-basically AM4 mobo VRM discussion-

[just_dave]

8 hours ago, rijzen said:

The asrocks in x370 flavor run much cooler though, like 33c at the back but that's a true 8+4 I think, I honestly don't remember, just poured through tons of data when deciding on my board. The MSI 4+2 on the carbon is a very large 4+2 using alot more physical components than the others. Not sure what Asus uses but I tried the x370 prime and it was a shit show of compatibility issues for me personally. But I would assume the strix outpaces the prime line anyways

X370 Prime Pro has the best VRM for great price. True 6+4 NexFET

[bowman]

11 hours ago, dave_k said:

Asus is better than msi 3+2 4+2 and gbt 4+3. MSI "4"+2 is better and asrock is somewhere the same as Asus

Its not about vendors being better of one another. Its abot certain VRM configurations each vendor chooses to use on a particular board design.
In this case, among all B350 boards the pro carbon's "big 4+2" (4+2 PWM with double the components on the Vcore side) is the best design you can get on B350 currently. Period.
Of course, Asuses simple 4+2 design is going to be better in current output than MSI's 3+2 but thats irrelevant to the point. The Strix uses doubled low side on a 4+2 design, so almost the same as MSI's big 4+2 but missing half the chokes and highside FETs per phase. Thats a difference though. Asuses design is doubled on SoC side however, which i find useless.
Also they use different manufacturers for MOSFETs, like ONSemi for Asus, Niko-Sem for MSI. I've always known ONSemi to be better, but surprisingly the datasheets tell a different story, where the NikoSem-s can handle more current.

 

Ultimately for anyone OCing an R5 both boards will do just fine. Into the future, regarding Zen+, the MSI's design seem to be a bit superior.

[just_dave]

40 minutes ago, bowman said:

Its not about vendors being better of one another. Its abot certain VRM configurations each vendor chooses to use on a particular board design.
In this case, among all B350 boards the pro carbon's "big 4+2" (4+2 PWM with double the components on the Vcore side) is the best design you can get on B350 currently. Period.
Of course, Asuses simple 4+2 design is going to be better in current output than MSI's 3+2 but thats irrelevant to the point. The Strix uses doubled low side on a 4+2 design, so almost the same as MSI's big 4+2 but missing half the chokes and highside FETs per phase. Thats a difference though. Asuses design is doubled on SoC side however, which i find useless.
Also they use different manufacturers for MOSFETs, like ONSemi for Asus, Niko-Sem for MSI. I've always known ONSemi to be better, but surprisingly the datasheets tell a different story, where the NikoSem-s can handle more current.

 

Ultimately for anyone OCing an R5 both boards will do just fine. Into the future, regarding Zen+, the MSI's design seem to be a bit superior.

I am talking about the second MSI 4+2 thats in Tomahawk etc. thats also a crap. Dont need to tell me, i know almost everything about these boards. Asus wins over a lot of Gigabyte boards (even K5 uses this shitty VRM) and over MSI 4+2 and 3+2, not big 4+2. ASRock has worse voltage control (3 phase vs 4 phase) but more components so they are the same as Asus

[rijzen]

7 hours ago, dave_k said:

X370 Prime Pro has the best VRM for great price. True 6+4 NexFET

Yeah the prime pro is good but I tried the x370-a prime (non pro) and it hated my memory (even on the qvl) and bios issues were annoying.so I sent it back

[just_dave]

Just now, rijzen said:

Yeah the prime pro is good but I tried the x370-a prime (non pro) and it hated my memory (even on the qvl) and bios issues were annoying.so I sent it back

issues were at the beggining with all boards, fun fact. Today i operated with Tomahawk which is called to have the best BIOS development from MSI. With latest BIOS, i couldnt get same modules as i run now over 2666, the BIOS resetted itself 3 times. It was a real nightmare, MSI never again. X370-A is the same as B350 Prime Plus (VRM wise)

[rijzen]

3 minutes ago, dave_k said:

issues were at the beggining with all boards, fun fact. Today i operated with Tomahawk which is called to have the best BIOS development from MSI. With latest BIOS, i couldnt get same modules as i run now over 2666, the BIOS resetted itself 3 times. It was a real nightmare, MSI never again. X370-A is the same as B350 Prime Plus (VRM wise)

I bought the Asus like 3 weeks ago and used the most current bios. The pro carbon I have now worked with my memory at 2933 out of the box (and it's not even on msi' qvl) and it's DDR 4 3000mhz so much better compatibility for my needs.

[just_dave]

1 minute ago, rijzen said:

I bought the Asus like 3 weeks ago and used the most current bios. The pro carbon I have now worked with my memory at 2933 out of the box (and it's not even on msi' qvl) and it's DDR 4 3000mhz so much better compatibility for my needs.

wat issues you had?

[rijzen]

2 minutes ago, dave_k said:

wat issues you had?

With the Asus my memory would not go past 2133 even with manual timings and tried the xmp 1 and 2 and the onboard audio was shit. 

 

The bios issue was that I couldn't set a manual overclock at all it just reset itself everytime it rebooted

Edited August 9, 2017 by rijzen
Forgot some stuff

[just_dave]

5 minutes ago, rijzen said:

With the Asus my memory would not go past 2133 even with manual timings and tried the xmp 1 and 2 and the onboard audio was shit. 

For a good audio you have to go for the Strix, the B350-F kills the B350 Carbon today. The SupremeFX is way better than the MSI ALC1220 non pimped up version. Can say from my personal experience.

The offset is a bit painful, the Prime Plus and A series are on the same par as MSI PC Mate, which i wouldnt touch because of the shit 4+2 vrm

[bowman]

On 2017. 08. 09. at 4:46 PM, dave_k said:

The SupremeFX is way better than the MSI ALC1220 non pimped up version.

How would you describe "way better"? Because the audio IC is technically the same on both boards. Is the difference in the software suite?

[just_dave]

8 hours ago, bowman said:

How would you describe "way better"? Because the audio IC is technically the same on both boards. Is the difference in the software suite?

The sound is fuller and basses feel more bass, they use different software than MSI

[just_dave]

On 7/26/2017 at 8:39 PM, KarathKasun said:

Thats where you are wrong.  At 25c PK616BA can push 50A per FET and the PK632BA can push 78A, and they can maintain nearly triple the current of the OnSemi parts at 100c.  They just are not as efficient, which is less important when you double up the FETs in each phase because you halve the on/active resistance.

 

Also, the MSI boards have double the chokes/inductors and caps as well.

@KarathKasun

Onsemi 4C06N(same as B) are rated at 69A at 25°C and Package limit is 80A. Where the fuck you got the information about 20A?

[KarathKasun]

9 hours ago, dave_k said:

@KarathKasun

Onsemi 4C06N(same as B) are rated at 69A at 25°C and Package limit is 80A. Where the fuck you got the information about 20A?

It depends on what kind of board and ground plane you use.  20A is the limit for standard pad size @ 25c (14.9A @ 80c).  69A is the limit for 1oz+ ground/power plane and larger than normal solder pads @ 25c (52A @ 80c).  80A is the current limit of the package itself with a package temp of 25c, and it will fail at that point.  See notes 1, 2, and 3 of the first chart.

http://www.onsemi.com/pub/Collateral/NTMFS4C06N-D.PDF

 

The datasheet I could find at the time only had the standard trace/pad dimension ratings on it, thus 20A @ 25c or 14.9A @ 80c.

 

You should be looking at the 80c ratings, not 25c ones.  25c is not a realistic temp for the FET junction when used in steady state voltage regulation on a motherboard.  If you look at the PDF for PK632BA it can push 49A at 100c, making it roughly equivalent to the 4C06N (with the extra conditions noted in its PDF).  Though with a bit more heat to dissipate and a lower peak amperage capability, the latter does not matter much in this use case though because the VRM will be used in a relatively steady state way.

[just_dave]

22 hours ago, KarathKasun said:

It depends on what kind of board and ground plane you use.  20A is the limit for standard pad size @ 25c (14.9A @ 80c).  69A is the limit for 1oz+ ground/power plane and larger than normal solder pads @ 25c (52A @ 80c).  80A is the current limit of the package itself with a package temp of 25c, and it will fail at that point.  See notes 1, 2, and 3 of the first chart.

http://www.onsemi.com/pub/Collateral/NTMFS4C06N-D.PDF

 

The datasheet I could find at the time only had the standard trace/pad dimension ratings on it, thus 20A @ 25c or 14.9A @ 80c.

 

You should be looking at the 80c ratings, not 25c ones.  25c is not a realistic temp for the FET junction when used in steady state voltage regulation on a motherboard.  If you look at the PDF for PK632BA it can push 49A at 100c, making it roughly equivalent to the 4C06N (with the extra conditions noted in its PDF).  Though with a bit more heat to dissipate and a lower peak amperage capability, the latter does not matter much in this use case though because the VRM will be used in a relatively steady state way.

So i was right but there is bigger difference on High side, not only 4C06B is bit more efficient and bit more powerful than PK632BA, but the 4C09B on High side is way better than PK616Ba, which has lower power and higher heat output, also high RDS(on) resistance, which means it will run hotter than 4C09B

[KarathKasun]

Strix only has 4x high side FETs vs 8x on the Krait.  So they are closer than you think.

[Guest]

On 7/26/2017 at 3:35 PM, dave_k said:

The MTB sensor reported 70°C on VRM while gaming 

How on earth man that's worse then the Tomahawk at stock you mean once it was OC right? 

[Guest]

On 8/8/2017 at 5:13 PM, bowman said:

I think you are refering to lower end MSI boards being worse than the Asuses as they use 3+2 designs if i recall correctly.
In his crash course video, buildzoid clearly marks the MSI B350 Gaming Pro Carbon to be the best B350 board VRM wise.

Also you're missing the fact that the B350 variant pro carbon board has the same doubled 4+2 design as the much more expensive X370 one.
Although the pro carbon is still technically the same 4phase design as the Strix, the doubling gives it twice the components altogether to supply current on.

Priced nearly the same, the MSI board might be a better choice regarding power delivery.
However, taking other factors into account like aesthetics, software, BIOS, the Strix holds on for its own. Im also in hesitation between the two of them...

All AM4 MSI boards use the same NIKO Crap, its just that the titanium has a great heat-sink and makes up for it plus the more true phases you through at the problem the better the temps will be.  

[just_dave]

2 hours ago, KarathKasun said:

Strix only has 4x high side FETs vs 8x on the Krait.  So they are closer than you think.

Main MSI problem are the century turning PK616BA. With R7 100A Asus has highside loss/fet 1.7W. MSI has 2.12 for single and 1.06 for doubled. Krait will output more heat but spread on bigger surface. Nikos suck.

[KarathKasun]

1 hour ago, dave_k said:

Main MSI problem are the century turning PK616BA. With R7 100A Asus has highside loss/fet 1.7W. MSI has 2.12 for single and 1.06 for doubled. Krait will output more heat but spread on bigger surface. Nikos suck.

With the highside doubled it generates less heat and has more area.  Loading is effectively halved, reducing the thermal dissipation by more than half (heat generation is not exactly linear with output).

 

Ill have a board in hand soon, but reports are pointing to ~60c for R7 @ 1.4v/100A.  Its nowhere near as hot (or as crap) as you claim, its in the same temperature range as the Strix.  Its performance is extremely good considering the other trash boards in that price range.  If you really want the best VRM I wouldn't be looking at either of these boards.  The X370 Taichi puts pretty much every other AM4 board to shame.

[just_dave]

2 minutes ago, KarathKasun said:

With the highside doubled it generates less heat and has more area.  Loading is effectively halved, reducing the thermal dissipation by more than half.

 

Ill have a board in hand soon, but reports are pointing to ~60c for R7 @ 1.4v/100A.  Its nowhere near as hot as you claim, its in the same temperature range as the Strix.  Its performance is extremely good considering the other trash boards in that price range.

Here are the pics. Strix is running at 1.44V R7 with no airflow

Pro Carbon is at 1.4V with 120mm 1400RPM, so lots of airflow. 

Now understand?

When High -side is doubled, there is half the current fer mosfets, not half the RDS(on), that applies to lowside

[KarathKasun]

3 minutes ago, dave_k said:

Here are the pics. Strix is running at 1.44V R7 with no airflow

Pro Carbon is at 1.4V with 120mm 1400RPM, so lots of airflow. 

Now understand?

When High -side is doubled, there is half the current fer mosfets, not half the RDS(on), that applies to lowside

I understand that they would both likely be at the same temps with the same airflow.  Unless you have data to make the case that the situation is somehow different?  1400RPM 120mm fan is NOT a lot of airflow.

[just_dave]

29 minutes ago, KarathKasun said:

I understand that they would both likely be at the same temps with the same airflow.  Unless you have data to make the case that the situation is somehow different?  1400RPM 120mm fan is NOT a lot of airflow.

I think (not sure) that with no airflow and R7 at 1.38V (max voltage i consider safe for long-term use) the temps would be the same, maybe the Strix would run cooler because the heatsink is more heatsink than the slab of aluminium on Carbon. I understand that Carbon has double the highside but PK616BA is terrible part. I have talked with AlphaC from overclock.net and there is no heat difference Carbon/Tomahawk, because with the highside, in case of Tomahawk, youre pulling 25A per highside, this means 2.12W of purely switching losses and approximately 0.2W of conducting losses (gonna count again, not sure). With Carbon, youre are pulling 12.5A per highside fet. This means 1.06W of switching loss and some minor conducting losses. But you have 8 mosfets instead of 4. Youre putting less load per one mosfet but the heat output is still the same and it does not really makes sense having 4 more highside when you have such abortion of a heatsink on the Carbons

[KarathKasun]

ALL B350 (and most X370) VRM heatsinks are utterly useless, the Asus one is no better.

 

As to the doubled high side FETs, you also get a larger power plane with them.  Shouldn't change much, but realistically it will change the thermal output of the design.  Also AFAIK the Tomahawk only has 4x low side FETs, so the Krait/Pro Carbon are physically doubled all around.

 

Regardless, surface area will make crap thermal pads less of an issue and will give more surface area on the board itself for heat dissipation through the power plane.

[just_dave]

6 minutes ago, KarathKasun said:

ALL B350 (and most X370) VRM heatsinks are utterly useless, the Asus one is no better.

 

As to the doubled high side FETs, you also get a larger power plane with them.  Shouldn't change much, but realistically it will change the thermal output of the design.  Also AFAIK the Tomahawk only has 4x low side FETs, so the Krait/Pro Carbon are physically doubled all around.

 

Regardless, surface area will make crap thermal pads less of an issue and will give more surface area on the board itself for heat dissipation through the power plane.

Nono, Tomahawk-class trash has 2* lowside.

The strix has 4 big "fins" while MSI has only block of aluminium.

Lets agree on that MSI compensates the crappy PK616BA with doubling and gets aporoximately to the level of Asus with heat output and efficiency while having a bit better power capabilities. I am gonna redo my calculations (for tenth time) and report back.

[just_dave]

@KarathKasun

check here.

Asus is 17-18W

MSI B350 in reality around 20W (another minor losses that i dont count)