10X IR3553 is pretty much going to handle 180A much like a 6x IR3553 solution handles ~110A

The IR3553 starts to become less efficient (<90%) at ~ 25A although ideally kept below 20A each if you want to only have to dissipate ~2W each. Gigabyte has increased the thermal conductivity massively versus x370 K7 such that it facilitates the use of IR3553 (40A powerstages) instead of IR3555 (60A). If you look at boards where the mosfets are separate instead of combined in a powerstage , 2W is more or less the limit you can dissipate on them without going into the > 100°C zone provided your total thermal resistance at the VRM heatsink is kept around 35°C/W §. 150nH Cooper Bussman chokes labeled FP1007R3 are used http://www1.cooperbussmann.com/pdf/7...9b2be9f2e4.pdf

(IR3553 data: https://www.infineon.com/dgdl/ir3553... ... e1767)

§ For example, the Onsemi parts being used on some boards list ~50°C/W thermal resistance when they are not heatsinked. Absolute maximum power dissipation (for the Onsemi mosfets) figures at 25°C are cited to be around 2W as well when given a 1 inch square pad while only ~ 0.8W when using minimum pad size.

The main advantages to a IR3555 used on the CH VII Hero are that it has a larger footprint which translates to more thermal conductivity along with a wider efficiency band > 90% (it is up to around 45A). Gigabyte's approach to the problem was different: they used a heatsink with massively more fin area and a direct contact heatpipe to utilize the full area of the heatsink more effectively. R40 chokes are used on the CH VII Hero, which suggest steady state performance was preferred over transients (chokes / inductors resist changes in current). Per IR3553/IR3550 datasheets for power loss vs output inductance, anything over 150nH has less than 5% normalization difference (i.e. if you lose 2W per powerstage with a R15 inductor you lose ~1.9W with a R40 one).

Where the TI NexFET powerblocks on the Taichi stand: the driver losses aren't accounted for on their solution , however 12X NexFETs have to deal with less current than 10X IR3553/IR3555 powerstages. The package size is 5x6mm versus the 4x6mm of the IR3553 and 6x6mm of the IR3555 ; thermally it is similar to the IR3553. While the efficiency curve on a CSD87350 NexFET suggests over 90% efficiency at 40A , it isn't a full picture of the situation. If you look at power loss at 15A without accounting for an increase in voltage (which would be around +3% losses at 1.4V , so if you lose 1W it would be more like 1.03W at 1.4V) it's already at 2W of power loss each. Their "typical" design point is 25A (3W power dissipation). A heatpipe is utilized at the board to increase effective surface area as on the CH VII Hero and GBT Gaming 7 , which allows it to compete. The Taichi's R22 chokes are suspected to be the 0.22µH Coiltronics ones http://www.cooperindustries.com/cont...427-fp110v.pdf
(see http://www.ti.com/lit/ds/symlink/csd87350q5d.pdf)

If you give your board some good airflow even a 6x IR3553 board can handle ~ 4.2GHz on the R7 2700X manually overclocked. If it wasn't possible then the Prime X470 Pro wouldn't be using 6 x IR3553. As you recall , the TDC (thermal design current) parameter for PBO (Precision Boost Overdrive) is 114A per Stilt's overview despite the EDC (electrical design current) of 168A. That's ~19A per IR3553 under sustained load. It won't run as cool but it will run , just don't expect to be doing any prolonged AVX instruction-based work on it without airflow.

I'm not going to belabor the MSI X470 M7 topic again because some people are so sensitive about it but if you look at the board it has no heatpipe at the VRM heatsink and 12 sets of discrete mosfets are used. The transient performance is going to be lower since the discrete mosfets' switching times are slower. In terms of power loss it will be close without accounting for driver losses ; efficiency suffers at low load if phase shedding isn't active. As output capacitance is similar to the Asus/GBT boards and R22 chokes are used I would expect steady state performance to come in around the Taichi's at best (the Taichi has more output capacitance and uses R22 chokes).

(see https://www.onsemi.com/pub/Collateral/NTMFS4C029N-D.PDF , https://www.onsemi.com/pub/Collateral/N ... D.PDF)


I don't think you can really go wrong with any of the three boards (CH VII Hero , Gaming 7 WIFI , Taichi) you mentioned as they are known quantities at this point (especially the Taichi which is a X370 rehash). They each have their own shortcomings and idiosyncrasies but that's a different topic.