Question How Important does VRM Amps Really Matter [50A 60A 105A]

[tomyhardz]

Power demanding processor need higher VRM in order to work efficiently,
After digging deep found there Amp rating for those VRM
so how the major role these Amps play... is it only for overclocking.
Is there any minimum/required for Amps 13 &14th Gen 600k,700k,900k series processors VRM....???

 

[drea.drechsler]

Power demanding processor need higher VRM in order to work efficiently,
After digging deep found there Amp rating for those VRM
so how the major role these Amps play... is it only for overclocking.
Is there any minimum/required for Amps 13 &14th Gen 600k,700k,900k series processors VRM....???

The current ratings of the VRM FET's or DrMOS power stages are critical but that is only part of the story for total power handling of the VRM. Equally important is number of phases and heatsinking on the devices. Also a factor is air-flow through the case so that the heat sinking can be effective; that is something a system builder can directly affect.

Very important, but much less obvious, is voltage stability for high current transient events. That's a function of the VRM controller and turn on/turn off characteristics of the FET's along with basic design topology of the VRM including filtering and even mother board design.

Power handling is important when overclocking most any CPU but many high core-count CPU's also need extremely capable VRM's simply to operate effectively at rated clocks, without long term damage to the VRM and motherboard.

 

[tomyhardz]

The current ratings of the VRM FET's or DrMOS power stages are critical but that is only part of the story for total power handling of the VRM. Equally important is number of phases and heatsinking on the devices. Also a factor is air-flow through the case so that the heat sinking can be effective; that is something a system builder can directly affect.

Very important, but much less obvious, is voltage stability for high current transient events. That's a function of the VRM controller and turn on/turn off characteristics of the FET's along with basic design topology of the VRM including filtering and even mother board design.

Power handling is important when overclocking most any CPU but many high core-count CPU's also need extremely capable VRM's simply to operate effectively at rated clocks, without long term damage to the VRM and motherboard.

Nice insight...
What about those Amps Ratings...you know any required Amp for those mentioned processor running on STOCK..or is there a way to calculate ourselves??

 

[drea.drechsler]

Nice insight...
What about those Amps Ratings...you know any required Amp for those mentioned processor running on STOCK..or is there a way to calculate ourselves??

Not really. The problem is pretty complicated: you can add up the totals of the current ratings of the FET's: like 6, 50 amp rated FETS in the VRM means a 300 amp total rating, then compare that to the total current draw of the CPU, assuming you know it.

But it's not nearly that simple. The first fallacy with that approach is you don't really know the current draw of CPU's, especially when overclocked.

The second fallacy is current rating of FET's are given assuming pretty specific test conditions: like when mounted on a certain size and construction of PCB material, with a certain size heatsink and at certain ambient temperature conditions and pulsed at a certain duty cycle. It's rare that VRM and motherboard designs and case cooling is well enough constructed to meet those conditions. This is just one reason why VRM's have FET counts that are (usually) way more than needed when you simply add up the total current handling capability of each FET.

Another fallacy with the approach is it's not really total current that limits a typical VRM' s ability to overclock: it's voltage stability. That's a function of the other variables that aren't very obvious. Since any overclock means trying to run with the lowest voltage possible to not burn up the CPU, stability is critical for usefulness.

The best approach to pick a motherboard is look at what CPU's the mfr's specs it for, and accept that it only means when run at STOCK clocks. Then build a system with great airflow across the VRM so that it runs cool and doesn't engage it's protection modes, which most all modern VRM controllers employ. Or read motherboard reviews that test with various CPU's to assess it's potential for overclocking.

 

[tomyhardz]

Not really. The problem is pretty complicated: you can add up the totals of the current ratings of the FET's: like 6, 50 amp rated FETS in the VRM means a 300 amp total rating, then compare that to the total current draw of the CPU, assuming you know it.

But it's not nearly that simple. The first fallacy with that approach is you don't really know the current draw of CPU's, especially when overclocked.

The second fallacy is current rating of FET's are given assuming pretty specific test conditions: like when mounted on a certain size and construction of PCB material, with a certain size heatsink and at certain ambient temperature conditions and pulsed at a certain duty cycle. It's rare that VRM and motherboard designs and case cooling is well enough constructed to meet those conditions. This is just one reason why VRM's have FET counts that are (usually) way more than needed when you simply add up the total current handling capability of each FET.

Another fallacy with the approach is it's not really total current that limits a typical VRM' s ability to overclock: it's voltage stability. That's a function of the other variables that aren't very obvious. Since any overclock means trying to run with the lowest voltage possible to not burn up the CPU, stability is critical for usefulness.

The best approach to pick a motherboard is look at what CPU's the mfr's specs it for, and accept that it only means when run at STOCK clocks. Then build a system with great airflow across the VRM so that it runs cool and doesn't engage it's protection modes, which most all modern VRM controllers employ. Or read motherboard reviews that test with various CPU's to assess it's potential for overclocking.

thanks for the elobrate explantion

then how would some one says you atleast need 10+ or 12+ VRM phase for certain processor to operate efficently...by running benchmark or what??...also nobody mention amps for certain processor to run efficiently..why is that so???

 

[drea.drechsler]

....
then how would some one says you atleast need 10+ or 12+ VRM phase for certain processor to operate efficently...by running benchmark or what??...also nobody mention amps for certain processor to run efficiently..why is that so???

In general, higher the phase count the better the power handling and voltage stability since the duty cycle for each stage is less at any given processing load. That's going to mean a cooler running VRM to allow lowest voltage operation. The hotter any device runs the worse its performance becomes for processing transients.

But equally important to phase count is topology: is it a discrete FET VRM? or DrMOS power stages? DrMOS is inherently far more efficient than discrete FET's because the driver and both the hi- and lo- side FET's are closely coupled and matched in one device. Many power stages (usually the higher current rated ones) have on-chip temperature monitoring, fed back to a controller that can dynamically control individual phase duty cycles to keep them in an efficient and safe operating range. You get such gee-wizardry with more expensive, high phase count motherboards.

One reason so few people might mention amps for efficient operation is so few have the ability to accurately measure it. It's usually something that's inferred by use of programs like HWInfo64 monitoring CPU and VRM telemetry which is frequently made inaccurate by improper BIOS calibrations from the manufacturer. Accurate measurement means installing a current loop of some kind in the motherboard. Some extreme high-end motherboards either have one pre-installed or are made to easily install one. Those are the boards to get for extreme sub-ambient overclocking.