Thank you for being so passionate about your job, this was extremely appreciated by me and everyone is this community.
And Yes i would love you doing a cryorig build with this chip running at 4.0 -.1 ghz, with maybe some emphasis on single-thread tasks, like games. Packing this thing in a S1 and being able to surpass/close the gap with the 9900k would be amazing
Never seen your channel and haven't watched the video yet, but liking for your humility. Thanks!
Thanks for tuning in!
Integrity of this channel is top.
Glad I'm subscribed.
Why can't more people be like you and admit mistakes were made and verify/correct any erroneous statements? This is refreshing compared to big name journos that can't admit they were wrong ever.
Lovely video, and massive respect!!
Really dig this video, not often you see youtubers owning up to their mistakes and explaining what happened, mad respect+
I don't see many things I know to be mistakes, and if a mistake gets pointed out in the comments I mostly see a reply from the channel.
No worries mate. Keep it up!
I subscribed for no other reason than you were willing to make this correction. I love following people that can see their mistakes and work to ensure the correct information is put out there. Thank you
Admirable to be honest and all this work to maintain integrity. I really wanted to believe your last video was right, but wishful thinking is as wishful thinking does.
Keep up the good work.
Subbed
Good shit. Respect for taking down and addressing + bringing up this topic.
you are a great man, always looking forward to your reviews.. of any sort. Keep up the great work man!
Good job removing the video when you were wrong, many UA-camrs don't and just upload their correction for $.
Mad props for correcting this, I'll reward with a sub :P
Such things happen. Thanks for updating us. All good. And we (you and all of us) learn from mistakes like this.
Good on you!
Thank you for updating and pulling video you'll always have me as a sub keep up the good work :)
Fantastic attitude. Amazing content. Keep it up!
Respect for posting this video correcting the earlier version!
Excellent video thanks!
Nothing wrong with admitting mistakes mate so props to you.
Well done! Thank you for the explanation as to why you removed the video and retested your cpu! :-)
I find that very few youtubers admit when they are in the wrong and it's refreshing to see you being this open and honest in this video.
Good stuff. Keep up the awesome work!
I use this info as a baseline, just a little more voltage but helped me a lot to obtain a solid starting point to experiment. Thanks !
Looking forward to some updates from AMD and board makers on this! I saw the video and got super hyped. Did testing and got a little sad. Would you mind doing testing between 3600+ ram speed and 3533 and below and how it effects the boosting of the chips? My experience anything over 3533 causes the chip to max at at 4.250ghz but lower than 3600 it boosts naturally.
Thanks for the updated video!
Looking up undervolting came across this video after hearing about your misinformation mishap, anyway owning up and making a follow up video about being wrong gets a follow, good job correcting
Thanks for retesting. 👍
Good that you do all the testings, so I don't make the same mistake when I get mine. Good job by the way. Everyone is learning about the new CPUs.
You are my hero. My temperatures dropped from around 50C to 60C at idle to 38C to 45C at idle. It runs much cooler, more silent, and my Cinebench R20 score went from ~7050 to ~7300.
I used 1.17V and 4.2GHz from your chart with my 3900X and TUF Gaming X570 MOBO. Thanks so much!
Respect for correcting yourself. Also just a heads up, you really don't want to use that 1.4V OC 24/7, as maximum safe voltage on high-current loads seems to be around 1.325V.
Oh cmon, you think will blow at 1.4V??? Those cpu's are made to sustain even 1.5V, same as ryzen 2000. May suffer in 4-5 years from degradation if you stay at 1.5V but who cares. You will change your cpu anyway after so much time, stop being a bitch, as long as your temps are not in the high 80's you are completely fine.
Being able to be own up to mistakes already puts you ahead of most techtubers. Keep up the great work.
nice work man
Kudos to you for a follow up video
fair play, refreshing to see an update, anyone who does these kind of tests knows that there is always glitches especially with new products. These things happen but again good to see a correction which was in itself interesting
many YTer .. are not like you.. they stay.. at their mistakes.. but Mistakes make great YOU ARE GREAT YT... !!!
I don't think it's a bug in the BIOS, it's just the way Ryzen 3000 seems to work with voltages.
Basically it never crashes with low voltages but instead adjusting everything else in the background to keep running.
The clockspeeds you see with monitoring tools are not the real clockspeeds, as no tool is quick enough to get the exact frequency when the CPU is changing it every 1-2ms.
No problem at all bro. Takes a real man to admit one's mistakes.
This is great content! I upgraded from a 1700 and have been struggling with high voltage and temperatures. To avoid issues until it is fixed I am running at 3800mhz and 1.1volts. It's still a great upgrade until the voltage issues are fixed in the coming weeks.
Having the same problem, what motherboard.
@@rudyhernandez6014 Asrock x570 gaming 4
You just earned a subscriber due to your integrity...
Props for your dilligence and thanks for exploring the topic deeper. I am enjoying very much, how the community is exploring the nooks and crannies of Zen2. And an undervolted R9 3900 still seems like a legit idea for some SFF-edge-cases. Fun times right now. I am curious how Zen 2 settles in over the next monrths, when BIOSes, drivers and software matured.
Whats up Ali! I've been watching your videos for a little over a year now. I've been wanting so badly to jump into the SFF scene for a while now, and I think we're currently in a buyers market, so now is the time I'm going to do so. There's a case that has very little reviews (due to it's failed kickstarter campaign I imagine) called the Nouvolo Steck. Considering you're in my opinion the most knowledgeable youtuber in regards to SFF PC building, I was wondering if you have considered offering your thoughts on it.
Regardless if it's video worthy, I think It's about time I commented to let you know how I feel about your content.
- Hands down, this channel offers me the most pure and comprehensive information in regards to modern PC technology. It is presented in the most visually stunning, and professional quality available. It is evident how much time and effort is put into your research, presentation, and production. And your opinions are always offered, but never overshadow the facts.
I have learned so much from your channel and can't express how much I appreciate your service.
All that being said, following up your previous video with this one shows me how important your community is to you. Removing a video that you labored over shows such a dedication to accuracy and honesty. That action speaks highly of your character. I sincerely hope you get as much joy and reward creating this content, as you provide me to consume it.
You rock! Thank you.
Props for correction.
no fuss, props and respect for discovering it.
Thanks for the correction, I did wonder... Would love to see power consumption (idle & full load vs stock). Thanks for sharing, I always enjoy your videos.
This is how any content creator should make corrections. Well done.
Thanks so much for this just dialed in the these in on my 3600 and actually got fairly similar numbers running on air.
Please stay on the topic and tell us when this is fixed! :) Appreciate your work very much. And: Everyone makes mistakes.
Quite interesting!
Will be looking forward to more data coming up about these CPUs.
You're one of the best dude keep it up.
I legit watched this video twice, very interesting!
I hope that my theory of a 3900X essentially being 2 3600Xs is correct. Of course they are likely binned as better chips, but if I can get one that performs similarly to this, I will be able to keep my chip and VRM cooler than my 1600X is right now.
Respect for admitting that mistake and doing a new video to clarify what was going on! I am looking forward to the Ryzen 3900X streamign video. I get heavy stuttering and framesdrops @1080p 60fps or 900p 60fps medium with the 3900X. Far away from that magical 1080p 60fps slow 6k bitrate. Not sure how they achieved that - maybe it was a stream only rig and gaming was done on a second one?
Wow, really impressed with your character, so many larger channels can learn alot from you, I'm looking at you the verge😑
Nice move, big respect.
I think the precision boost algorithm actually lowered the clock speed when voltage is not enough, just that he monitoring softwares are not reporting the correct numbers.
Undervolting is still possible while using precision boot. I set a -0.0125V offset to Vcore and used a higher LLC setting. Result was about 5 degrees temperature reduction under full load. The multi core and single core performance was not affected. It’s not much but it works.
please do test the 3900x with new voltages thermals up to 1.15V with LP coolers !!
I'd particularly like to see it tested with a Noctua NH-L12 with the included 92mm bottom fan and a NF-A12x15 top fan. That combination fits a number of SFF cases and is likely to have enough cooling power for the 3900X, though undervolting may be helpful.
good job ptimum tech!
Respect dude you have dignity and honesty something very rare this day and age
you are an example of honesty on You Tube. Thank you for all your hard work.
thanks so much bro I did a complete new build but from the first day I was getting high temps on R5 3600 like when gaming it got above 95 C, when I saw the Ryzen master utility it was running 4.2ghz on 1.3 v from default , and I am using stock cooler, watched your video , then set it to 3900mhz on 1 v for normal temps and you would not believe it am get not more than 75 C even when stress testing on AIDA 64, TYSM BRO
its known that the lower vcore increases clocks BUT it reduces the performance, makes no sense, the Ryzen chips wont crash, the cores just perform poorly at a reported "higher" clock.
Good for you. Takes a real man to admit when they make a mistake and address it
Video is exactly what I wanted/needed. thanks.
I'm late to the party, but since no one else mentioned it: If you want to undervolt 3rd gen Ryzen CPUs, don't bother setting manual voltages or frequencies, lower the package power target (PPT) instead. The CPU will then figure out the best combination of voltage and frequency on its own and you'll still get single core boost. For example, my Ryzen 5 3600 has a default PPT of 88W. When I lower this to 50W, I only lose about 10% multi-threaded performance, single threaded performance stays the same and my fans can run at roughly half the speed.
See also www.reddit.com/r/Amd/comments/ceakbs/if_you_want_to_save_powerreduce_thermals_reduce/
Dude You deserve m00re subscribers !!!! Keep it up !!!!
It's really helps, thank a lot
Thanks for the video. Can't wait for the laptop variant of this. ULV with good silicon should be great on Zen3
Good findings. I found this out on my 3700x about a year or so ago. Ryzen doesn't actually crash or blue screen due to low voltage it will run and manage like nothing is wrong but still trying to get the voltage it needs. I would laugh when ppl would it was stable at some low voltage and scores come back horrible.
stability isn't exactly black and white, you can be stable at higher clocks and lower voltages for everything except AVX workloads which is ok for people not doing video editing and just gaming.
What I mean is that you'll either find workloads that will crash it / generate errors or you won't.
Here we are seeing what appears to be internal throttling because the CPU isn't happy with the Vcore - instead of just crashing.
In a way this is potentially good as a failsafe, but in its current state I think we can all agree this is unexpected behaviour - especially at a moderate vcore of 1.25V.
I can see Apple using this to get a 12 core chip into their Macbooks
Then they try and make it look minimal and 'innovative' and end up making the device overheat and kill itself in the long run all in the sake of business, ffs.
@@Oliver-bn7jt and make you pay $4000+ while doing so
Well this didn’t age well.
My R5 3600 works at 1.125v@4.2ghz Flat CinR20- 3750
When i put 1.25V and 4.2Ghz. I noticed that now every single monitoring programme thinks that cpu always runs on 4.2Ghz and 1.25V Always.
Any comments on this ?
P.S. had Stock settings. 86deg in aids 64 stress. 3587 in C20 test.
At 1.25V My ryzen 3600 shows 79deg in aida 64 with constant 4.2Ghz and 3748 in C20 test.
At 1.2375V I got 76deg and 4.2Ghz and 3763 in C20. Single core 483 - same as stock 481
I have done quite a bit of research and testing and I want to share my settings with you to get your take on Vcore safety as I have been reading A LOT about static Vcore voltages.
I currently have the following settings resulting in stable bench marking across R20, R15 and various others, as well as with my gaming sessions and studio work for design. I am using the Kraken Z63 and I idle around 45C to 55C and hit 79C in R20 with a score of 10100. While gaming I am usually in the range of 50C to 60C.
Below are my OC settings:
PBO disabled
Global C state disabled
LLC 3
Vcore 1.36
RAM 3200
Fabric 1600
Stock AMD Ryzen High Performance Power Plan
CCD0
CCX 0 4.425
CCX 1 4.425
CCD 1
CCX 0 4.325
CCX 1 4.335
My main concern here is that many forums suggest that the static Vcore should never be set higher than 1.32 with many suggesting that 1.3 should really be the highest or else I will degrade my chip very quickly. In your video I believe you set the Vcore to 1.375 on the 3600. Has this impacted the CPU over time? In R20 my Vcore sits at around 1.250. While gaming it is always at the max setting of 1.36. When Idle, it's also at 1.36 all the time.
I really want to stick to the manual OC because I do not like the heat production when using stock with PBO because the voltage jumps so much as does the heat. I really don't want to run OC settings that will degrade my CPU in less than a year so any feedback on the topic would be greatly appreciated.
2:36 Because Ryzen Vcore range is huge between lowest stable and highest boost. It's only logical at least to me.
The best stability test for me is playing "The Division 2".
The game is so bloody sensitive and kicks out while all other tests run stable for some time. That's matching for CPU and RAM overclocking.
One thing you could add was how much voltage you can reduce from stock in BIOS without reducing performance. Seems to be somewhere between .05-.1v reduction.
1.3V or maybe even 1.2V seems to be a reasonable voltage that doesn't really lower performance, while still being very stable and lower than stock voltage.
Also just a note that 4.3GHz is a no go for games like CSGO. This game has a bug where things in the game will begin to look weird and "damaged" UI will appear and the console windows will display continuous errors.
Hope my future 3700X runs stable all-core 4 GHz 1V - 1.1V. It's enough horse power for me in 1440p ultrawide games and daily use.
Nice video, What brand are those RAMs?
New sub here just for admiting and correcting your mistake, respect.
Dude you are a legend, a little late to the party but thank to this i solved my temp issues and dropped 20 degrees of my max temp! THATS NOT A TYPO!!!!!!
What about single- and multicore performance drop?
Integrity rules👍
Since CPU temperature matters a lot ... what were the CPU temps during your testing? Each 20C changes the frequency margin by 100MHz. It's a big deal. More importantly, depending on the load, when you use fixed settings like that and you happen to have a load which heats the CPU more than usual... now suddenly your stability margins disappear because your temps are 10C higher and you wind up with an unexpected crash. This is the down-side of using fully-manual OC settings (in addition to the single-thread performance downside).
Second item... these BIOSes are very broken, for sure, but another way to approach this, and I know I'm probably beginning to sound like a broken record... but another way to approach this is to set the wattage for the socket in the BIOS to a lower-than-stock value and then maybe try adding in a negative VCORE offset. Assuming the BIOS implements the parameters properly. When you do this the system will run relative to the temperature you manage to get the CPU running at, will continue to boost single-core workloads effectively, and most importantly the system should (I'd like to say 'will' but I can only say 'should' at the moment) compensate for temperature. All automatically.
Lastly, in terms of stability... that's a difficult question to address. Benchmark stability is very, very different from day-to-day stability. I get benchmark stability all the time but if I throw a real 24-hour workload at the system, such as bulk application compiles, settings which are stable for the benchmark wind up not being sufficient for the real workload. The system on a cool day verses the system during a heat wave... there are a lot of variables that manual settings cannot compensate for.
-Matt
You need to measure the SVI2 TFN sensor to get an accurate voltage reading. Fact of the matter is, not one of those sensors measured by software is accurate, at least not compared to actually sticking the probe of a volt meter into the socket. But AMD engineers have confirmed that the SVI2 TFN is the closest approximation to what voltage the CPU is actually running at.
Thanx, Man! I am always for undervolting both CPU & GPU. I'm ready to sacrifice some raw power for the sake of cool silence)
Nicely done.
Same thing happened to me with my 3600. I was running a golden R5 1600 @ 3.9 GHz 1.21v (sff case, didn't want to go higher with the voltage for thermals). I thought I could quickly undervolt the 3600 and be on my way. Spent that whole night fiddling and trying to figure out why my performance numbers were absolute garbage. Set everything back to auto and it was fixed. Hopefully some improved BIOS versions come out soon!
Thx for the video itry this with my r5 3600 & work without issues until now : 1.25 v for 4.2 ghz my Mobo is B450 aorus elite....everything is cool