Well that could also be verywell a false readout really. Software isnt really an appropriate manner of measuring mosfet temperature. Sensors can be wrong, or software can have readout bugs. If you want to measure actual mosfet temps you have to use something like a thermal laser to determine the exact mosfet temperature.
Anyways, allthough the vrm implementations on the Msi boards arent that great, because of the cheaper mosfets. However to be pretty honnest, if you want to overclock the max out of an R7 cpu, its just better to invest in a decent X370 board really. i cannot say that enough. Still wendell reached 3.9Ghz on the board without much of an issue if i remember correctlly.
But yeah if you want the best overclock experiance, then people should simpley invest in a decent X370 board. And then at this moment, Msi is probablly not the brand to look at. But yeah i do researches on vrm implementations pretty much since the X58 era of motherboards. And i kinda know what to expect from brands.
Well basiclly if the mosfets / powerstages used for main vcore were IR3555 60A stages or similar. Then a 4 phase vrm for main vcore would more then eduquate enough. But thats of course not the case with these cheaper B350 boards. Raw phase count doesnt really tell the whole story. The story goes deeper then that. Of course more true phase does matter in terms of ripple current reduction, for a cleaner power delivery, and better efficiency. But thats a deep story for another time.
B350 boards in general arent really the greatest choice if you want high overclocks like 4.0GHz on R7 cpu´s 24/7. For those kind of things,. people should really look at board like the Asus crosshair Vi hero, Asrock Taichi, Asrock professional gaming, Aorus Gaming K7. But hey, of course B350 boards are cheaper for a reason.
I'll say this for this board... I'm running 3.8 comfortably, but it will not run stable for anything above that. The fact that I'm air cooling may have something to do with that, but 3.8 has been solid for me with 2933 ram. Cinebench of 1663 is more than good for me.
That said, I do agree that this board runs really warm, I've seen some really high temps during AIDA64 during load test but I think for my overclock it certainly serves its purpose. BTW, coming from a i5 2500k....this thing is impressive. It's immediately evident in lightroom where I do some photo editing.
The reading is correct. Check the video Pholostan posted yesterday. That guy makes the calculation. Around 22 Watts of heat are generated at the VRM stage. The cooler design cannot dissipate that amount of heat.
After my finding yesterday I adjusted my case fans, I'm running the board inside a Mastercase Pro 5 maxed out with 5x 144mm fans and also have 2x CPU fans. So now the back and back top fans are basically in sync with the CPU coolers to get at least some cooling going on the VRMs. Noticeable but still too little effect on the VRM temp, while I pay for it with increased noise.
Well the board "only" cost around 120€, so I guess I'm gonna run it until it burns and buy a better one next year. Still a pity, if only I knew before, all those problems, all the time invested..
I've also now flashed BIOS version 7A34v14 released two days ago. Gave me a minor heart attack cause afterwards I tried some higher mem frequency and was rewarded with nothing but a blank screen and a cursor at BIOS stage. After a reboot it worked again, but didn't really change anything for my DRAM compatibility. Still locked at 2667Mhz and still having the cold boot issue after the PC was unplugged for the night. So the first issue is due to me running 16GB - sticks, the later is a Ryzen bug not providing the adequate DRAM voltage directly at turn-on. Explained in the reddit I posted above. VRM related I guess, too. Also I didn't see a new option to either adjust SoC Voltage or provide DRAM boot Voltage. Well at least they tried, and have reacted pretty quickly, some respect for that.
I don't think so, those mosfet's will run hot under a load like that thanks to them not being very efficient. They need to dissipate quite some heat through that small heatsink. And IIRC they are rated to run at 125C continuous, so if they are cooled they'll survive. The capacitors right next to them will probably not be as happy though, as caps are usually not rated for that kind of temperature. More expensive Rubicon top out at 105C IIRC. Will it be a problem though? I don't know, but I wouldn't bee too worried. The mosfet's will be alright and caps usually have thousands of hours in their ratings, you need to run your CPU at high voltage for quite some time to get to those numbers. Get some airflow over the heatsink and call it a day.
The cheapest MSI B350 board has three phases. All others seems to have four with one less high side mosfet except the Carbon that looks like it has doublers? All the X370 boards except the Titanium seems to use doublers too. Not sure how they do that if they all have that same controller, the RT8894A. The Titanium seems to be the only one that has an IR35201 controller (six phase).
Not that other makers B350 boards look fantastic, it is the lower tier after all.
To leave some positive feedback after all: The Ryzen 7 is a low-cost beast when it comes to 3D rendering. Also Blender CPU rendering was highly optimized last year, my oc Ryzen beats my freaking GTX 1080 by around 10 minutes on a complex Cycles scene, wtf? Ryzen OC: 37min GTX 1080: 45min EDIT: Updated the numbers, read wrong. Still impressive!
So mileage will definitely vary depending on the scene, but considering how much cheaper it is to build a CPU network rendering node, rather than a GPU rendering node + the fact that I can completely use my system, even game while rendering on only 8 CPU threads + OSL only available on CPU + the much more RAM available on CPU rendering makes me want to push harder on the OC. I'm gonna measure the VRM cooler now and build some custom aluminium cooler which I will glue on top, just as an experiment.
EDIT2: So 3 little Aluminium Heatsinks (ca. 15mm x 30mm x 20mm) + a bit of thermal glue dropped the VRM temps around 10°C, which doesn't sound that impressive at first, but the VRM temps are also rising a LOT slower now. In combination with the adjusted fans I'm now not exceeding 94°C on the VRMs on full load.
That's great cause not only I expect for the board to live much longer, but also gives me a few Mhz / Volts more headroom.
Success. Cost: 10 minutes of work, 15 minutes of walking to my engineer-dad and back to gather the parts.
Sure, I'll try and make some tonight. It's very little space in there, gonna get my wifes camera with flashlight.
Neither the one nor the other in my opinion. I'm basically glueing together two metal parts with a glue that can be removed completely using rubbing alcohol. No electronics harmed. In the past I've even sold custom modded parts like that on Ebay, buyers where pretty happy.
Replacing (soldering) the capacitors would be another story, I'd still do it if I had an otherwise fully functional board with e.g. a blown capacitor. Still I wouldn't call it reckless. I've grown up with electronics and computers, sometimes I buy stuff broken cause I couldn't afford it otherwise, then try & repair it. Usually it works.
I don't mean reckless in the sense that it could go wrong (that would be a given). I mean if you ever need a warranty, even with the stock cooler put back on, MSI will probably deny warranty claims due to unauthorized modifications, no?
Yeah those Nikos fets arent really efficient. They are very temperature sensitive, and its kinda sadd to see Msi using them on their premium Xpower board awell.
But yeah B350 is lower tier afterall indeed. Like i said above, of course certain B350 boards will be capable to overclock R7 cpu´s fine. But if you really want the best overclock experiance for R7 cpu´s, you just better off buying a more higherend X370 board like i said above. The quality diffrence between used components is significant.
The Richtek RT8894A is a 4+1 phase pwm So with doubling you can create 8+2 like the Msi X370 Krait. Or 3+2 like with the lowest end Msi B350M boards.
The hardwareluxx list is a nice list indeed, they still miss some bits and pieces, but they get there. I had most of the info for X370 boards allready from my own analyses. Nice to have an additional list for manny B350´s aswell.
Manu, is this 3800Mhz OC limited by the heat, or is it otherwise unstable? Agreed that this Tomahawk isn't very good quality VRM, but I really think your board is performing worse than others that I've read user experiences about. One should not have to add additional aftermarket cooling to get a mainboard OC to work at what is only 100Mhz over your All Core Boost!
First of all they would have to prove that, wouldn't they? I don't damage anything so there's nothing to prove. And further more:
tl;dr I don't give a f***
Long story: MSI dropped me already when I complained about the "cold boot" issue and other issues with the board and the BIOS. I asked for a warranty replacement and they denied. Similar story with that hardware internet shop which from now on I will avoid like the plague. I asked for a replacement, they kinda agreed after 3 calls and 2 emails, but with a lot of additional bother, me paying all of their overpriced shipping cost (basically double the regular shipping cost) back and forth, paying another "payment fee" (yeah, that's right), sending a lot of emails just to get my money actually back and not just a voucher for their shop... For a ~100€ board? All this action can be considered illegal in my country, but what can I do? Get myself a lawyer? Again for a 100€ part?
I certainly believe, at least inside any EU country you can put on any aftermarket cooler you want without voiding warranty. CPUs have overheating protection built in. E.g. Zotac stated for their GPUs you can even replace the complete cooling solution, overclock like crazy, still warranty will be intact, as long as you don't do something funny, like that kind of stuff:
You're right for sure, I really tried to get a X370 board with a PCI slot, there simply weren't any when I bought my new rig.
I'm running 3.9Ghz +- a few Mhz, experimenting, 24/7. Rendering with 16 threads is now peaking around 78°C on the CPU and 94°C on the VRM (before +-120°C and rising).
I still keep getting those greenscreens occasionally. Not sure if it's a Windows 10 crash, not enough voltage. I think I got it one day but they keep coming back.
What limits it? Well basically voltage I'd say, I never dialed in more than ~1.41Vcore and that wasn't enough to reach 4Ghz so I never tried harder cause I cannot run that much Voltage productive anyway without buying new CPUs all the time. And I don't really care about 100Mhz more or less.
Here some images, sadly I couldn't make another temp shot under full load cause yet another greenscreen, so I have to ask you to take my word for it.
Only ten minutes cause the last one crashed and I need to continue with other stuff. Also "only" 15 threads, that why the CPU not fully maxed out. But check the wattage. Take my word VRM and CPU temps are not rising any further than in the image, not after 2h or more, not with 100% load on all cores.
)
