Originally Posted by dansi
Regarding the hardware info test, to be fair, they are using the 12 core part. maybe asus design will play better with the 16 cores.
Maybe for the lower boards vs boards with doublers (due to the doubler delay *), but most definitely not their flagship. The math definitely doesn't check out unless you're pumping water through the Formula.
Assume you have 3 boards.
One (Xtreme) is driven by a state of the art (for computing) 16 phase direct PWM with fourteen 70A TDA21472 powerstages (peak efficiency ~95%) for VCORE and has a finned heatsink , thermal backplate, and 95% board coverage with metal. On a 8 layer PCB with 2oz copper. Keep in mind having actual 14 phases with current monitoring means you can achieve higher efficiency across the whole powerband (across more current values).
The other (Godlike) has 14 phases doubled from 7 to fourteen 70 TDA21472 powerstages (peak efficiency ~95%) for VCORE, 8 layer PCB, and minimal fin area.
The ASUS one has 7 phase "teamed" to 14 60A IR3555 powerstages (peak efficiency ~93.5%) for VCORE , minimal fin area compensated by using a backplate, and likely 8 layer PCB.
Guess what? That's exactly what the hardware info NL test shows: Xtreme ahead of Godlike which is ahead of ROG Formula/Hero.
The only way that the direct 14 phases is worse than 14 phases from teaming 7 is if the voltage being pumped in is absurd for the direct 14 phase while being a moderate 1.3V for the teamed design or if they're running at an operating point where the efficiency is lower for TDA21472 than IR3555. That still
doesn't account for the fin area either and if you're not running water through the Formula the air inside the VRM waterblock acts as an insulator.
Until we have 16 cores in September I'm more inclined to believe it is ASUS manipulating BIOS settings than the ROG boards being superior to the Godlike and Xtreme. If you believe ASUS then they're preying on people's non-engineering gullibility.
* IR3599 ~8ns rise/fall transition ; propagation delay ~9-10ns https://www.infineon.com/dgdl/ir3599...5355cdb56a176f
* ISL6617 ~4.5ns rise/fall transition ; propagation delay ~35ns https://www.renesas.com/us/en/www/do...et/isl6617.pdf
P.S. in videocardz' comments section (older b450 comparison):
this to be distinctly false and physically impossible.
Tomahawk has better mosfets (4c024N+4c029N instead of 4c09n +4C06n) and 2x the high side fets along with a bigger extended heatsink