To give you some idea, aside from me going into vrm details too much.
Hardware unboxed actually made video on X570 boards tests with a 3900X.
The Msi X570 Gaming edge handled the 3900X pretty terrible,
with vrm temperatures well around the 115C.
Above the 105C is actually a number where i personally would get a bit concerned, wanting discrete mosfets to be.
Because that is getting close to their max trashold.
Now to get a bit more into vrm detailes.
THe Msi X570 Gaming Edge is a 4+2 phase design doubled to 8+2 using IR3598 doublers.
Mosfets on the highside 2x Onsemi 4C029N’s and low side 2x 4C024N’s for each phase.
Now lets take a look at the Msi B450 Tomahawk because why not.
4+2 phase board controlled by a Richtek RT8894 pwm, which cannot be doubled from.
For the mosfets they use the same configuration and mosfets like the Gaming Edge.
But the Gaming Edge has twice the amount of mosfets and phases powering the Vcore vrm.
And the Msi X570 Gaming Edge was already struggling with the 3900X.
Those tests were done with just autoboost pbo enabled.
A 3800X can easally pul about 110A ish of current, the 7nm procedee,
is slightly better on efficiency and current draw then the previous gen 2700X.
Which can easally do 130A ish.
But a 3900X can easally pull 160A / 170A ish of current so keep that in mind.
Yes, I did some homework on this and I see what you mean; the B450 boards are often simply not built for the 3900X or >130A CPUs. While some can handle a 2700X like the Tomahawk, they are pushing it quite a bit.
With that said, I think I read somewhere that the Tomahawk MAX upgraded their VRMs, so, one thing I would like to see is a lineup of how hot the rev2 B450 boards get vs the cheaper X570 boards. While the B450 are not top of the line, I still attest they are quite great workhorses for the price they get.
I agree with you that anything above 3700X is dumb to attempt on the “gen1” B450s though, and even the 3700X is pushing it. 3600X and below, quite reasonable.
@wendell, is this a good video idea to clear some misconceptions? Perhaps phrase it as “do you really need that x570”?
I agree this is an offtopic discussion; I want to know two things though.
Has this been extensively tested, or is this just theory?
Is there a thread where this has already been discussed?
I perceive your reasoning to be something along the lines of “B450 is cheap, therefore it must be trash.” This is not an unreasonable opinion, but if you have nothing else to back this up on except “It’s last gen tech”, then yes, I will take that opinion with a large grain of salt.
I want benchmarks, use case tests, how much does this really matter in the real world? NVMe drives are superduper fast yet there is pretty much no advantage to go PCIe 4.0 over PCIe 3.0 in real world computing right now. I suspect this is a similar issue.
Your observations about the 9900K are very spot on, however, and yes they are the more stable platform for now. Give it 6 more months, perhaps not so much.
If we are to continue this discussion let’s move to a new thread.
Ok, so researching and looking around some more, I think I have reached a conclusion on this;
The MSI B450 Tomahawk Max should most probably not be used for 3900X, and I was wrong for suggesting that. Based on all the evidence, VRMs should end up around 100-105 degrees celcius which is a bit much.
However, there could be a silver lining here - IF you put in a beefy cooler and blow air directly on the VRMs, temps could go down to as low as 85-90 degrees, not ideal but still within range. Do not watercool, you need to cool those VRMs as much as possible. An undervolting could also be possible on this board, which may help things.
If anyone out there is brave enough to try this and confirm, let me know your findings. Also interesting would be the 3950X and if that performs any better.
As for 3800X and below, the Tomahawk MAX should be powerful enough to support them. Obviously not for an OC scenario, but definitely at stock.
I agree on the second one. And really, if you’re spending that much on a CPU, get a proper Mainboard to support it. If you’re getting a 3900X, you’re clearly not strapped for money and aren’t exactly looking in the budget range.
But going from this to stating “Ryzen 3000 clearly needs X570” is a long shot. There are reasons to go with the new Chipset, but generalizing that any Ryzen 3000 CPU definitly needs an X570 Board to be “safe” and function properly is sketchy at best.
I see no reason to go for X570 when building a 1000$ System with an 3600X…
I still think you could do something stupid like, put it in a fireproof chamber and run a 4 hour torture test, see how bad it will get and how bad temps will be and so on. From that you could get a good idea on how far you can extrapolate.
Maybe “But will it burn?” should be the title of the series instead. Or it’s just a terrible idea in general but I for one would find it interesting to see how bad it gets.
The best B450 boards vrm wise are the Msi B450 Gaming pro carbon AC,
and the B450 Krait gaming i believe out the top of my head.
They both have the same 4+2 phase vrm implementation as the B450 Tomahawk.
However they doubled up the components on each phase.
So those boards look like a 8+2 phase design, but they definitely aren’t.
But still given Hardware unboxed results on the 3900X + Msi X570 Gaming edge.
I don’t think that those two particular boards would handle a 3900X well either.
Because in terms of max current capability handling those two B450 boards,
and the X570 Gaming edge are similar.
The only main difference is that the X570 Gaming edge is a 4+2 phase doubled to 8+2,
controlled by an IR35201 pwm and IR3598 doublers.
For the rest they use the same ONsemi discrete mosfets,
which aren’t particularly bad for a budget board, but for a $200,- + X570 board,
seeing those Onsemi fets used is kinda unacceptable if you ask me.
This is unfortunately one of the reasons why Msi does not have any interesting compatitive board under the €300 range.
What is a good load to test VRM?
P95, aida, occt? Load on gpu and CPU or just CPU.
The tests I’ve seen doesn’t really specify, and most are no airflow open bench. Which while good at showing the effectiveness of the VRMs basically ignores cooling solutions. And convection is a really shitty air mover.
As long as it’s cool enough in a real use case, with airflow, I don’t much care if the temp is 90C or 40C.
Well you could use P95 or occt for example indeed.
Or you could do a long video render project render it on the cpu.
That will also be a pretty good test i think.
However keep in mind that temps measured on the vrm it self,
isn´t the actual temperatures inside the mosfet.
If we take the video that HW unboxed did on the Msi X570 Gaming edge as an example.
The vrm was hitting 115°C in his tests, which for a discrete mosfets setup is pretty toasty.
Because the actual temperature inside the mosfet will be higher.
Still of course i agree that airflow inside a case will also be a factor in play in comparison to an open test bench.
But aside from all that, the vrm implementations on pretty,
much all B450 boards are mediocre at best.
Like i said the Msi B450 Gaming pro Carbon AC, and the Krait Gaming i believe are the best B450 boards out there vrm wise.
But i still wouldn´t consider those decent enough for a 3900X or 3950X.
Will it work if you don´t push a 3900X too often?
Sure it will, but yeah i personally wouldn´t really recommend it.
I understand that money is a factor in play.
However there are decent X470 options as well, like the Crosshair VII hero or Asrock X470 Taichi.
The MSI B450i board is a good one, it has native 6+2 IR 3555M 60A and IR 35201 controller, I was pulling over 200W running prime95 overnight with my 1700 at 4ghz 1.4v and the highest temp recorded for the vrm was 75C. That reading comes from the chips too, being read from the controller using hwinfo.