Answer here 29:49. "105C I wouldn't worry about, 110C I still woudn't worry about" And here for 24/7 31:18 "if you're running something 24/7 I'd want to have the VRM at 95C or 85C" You're welcome
@@MJ-uk6lu This video is so educational it's mind blowing amazing, for those who use brain that is, and want to be smarter. Guy explained in detail what is what and why some temps don't mean what you think they mean.
Laptops throttle when vrm goes above 75 , speaking from experience with Dell laptop, ofc this is not the silicon limitation, it’s the limit set by Dell , so your manufacturer can screw you up via a bios update even if your hardware can handle more ,
yeah but he's talking about a gpu vrm so it would be pretty close to an encoding workload i think. but i dont care anyway. ill be on to a new card before this one dies. i actually mined the purchase price plus about 50% (around 700 usd) so everything from here on out is just gravy baby!
7970 E-power ran into some issue caused by me being a moron. I can probably still get it working just need a hot air station. 1080Ti is planned for LN2 I've just run out of thermal paste.
yeah, Blender uses AVX and 3delight uses SSE2, and both pound the CPU rather hard. And it is not a rare thing for 3D renders to take days to run. So yeah, Different realm from gaming. 208 days of 3D rendering on a 5k hour cap just isn't cool for a motherboard that can cost a few hundred dollars. A render box is not a cell phone that gets replaced every year, it's an investment that many expect will last a few years. skimping on the VRM cooler just because it can run at 125c in a 20c room is no Excuse, because not everyone lives in a walk-in freezer, lol. Arguing that a consumer part should be exempt from being required to run any workload for the expected life of the part, is like a car dealer telling you the car is only warranted if you do not use it at all for the entire duration of the warranty period, lol. The CPU is advertised to operate at a particular performance level. And if a motherboard advertises it is made for that CPU, it should be able to run that CPU at it's rated performance regardless of where on earth someone lives or what workload they use the CPU for. Great vid BZ, B)
If you're running extended period renders you shouldn't be using consumer hardware if it's your main income. That's why proper workstation hardware is set to a higher degree of stability. Consumer hardware is a race to the bottom to make as many sales with the least margin, making the highest quality possible shouldn't be expected.
I can sort of understand the point, even tho I still feel that if a motherboard is advertised to support a CPU, it should be able to run that CPU at the CPU's advertised performance level for the expected life of the motherboard, without being told that I can only use the CPU for so many hours a day or only use a small portion of the CPUs potential stock performance. lol. Yes, I think we all agree that 'Consumer' stuff is junk, and anything for i7, i9, TR, and Ryzen, just isn't something a 'Pro' would be able to rely on, because the motherboards are junk. lol.
He's using what I call the IROC-Z excuse. The first year IROC-Z cars had a really really bad shimmy if you go over 110 kph. I met someone who took his brand new car to a dealership and asked them to fix it. They kept telling him there was no problem then when he said, "you have to go over 110 Kph" they said It's illegal to go that fast, there's no problem with the car.
100 kph is 62 mph. There are no roads I'm aware of where the speed limit is higher than that. You can go up to 125 without getting a ticket but they are technically right. It's still no excuse to screw somebody with a faulty new car. You can take it to a track and drive it.
I set a limit of 90° for any component reading, however I think a lot of VRM fails are poor quality not heat. Hey, coil (inductor) whine, can you do a video on that subject.
I consider every temp in the 3-digit range dangerous, even for 150°C+ rating mosfets. Although I haven't seen such temps since the GTX480 and the HD4870, I'm always keeping an eye out when I'm going for score runs.
Thanks for doing this vid Buildzoid. Some of us have electronics backgrounds so we understand the spec sheets you reference. We can remember thumbing through TTL logic books and such for data sheets. But most people don't know these things and this video is a good intro to how some components work and why temp is important to them. Good job!
It's a 4-phase VRM with RT8894A running in 4+2 mode. 2x 4C029N high-side and 2x 4C024N low side mosfet (per phase). That's about what I can find on the interwebs for the B450 Tomahawk . The carbon pro has a really similar VRM (for CPU VCC) just with more inductors. Same VRM for SoC. Still a 4+2 phase and same mosfets and controller.
Very interesting about the capacitors. Makes me wonder about the caps on my two x58 motherboards made in 2010. After an upgrade to a newer system it could be fun to desolder the old caps and replace them with new ones and see if it becomes possible to get higher overclocks or reduce the vcore setting.
This video is quite informative and certainly makes the case for monoblocks and full cover GPU blocks in any high end application where you are going to stress the VRM. I never really expected component run time to be impacted so significantly when bringing the temperatures down from even 90c to 70c. Really good stuff.
I would like to encourage you to do a favourite motherboard VRMs in the sub $100, $150, $200 price ranges for Intel and AM4. Just so we can see what you would prefer in each category of what you have tested. We always see the "this is crap" videos, but never really see which ones have made fair compromises that lead to an above average product.
Hi. I want to mount a hybrid water cooler only on the graphics processor on a video card. And I want to know on which components I have to put passive radiators and which I don't. From what I see on Inductors does not require additional cooling if a fan blows on them. So on which components should I put radiators ??
Wrong on the inductors. Look up thermal aging on powder core inductors by Coilcraft. Depends on the binders used in the inductors. If using a ferrite core, yes indestructible but saturation happens more quickly. Love your videos. Also note that not much different between semiconductors made for commercial use. So generally very reliable. Board temp >105C starts to damage board organic materials.
Hey Buildzoid, I recently bought an ASUS P8Z77-i deluxe. Should arrive soon™. The VRM is on a daughterbord. Would you be interested in detailed pictures?
Luis Scheurenbrand Deluxe are the ultra nerdy boards nobody needs, no Thunderbolt Display?, apple deluxe, wifi onboard? What does it need to do, Data, Computing, Crypto?
I am not quite sure what you want from me. I will most likely will use the Board in a small PC (~7l) have a powerful small PC that I can bring to my friends to test VR. The board should be fine for that and I just bought it because it was very cheap for all the stuff it has.
Since you probably know more then me, I need your help with something. I have an issue with my msi b350 pro vd plus motherboard. I updated my bios to fix the 0.55ghz issue when booting up. But now when i put my computer to sleep, and wake it up. It boots to 0.55ghz. Anything you would recommend?
Stay Humble I got this bug when I raised voltage over a certain point on a couple mobos. It was incredibly frustrating and answers were hard to find, honestly if you can, I would just switch motherboards.
Well you're safe to go quite high but for 24/7 operation or intention that it will run 12h/day but for many years, then you want to go lower if possible.
Not at all. These boards are soldered at 245 C (+/- 15 C) albeit for 60 - 90 seconds, twice. They use SAC305 alloy, so 120 C-ish is all you want to see as a Tj (solder). Above that, creep starts to be a problem and you could cause a premature failure of the solder connection with extended exposure times. In other words, the device should outlast the solder connection at the limits of the FET's rated operating temperatures.
Thanks, it's good to hear someone knowledgeable making a good case for underclocking. Even better with a little brogue. Guys overclock with the CPU in mind but not necessarily the life of the board. Seems foolish. If you want higher clocks, save up and buy a CPU that runs at one. Bottom line remains what it's always been: heat kills. Cool the f*%k out of everything.
@Ggchb Gig ghb U are wrong. I personaly tested 8700k OC on Gigabyte Z370 HD3 1,38 V core 5000 MHZ. On this setup motherboard hit 122 C and it trigger temp protection (on 123 C) what downclock the cpu to 800MHZ. For 100% 5ghz uptime i need to put a Fan over the VRM and it still have ~ 111 C . (AIDA 64 stress test) I am 90% shure that this motherbord also have the 5GHZ auto oc mode so even absolute beginner can overheat the VRM with one click. Now Intel announce that this boards will support incoming 8 Core cpus. Rumored 9900k may even trigger temp protection on this motherboard with stock + enchance all core option what are ususal set default.
So my 8gb 470 Nitro is fine with the VRM's hitting 95-99C? That's at stock boost (with power slider maxxed so it doesn't throttle), I undervolted the card ever since I got it because of that, I wasn't comfortable seeing that temp on anything let alone the VRM's of my GPU.
The reason VRM on MB and GPU's are often not cooled well is because of planned obsolescence and manufactures not wanting you to keep your non premium component for a very long time, but rather purchase new hardware after 2-4 years. It is just good for business.
you are to blame if you put a premium component on a non-premium mobo. A basic mobo is ment for basic 65W (or less) CPUs that is cooled with a top blower style cooler to actively cool the VRMs.
I'm adding an AIO to cool my GPU with the Kraken G12. Anyone have advice on what VRM or VRAM components I should add heatsinks to? I'm thinking the mosfets for start.
It would be good to add heatsinks, although G12 has that little fan that is blowing directly on mosfets and most people say it is enough because that fan doesn't blows hot air on them like the fans of regular coolers do by pushing air that passes first through cooler fins than hit the gpu board.
Oh man, this is so useful, you have no idea. I was kinda freaking out because the VR MOS reading on HWinfo64 was reaching 101C during Prime95 on a B450 Tomahawk. I knew it wasn't that bad because it was stress testing but I felt a little unconfortable with it. Thanks for helping us noobs! Now I'm running my almost stable 4GHz OC (R7 2700) @1,35V. It crashes at almost 1 hour but idc aslong as I can use the system to surf the web, play dota 2 and college stuff.
why do you OC in the first place if it is just about "surf the web, play dota 2 and college stuff" lol I literally disabled the CPB on mine altogether :p
How about a video discussing the different commonly used MOSFETs in VRMs, their specs, pros and cons. A better informed consumer can push manufacturers to use better components and design.
The X470 Master SLI didn't have temperature sensors but the AUXTIN1 sensor showed the VRM temperature, so maybe check HWinfo for any values that go up when running Prime95.
So I assume for very short durations on the order of seconds or minutes when the parts are not in use, higher temperatures are fine (like for soldering) right?
"Nothing happens to the wires" Don't try to be a smartS or know if it's relavent. But What about "electromigration" ? degradation which associated to higher frewancies that comes along with higher voltages ?
5,000 hours is just over 200 days, that's a ridiculous lifespan for a PC component! My Gigabyte P35-DS3L motherboard is 12 years old, and I still use the PC. I've just built a new Win10 PC with a Gigabyte Z390 MB, I plan to keep this PC for at least the next 10 years.
I have a Zotac RTX 2060 AMP rated at 190watts max and I have flashed it with a Gigabyte Extreme 2060 BIOS and now i can draw about 240watts max (power limit from 112% to 126%) - My card has 1 x 8-pin connector. Is this all fine? - I have a probe stuffed as close as I can get it to the VRM area and it reads about 85c after FurMark/OCCT for about 10 mins. So assuming the temp is higher internally I still feel its not too hot. What do you guys think?
Back in 2019 I bought an Asus TUF Z370 PRO motherboard, and didn't realize it had crappy 4 phase VRMs. It looked like more phases because of component doubling, yet it barely handles a continuous 95 Watt load without VR Thermal throttling. Gotta set a power limit on Intel XTU of 80W to keep the mosfets under 95ºC (when the GPU is also pumping out heat). Even then, it manages to sustain 4.1GHz continuously which is more than enough. The last 600Mhz of boost make the CPU so inefficient, by requiring almost double the power to run.
Solid caps last much longer than electrolytic caps. Electrolytic caps can go bad even not being used. Motherboards use to have lots of electrolytic caps 10 years ago. You will probably see 5-year-old boards with solid caps lasting longer than the 10-year-old boards with electrolytic. I glue heatsinks on the VRM with heatsink cement. I kinda expect to throw the board away instead of repairing it in the future. If you have a fan in the back turn it to blow air into the case and make a shroud to blow air over the VRMs. If you have good enough cooling using the rear case fan to blow air is the best solution. Someone sells fan shroud kits for this on eBay. You probably can download a file and 3D print it. I only worry about it because I can render video or render 3D for some time.
Yeh but ... it's still a good idea to limit thermals and thermal cycling and expansion, all of which cause things like bad solder joints ('drying out' or cracking) and PCB's de-laminating. I don't like the idea of any component on a PCB (particularly surface-mount) regularly going over c. 80°.
I just measured the temps of the VRM's on my X470 ultra gaming, running a 2600x at stock settings. Software (HWInfo) says VRM Mos is at 51°C ( gyazo.com/72ea76027ebe1170a6fc77645aba033c ), I also measured the indcutors and the capacitors using an infrared thermometer reading 39°C for the capacitors and 41°C for the inductors. Judging from what I've just learned my mosfets and capacitors should last a very long time. How much can I trust HWinfo and an infrared thermometer though?
Likely a niche issue; but what about the high VRM temps affecting the mosfet's maximum current draw? When a board uses cheap mosfets like this: www.a-power.com.tw/merge_pdf/AP4024GEMT_Datasheet.pdf where it can easily derate to less than half its max current output; could using a high power draw product (2700X in cheap AM4 mobo for example) and running a hot VRM potentially cause a situation where the fets can't push enough current for the CPU's load and fail/burn out?
My vrm mos hits 93 celcius on my gigabyte b450 motherboard when under cpu stress test for a long time. The CPU is overclocked. Is that temperature safe? I'd guess it's not that high when gaming, right? Edit: The vrm mos maxed out at 55 celcius when gaming lol
my hmm they are not as Flimsy as i thought! my gpu gets to about 66.. there is no way my VRM or Mosfets could get up to 130c! even without cooling pads? right
Buildzoid could you please talk about mounting smaller waterblocks to motherboards for things like VRMs? The market doesn't produce much in the way of Monoblocks and other custom blocks so you're left with generic blocks where screw holes don't line up or in some cases there are no screw holes at all. So how do you mount and use these miniature waterblocks properly?
As with any semiconductor device, its performance degrades (derates, thanks for using this word and reminding it to me, BZ) above room temperature and that's it. The device can function up to 80, 90, 110 or whatever, BUT if you want it to perform its best, keep it at room temperatures. (simplifying here)
Hmmm...so when a card degrades, is it likely the chip or just the caps? Is there a way to check without a scope? If you could bring an old degraded card back to life cheaply by just replacing some caps that would be neat!
Do any GPUs have a VRM thermal monitor? I have a miniATX 1070 and the VRM cooler runs real hot, but I can only seem to find the core temp in Afterburner
The lead question is wrong. What matters, is dissipating enough heat fast enough to avoid a thermal breakdown, of any affected component (fets, coils, board, etc).
It seems like these types of topics are aimed for those that go for overclocking or for some reason extreme nit-picky when it comes to temperatures. Pretty sure normies who buy parts or pre-built, plug n play fortnite wouldn't need to worry about this shazam.
As long as you run the hardware stock in almost all cases you're usually OK. Some Intel X299 boards(can't cool enough when running high current CPUs) without enough cooling or Powercolor's Devil RX580(Overheat when running furmark) are examples of poor VRM system designs. The PC Master Race Cult is pretty stupid at times, every community has that.
I agree with stock/factory settings. It's just I've seen many posts about people nitpicking VRMs to the point that they want everyone to buy high-end motherboards. lol.
Like Buildzoid said this is channel is not called @stock. Some of us like this stuff and if you are going to run a system with a high oc for years. The electronic parts are quite important (running a 6 core Intel on X79 @4.6 - 5GHz daily for almost 8 year now) so the motherboard i chose back then still serves me today ;)
I still run an overclocked Core 2 Duo system 24/7 with absolute stability(12 years). I carefully selected quality components and kept the heat out. "Normies" don't understand the significance of paying more for better components. "Normies" think the cheapest MB works just as well as the most expensive just with less features. I personally want a minimum 20% headroom in my power supply components. Think if every time you drove your car it required bouncing off the rev limiter the entire trip. How long do think the car will last? The question isn't will it work but how long will it remain working. There is also the question how long will it worki with zero problems.
hi, im thinking about replacing my 980 Classy with a 1080 or a Vega 64, now that cards are reasonable cheap again. You think its worth it? i dont really feel like its a great jump. At least not for so much money.
Funny. I was just looking up on my other screen the specs of the Vcore VRM in my Z170 laptop to see whether it could handle an 8 core 9x00 right at the moment you mentioned CSD87350Q5D. It's got eight of them in a 4+1 phase doubled setup BTW (Clevo P870DM3). Should be ok FWICT, temps on the heatsink designed for "95W" 6700K will probably be the limiting factor to clocks/volts anyway
My fully unlocked R9 Fury Strix had the capacitors fell off, soldered them back on, still works fine. Asus made an horrible job with these cards. They used a "high" performance thermal pad for the gpu/HBM heatsink contact and on mine half of the HBM was not in contact with the thermal pad.
I guess this gives me some solid, specific reasoning for why my graphics drivers crash unless I underclock and undervolt my HD7770 by at least a 100Hz from stock. RIP. (Still, could also be the equally old/shit mobo it's on.) Ah, NEET life. No funds to replace any hardware.
Thanks for the explanation, Buildzoid. Is Magnachip a good mosfet manufacturer? I'm interested in this SOYO (now called Maxsun) Z370 Board that uses ISL 69138 as controller configured as 7+0, and uses 7 ISL 6596 as doubler and using Magnachip MDU1516 mosfets. Is this a good config by the board manufacturer? Thanks Buildzoid
@Actually Hardcore Overclocking So if you could further my education here, i am running a 1500x @ 3.8ghz @ 1.35v with a arctic freezer 33tr on a asus prime plus b350. i don't do production work at all, all i do is game sometime for long periods would you recommend that i still actively cool my vrm's? For reference my cpu temp's are 29-32c idle and 47-50c while gaming. Thanks really enjoy your content.
Gaming is not very demanding on the CPU or its VRMs. The only thing you might need to be concerned about are the GPU's VRMs if it is running very hot for long periods.
The VRM on my X79 DARK are at 32c idle, and I have never seen them above 60c under load with a heavy OC. The reporting is by HWInfo. BIOS gives me the same idle readings of VRM temp. As soon as I do some maintenance on the system, I'm going to design a VRM waterblock for it. I may do one for the chipset, but I don't know if it is necessary. My GTX 1080FE GPU temp stays at 25c idle, and since it has an EKWB-FC block, I have to assume that the VRM's are close to the same temp. Again, I don't think the card has ever gone above 50c under load.
It's not really about the money. I'll have to design and build them myself, because there are none available for my board. And since I have a massive cooling loop, with the CPU, GPU, and RAM on it, I figured why not! I like design and fabrication. Who knows, I may even be able to sell them once I have the final prototype.
What is the best non-electrically conductive TIM Pads you can buy today for a video cards VRM's, GDDR5 Memory, etc? I am redoing all the pads and and also converting the backplate to a thermal solution as well. Thanks
Would also love an answer to this, it's had me bothered that my vega 56 is hitting 112-115c sustained while playing games at an uncapped framerate. Seems defective, because the core is rarely going over 75c. But in the video he says common gpu caps have a recommended temperature of 125c. Really wish it just had better cooling.
31:00 ish you said at 12 hours gaming a day you might wear out your gpu - that's assuming 95°C on the mosfets though, right? At say 75°C (I just measured 72°C the backside of the PCB, exactly where the mosfets sit of a GTX 1070 with my infrared thermometer), according to the data on screen, the card should last about 10 times longer than it does at 95°C right?
I haven't quite finished watching this yet, but I saw him talking about stock voltages having margins that help with VRM degradation. And it instantly reminded me of my last system that ran an FX 8350 that I had at 4.7GHz at 1.475v on the vcore, and after 2 and a half years it got to the point where I needed 1.45v on vcore to handle 4.4GHz even, which I could do on stock voltages when I bought the system, around 1.05v vcore. Pretty sure I have dead caps on that VRM because of how bad that degradation is. Because in a totally dogshit board I can still get that chip to 4.4GHz at 1.05v vcore. Which kinda surprised me because it literally happened overnight. One day it ran at full load with no issues, next day it refused to even get pass the Windows login before crashing. On an Asus Crosshair V Formula-Z. Most likely was my case having terrible airflow that allowed enough heat for the VRM to degrade as much as it did.
i have a friend that one day, his pc litteraly exploded, every capacitor of the motherboard (Ashrock) litteraly blew up, his cpu melted and destroyed the ram and i think also the gpu, someone have a clue of what happened there?
