AMD Ryzen 3000 Undervolting Offset vs. Override | Vcore Voltage
HTML-код
- Опубликовано: 21 июл 2019
- In this content, we benchmark the Gigabyte X570 Master and MSI X570 Godlike to demonstrate undervolting behavior on the AMD R9 3900X & R5 3600.
Support GN via the store: store.gamersnexus.net/
Article: www.gamersnexus.net/guides/34...
Undervolting is a valid concept that can net real performance improvements in GPUs and CPUs alike. Those improvements are typically in the form of reduced power consumption (watts * amps = power), reduced heat load as a result, and increased frequency boosting headroom under Precision Boost 2 parameters. Today's video looks at valid and invalid undervolting results between two boards and two CPUs, showing how to know when your undervolt is working (or not). We also demonstrate offset vs. override voltages for this test.
NEW RYZEN CPU LISTINGS:
AMD Ryzen 5 3600 (Amazon): geni.us/vtPUE
AMD Ryzen 5 3600X (Amazon): geni.us/uxr8
AMD Ryzen 9 3900X (Amazon): geni.us/uPxxADC
AMD Ryzen 7 3800X (Amazon): geni.us/kHxcq
AMD Ryzen 7 3700X (Amazon): geni.us/sdVDHO
AMD Ryzen 5 3400G (Amazon): geni.us/IWN0c
AMD Ryzen 3 3200G (Amazon): geni.us/AxJmDW
Or grab an affordable R7 2700X on Amazon: geni.us/HPeV or R7 2700 geni.us/CHQ0B
Intel's i9-9900K (Amazon): geni.us/eNRPu
Like our content? Please consider becoming our Patron to support us: / gamersnexus
** Please like, comment, and subscribe for more! **
Follow us in these locations for more gaming and hardware updates:
t: / gamersnexus
f: / gamersnexus
w: www.gamersnexus.net/
Host: Steve Burke
Editorial, Testing: Patrick Lathan
Video: Andrew Coleman
Links to Amazon and Newegg are typically monetized on our channel (affiliate links) and may return a commission of sales to us from the retailer. This is unrelated to the product manufacturer. Any advertisements or sponsorships are disclosed within the video ("this video is brought to you by") and above the fold in the description. We do not ever produce paid content or "sponsored content" (meaning that the content is our idea and is not funded externally aside from whatever ad placement is in the beginning) and we do not ever charge manufacturers for coverage. 
Игры
Article here! www.gamersnexus.net/guides/3494-amd-ryzen-3000-undervolting-offset-override
Watch our AMD R7 3800X review here: ruclips.net/video/PAGQwWDyURI/видео.html
Do a
Ryzen 9 3900x SMT ON VS
Ryzen 9 3900x SMT OFF VS
Intel I9 9900k Hyperthreading OFF VS
Intel I9 9900k Hyperthreading ON
would be interesting!
Could it be that the infinity fabric frequency in zen 2 is affected negatively by undervolting in a way that affects performance in AMD case.
Steve, If the clocks do stabilize with u/v I suppose there can be a benefit to .1% lows in games from a mild u/v, any thoughts?
Hi team, wouldn't be better if we were to use PPT and to lower it in the precision boost option. I think its under AMD CBS (in Gigabyte BIOS) and change it to a lower stock value. I.e. change from 88W in a std value for 3700x (65w tdp chip) to 70, 65 or 60W.
Now with PBO and Loadline Calibration: ruclips.net/video/YQ8zdprzEjI/видео.html
Thanks for the follow up! I'm not sure at this point whether the 'clock stretching' is closer to a bug or a failsafe to keep the CPU running.
Great content piece as always and hopefully we can all get some rest soon lol.
This is the result of more "smartness" in the new modern CPU world. They are behaving more like GPU's now with sophisticated firmware. When the CPU require/requests more voltage then the current value offset set in BIOS the motherboard will attempt to override the current setting and this causes strange behavior we are seeing.
I love your channel! :D Such high-quality content backed by someone not willing to publish false results.
It probably works on 1v because it's also optimized for servers at around 1v. A consumer board should probably chrash though. It's just not running properly
What i have heard, it is not biosbug but a processor AVFS feature
Optimum Tech, your journalism and integrity is truly commendable- and I would like to acknowledge that. Also, thank you for the rich and professional excellence of your content. Keep it up!
2019 Belle Delphine sells bath water
2020 Steve sells GN overclocking water
"GN overclocking water" doesn't sound bad for an open loop liquid.
Hell yeah!
She got murdered for that😂
If you look on the ingredients, the bottle is filled with the tears of JayzTwoCents after loosing the OC battle.
Steves hair sold as a compounded thermal pad. We know he shampoos with Grizzly Kryonaut anyway. Would buy.
I just watched this whole video and didnt understand one thing! So proud of myself lol.
I only understand that if I slightly undervolt it the temps go down and my junction temp isnt going BAT $H!T MAD!! 😆👍
dude I don't care about stupid tests, I just want to game on a amd cpu and Nvidia gpu LAPTOP without the cpu reaching 95+ Celsius for no reason on performance mode and just destroying my laptop with thermal throttling! GET IT?
I just want to say thank you for all the hard work you guys do, really impressive what you managed to test so far and i really highly appreciate that.
So thank you.
Thank you!
This and a "Thank You".
Agreed. The Steve is LEGEND
I see this as a good thing, instead of crashing, it slows down, which is a much better alternative for most people. They just need to implement a way for you to tell that his is happening.
That would be a good idea.
so do overvolting too, detecting micro stutter is a bit annoying
I see why that can be good for many people. But, for experienced overclockers, there should really be way to disable that behaviour, because it looks like it can limit before it would actually get unstable, many say that they can get higher manual all core overclock at lower voltage than stock or PBO.
@@Ph42oN ya it would be nice to have it both ways. But I doubt it ever happens.
I noticed CPU utilization being capped off when undervolting using an offset (task manager showed 87% utilization on all threads). That's one way to tell I guess...
does any of this (or anything similar to this) apply to other CPU architectures like skylake? when I was undervolting mine I had this thought that perhaps even though the frequency was stable and it wasn't bluescreening, the scores might have actually been getting lower and/or the CPU might be encountering some micro faults or other interesting issues like increased DPC latency. the reason I had this thought was because hitting the lower volt limits (I got to 1.21v on 4.2ghz all core on a 6700k) seemed to give me the occasional quick 1 second "bzzzzzz" system hang (not a bsod, it would go back to normal right after) whereas with increased voltage this wouldn't happen.
heyyy g mannn how's it goinnn mateeeeeeeeee
Wtf why no comments
Noticed a similar behavior on my undervolted i7-9700 when I had it. It drove me crazy back then and I didn’t exactly knew why it would occasionally hang for a second.
Not really, it’s just accentuated VDroop in that case.
The blue bars on the side are a nice touch.
People are just now picking up on those?
@@SirNickyT I think it's a GN meme now... (goes out to shout at the clouds)
@@SirNickyT Being condescending just makes you sound insecure.
@@iliilili310 sounds like someone pretty familiar with insecurity. It was a genuine question because steve pointed them out like 4 times when they started using them but they're so subtle you could easily not notice them. I can't help it if you can't tell the difference between being condescending and genuine curiosity. Get over yourself. 😁👍🏼
@@SirNickyT Desperately attempting to be condescending, again. You really are insecure, huh?
We're like cavemen poking and prodding alien tech monoliths in trying to figure out how these Ryzen chips behave.
appreciate the movie ref
@@fredEVOIX Talking monkeys are real, dude.
@@CapaNoisyCapa nope 2001: A space odyssey
Not really, the CPU has some throttling status registers that are not being reported by most tools - but are documented in CPU documentation.
@@metacritical42 Indeed.
Good job once again. I particularly like the way you were clear in stating this was not meant to shame the person who did the original video, and kudos to them for taking it down once they realized the data was wrong.
Guys please help,
I built my new pc with 3800x and asrock x570 taichi.. and the idle temp is almost 40c and I don’t think that is normal.
Applied thermal paste properly and reduced LLC to level 2. Should I lower the voltage from 1.45 (default) to any value less than that?
try 1.35 and 40c is normal
@@Noooo23523 3600x here at 1.325 volts 4.3ghz all core idles at 40c max temp under stress testing was 63c
@@zackmandarino1021 What cooler?
@@_thechosen hyper 212 black edition been a really good cooler
Ryzen 5 3600 here and I get around 60°C idle. CPU cooler is an Aigo Lair. I'll try changing the top fans from exhaust to intake (I have a dust filter) and undervolting tomorrow
Kudos for the way you have the time bars on each of the slides. It helps me find the slides I am looking for when searching back. Even more important is the way you do a time bar for each question on Q&A sessions. Very helpful. Thanks
Just want to say thanks to the Gamers Nexus team for all the awesome detailed content you have been producing lately. Keep up the awesome work guys! It great stuff to watch a learn more from.
Also see ruclips.net/video/5JTuxMnrqds/видео.html How to lower Temperatures on Ryzen 3000
using Ryzen Master.
Thanks for the explanation! As always, I love the deep, realistic done into the performance hinders the CPU BIOS changes. Thanks!
This is very interesting, I honestly never really thought about it like this. I want AMD's explanation follow up for sure.
Awesome work! Love your content. I send my friends to your channel when they are looking for info on what hardware to buy. You and buildzoid are some of the hardware youtubers I actually trust.
Great set of tests. I was exactly looking for something to do with performance and temps when the Vcore is lowered for reduced power consumption. Looks like undervolting does wonders for cpu temp and power!
One question, why are you not at a million subscribers yet GN? The quality of the data you provide and nearly flawless presentation allows for you to easily attain these viewership numbers. Like always, best of luck's and fantastic work for the entire GN team.
@Blokka Nokka ikr? it is well deserved
This is entirely over my head, but fascinating to watch anyhow 😁
Just make sure your getting proper rest. your smashing it out of the park and really dig deep. cheers mate.
I can't believe that Ryzen 3000 has been around for so long, holy crap.
Same, already 1 year... not long from now it’s gonna be suddenly 8 years ago
I've been using a -0.125 offset with an LLC of 2 on my 3900X in the Crosshair VII Hero since I got the CPU. My Cenebech scores improved slightly and I got 15c lower idle temps and 5c lower temps under full synthetic load. I started doing this because the Asus bios on the C7H seemed to be putting more voltage through the CPU (up to 1.5v at idle). AMD has since explained the high idle voltage on Reddit, but I'm happy that it allowed me to stumble into the benefits of using a negative offset on the 3900X.
Even with the performance loss which looks to be 20% at 1v on the 3900x it still seems the improvement in thermals is something that in certain builds might be worth doing if you want to put it in an ITX build with limited cooling. It would be interesting to see more numbers on what the performance loss vs thermals improvement looks like at other under voltage amounts. Interesting video as always and keep up the good work.
If i recall correctly, setting a power consumption limit (PPT limit, cTDP, whatever) works better than limiting voltage, at least from what I've read from 1 or 2 reddit comments. And the thermal results are similar, without the risk of possible instability.
@Advocatus Diaboli I'm sure they hate you too.
Far better to use a SOC which is designed to run to a low power budget, 10W passively cooled is not unreachable.
Simply buy a cheaper CPU with a lower TDP rather than buy a Ryzen
@@RobBCactive Nah, that's shit.
I maxed out PBO and put about a -.08 offset on my zen+ / 2700x system and get insane results. 4.3 ghz all core with any submaximal load. You said these tests were done with PBO off right? I'm curious if buildzoid's offset instructions still apply.
Optimum Tech is a great channel! You guys would love working with him!
This is super interesting stuff! My overclocking practices generally involve simply running Prime 95 for 10-20 minutes to make sure my voltage is stable for the clocks I'm getting. I'll have to keep this in mind as I start tuning clocks on my 3600.
20 minutes of stability testing isnt enough...
One of the few no bulshit channels on RUclips
interesting glad you made this video cause i planned undervolting my cpu due partly pairing with a b350 gaming f and hoping it would assist vrm temps. in the past ive undervolt fx processors but never verified if performance degraded
I got 3900x with a x570 taichi at -160mv offset completely stable, my temps decreased about 10-15° and i haven't noticed any performance loss
Recently I noticed this behavior while tuning my Vega 56. Voltage at 975mv p7 seemed stable, clocks were at ~1620mhz while monitoring, but during games I was actually losing 10-15 fps compared to 1075mv.
Temps definitely improved but at the cost of performance, even though the numbers all looked good during stress testing.
Could you please do a deep dive into memory Frequency & Latenzy scaling inkl. IF OC too - that would be so awesome!
I've also just been patiently waiting for memory scaling. GN's initial Ryzen Memory coverage was the best by far, and curious to see how much of that behaviour carries over to Zen2
Ya, new too. GN does great coverage on the original Ryzen memory frequency vs latency. Like to see if Zen 2 follows in the same way.
lab501.ro/procesoare-chipseturi/amd-ryzen-3000-part-iv-ddr4-scaling-english-version
Best article I have seen so far.
Pls
I'm sure hes working on it though it seems obvious
www.techpowerup.com/review/amd-zen-2-memory-performance-scaling-benchmark/
this is good to know, thanks. I have a -100mV offset for my 1600 and I believe I'll need to remove or reduce that when I plan to upgrade to a 3700X
Power figures would have been nice so we could see how the performance per watt changes when severely undervolted.
"Its not wise to undervolt Matisse, unless you enable the OC-Mode (fixed frequency).
That's because the CPU is constantly monitoring its supply voltage.
When it sees the supply voltage below the desired level, it starts increasing the VID request from the VRM controller.
When the VID request increases above the maximum voltage allowed by FIT, the clock stretchers will kick in. Your performance will decrease, despite the change is not visible
in the frequencies diplayed by the usual monitoring software.
For example on the 3900X sample I have, the default VID request during CB20 NT is 1.3125V.
With the offset set to -25mV the VID request raises to 1.33125V or so. At -75mV the VID request is already 1.36875V and increasing the negative offset any further will trip the stretchers.
There is an undervolting margin available, but it is tiny.
The exact behavior will depend on the silicon specimen and the workload.
Its also the very reason why this guy thinks he is running at > 4.2GHz with just 1.00V..."
Source: www.overclock.net/forum/28044514-post7.html
Here's a user's own UV testing: www.reddit.com/r/Amd/comments/cfdog5/realworld_3700x_analysis_and_other_ramblings/
Nice copy pasta
There is one guy that ubdervolted his 3600 and is happy. He shows me benchmark results. c20. aida64... temps dropped to 55 in game and benchmark scores became bigger.
@@Kern1909 and who's that guy?
@@syedsaadnasim You can count me as well.
@@Kern1909 tell me your settings in detail please
Hey guys sorry to nitpick, but volts * amps = watts (power). You have it messed up in the description.
Thanks for another eye opening segment . Had a strange feeling about things when a Ryzen 5 3600, already overclocked to 4.5 GHz and started at a Vcore of 1.38V, seemed to gain some serious thermal headroom with each decrease of the Vcore value. Ended up not going any lower than 1.5V ( in Override Mode) before I started having concerns and questioning a too good to be true scenario. Then I stumble upon this. Thanks for ALL the great research and crunching GN & team.
My 3600 is a dud. It can barely handle 4200ghz at 1.38v on a tuf x570 plus MOBO, seems to be a bronze sample 😂
you've degraded your CPU badly if that's at a static oc. the limit for matisse is 1.32v anything over that will cause degradation. I did the same thing.
I've had the best experience limiting the watts limit. Even a small limitation of this option has a very positive effect on the temperature. Without losing much performance .
Why oh why?? Temperatures are a byproduct of doing work.
Instead of limiting peak watts you should just buy a cheaper CPU or use power management to throttle the CPU.
Under-volting has the potential to reduce waste heat, transistor gate leakage and improve efficiency, leading to better performance with that specific die.
It all depends on the silicon quality being high enough to operate at lower V and hence reduce power consumption.
But if dynamic chip management regards under-volting as an error due to bad p/s or mobo then the whole technique is invalidated.
@@RobBCactive my dude this solution / tip is over a year old.
today there are good power plans or CTR that help WITHOUT any loss of performance
so thank you for your time for the comment but "context" and "time" would have saved you these lines!
I'm running my 3600 at 1.168v 4000Mhz all cores, using the Ryzen Master software. If I go one step below at 1.1624v, the system crashes
Cinebench R15 Scores
Stock Profile - 1577
Undervolt - 1588
I'm using a stock cooler btw
My explanation would be that during stock settings, the V rises to 1.4s to touch the maximum boost clock but then it heats up the chipset instantly to 75 degrees or so and it throttles it back.
Using the undervolt, the temperatures don't rise as much and that's why it's able to stick to 4000Mhz. Also that people would say the software reports incorrectly, well if I underclock it a little more it doesn't send. Which means that it is in fact sticking to the voltage I'm giving it
I started messing with Vcore on my 3700X literally a day ago, this is precisely the video I needed.
whats the best settings for cool temp n Voltage for anything above 4ghz on all core
yeah, nvidia pascal did the exact same thing. Truthfully Radeons do not, but stability flies right out of the window though... even if performance 'actually' improves.
I personally have found a sweet spot for the old 2700x at 1.2v fixed voltage with 4ghz on all cores. Temps and consumption are down (the spikes of the auto boost are crazy), multicore performance slightly better, single core workload slightly worse!
For me it's a win-win better performance with locked 1.2v instead of 1.45v spikee
Same thing on my 3600 :)
My sweet spot for my R5 3600 was 1.275Vcore, 1.25 VID, runs stable it seems at 4300Mhz
@@Jacob_Overby hilarious how loads of 3600 chips clock another than my 3600x
@@nolef7nu7 I couldn’t get mine to go over 4050Mhz, these guys’ chips are just built different
AMAZING! The new Ryzen CPU's are really Cool for all Tech Nerds. Every Day we find some new things
Nice to see everyone is enjoying their AMD experience.
Thanks for this video
Would be great to also see performance with PBO and voltage offset, as I'm pretty sure that's how you get the most out of it. Ask Buildzoid as he talked about it in a past video. Thanks.
Undervolted R5 3600 with Ryzen Master 1.18125V @ 4.2Ghz and it works like a charm (air cooling). Temps down few celsius and Cinebench R20 went up 3460->3810
i saw that video, seemed legit with all the comparison tests. but yea with new devices being tested in different configurations it can get a bit tricky to work in ways you want it to.
I am curious if small small offset undervolts will have an affect on PBO behavior. Would it enable higher boost clocks? Or longer duration boosts? If there is a benefit, is it solely based on lower Tjunction?
Do DDR4 speed dependancy benchmarks next
Reminds me of gpu memory overclocking. At upper limits the higher clocks tend to diminish performance contradicting the gains seen in lesser clocks.
So decreasing it down to 1 volt lowered heat output rather dramatically, and also lowered performance as well.
Though, did power efficiency improve or worsen?
Taking the "CB Mark" for CineBench R20 nT score difference, where the 1 volt undervolted one performed 80.8-82.5 % of stock.
And the corresponding graph for the temperature Tdie, where we can see a reduction from about 60 C down to about 30-40 C, and if we say that the ambient temperature were a constant 25C, then we could say that the power dissipation reduced to only about 30%, though extremely rough calculation based on the idea that a heatsink has a linear delta-T/watt figure through its operational range, this would mean that the CPU's power efficiency is about 2.7 times higher. (I would say that seems unrealistic.)
So an actual power consumption comparison plot would have been far more interesting, then the Tdie plot in my opinion. (Since Tdie will greatly vary with what heat sink one has, the power consumption would mostly be effected by clock speed and feature utilization on the die itself. (leakage currents are though temperature dependent.))
In the end, how much did the power efficiency actually improve? If it improved at all. Could be useful to know if running these in servers or the like.
extremely likely that efficiency improved, most consumer chip products are on the upper end of the scale, around the nosedive in efficiency
my ryzen 1700 runs at 1.08v all the time at 3.2 ghz under boost automatically and has very low temps and 45w tdp and the stock cooler thats 95w can overclock good with noctua normal silicon paste before it dries after 3 years and goes back up to 80c to 90c from 60 - 80 c and i overclock to 3.8 ghz and 3200 16 16 16 35 cas ram on 1.350 v core and 1.360 v ddr and i tested the cpu in burn benchmarks not only games where everytimes stays under 60 but crashes when the paste is dried because rhe temps rise up too fast
I wonder if Steve is planning an individual ccx overclocking video. Der8auer had some success that way. Seems like the best method of overclocking so far. I'm just hoping board partners start implementing it into their bios.
Yeah, but this tool Roman used to CCX-OC isnt widely avalaible.
Keep it up! Still much to do like mem scaling and fclk!
Thanks to playing with this I was able to set my Ryzen 5 3600 at locked 4Ghz at 1.15 volts and not only I got rid of the saw-tooth reports in monitoring software but my CPU runs stable at freaking 18'C cooler. All tests like furmark and cinebench, 3dmark report same or slightly higher values. Still worth it for almost 20'C reduction of temp. Either it actually works or I was very lucky with the chip.
I tried a 3600 at 1.325V and 4.125GHz, it reaches 72°C on booster. In Stock it was 85ºC
It gets 1567pts on CB15 and it was 1508pts in stock.
I used AMD Ryzen Master program to test.
Good vid noticed on my 3900x goes for soc voltage to.
great you bring undervolting to attention, can you do some tests with the Treathripper cpu`s? i will try things anyway as building a TR4 2920 server in a 4U case , so limited cooling and 24/7 use it might do some good to have a slight undervolt to limit heat and poweruse.
you should also try the positive offset to compare, it will show how the cpu behave in certain condition
I noticed same with my 2700x before. I don't think it's a bug, more like precision boost works. Clocks are set based on actual temperature, power limits, voltage delivered! Etc. You deliver less voltage, frequency doesn't jump as hight.
Same with PBO - all it does is set much higher power limits AND raise CPU voltage. Result are higher frequencies set by Precision boost.
Its like that since Zen+.
Optimum Tech, the KING of compact builds!!! Excellent youtuber.
Yes, props to him for the quick response on the correction. He admitted he was fooled by the numbers and pulled the video as soon as realized what was going on.
I'm not into SFF, but subscribed and watch his videos.
-.1v offset looks really good, not a fan of 1v though, thanks for the work!
I flashed the latest MSI B450 Tomahawk bios for 3000 chips even though I have a 2700x.
I gained 500 points in Firestrike and system is very stable.
I use another method, the kingdom come deliverance initial screen, in window mode at the lowest resolution possible with the lowest settings possible.
This is my cpu vs my cpu in various power plans, so no problem related wit that one "shilling situation". The monitoring tool is the HWinfo.
I've tried all possible overclocking scenarios and the voltage is still spikes like crazy without any meaningful effects.
So the best scenario is to let everything stock in the bios and overclock only the ram and put the Ryzen power saving plan in windows. The performance is nearly the same, without seeing the voltage and temperature spikes bouncing all over the place.
I use the Vega 56 and the fps is 270 with the low power and 286 with the maximum performance.
Also the voltage stays at 1.31 under 100% load and 1.38v boost.
Best balance I've been able to find for my 3700x is setting a negative offset of 0.09 while also enabling PBO with 200mhz. As I understand it, the PBO setting kind of counters the negative offset, but not the extent that it ever really over volts like it did stock. My out of the box vcore was damn near 1.48 under load on an Asus x470 gaming-F, and temps were in the mid 80's in cinebench r20. Now my temps don't exceed 76 in the same test, and my vcore under load hovers between 1.34-1.39. My cinebench score actually slightly improved, and my multi core results are consistently between 4,865-4,900+, while my single core is 495-499. I think this slight improvement in results and thermals mirrors your results on your 3900x with negative 0.1 offset.
Idle temps also dropped from mid 40's and 50's to low to mid 30's now. Gaming temps in really cpu demanding games like BFV are in the low 60's , and most games stay in the 50s.
That's great! Are you using the prism cooler?
@@KapanthPT yep. Stock cooling. Coming from a delided 4790k with a noctua I wasn't sure what to expect with the wraith, but I'm honestly impressed. It's in my old HAF 922, and that case has good airflow, so I'm sure that's helping a bit too.
I'd like to know the real reasons the 3000 series is volted the way it is. I personally and hav seen from countless others, that lowering voltage to around 1.25-1.275 (peak) results in much lower thermals and as good or better cinebench scores +/-2%. AMD must have known this. What is going on there? Something to do with peak/consistent max clock preformance?
Seriously?
Compatibility. This is why OC'ing exists at all, companies clock their CPU's conservatively so that when you drop them in a system they just work.
What you're doing is just dropping the voltage to match their conservative clocks.
Factory settings have to work on every CPU in that skew, locally tuned settings only have to work for the hardware/software actually used.
Seriously people oc GPU and it seems to work fine but months later a different game has crashes until the boost is turned off.
People don't like dynamic CPU management because they want to say the RN YZOK has 3.7 Ghz base clock with 4.2 precision boost, with this benchmark score.
Imagine if the numbers changed greatly because you dialed the a/c lower.
My only question is if you had someone who cared most about reducing noise, would doing this extreme Vcore voltage help them? As your results indicates a lower temperature which implies lower fan speeds needed?
Great video! I've been experimenting with this a couple days now. After doing what you've done do you recommend setting a V offset or adjusting the Vcore to a single value like 1.3?
My cpu seems to be ignoring my offset values
You shoudl do a test with eco mode. I changed my 5800x on the master to eco mode 65w and it still runs 15661 in cinebench and while gaming with a 3090 it runs at 5 to 10% and even lower. So for gaming eco mode is a must use in my opinion
I'm gonna need all of those graphs.
I dropped 5 degrees from my idle temps with an undervolt to 1.35, from 1.45 on my 3900. Also took the system to 4200mHz all core, and saw a 200point cinebench improvement. Will try with the voltage at 1.45 4200mHz just to see.
Oh nice you updated the large mat to have to dual silhouettes, wish my medium mat was bigger, large mat is too big and medium mat is too small
After watching this video, I decided to do a performance validation on my undervolted Ryzen 5 2600X. Running Cinebench R20 twice with stock auto voltage and my -0.1V offset, I noticed that the undervolt led to a 2% increase in all-core performance with no change in single core (with a noticeable change in boost frequency from 3975 MHz undervolted down to 3850 MHz at stock auto voltage), a 2-4 degree reduction in CPU peak temperature, and a 3-5 watt reduction in power consumption as measured by my UPS.
You can increase efficiency further by modifying the ryzen balanced power plan to use a minimum processor state of 5% instead of 90% and raising BCLK to 102mhz or higher, however do note that this might bring potential instability.
@@thetrashman5252 I don't have a Ryzen Balanced power plan, even after installing the latest chipset drivers.
@@b0bsaget007 Strange, try making the same modification on the regular balanced plan instead i guess, the results should be similar.
@@thetrashman5252 it seems to already be set to 5% minimum by default.
@@b0bsaget007 Interesting, well in any case if you want to still increase performance you can try raising BCLK, however that might require a less aggressive undervolt and might not function on some boards.
maybe this is explaining why when overclock to lets say 4.3-4.4 on all cores and manual volts 1.35-.1.39 cpu performance is lower. because 1.35 is lower than auto 1.47! maybe this is a key problem of manul zen2 OC?
This is probably tied to Ryzen throttling behavior also seen on Threadripper Zen+. When CPU overheats, Vcore is dropped like a brick (to about 1V actually) and performance obviously tanks. So the flat value probably invokes throttling, specifically maybe EDC throttling in this case? (Low voltage, CPU still wants similar power but can't get 50% more current.) Or special separate failsafe mode.
This is for use in laptops, I suppose. I'd check if PBO affects the behavior in this scenario. It might.
Been playing with my 3700x since getting it, I've found 2 things that have helped performance, thermals and voltage with my ASUS X470 PRO. Simply setting the vcore to an offset of -100mv kept things quieter, cooler and performance was almost identical, another way was I had it running at 4.2Ghz 1.352v, I've noticed at stock it will run at 1.331-1.362v during an all core 4.2-4.3 boost, I've never seen it actually hit 4.4 yet, I just can't stand those voltage spikes while idle which cause the fans to ramp up like crazy for no reason, even with a very strict fan curve so I think I will be keeping it running at 4.2 1.352v, I know there was a post saying Ryzen 3000 series shouldn't go further than 1.325v for all core but I honestly don't believe that at all seeing as it is doing higher than that at stock when all core boosting, also cores can still sleep and slow down even with a manual overclock which is very interesting, even though HWiNFO doesn't display that, Ryzen Master does show lower core clocks and sleeping cores even with a manual overclock set
That was my concern: cores not sleeping and always run at 4,2ghz. I will look into it, since my 3700x runs 4,2ghz on 1,25volt on aida64 90min stress run. Temps dropped 10 degrees and especially the max temp spikes didn't exceed 90 degrees celcius.
I have the same processor but a different asus motherboard , how do i set the offset in mv instead of v? I can only do 0.0625 adjustments
I've been messing with undervolting my 3900x with a negative -0.1000v offset, now while my machine's performance was sort of on par, Cinebench etc were fine, did some x264 encoding , but it was completely unstable on idle, it would bluescreen as the voltages dropped to low.
That wasn't straight away, this was after a good 30 minutes of idle.
A -0.050v offset is stable on both load and idle, but temps jumped up by a good 7 - 9c in Cinebench and AIDA64 stress test.
Coming from Intel it's been a whole new learning curve with Ryzen 3000.
i got a 3900x too and I'm missing my 4790k. Intel just works rock solid. I'm regretting not building a system with 9900k.
Thanks for the work you all put in. If building a 'compact' ITX travel PC would undervolting be a viable means of mitigating temperature throttling?
yes
and it also uses lower power consumption
Could you revisit this ultra low voltage stuff but for the APUs (65W) I'm interested to see if it can be made to run stable and passively cooled.
I'm talking 20-30W heatsink(s) that fit in a htpc case not massive tower coolers
isnt there a tdp setting instead? or maybe it would just auto throttle
When I offset-undervolt Ryzen 5 3600 the clock speed goes up a little bit, BUT the max cpu usage CAPS OUT at 87% on all cores (clearly visible in task manager). Resulting in less power consumption and less MAX performance (like cinebench), but non-CPU intensive use scenarios (like gaming) might actually see a benefit still. Only validating with cinebench has its downsides too.
The one time I messed with the Vcore override (R7-2700 @ 1.2V) my system crashed almost right away. I just kind of assumed it's because I needed to do something else and it was beyond my knowledge. But I've had good luck using the Offset mode and reduced temperatures pretty significantly.
Hey Steve, do u plan to make a video about this topic to ryzen 5000 and the curve optimser?
But did you set a fixed CPU Speed or did you leave the speed at auto? Im thinking that the cpu speed adjusts to the voltage..
"Poking them with sticks and throwing rocks at them." 😆😁😂
Hi Steve, first of all, congrats for the videos.
May I give a suggestion for an upcoming video. Could you please show if the PCI-Express extension cables still works when using a PCI-E 4.0 card and motherboard?
Thanks a have a great day
i think so, there's no physical change on slot itself
@@ouwkyuha yes but the requirements for the signal are way higher. I watched a video showing these extensions were not working anymore
Just got a 3800X for my old B350 TUF board to replace my 2600 and noticed it was super hot in comparison - offsetting core voltage by -0.1v lost about 4% performance in Cinebench multithreaded and less than 1% single thread, but -0.075V lost no performance at all and dropped my max temps under load by 6C.
Great info and testing. Now what I really want to know (and this may only come from AMD) is why? Error correction? I don't really know what could allow the clocks to stay at full speed at that voltage, but still show a massive performance dropoff. Lowering the clocks behind the scenes then misreporting? I would hope AMD is smarter than that, because that would be ripe for controversy.
Any effect on power consumption? I'm gonna re-purpose my X570 Aorus Master & 3700X as a home server (proxmox with a bunch of VMs), so I'm probably going to use ECO mode, but maybe undervolting the CPU (and maybe other components?) would help further. Was hoping this video would shed some light on that. Oh well.
So between this video and the one testing boost clocks and temperature. I finally have my 3800x boosting past advertised max. Before I couldn't get it to boost over 4.3 on any core. Even under water cooling CTR identifies my chip as a bronze sample, it doesn't OC well at all.
I picked up a Freezer II 420 and fitted it in my case with some modification. Then started playing with PBO and voltage offset. Now I'm running a -0.1250v offset and hitting 4.6GHz on any given core. Holding 4.4GHz in a 20min CB20 loop never going of 61°c. Not only are my benchmark scores finally equaling the 3800x averages they're even a little bit above average. Finally feel like I'm getting what I'm supposed to out of my CPU.
When I set the voltage to a fixed value, I noticed that the task manager shows that the cpu is not utilized 100% anymore
Do HEDT cpu test please (Intel 1MB L2 cache per core) whenever you have time for it.
great content.
A -.13V offset bumped my single and multi core scores in C15 and Passmark on my 3900X - better than forcing a 1.352V
I'm on an ASUS Crosshair VIII Hero which appeared to have high voltage at idle and load.
thats nice, but try stability testing instead
@@jscb7003
How much more stable do you need it?
thats happening on others also Taichi has some voltage scaling issues at idle you would think before releasing this platform they would of figured this crap out
And I thought I was cool with my 102mhz BCLK -0.1v 2700X.
Your mileage may vary depending on your setup. On Asus X470-F Gaming + R7 2700X + Wraith Prism combo undervolting with vcore offset is beneficial. Because on default Asus X470-F Gaming outputs ridiculously high vcore and that heats up the cpu too much with Wraith Prism boxed cooler. When setting vcore offset to -0.1 you get higher boost clocks that stay on longer and your idle/surfing temps decrease thus leading to quieter system. After vcore offset of -0.1 the CB15 score went up quite a bit.
Exactly what i done on me asus strix 470x-f hooked with Ryzen 3700x ,after i put undervoolt offset -0.10000mv, i get lower temps like 35C idlle and 65C load max,with pbo,xfr and smt all on,processor still able up on all cores 4.395 gz and that result in Cinebench r15 score=2100,wich i think is good.
Has Everyone forgot to get good temps on over half of the 7700k's these voltages had to be reduced which would return a < 20'c drop, so it would make sense if there is improvement over other chips, just not as great of an improvement
Although I enjoyed the follow up, I have a question that wasn't answered:
Is the 1V option viable? Meaning, if someone wants to use his cpu for very light workloads for a few hours, doesn't care about clocks and wants ultra low power consumption and passive cooler temps, can it stay stable?
Mine does stay stable at 1.100volts at 3.6 ghz with the boosting things off
The best thing I got from this video was that unlike mine, your 3900X actually consistently stays at max boost clock during a 1T R20 run.
Mine at default settings runs about 300 MHz lower and I can't improve on it at all.
I sure hope a BIOS or X470 chipset update will fix this.
Is your cooling solution working correctly? Might be thermals that hold it back from boosting
@@asmodin88 I have a Noctua NH-D15 and my case has very good airflow.
it seems a board problem rather than cooling problem. Maybe VRMs overheating?
@@user-po8vd5gp5w X470 Aorus Gaming 7 Wifi. VRMs not getting above 50 degrees.
why use default settings? set vcore 1.35v and run again everything. Also you can try to use Asus tool to CCX OC check der8auer video.
Yep, undervolting outsmarts the pbo, because temperature becomes at some point most important factor on boost
Interesting info and seems to pretty much mirror what I have seen, my 3900X was running at 1.42v out of the box (ASUS Strix-F, BIOS 4012 - still not updated). I manually set it to 1.2v - worked fine but only boosted to 4.1Ghz. Now running 1.375v (shows as 1.38v) and it boosts to 4.55Ghz - still can't get that 4.6, might have to try 1.85v.
Temps at idle are around 36 now (was 49 at stock - but I also had a crap mate with the water block which did not help), but does climb into the 70's under god load - running Handbrake 1080p encoding. If I run the radiator fans at max speed it never gets above 76 though.
Not going to update the BIOS till a lot of the issues are worked out - but I do wonder if most of the behaviours (temperatures, voltages, boosts etc) are all down to very immature BIOS/AGESA? Either way, the 3900X makes light work of most things :) I should also point out the the 4012 BIOS for the Strix-F mucks the RAM voltage right up, so had to fix that manually as well.
1.85v? are you trying to cause a house fire or something? we're not on 32nm anymore my guy
And that is why you should always proof read before posting ;) I meant 1.38 naturally ;)
I undervolted my rx 570 all the way down to 970mV at its stock speed of 1250MHz, and got a very, very slight possible improvement by +1 fps on average in benchmarks, but also about a 20C drop in temps, with the max temps eventually settling around 67C in furmark without even having to spin up my fan super loud. initially I'd tried overclocking it, but had pretty abysmal results, and i could barely even get it stable at 1300MHz. so i tried undervolting instead and I guess I just got a golden undervolter. I've decided im pretty happy with the stock performance at this point, given how little power its sipping and how little heat its producing.
and ur gpu die very fast;p
@@Venom-gs4nz I've had it like that for over a year now and it's never had any problems. Still performing great. In any case, undervolting doesn't cause increased wear. It's generally best to run it at the lowest voltage that is still stable if you want to increase the life of a processor.