Rambling about my frustrations with DrMOS/Power stage nominal current ratings. VERY ROUGH VERSION
HTML-код
- Опубликовано: 3 июл 2024
- My Patreon: / buildzoid
Teespring: teespring.com/stores/actually...
Bandcamp: machineforscreams.bandcamp.co...
Twitch: / buildzoid
Twitter: / buildzoid1
Facebook: / actuallyhardcoreovercl... Наука
15:50 "you never see them (TDA21490) on motherboards" - False. MSI B550 Unify / Unify-X uses TDA21490 for both vCORE and vSOC.
Well this is embarassing you'd think I'd remember that since I have both of those boards.
bz rant video, fuck yes
oh yes ... rant:D why is not marked with !rant or so xDD fk that means i didnt find all rants in the video pool yet:D fk
peak UK election content
i love your Rambling vids
Why not build a power stage testing PCB? I think getting a controller to run a single or dual phase "epower" would be interesting this way you have control over the thermal characteristics of the board itself, Get one of the overseas mfg guys to build it
Sure but how does it fix overheating power stages when soldering
@@night_light2867make the board small enough to fit on a cheap heatplate like the miniware thing and use a stencil with leaded solder, there's no getting around putting heat on the thing but if the heat comes from the board then the solder then the powerstage its less likely to mess with it
Warning: thermal rant. Never quite clicked just how baked these people were when they assigned these current ratings. The chips can barely survive them without extreme measures. Take the MP86998: Extrapolating, it should be burning about 20W at 80A. The thermal resistance data illustrates the engineer's nightmare. 12.4°C/W to the package top, but only 2.7°C/W to the bottom. Tjmax = 125°C, so I'll say Ta = 25°C, and no sub-ambient cooling. The physical limit for heat dissipation out the top is ~8W, so the majority of your cooling must occur through the PCB. A bunch of well-placed vias can get the thermal resistance to the bottom of the board down to about 6°C/W.
So if you can pull 4W out the top, dissipate 4W into the surrounding board horizontally, and pull another 12W out the bottom of the board, you can hold the damn thing's junction temp somewhere around 100°C, depending massively on your heatsinks, TIM, airflow, etc. I don't even want to think about the 100A+ parts...
Having the exact same mosfet rated for different current DOES make some sense, technically, if it is in a different package, with different thermal characteristics. For example if it has a metal top vs a variant with a plastic top, the metal one would be easier to cool, but it might introduce other problems like EMI. or potentially not being able to use a common heatsink for all of them since it would make unwanted electrical connections.
From my understanding that is usually the justification from manufacturers for these ratings in general. They are giving you the absolute maximum rating, which should never really be used, instead of something that actually makes sense to use.
We need 80 plus bronze/silver/gold rating on motherboards.
And what would that do? 80 plus rating isn’t exactly good with PSU’s either.
Not watched but I'm guessing it's about the nominal current and Ptot being specified for Tj 25°C, which is of course rather unrealistic, but only intended as a common design reference point.
no it's actually that a 70A power stage from 1 manufacturer can run significantly cooler than a 110A power stage from another. And that in some cases parts from the same manufacturer with different nominal current ratings have exactly the same efficiency(at least according to the datasheets).
@@ActuallyHardcoreOverclocking the main issue is that the big current rating is just a marketing number that has very little say on what the performance are in practice
I've found that most mosfet manufacturers list the mosfets by their 25°C case temperature maximum current. It's not realistic unless you have a refrigerated water cooling system. I wish mosfets were marketed by their 100°C max continuous current rating as that number is applicable in an air cooled system
Well to be fair you can heatsink them and with airflow they can be kept at a reasonably low case temperature.
A good data sheet will have figures for Tc=25 and Ta=25 (no heatsink, just PCB to draw away heat).
The data sheet should also have derating factors and maybe even a graph of max current vs case temperature so you can pretty easily figure out the limit for your particular situation.
I wonder if one day we will have motherboards where the CPU heatsink also contacts the VRM Mosfets...
@@nicholasvinen Yes, a good datasheet will have all of the information needed to estimate performance. My point was that without those extra details, the 60A or 80A number is meaningless. For instance, the 80A mosfet may have a worse junction-case thermal resistance that causes worse performance compared to the 60A mosfet with the same cooling system.
Hey BZ, your video was sure packed with so much information! You were trying to justify why the nominal ratings were being used instead of a more reasonable one. It's probably because they want to make it sound better than the ones their competitor makes. Let us know more about it in another YT video.😅😊
So we need a standard for it? 😅
Was interesting! I have a Gigabyte B650 Aorus Elite AX which also has the same VRM minus 2 stages as the X670 I believe. No problems since day 1 here
Doesn't asus have a deeper llc setting? If the cpu input voltage is not calibrated, asus boards provide a lot more voltage, which can explain it being hotter
the CPU was getting the exact same die sense voltage on both boards.
Would like to see the hynix ddr5 basic oc revisited, since that would cut some time off of my new build overclocking
babe wake up, new rambling just dropped
One common DrMos for back on the 115x boards was the CSD87350 from TI
That's a dual nfet not a DrMOS
@@ActuallyHardcoreOverclocking I seem to remember them having an integrated driver but just no monitoring circuitry, but I guess I am confusing it with something else from back then...
would be useful to have nominal current rating to say at which amperage it reaches peak efficiency, or closest 5/10 multi.
but big number sells ig
yeah that would be somewhat more useful
Sir I have TeamGroup T-Force Delta RGB 32GB (16GBx2) DDR5 6000MHz cl 38 ram and 7600x processes could you please recommend me some super tight stable 6000 mhz cl30 ram timing sir .I am currently using your timings from your video sir
First!
First