No video :(

How to Kill Your CPU with "Safe" Voltages

Подписаться 2,3 млн

Просмотров 223 тыс.

50% 1

We demonstrate how you may be pushing more voltage than is safe into your CPU or SOC, done by measuring at the socket of an R5 2400G Raven Ridge APU.
Ad: Buy Thermal Grizzly Conductonaut on Amazon goo.gl/pygud5 (or their standard Kryonaut on Amazon goo.gl/8ySAVs)
Article: www.gamersnexus.net/guides/32...
How to Kill Your Motherboard: • How to Kill Your Mothe...
What is LLC video: • Explaining Load-Line C...
This content follows-up on our "how to kill your motherboard" video, focusing instead on the process of killing an integrated memory controller in an APU or Ryzen CPU. This also applies to Intel CPUs, though the voltages and motherboards are obviously different. The main concepts remain the same, though, so everyone can benefit from this topic.
Anyway, we demonstrate how the HWINFO (software) reading of voltage is inconsistent with the socket reading, and that the two can be a large enough gap to cause damage or degradation of the IMC over time. This is done by checking SOC voltage with a multimeter vs. with software. We also use this to demonstrate the difference between SOC voltage and APU GFX voltage, further leveraging the point to talk about "safe voltages" for Ryzen & Raven Ridge overclocking.
We have a new GN store: store.gamersnexus.net/
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
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.

Опубликовано:

27 фев 2018

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 738

@GamersNexus 6 лет назад

Article: www.gamersnexus.net/guides/3251-raven-ridge-soc-voltage-guidelines-how-to-kill-cpu-with-safe-voltage How to Kill Your Motherboard: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-vLMd-5yxTAc.html What is LLC? ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-NMIh8dTdJwI.html

@JubulburbWuldrd 6 лет назад

So high voltage can be dangerous, and as you name only voltages(and i can see that's pretty universal in pc hardware) i assume what vrms driving soc outputs constant current? If no, why do we care about voltage but not about current?

@redhawk3385 6 лет назад

Use the killed cpu as a box opener

@isoinebriation5060 6 лет назад

@Gamers Nexus So my question was partially answered with the LLC video with buildzoid, but I'm still unsure. The short version is simply, since voltage reporting is mostly highly accurate "estimations", how possible is it that the reported cpuz voltage is higher than the actual voltage, the opposite case of which you're showing here. Reason I ask is I have a R7 1700 under Arctic's 240mm aio, and was playing around with the OC seeing how far I could push it. At one point I had cpuz reporting 1.51v and yet prime95 only got the cpu to 75c (this wasn't an everyday clock, I turned it way down since, calm down comment section). Since this performance is much higher than I've expected compared to multiple other reported overclocks, I'm curious if that 1.51v figure reported by cpuz is accurate or not, and by how much.

@thefirehawk1495 6 лет назад

This is a below standard quality from what I've been used to from GN, all I've seen in this video is wild speculation that you may be putting more voltage into your CPU than you actually are without knowing what sensors (and where they're placed and how they work) are reporting what voltages to what software and then comparing that with non-representative measurements taken from the socket (because socket voltages aren't necessarily the same as what actually goes into the CPU due to series resistance and even though one of the probes was actually on a molex and not in the socket. If you had 6v going in and 5v going out of the cpu the cpu would have 1v, but your molex-socket measurement would report 6v, I'm not saying the measurement is necessarily incorrect of gives necessarily incorrect results but it does introduce variables that could be easily removed or controlled for.).

@TheJohdu 6 лет назад

very good video. especially after most YT reviews showing their overclocking procedure setting both soc and gfx voltages to 1.35V :-o

@aidan4672 6 лет назад

So... I shouldn't round it up to 2V?

@VarietyGamerChannel 6 лет назад

Unless it's a 486, no.

@vnyggi621 6 лет назад

5v is ok too, just dont go further ;)

@kendsplaining 5 лет назад

lol 10v is pro gamer mode 😎

@theunholybakery1990 5 лет назад

85v is safe for celerons. They deserve that voltage.

@somebody2978 5 лет назад

Set VCore to 220v in BIOS and will enable FUTURE mode in BIOS. Make sure your CPU has a flux capacitor and a cooler compatible with TDP 1.21 gigawatt.

@batmangovno 6 лет назад

Oh shit *runs to downclock his IntelHD*

@Acenis 6 лет назад

FeelsBadMan

@TheRguru1 6 лет назад

That IntelHD Graphics is going to overpower your monitor and cause it to detonate.

@coolcat23 6 лет назад

The speedometer on my car shows "50" when I'm really driving 75 mph. That's not a fault, just a behaviour.

@bizzy156246 3 года назад

you should buy a new gauge motor or whatever they are called unless its digital then it might be the instrument cluster thats bad

@barrymckocinher4649 3 года назад

@@bizzy156246 I'm pretty sure they were joking...

@overclockeverything2310 6 лет назад

Still waiting on my 4690K to die its been running 1.51V for over 3 years.

@aidan4672 6 лет назад

If it's cool, it'll last quite a while.

@overclockeverything2310 6 лет назад

Yea its delidded with LM tim on a custom loop.

@Mikey-lx9xs 6 лет назад

No degrading?

@overclockeverything2310 6 лет назад

Haven't had to do anything its been fine. I push all my hardware to the limit. Need an excuse for upgrades :P

@AssassinKID 6 лет назад

CPU are made to last for a couple of decades when running within specs, even with mild voltage degradation their useful lifespan won't be affected. Most of the time the motherboard is more likely to die before the CPU does.

@Qyngali 6 лет назад

I dare say that the MB manufacturers ARE responsible, they should do everything in their power to tune the boards so they report correct numbers. At the very least make sure that you don't underreport it, that's a recipe for disaster. Which incidentally is why I never recommend Gigabyte to novice builders... Hey Gigabyte, go buy your engineers some multimeters ok? (Well the other guys as well of course...) Edit: I was intending to add something but I can't remember what it was. Oh yeah, the state of bios support for the APU's are still quite rough, that the MSI board got non functioning GPU voltage control right now is probably just that. Let's hope they improve on it on current boards and don't gloss over it to push sales of the new chipset boards that are coming...

@fartonaut2291 6 лет назад

D Bucky You should be able to expect them to be somewhat accurate.

@Dracossaint 6 лет назад

Fartonaut resistance varys from board to chip though even. So getting it exactly accurate or llc perfectly on the nose would add significant cost due to the fact it would take a hardware monitor similar to volt metter except it's inlined with Internal wiring which ya also would have to either run it to a standout point and hope a user would properly ground it to the standoff when they install it in a case or have two wires coming off it. One to be taped on where ya wanna get x voltage reading and one for ground via molex or some other way. This will not work and adds extra cables. It's best to just use a volt metter and take off the side panel of your case. Easy peasy, what they should do is clarify such things as Steve has mentioned since overclocking seems to be championed (I'm unsure if that's too dramatic or not?)

@KenS1267 6 лет назад

That's simply absurd. You're basically asking that every MoBo have a high accuracy voltmeter built in. Go price such. Would you be willing to pay that premium on every MoBo? They use reasonably accurate techniques that are sufficient for 99% of end users needs. If you need more accurate readings then get a good voltmeter and take direct readings.

@leexgx 6 лет назад

They should be more or less be around the correct voltage (most are what +/- 5%) not 0.1v higher then what you entered

@RyuFelinshire 6 лет назад

Except... the motherboard is reporting the correct voltages... the CPU isn't. You can see it in the video if you're familiar with the sensors on HWinfo64. The motherboard VCORE SOC sensor, which is cut off to read VCORE... is showing the correct voltages.

@Bakamanzz 6 лет назад

This is some fantastic content, this will be coming in handy today as my Apu just arrived pretty crazy the differences between software and the actual reading did not expect that huge of a jump but will keep that in mind going forward thanks guys!

@kaos. 6 лет назад

This is the kind of video you would expect to see on a enthusiast/technical channel. But only Gamers Nexus does it. Thank you Steve! ♥

@d3c0deFPV 6 лет назад

I know the input values aren't exact, but I didn't expect this much of a difference. Interesting insight, thanks Steve!

@cozacmihai 3 года назад

You should take the mainboard ground as negative reference, not the power supply cable. There is a voltage drop across the power cables so the voltage read will be higher than actual Vcore voltage. Or even better measure directly on a capacitor of the respective voltage.

@soundspark 6 месяцев назад

Excellent point. When I was trying to scope out my motherboard I did clip the ground to a grounded piece of metal on the motherboard, but I was only checking the outputs of the VRMs for stability.

@dm223 6 лет назад

awesome work, addressing the real issues of overclocking.

@stereodark 6 лет назад

A great research piece again! Well done!

@keithberwanger6595 6 лет назад

Next up: how to kill the GPU you are going to sell used to someone after block chain mining becomes obsolete. FT bitcoin and 95+ efficiency.

@keithberwanger6595 6 лет назад

To be fair, you kill power components on the card before the GPU dies, but it's a joke so I'm sorry

@gen-X-trader 6 лет назад

what will make it obsolete?

@keithberwanger6595 6 лет назад

Kyle Van small scale mining operations fronted by individuals will die out once cryptocurrency and blockchain become so widely accepted that "casual" individuals will be using it. Once the general populations begin using something scams become more widespread and soon enough federal regulations are placed upon it. Regulations that can track usage of coin and will eventually find a way to track individual transactions then the big original advocates of it drop off and then it becomes an everyday kind of thing that is no longer feasibly "mined" on a GPU array due to exponential growth, taxation, and regulation. Blockchain and cryptocurrency are logical evolutions, with evolution come changes.

@gen-X-trader 6 лет назад

Keith Berwanger you seriously think that's happening in the next few months? I have a hard time visualizing major changes before the next halving. No one really cares what's going to happen in 2025. People are going to be buying things based on this year. Bitmain owns such a crazy amount of crypto they by themselves can fluctuate the market and accelerate or reduce demand. I don't honestly understand why so many miners are frothy over the video cards anyway. You can get the T9 and S9 all day long and you don't have to worry about multiple motherboards or PC setups to run video cards

@keithberwanger6595 6 лет назад

I don't /seriously/ project anything happening. I do /seriously/ think anything can happen with such a volatile concept. I won't pretend to know the nuances of it all, but anything that can grow exponentially can mutate exponentially.

@TheGuruStud 6 лет назад

No one is doing these kinds of vids! You guys deserve FAR more credit.

@JP_Stone 5 лет назад

Yea almost kill my CPU until I was informed about this. Running my 2700X at what I thought was 1.75V but the MoBo was actually pushing closer to 1.4V. Realized I had followed bad advice and 1.25V (which is pushing closer to 1.35V) is enough voltage to sustain a stable 4.2GHz overclock. To many people in forums and online in general recommend just whacking up the voltage to "stabilize" your overclock.

@mattsmechanicalssi5833 6 лет назад

Great stuff as usual, Steve! You have to be THE MOST TRUSTWORTHY GUY ON THE NET. Your level of information far outstrips most people out there. Your dedicated, and accurate, and you have great hair! Keep up the great work, Matt

@astilen5647 6 лет назад

Calm your ti ts.

@MyD69 6 лет назад

Thanks for existing. Keep up the great work!

@GurtTarctor 6 лет назад

Ok then, now I need to look up how to measure voltage from the back of the board. GN the best tech channel on RU-vid strikes again!

@jcoc8877 6 лет назад

As an XOCer myself. This was a good watch. Lots of good info here.

@paskowitz 6 лет назад

This is why I like top ROG boards. Voltage contacts on the front of the board and cleaner/more consistent power delivery. It would be interesting to see a board like the Apex and the most basic Z370 board. It would be interesting to add this metric to mobo reviews. Although, this doesn't really matter for discrete GPUs since all modern ones are voltage capped and don't perform better with higher voltage since clocks depend more on temps.

@ReanuKeevesAus 6 лет назад

paskowitz when I built my Ryzen build around cyber Monday deals last year the reasons you stated about Asus ROG boards is why I went with the Asus ROG X370 Hero board. Before that build all the ROG components I had or used in builds for friends and family were absolute beasts and were parts I could rely on!!

@snoitseuqpi1119 6 лет назад

I must say, the Z270 Maximus IX Apex is the best. So excessive for sure, but also for what it was built for, I don't think it can be beat. I also have the z270e strix board, and to be fair, if you want to OC, the heat sinks are decent on the VRM, and the VRM works great, with most of the options that other ROG boards have.

@BLOEDVLEK 6 лет назад

Yeah I love my ROG RVE10 and I'll always buy high end ROG boards for all my builds, the only thing to be careful of is that you make sure you go in an manually set ALL voltages as on the auto settings ESPECIALLY with x.m.p enabled, these boards do push a lot of voltage and there are lots of reports of these auto setttings killing cpus.

@berni8k 6 лет назад

That is not the right way to measure the core voltage !!! You are not only measuring the core voltage but also adding in all the voltage drop on the cables from the motherboard to the PSU. Might not sound like much but there is easily 20A flowing back trough the ground cables (12V 5V 3.3V rails combined) and there can be easily 0.01 Ohm of resistance in the cables this can cause an extra 0.2V to appear (Amps*Ohms=Volts) The correct way to do it is to touch the other multimeter lead to a grounded pin ON the actual motherboard, ideally touch it to the other side of the capacitor you are measuring. If you repeat your measurements like that you will see a difference.

@thequantaleaper 4 года назад

Agreed. And also, the recommendations are surely based on the board's own measured values, not externally measured values on a multimeter. I'd have to see long term failure rates to determine if these risks are accurate.

@vidiveniviciDCLXVI 4 года назад

@@thequantaleaper Yeah this way of measuring didn't sit right with me either. Plus the guidelines are there and they take everything into account.

@happygimp0 4 года назад

The correct way is to connect the power inside the CPU to two unused pins (has to be done by AMD), which don't draw any significant current,

@tcclaviger 4 года назад

Absolutely right, thanks Asus for the probe points that do it right! In regard to the comment by Casey Goode, the voltages quoted are ABSOLUTE voltages, not "read by software or MB", absolute. The meaning is not ambiguous, the chip manufacturer cannot account for the different MB constructors voltages, they can only quote what their own equipment will take. If you have a board that delivers 200 mV over target, well, take it up with the MB manufacturer. What is FAR more important... over/under shoot. How far over and how far under will do serious damage over time.

@josepholsen9831 4 года назад

berni8k You’re right. With those low voltages every little bit added counts. Wondering if he has a BSEE. Nothing against him if he doesn’t. Just want to see what passes for engineering these days.

@WestCoastMole 6 лет назад

Excellent Video Steve. You illustrate clearly why Overclockers should not trust the voltage readings that come from Software Apps. The excessive fluctuations in the A-to-D converters used in the hardware on the board. I know I would like to see, at the very least, a standard adopted where all Mobo Manufacturers supply overclocking motherboards with the SOC Voltage and Ground brought to a clearly labeled Hard Post on the top side of the board. In addition the more diligent manufacturer should supply a precision Reference Standard set to the CPU Manufacturers "never go above" SOC Voltage. Good quality DMMs run anywhere from 80 to 200 bucks. That isn't a great deal of extra money to spend considering the user will only have to do it once.

@SaccoBelmonte 6 лет назад

I've been taking CPU vcore measurements out of the socket since Ryzen launched (Prime X370 Pro) While they vary compared to HWinfo some say that there could be a drop from the back of the socket to the actual die. So things have to be taken with a grain of salt. I generally recommend not to go over 1.425v for a Ryzen chip. SOC never over 1.25 and generally is better low (here is at 0.975v) If you wanna do a truly helpful video grab an oscilloscope and try to find peaks at LLC4 / LLC5. There is a huge debate around Ryzen about this.

@SaccoBelmonte 6 лет назад

There is a debate whether LLC4 and LLC5 are harmful or not and to determine that we need to know if there are peaks while transitioning from load to idle. So far only one person measured and found no peaks but we need a definite confirmation. For that Gamers Nexus needs a high quality oscilloscope. (maybe rented) I'm on a 1800x at 4.05Ghz 1.418v vcore LLC4, 0.975 v SOC LLC30

@berni8k 6 лет назад

If the peaks are short enough to need a good oscilloscope to see it (I take it >100MHz) then they don't matter because there are extra caps on the CPU package that can clean up blips in voltage that brief. The problem is that he is stabling the black lead of the multimeter into the PSU, this adds the voltage drop of the PSU cables in and throws off the reading by as much as over 100mV. Doing that is fine for testing if your 12V rail drops. But its NOT fine when you are mesuring voltages so tiny.

@gillianseed4419 6 лет назад

yeah, you really need to sense ground near the cpu, preferably on the other side of the cap when there are triple digit amps flowing milliohms matters

@kuhrd 4 года назад

One thing that I would point out just for new users to also understand is that the meter they use could also be several % off from what it is reading/displaying and that accuracy varies by range. An inexpensive hardware store meter could be out as much as half a volt or possibly more. This is especially common for cheap meters that cost less than $250-500 US and don't come with a lab calibration sticker or statement of calibration from an independent lab. The lowest precision lab calibrated meter I would consider using for overclocking would have a 4.5 digit display and 20K counts but a 5.5 digit display and 200K counts would still be a better option for seeing problematic voltage variations during operation.

@anonamouse5917 6 лет назад

Very informative !! Thanks for this.

@HappyBeezerStudios 6 лет назад

Reminds me of my old P35 board where I went with a pencil mod to counter the immense VDroop woth a quad. Still find it crazy when people put something like 1.35-1.4v onto recent chips. That was considered high to threshold 10 years ago and todays chips are manufactured way smaller.

@fellthesky8285 6 лет назад

Thank you for this video. People were going crazy over the internet because it was so easy to get 1600 mhz with 1.25v. At least 1 channel recognize this and explain why is particularly dangerous to believe in software to measure voltages.

@joshuagomez3600 2 года назад

Great job gettin in on the details. Interesting the variance between manufacturers and from different measurement points.. I can't help but wonder if there is some sort of unspoken industry standard, or if they really just sort of all do their own thing. Kinda makes you wonder which voltage is really the important one, because even if you are seeing a certain voltage on the fet if there is loss on the way to the sensitive component then I would think that the loss would have to be kept in consideration and compensated for.. But it seems like they all act so differently its almost impossible to say..

@astilen5647 6 лет назад

Learned a lot, thx GN!

@johncaccioppo1142 4 года назад

Had a mild yet annoying audio stuttering problem with my new Lenovo, the support team tried to help but couldn't, however we did find in the user forums some discussions that led me to conclude the problem might be an undersized power supply, 400w not nearly enough for a gaming desktop. (No wonder it was such a bargain!) What made this more credible is that I read numerous solutions from people that only occasionally worked. When I disabled my WLAN I found the audio stutter I was fighting diminished by about 90% further making me conclude that this is a power issue.

@alb9229 6 лет назад

Great video Steave ! As i see things MB manufacturers should ONLY provide "safe" numbers that correspond to what you type in , since having to verify those numbers with a MM (as it is now ) makes litle to no sense for the average customer who wants to OC .

@GrulbGL 6 лет назад

dude, that was insanely important! i need to make sure the voltage of my cpu (currently a fx8320 at 1.39) if that was really 1.39.

@donkenta3634 6 лет назад

Sorry, but I disagree with you, HWmonitor, in the CPU section read 1.1V from the Sensor in the CPU, below, in the mobo section (CPU Vcore... SOC, Sensor in the mobo) read 1.368V, VRMs Output, equal of the measurement from the multimeter. The difference between both values is the Vdrop, compensate by the LLC setting in the BIOS. Values taken in the minute 12:58 . (Sorry my English)

@MuttleyGriffin 6 лет назад

So we should focus on mobo section "CPU VCore SOC", correct? (Mine is at max 1.2v so I think I'm good).

@HappyBeezerStudios 5 лет назад

I had boards that were set to 1.48v and had a hard time to let the sensors hit 1.42 idle/1.40 under load Got better after some vdrop pencil mods. Agree that the important voltage is those actually hitting the chip, but voltage sensors are just as inaccurate as temp sensors.

@justaperson8815 5 лет назад

True, not to mention using a cheap meter and can the motherboard.Noticed in my Msi bios the voltage is 0.01 higher then whatever voltage you set it to.

@HappyBeezerStudios 5 лет назад

Don't really know if they make those, but a multi do uses on

@brad8122 Год назад

So he read vddcr CPU voltage with the multimeter and mixed them up? is he sure that he is taking soc voltage measurement? if he showed the CPU Voltage and show thats not it ( a higher value... Over there). We need a captain

@sitordan 6 лет назад

That wobbly monitor is a constant in the latest videos :)

@justinsugay1149 2 года назад

In general, other than getting a higher benchmarking score it's usually not worth the extra voltage or higher frequency to run your components at and the cooler the better as long as it's not affecting your overall experience using the computer. I try to run my computer as quiet and as cool as possible I do test out how hard and far and extreme I can get it but for 24/7 use my realistic settings are much more conservative. This whole video definitely confirms my ideology when it comes to building and running one's PC for benchmarking versus daily use. As always Steve, great work!

@HellGatefr2 Год назад

Video is a bit old, modern CPUs seem more robust, and you will notice they can run at very high voltages stock

@Poofu 6 лет назад

one thing to point out is due to the distance between that molex gnd and the vcore pin the value could be slightly off due to the molex gnd not actually being 0V

@jakegarrett8109 6 лет назад

Thank you so much for making this video Steve! I've been telling my cousin for over a year now, don't trust the software, its usually not accurate. So many people believe it like its written in stone, even the impossible (like sub-ambient temps on FX, just physically impossible), while I'm laughing at the 1 million volts on one of my GPUs show. Always check with a multi-meter to know for sure, never assume what you type is accurate with what happens. Great video Steve, keep up the awesome work!

@astilen5647 6 лет назад

The HWiNFO64 software is accurate, this video proves that. Steve used a digital multymeter to cross check and validate the readings. Fo example if you set your OC software at ~1.1 volts, check the HWiNFO64 motherboard section to see if the voltage on those components exceeds the safety numbers the manufacturer provided

@jakegarrett8109 6 лет назад

Ok, so 1 million volts was correct... And 8c CPU in a 20c room is also correct. And even though I measured with a multi-meter, and it showed Vdroop of 0.01v, software was showing 0.1v (order of magnitude different), as well as 1.3v instead of 1.55v like I typed and was reading with DMM, then you're saying software was right the whole time? Mind blown! I just invented a CPU that runs cooler than its surroundings (so it must be absorbing thermal energy, and creating electricity), while capable of mining. I just solved global warming and mining at the same time! Where do I sign up for my Nobel Peace prize?

@astilen5647 6 лет назад

Calm your t its

@jakegarrett8109 6 лет назад

But how can I? I just invented something better than perpetual motion! I'm off to save the world, providing energy for 3rd world countries, Hash power for miners, and solving global warming. Its a miracle!

@astilen5647 6 лет назад

Stahp

@davidwithey6036 2 года назад

my old Ryzen 3600x was at 1.4v on for time before I wanted to upgrade. Never missed a beat. I gave it to my mate cheap and his still using it on the same board with same OC still going strong.

@trahim2 6 лет назад

Where you measure these voltages is imperative. There are resistive losses between the voltage source and the device being powered. I believe the reading in hwinfo is coming from the ADCs on the cpu itself. Therefore it's getting a more accurate reading of voltage that arrives at the silicon. This is why all the hwinfo voltage readings are lower than what your meter is reporting. There's lots of variables that affect measurement, but the main issue is the difficulty in accurately measuring voltages in the millivolt range.

@GamersNexus 6 лет назад

trahim2 That was the entire point of the content, yes.

@trahim2 6 лет назад

Gamers Nexus hah, you're right. I didn't watch the video all the way through. Gamers get this stuff wrong so often that I just assumed you weren't aware of these factors. You've done your research. You're no Linus. ;)

@AngryChineseWoman 6 лет назад

trahim2 So basically hwinfo's reading is accurate ? That's a relief

@Cafe_TTV 6 лет назад

trahim2 He's not Linus

@geraldalonso8232 6 лет назад

Very good video as usual. I would have liked to see results from a good multimiter.....

@YuRaLL 6 лет назад

euhm. the correct voltages the mobo puts out are in the motherboard section of HWinfo (it shows 1.2 when you first pointed it out at CPU VCORE SOC.) the CPU shows lower numbers because of internal resistance I reckon. so that just shows the vdroop between what the MOBO delivers and what is actually used by the chip.

@ludoflame 6 лет назад

noticed that too

@derx6666 6 лет назад

Indeed, saw it too. Strange that Steve didn't notice this.

@MuttleyGriffin 6 лет назад

So when monitoring that SOC on my Ryzen 1700 is at max 1.2v, I'm correct at looking at the "CPU VCORE SOC" under mobo readings in HWInfo? This video made me second guess myself by checking the CPU readings of "SoC Voltage (SVI2 TFN)" which reads at 1.09v.

@ludoflame 6 лет назад

for good or for bad my mobo(gigabyte ab350 gaming 3) shows both sensors at around the same voltage with differences around 0.05 volts

@MuttleyGriffin 6 лет назад

I have the Gigabyte AB350M Gaming 3 (the mATX variant of your mobo). My avg voltage difference between the two is around the same at 0.048v, and max voltage differences 0.094v. Granted not a lot, but to certain users, could be enough between sweet spot and degradation.

@b28uby 6 лет назад

@Gamer Nexus if you look at cpu core voltage, it is a correlation between it and the one that you modify in amd master and what you measure.

@BornaPrpic 6 лет назад

Great video. I've got an Asus B350-I with an 1700 in a Phanteks Shift case (gonna need to print a top extraction fan mount at one point). After seeing the thermals on the board. I'm sticking to 3.75 even with a 120 Coolermaster AiO. Higher I could probably put it horizontal and grill sausages. Thankfully it's only purpose is for HTPC and as a rendering slave for v-ray distributed when needed (cheaper to have a few 1700 than less Threadripper systems). Next up is a 1700 NAS system hahaha. Can't wait for kill the board series. I need a video so I can shut up and just show people :P

@johnlacquey7240 6 лет назад

What is to keep me from soldering on lead to the SOC, another to ground, and keeping a cheap multimeter on full time while OCing?

@jc265 6 лет назад

How to kill your CPU, aka how to take advantage of Micro Center's in-store replacement plan and get an upgrade every year or two.

@akalion213 6 лет назад

OC voids the warranty though

@jc265 6 лет назад

Akalion Manufacturers warranty, sure. But Micro Center's in store replacement plan isn't the same as the manufacturer warranty.

@vnyggi621 6 лет назад

Hodor they cant tell that its been overclocked anyways

@alien_man1669 6 лет назад

Haha thats exactly what i just did to my asshat *cough* asus cheap prime b350 plus with my ryzen 7 1700 and my friend accidently let ryzen master kill his tomahawk ab350 pro with ryzen 1700 so we have to do it again. And just shoot the shit with them and dont say you overclocked it. If anything blame a bios update for changing voltage lulz

@harunmeta2258 4 года назад

@RectalDiscourse thank you

@ThisRandomUsername 6 лет назад

I haven't watched the whole video yet, but by using a disconnected Molex ground line as the reference, you're adding in the voltage droop across the other ground lines that are connected to the board which are carrying some current. This happens because every wire has some resistance, and there's current passing through them to get to power supply ground, so they generate a voltage. V = I * R. This won't be much (probably less than 1/10 of a volt), but it can add to it. If you want to measure more accurately you should probe as close to the CPU as possible.

@shankly1985 6 лет назад

Great video can you do another using maybe an i7 8700k?

@stolz_ar 6 лет назад

I love this channel so much...

@Capta1nSni9er1 6 лет назад

I wonder does this also apply to ram ocing?

@brovid-19 3 года назад

Shit.. and now I gotta use my multimeter to overclock. . and it's covered by the mounting bracket. This makes me so confident

@LimbaZero 6 лет назад

recommend to use GND from MB or measure ground reference from MB.

@Drinnan 5 лет назад

Although I've since owned a q6600 and 6700k, my first self built pc was an AMD Athlon XP 3200+ @2.2ghz. back then I didn't know how to overclock and I never did in the time I owned that but when I bought my q6600 and sold the AMD to a friend, he fried it within 2 weeks, just ramping up voltages because I told him how good the q6600 was. I kinda wish I kept it for sentimental value

@kristiangronberg3150 6 лет назад

Hi man, I'm just curious. Did you check the voltage difference between the 8 pin ground and the mesurments capasitor ground? the difference is still so much that there is over volt, but if the ground points is (because the resistance in the motherboard) higher than the 8 pin ground, the cpu will see lower voltage. It might be marginal, but like I said, just curious All the best and thanks for the great work!

@wont8280 5 лет назад

thank you so much love you guys

@chrcoluk 6 лет назад

Not surprised this is happening, many people on forums have been advising me to run 1.4v as 24/7 use on my new coffee lake, and you got specialist companies selling pre binned chips that are rated at really high voltages like 1.37-1.42v which is crazy for daily use. Even worse when you consider that chips actual voltage tends to be higher than whats set in bios and what software tools report. There seems to be a big misconception that as long as you can cool a high volted chip then its all good, as if only heat kills cpus.

@teardowndan5364 6 лет назад

That isn't the correct probing method to get accurate CPU voltage: it doesn't account for voltage droop across the ground plane between the VRM and CPU, possibly some more due to droop on wiring between the PSU and motherboard. Your readings may be off by 100mV simply from taking your ground from so far away from the load.

@GamersNexus 6 лет назад

No. The difference is about or less than 20mv.

@teardowndan5364 6 лет назад

On all the boards? I'd expect that to vary a fair amount between cheaper and more expensive boards with a different number of ground planes, different plane thickness and possibly split planes. BTW, when chip manufacturers specify supply voltages, that's usually directly at the chip package's pins/pads. The closest you can get to measuring that would be to solder jumpers to one of the capacitors either under the IHS or under the CPU unless the motherboard is re-purposing a pair of VCC-GND pins for VRM feedback to read package voltage.

@GamersNexus 6 лет назад

On the boards we tested.

@Vishalkumar-mu5hy 3 года назад

If you deeply know about mosfet drive controls and inverters.Because after the constant drive voltages and frequencis you can increase the source voltage on mosfet because the source voltages and drive voltage on mosfets are not that interlinken with each others.But finally i want to say that these things are easy if you are able to recircuit and resolder your whole motherboard.

@crackny4n 6 лет назад

Check the CPU Vcore at the motherboard's readings...

@kasimirdenhertog3516 4 года назад

I remember casually setting 5V on a CPU that’s designed for 3.3V, just because the BIOS didn’t support more options. How far we’ve come! 👨🏻‍🦳

@VampirkoOrus 6 лет назад

What about a voltage drop on a path from cpu to ground? I don't know how much resistance in between them, but they may affect your results. Could you find a pin that located at exiting current from cpu and make your results a more precise? Ah, and also voltage drop from resistance of a socket.

@bozoc2572 2 года назад

Indeed, this is how not to measure Voltage

@Fr3akyPro 6 лет назад

Hmm I think I should dial back my 1.52V Overclock... 🤔

@Fr3akyPro 6 лет назад

Momi ye got the 8150. Won't stay at 4.2 without at least 1.5V

@triniox3186 6 лет назад

Momi u didnt get the 420 joke did u?

@angryman9333 4 года назад

lol

@LudwigvanBeethoven2 5 лет назад

Cant they put an integrated multimeter for critical parts? I would like to know the physical reason about this difference, why motherboard cant read nor write correct voltages.

@Michael_NL 5 лет назад

Can you do a video on transient voltage spikes when using high LLC? For the new z390 gigabyte boards there is no know info about how much the voltage spikes when coming off load with high LLC. Im running 1.345V on my 8086K for 5.1 Ghz, but I am afraid that my voltage will spike +0.200mv to 1.645 for a millisecond when coming off load. That will slowly kill my cpu in a year.

@ljfaag 6 лет назад

Can you show us which point you're measuring with the multimeter probe or how to determine that point?

@VarietyGamerChannel 6 лет назад

Good old 80486 handles 3.6v.

@v1asec 6 лет назад

Thank you Steve, you are doing a great job at explaining things. I watched your video in a discord channel and had to go to the browser so I could upvote. Currently I'm not doing any OC except for XMP on RAM. I pushed 1.36V in BIOS to make it more stable. I even had to push 1.38V on a friend's rig because it only POSTed at 1.36V, but it could still be unstable. I hope it is "safe".

@snekbaev 6 лет назад

Is it just me or the measured voltages correspond to the "CPU VCORE" (not Vcore) entry in HWiNFO64?

@james2042 6 лет назад

*"In my motherboard's UEFI"* I can control the load line calibration of the SOC vrm as well as the CPU vrm. In testing I could actually sustain higher memory overclocks by tightening the SOC voltage.

@NAmania 6 лет назад

What about VCORE hwinfo readings?Are they accurate?

@forkbomb_ 6 лет назад

Steve really seems to be into killing hardware, or at least making them suffer! 🤔

@DaveBoxBG 6 лет назад

Any info if Asrock increases the voltage like GB?

@tristenturner832 6 лет назад

i would have thought that the chip readings would be the more accurate voltage. Where you're reading is before the contacts and there could be more resistance between the socket and the back of the silicon chip compared to where the chip is reading. that doesnt change the fact voltage is still higher than the setpoint, but it might not be as bad as you are seeing.

@OVERKILL_PINBALL 6 лет назад

I think there is a lot of confusion about just saying "this voltage will kill your CPU" because the truth is that the answer really depends on thermals (and if you're skilled/knowledgeable about overclocking ie: you don't try to use 1.5v+). I'm not saying it is impossible to to kill a CPU as I realize that it is possible however I disagree that it is "easy" to do. I think there is honestly way too much "Myth" about "CPU degradation" over time even with CPUs as old as 45nm @ 1.4x volts. If you know what you are doing and have the proper cooling for high voltage (at least an AIO with a decent CPU block) it is actually difficult to harm a CPU with voltage less than 1.5v. This topic has been beaten to death long before RU-vid. The only thing left is for it to be proven in a controlled test as there's not much room left for words at this point (it's just a really old debate). I would love to see a controlled test with various identical systems running 24/7 with different voltages applied along with air cooled vs water cooled with a minimum ONE YEAR of running the systems. Run an extensive suite of tests with the systems including overclocking and then run all the same tests again at the 1 year mark. Keep it going for 2 years.. 3 years.. let's prove this myth one way or the other once and for all. GN might be the one to get it done :) IMO, the CPUs themselves, if cooled properly with less than 1.5x, will be just fine and will work forever.. I am someone that has applied 1.4x volts to many many CPUs over the years (always been an overclocker) and ran them for YEARS at a time and never ever saw any of this "mythical degradation".

@mykeprior3436 5 лет назад

You're speaking out your ass. It is no myth, and as junctions get smaller and smaller the problem WILL inherently get worse and WORSE. Which is likely the real reason you're seeing "old" CPUs doing just fine today at near the same frequencies. Voltage does impact transistor junction breakdown (thermal breakdown is not NEARLY as much a problem) and as opposed to thermal meta-stable changes, voltage breakdown causes permanent stable changes, aka. permanent shorting. A lot of it is the chip lottery because the doping process is not "perfect" for better or worse. Not only that, but a lot of VRMs have difficulty maintaining "stable" high voltages, where the margin for error is not linear as you increase it, and you can easily over-volt up to hundreds of millivolts higher than intended when overvolting, very potential breaching breakdown of the dielectric doping...and thus you have permanently damaged the chip. Will it not work at all anymore? No. The breakdown usually isn't that catastrophic. If the VRM is very stable and you're under 1.4V? Most CPUs will be ok. Unstable VRM and above 1.4V? You're asking to ruin it...no matter your cooling. Can you ever overclock that high with that voltage again? Will it get worse over time now until it IS catastrophic? Have you shortened it's lifespan from 10 years to 2? YES. Your cooler isn't a superconductor, even running at sub zero temps overclock extremers are ruining the chips, but for that sport longevity is inherently not the point. Besides, on the whole heat and voltage is bad. But voltage is markedly many times worse and riskier.

@ZZstaff 6 лет назад

Somehow I missed this video. I have i7 7700K with a Gigabyte Z270X-UD5 motherboard and have the Vcore set at 1.235 and onboard graphics set to 0.600v. I do not use the IGP, I have a MSI 980Ti Lightning. These voltages seem to be fine as I have had this build for just under 1 1/2 years.

@TheRguru1 6 лет назад

My ASUS motherboard automatically set the Vcore to 1.5V, and that was running at stock freq with all default bios settings. I didn't realise until like a month after owning my PC and I turned it down to like 1.2V immediately.

@CHAOSMOVEMENT 3 года назад

Thanks for the warning. I double checked my Asus Crosshair VIII Hero, and my manual OC at 1.225v is 100% on point using a multimeter.

@retroman2884 3 года назад

I don't doubt that there are differences between the voltage reading in the settings and the "actual" voltage. But I doubt whether the voltage meter is calibrated or not, and when this was done.

@Bourinos02 6 лет назад

Is this behaviour observed on Intel as well? (and for any APU/CPU?)

@Djhg2000 6 лет назад

What if you knock the 12V rail down a notch, does it change the other readings? Basically the question is wether the motherboard measures against an internal reference voltage or an external presumed stable voltage.

@SoundDesignerslabel 5 лет назад

Have you checked that the voltage you read in the socket with the multimeter does not pass the voltage limiter( MOSFET) before feeding the cpu? The voltage limiter may limit the excess voltage that enters the socket. Of course, measuring in hardware is much more reliable than software, but if you go through a voltage limiter the difference may not be that big. As you know these little details can make all the difference. Love guys, Keep it RAW!

@ev0rn 6 лет назад

could you make a video of how accurate are the z370 mainboards are messuring the vcore of the cpu? would be very nice. i do not have a multimeter to check that. important are the models of asrock (taichi and fatality itx) and asus maximus x hero and apex.

@codercable9088 6 лет назад

Either HWMonitor is mislabeling stuff or you are misreading it. HWM was almost exactly matching your multimeter readings, if you looked at the CPU VCORE voltage. Take a look at 14:53. VCORE voltage: 1.284, measured voltage: 1.28-ish.

@wii166 6 лет назад

Take the crappiest B350 board and set the SOC to 1.2V and v-core to 1.425V and keep the CPU cool and i bet you can break the VRM and then break the CPU. I say this before i watch this video. I love gamer nexus they really do great tests and i'm happy they do it

@invalid8774 6 лет назад

Might, sure but most decent power supplies would turn off the moment your VRMs burn out. Cant guarantee that your cpu wont get damaged in that time though. High voltages roast it really fast.

@westnylefx 6 лет назад

I never see my soc voltage go above 0.9v on my x370 pro. I do however see 1.4 sometimes hit on my 1800x. What is the usual 3.9 or 4ghz voltage 1.275v?

@100YardFumble 4 года назад

First day overclocking. Pulled my PC out of storage because I'm getting back into gaming. My PC is almost 10 years outdated. Got it up to 4ghz at 3000mhz was looking at other people's statistics and I think my PC did well for being old. I'm not going to lie I was running voltage pretty hard but I don't think anything was damaged. Trying to figure out how to calibrate the cas, trpc or whatever ii can't remember the exact names but ya know what I'm talking about.

@Vafarien 6 лет назад

oh no... my i7 7700K is doing some kinda weird thing with voltages where it goes as low as 0.885 and as high as 1.395 . i'm worried.

@JohnDCrafton 6 лет назад

Hey Steve, CPU VCORE was showing accurate voltage compared to your multimeter.

@dieseldiesel9292 5 лет назад

Decided to oc my 3770k and set my voltage to 1.2v. Little did I know it was voltage offset and It ended up maxing the voltage out at 1.8v.

@hugo9016 4 года назад

What is voltage offset?

@opoxious1592 3 года назад

I have my i7 990x overclocked to 4.62Ghz at 1.448 volts, for over more than 10 years. Still runs like a charm.

@ParallaxEffect 2 года назад

He says a lot of things but I waited and waited to hear data or even an explanation about why high voltage actually kills a chip and.. like every other neckbeard, thats where the details get suspiciously thin.

@Bnryzombie 6 лет назад

Does gnd resistance play any part into that number from the processor? IE the Resistance from the CPU to the PSU gnd?

@StudioBrock1337 6 лет назад

I love stuff like this. Also makes me glad that although I'm an awful overclocker (for example when OCing my 6600k I first shot typed in 4.4 ghz, and then barely tuned it to a stable 4.6. My RAM I literally typed 1.35v, 1.2(or 1.1 can't remember SOCV), 15-17-17-36 and 3000mhz. And first shot it booted so that's what my RAM runs at because fuck it it's good enough.) my conservative nature with voltages would appear to be my saving grace. Like if you say I need 1.25v for x clocks I'm going for 1.2v at those clocks. If 1.25 is a safe maximum then I'm not exceeding 1.2v and at 1.2v I'm probably not touching LLC. I've never liked LLC because it either does nothing or causes massive voltage spikes (by massive I mean .05-.1v).

@kybiras1 6 лет назад

i ran an 2010 i7 870 on a entry level Gigabyte board without any heatsinks and lowend VRMs/ power phases. It took me around 1 year to blow that board with 4.1 Ghz. Not bad I say !!

@ZZZXYZ 4 года назад

@9:28 How do you know which correct cap will read the soc voltage? multimeter is blocking view. Also does it matter which molex ground?

@einherz 5 лет назад

could be board voltage some higher than inside cpu? by stabilization room, may be?

@vega1287 6 лет назад

i've got 2 questions 1 how low would you asume mx cpu will get if i direcht die cool it with a pleter witch is cooled by a 65mm quad rasiator but there are dual 1080ti's on the same loop and 2 at those temperatues what is a save core voltage (cpu=i5 8600k)

@wladewig 6 лет назад

It would seem to me that 1.2v is a good empirical (middle of the road) value for SoC in that there so much variance between manufacture architecture, and electronic component quality. That said, t would seem to me that the absolute or max SoC voltage allowable is a need to know requirement - would that be 1.35 volts? Question: Curious to know, to stabilize SoC voltage fluctuation, did your testing include changing LLC - I saw from your excellent presentation that there was quite a bit of volt flux?

@johnhenryscott5319 6 лет назад

If you just enter and Overclock and don’t change the voltage, can you still have issues (lack of voltage instability aside)?

@memesforgenz4736 5 лет назад

Thanks for making this, my voltage was WAY too high. If it wasn't for this I would of killed my system

@johnbrowne3964 6 лет назад

I have a question, many OCers just raise the speed only and leave the voltage at "stock" or auto setting- so how accurate/safe is the stock/auto option regarding actual delivered voltage? It seems with these apus you can get 1.4 or so more speed, but only 1.2 x in performance (the square root). It may not be worth the last 100-200 mhz if you need to raise voltage for stability with the really unknown actual output sent to the socket. Not everyone is going to pull mobo and measure things- I think there might be a shorting danger with the small and crowded under-socket area.