Ryzen 3 Mobo Update

[Seabosz]

hey, Just a quick question my friend wants a pc and ive picked an ryzen 3 1200 for him with a Msi b350 PC MATE but i need to update the mobo to support the cpu do i need to bring my ryzen 7 to update this or is there an diffrent way to do this?

[LienusLateTips]

ya

 

you do

 

But I wouldn't recommend MSI B350 boards

[Seabosz]

Just now, JDE said:

ya

 

you do

 

But I wouldn't recommend MSI B350 boards

he already ordered it and why not? mine is working great

 

[LienusLateTips]

1 minute ago, Seabosz said:

he already ordered it and why not? mine is working great

 

mediocre VRMs compared to others in it's price range

[Seabosz]

Just now, JDE said:

mediocre VRMs compared to others in it's price range

which one would you have bought?

[LienusLateTips]

Just now, Seabosz said:

which one would you have bought?

ASRock AB350/AB350M or ASUS PRIME B350

 

 

It's fine though. If you already have it or already ordered it, it works perfectly fine. It's the type of thing I would recommend switching if you haven't bought but keep it if you have already ordered it.

[Seabosz]

1 minute ago, JDE said:

ASRock AB350/AB350M or ASUS PRIME B350

 

 

It's fine though. If you already have it or already ordered it, it works perfectly fine. It's the type of thing I would recommend switching if you haven't bought but keep it if you have already ordered it.

okay is this like an issue with msi motherboards or like every msi product and how will it effect me?

 

[Seabosz]

Just now, dave_k said:

MSI has crappy memory support and VRMs suitable only for 4 cores

so is it going to lower my r7 1700 performance?

 

[Seabosz]

Just now, dave_k said:

Wow, putting R7 on MSI B350 is pure fire hazard.

No, but expect much shorter lifespan of the board

i even have it overclocked to 3.9.......

[Seabosz]

Just now, dave_k said:

Yeah. Expect failure in next 6 or so months

VRMs are baking over 100°C

are there meters to see this?

 

[Seabosz]

Just now, dave_k said:

You need to change your mobo. There is not chance that mobo holding up for more than year

I dont have the money to do that i've spend it all on this combo

[Seabosz]

if the mobo dies on me will it also take the cpu and or ram with it?

[PCGuy_5960]

20 minutes ago, dave_k said:

VRMs are baking over 100°C

Can be fixed if you remove the heatsinks and use a fan to cool the VRM. As der8auer said in this video:

19 minutes ago, Seabosz said:

are there meters to see this?

You can use an infrared thermometer.

Edited September 4, 2017 by PCGuy_5960

[DocSwag]

4 hours ago, dave_k said:

The MSI B350s have to bad heatsink it wont really help.

Ryzen 7 overclocked on those boards is still a painful overload for the high-side mainly

A fan ALWAYS helps.

 

Just having a fan blowing on the vrms could potentially lower temps by 20-30C

[mariushm]

If you don't do overclocking, the MSI boards with b350 chipset are perfectly fine.

 

If you do overclocking, the VRM circuit - due to somewhat undersized heatsinks - will be a bit hotter than what's generally agreed as "safe long term temperatures".. think 70-80c max.

 

Normally you don't want the circuit to go above 100-105c because above that threshold, in time (weeks, months of 24/7 use at near 100% loads ) the circuit board can be affected, the glue that keeps the layers of glass fiber can go bad and the circuit board then can be affected.

The chips under the heatsinks can survive up to 125-150c.

 

The motherboards already should have a temperature sensor under the heatsink and they should throttle the cpu if the temperature gets close to 100c - basically your cpu will lower its frequency a bit so that less power will be used and therefore less heat will be produced by the VRM (giving it time to cool down)

 

So you can leave it overclocked and don't panic, people in this thread exaggerate a lot, it's not THAT bad. In all likelihood your VRM even with overclocking isn't reaching 100c when gaming (as most games don't use all 8 cores) and definitely won't use your cpu at 100% in Windows or when watching movies.

If you're really worried it's enough to just get a 80-120mm fan and position it above the VRM heatsink to move some air over it and that should be enough.

[Guest]

9 hours ago, dave_k said:

The MSI B350s have to bad heatsink it wont really help.

Ryzen 7 overclocked on those boards is still a painful overload for the high-side mainly

Ok now i can say this is not 100% right i see a massive decrease in my VRM temps when using the stock fan that comes with my 1700 over my VRM. Temps drop 25C from max load. 

 

What the user needs to do is set a fixed voltage of 1.2V this equals the amount of amperage of an 4 core at 1.425V. 

 

Stick with stock speeds or overclock but limit voltage to 1.2V(3.6Ghz should easily be achievable) 

 

 

Also my board with stock temps is perfect its when you OC at stock under prime95 for 4 hours the VRM gets to 68C which is fine. CHEW has said that below 100C is fine and try and get VRM to be below 80C during 100% load for a long lasting board. Also i will keep saying this MSI>>>>>>>>>>>>>>>>>>>>>>>>>>Gigabyte B350 as they have better VRM but basically useless heatsinks yes even worse then MSI's that is why users are seeing 100C+ at stock with a R7 so if MSI B350 is terrible Gigabytes is a total nightmare. 

 

Techspot has also said this on basically every review of a board 

"Anything under 60C is great, 60-80C is acceptable, and anything above 80C is a bit worrisome (if at stock)." 

Read more: https://www.tweaktown.com/reviews/8099/asrock-x370-taichi-amd-motherboard-review/index10.html

 

KEEP VRM below 80C and you are fine. Monitor system+motherboard temps in hwmonitor as that is what MSI uses for their VRM sensor on B350. 

 

 

[79wjd]

1 minute ago, dave_k said:

These mobos are trash and they will overheat.

You have to understand.

If you measure 70°C on the heatsink, it doesnt mean the mosfets have 70°C.

They have 100°C

This sounds a lot like the "CX Green will kill your computer and burn your house down".....

[Guest]

Just now, dave_k said:

These mobos are trash and they will overheat.

You have to understand.

If you measure 70°C on the heatsink, it doesnt mean the mosfets have 70°C.

They have 100°C

Not sure where but in the VRM thread at overclocker.net a user with a MSI b350 board pointed a IR gun at the back of the board right where the mosfets are and the temp was the same as it was in software. 

 

VRM's on all B350 are basically trash and underpowered for a 8/6 core OC

[Guest]

Just now, dave_k said:

You still have the Tj. The mosfets are inside most likely ceramic casing which does isolate it. When you measure hot on heatsink, the mosfet has be frying

I actually haven't thought of that. Still i think 80C max in software is fine for stress tests. I was looking up the mosfets that are being used in the board and they can handle 150C max

 

Mosfets being used 4 NIKOS PK616BA and eight NIKOS PK632BA MOSFETs

 

 

[mariushm]

1 hour ago, dave_k said:

These mobos are trash and they will overheat.

You have to understand.

If you measure 70°C on the heatsink, it doesnt mean the mosfets have 70°C.

They have 100°C

No, the difference in temperature is not that big.  The thermal pads will not transfer temperature efficiently (or in better words, not as efficient as thermal paste), but you'll probably have only around 10-15c difference between the top of the mosfets and the heatsink.

And even if the mosfets are at 100c, they're still rated for 125-150c (125c is common for lots of mosfets, 150c less often)

 

[mariushm]

51 minutes ago, dave_k said:

Yeah, but the heat reflects into to the capacitors, on MSI especially shitty ones.

 

The polymer capacitors used on MSI motherboards are no shittier than what's used on other boards. Even cheapest 2000h@105c cpacitors will last years.

Even if the mosfets are at 100c, the copper fill in the motherboard and the inductors and the heatsinks over the mosfets are so good at sucking up heat that by the time heat reaches the capacitors, you'd be at less than 60-70c near the capacitors. 

 

Polymer capacitor's lifetime isn't calculated like you'd calculate electrolytic capacitor's life, where you double lifetime for every 10c drop in temperature from the rated temperature (usually 105c). With polymers, a 10c drop will actually increase the life of a polymer by a huge amount.

 

You can basically use the formula L = Lorig x 10^z where z = (Tmax - Tambient) / 20

 

So let's say the capacitor runs at 80c and it's rated for 2k hours at 105c .. then the capacitor should last at least  2000 x 10^(105-80)/20 = 2000 x 10^1.25 = 2000 x 17.78 = 35,560 hours or 1481 days or 4.05 years.

That's 4 years of continuous 24/7 use at 80c which not gonna happen in real life.

If you do video encoding or prime95 or rendering 10 hours a day out of 24 hours in a day, your capacitors will last 3556 days or ~10 years - you're gonna throw away the system in 5 years because by that time you're gonna have 64 core processors.

 

 

51 minutes ago, dave_k said:

 

Any B350 with overclocked R7 will die within 1 year

You can measure it, ROJA (junction-to-ambient) for Nikosem PK616BA is 50°C/w i think. The mosfet outputs 2.55W of heat. . .

 

2 hours ago, jdwii said:

I actually haven't thought of that. Still i think 80C max in software is fine for stress tests. I was looking up the mosfets that are being used in the board and they can handle 150C max

 

Mosfets being used 4 NIKOS PK616BA and eight NIKOS PK632BA MOSFETs

 

 

 

Oh ffs ...  A Ryzen at 3.8..4 ghz will use at most 150 watts in some particular apps.  Tom's hardware measured up to 140 watts at the 8 pin cpu connector. 

Here's link if you don't believe me : http://www.tomshardware.com/reviews/amd-ryzen-7-1700x-review,4987-8.html

 

With Ryzen, even if you set the voltage to maximum allowed 1.42v you're not going to go much above 100 amps (100 x 1.42 = 142 watts) and that's split across four phases, so you have 25a per phase. 

 

That 140 watts is again measured at the connector, and on a good motherboard the VRM would be at best 95% efficient ... so basically the CPU itself uses around 130 watts and maybe 10 watts are wasted as heat.

 

Even if you assume the VRM on MSI boards sucks and it's only 90% efficient, the VRM will have to dissipate around 20 watts as heat. Those 20 watts will be spread across 4 phases, each with 3 mosfets (so 12 mosfets overall) AND across the inductors and a tiny bit of waste will be in the VRM controller itself as well.

Of course the big part of the wasted heat will be in the eight mosfets (4 pairs of 2 mosfets) so let's say those will dissipate 16 watts out of 20 watts ... you'd still have only 2 watts of wasted heat per mosfet.

 

And we can simply look up the mosfet datasheet: pk632ba.pdf (seems to not be Nikos but some clone, but on other sites i see the same specs mentioned)

You can see in datasheet there, the mosfet works up to 150c, it has continuous current of 20A and max 2.6w power dissipation  (both at 25c ambient, which ain't gonna happen). But they're also rated for 16A and 1.6w at 70c

 

So if we go with 100 amps across 8 mosfets, we have 12.5A per mosfet (less than either maximum, 20A or 16A) and we have maybe 2w per mosfet.

 

Another way to estimate how much the mosfet will dissipate ... Rds (on) is 3.3 mOhm but we can use 5 mohm for easy math ... so at 12.5A per mosfet ... P = I²R  = 12.5²x0.005 = 156.25 x 0.005 = 0.78125 watts dissipated just due to Rds(on) ...

 

The mosfets will dissipate some heat through the circuit board where they're soldered to big copper fills that will suck some heat (and that's where the Junction to ambient thermal resistance matters) and then the heatsink helps cool the mosfets more. The heatsink doesn't do all the cooling, just a significant amount of cooling.

If this was the case, we wouldn't have motherboard without any heatsinks at all.

 

It's FINE.   Don't fu@*ing worry so much about it. 

Yeah, like I said, it's weaker VRM compared to others but it can do Ryzen 7 with some overclocking just fine without dying. Stop spreading crap.

[Guest]

^^^^^^^^^^^^^ Ok so you are willing to replace my parts if i set 1.4V 3.9ghz on my setup and have it run Prime95 blend for 24 hours or hell do some heavy encoding that i need to get done? Just like to know why you think 150+C is fine? Also note H.264 encoding took my CPU to higher temps then even Prime95 blend so you are telling me i can do this safely? 

 

 

If you are willing to replace my setup i will push it at that and have it do folding@home already read several users who broke their R7+B350 setups doing this but sense you say its good lets give it a good old try. 

 

 

If you aren't willing to do this then i can say its false many FX users had issues where their board will actually start on fire do to crappy VRM. 

[mariushm]

I don't say 150c is fine. I just said 150c is the maximum temperature the mosfet can tolerate. It will blow up above that temperature.

Designers will try to keep the temperature below 100c and maybe will allow the mosfets to get to around 120c for very brief periods of time (let's say seconds). Like I explained, continuous temperatures above around 105c is bad for circuit boards, so designers try to avoid that. Making the mosfets go up to 120c+ will definitely make the circuit boards get close to 100c or higher and nobody wants that.

Your mosfets won't ever reach even 120c, there's thermal sensor near the mosfets (most likely on the same copper fill on the circuit board which acts like a heatsink) and the bios would throttle your cpu if the mosfets go over 100c or something like that.

The Ryzen processors just can't achieve high enough frequencies to use a lot of power. You can give them a budget of 200-250 watts and they'll still only do 4.1ghz.. maybe 4.3 ghz.

At 3.9..4.0 ghz they just won't use more than let's say 150 watts ... and like i said, that means less than 20 watts of wasted energy across 8 mosfets, so the mosfets simply can't get very hot.

 

As a comparison, here's a msi board with same mosfets but a 8 phase vrm but which uses just a pair of mosfets for each phase (not 1 high and 2 low for each phase): https://www.tweaktown.com/reviews/8171/msi-x370-sli-plus-gaming-motherboard-review/index10.html

So really not much better vrm, just a bit better because mosfets are spread across a wider surface and naturally the heatsink will be wider as well.

 

You can see at 4 ghz measuring 174w at the cpu connector (8 phases means a bit less efficiency, more wasted power in vrm) and you can see in the thermal images that the hottest component is the VRM controller itself at 75c (during heavy load) - I think that chip can work 24/7 at 90-100c so if it reaches a max of 75 watts, it doesn't need a passive heatsink to stay cool.

 

The heatsink itself across the mosfets and the circuit board right near the heatsink are not even close to 60c. Like I said, this particular x370 board is not much better in the vrm department compared to  vrm above, so the above vrm will not be much hotter.

 

[mariushm]

I don't say 150c is fine. I just said 150c is the maximum temperature the mosfet can tolerate. It will blow up above that temperature.

Designers will try to keep the temperature below 100c and maybe will allow the mosfets to get to around 120c for very brief periods of time (let's say seconds). Like I explained, continuous temperatures above around 105c is bad for circuit boards, so designers try to avoid that. Making the mosfets go up to 120c+ will definitely make the circuit boards get close to 100c or higher and nobody wants that.

Your mosfets won't ever reach even 120c, there's thermal sensor near the mosfets (most likely on the same copper fill on the circuit board which acts like a heatsink) and the bios would throttle your cpu if the mosfets go over 100c or something like that.

The Ryzen processors just can't achieve high enough frequencies to use a lot of power. You can give them a budget of 200-250 watts and they'll still only do 4.1ghz.. maybe 4.3 ghz.

At 3.9..4.0 ghz they just won't use more than let's say 150 watts ... and like i said, that means less than 20 watts of wasted energy across 8 mosfets, so the mosfets simply can't get very hot.

 

As a comparison, here's a msi board with same mosfets but a 8 phase vrm but which uses just a pair of mosfets for each phase (not 1 high and 2 low for each phase): https://www.tweaktown.com/reviews/8171/msi-x370-sli-plus-gaming-motherboard-review/index10.html

So really not much better vrm, just a bit better because mosfets are spread across a wider surface and naturally the heatsink will be wider as well.

 

You can see at 4 ghz measuring 174w at the cpu connector (8 phases means a bit less efficiency, more wasted power in vrm) and you can see in the thermal images that the hottest component is the VRM controller itself at 75c (during heavy load) - I think that chip can work 24/7 at 90-100c so if it reaches a max of 75 watts, it doesn't need a passive heatsink to stay cool.

 

The heatsink itself across the mosfets and the circuit board right near the heatsink are not even close to 60c. Like I said, this particular x370 board is not much better in the vrm department compared to  vrm above, so the above vrm will not be much hotter.

Quote

 

If you aren't willing to do this then i can say its false many FX users had issues where their board will actually start on fire do to crappy VRM. 

FX processors were made on 32nm or 28nm or whatever and had a TDP of up to 125w (excluding those fx9*** abortions).

Ryzen was made at 14nm so it's more power efficient , it peaks at 95w tdp with 1700 and others at a TDP of 65w. They just won't use as much power as the FX processors.

[mariushm]

16 minutes ago, dave_k said:

Hey look, the PK616BA will do 2.55W of heat under R7. Thats no fine, it can do max around 20A and you are pulling 25A.

How much current is 150 watts / 12v then divided by four phases ?

 

ps. and stop saying "2.55w of heat under R7" because I'm sure you didn't actually measure it, you're just repeating some bogus number from somewhere.

 

tldr : high side mosfet doesn't see the same current as the low side mosfets (that's why there's two low side mosfets per phase in the vrm and just one high side mosfet per phase)

 

 

 

 

 

A bit old but explained relatively well.