Why the ASUS Equalizer is actually good and current imbalance is not what really matters
Posted by IntoxicatedHippo@reddit | hardware | View on Reddit | 23 comments
In the videos and posts examining the ASUS Equalizer cable that I've seen recently, people have made a fundamental mistake by considering only the current through the wires/pins instead of the thermal dissipation. We care more about thermal dissipation than current because dissipation is what causes connectors to melt. Looking at the ASUS cable, it's obvious that it reduces the risk of a fire, as I will explain below.
Let's take a look at some realistic numbers for a bad scenario that could result in a fire with a regular cable. At both ends of the connector let's assume we've done a really bad job plugging it in and mangled 3 ground and 3 power pins on one side of the connector but we have a good connection on the other side, say 15 mohm and 1 mohm for the bad and good pins respectively. On each end of the connector we've mangled opposite sides so each bad pin connects to a good pin on the opposite end of the cable. It's realistic that half your contacts are significantly worse than the others if you consider people shoving a panel against a cable and folding it over. We also just have a pretty normal 45cm cable for 6 mohm resistance in each wire. Something very important to note here is that every pin-wire-pin connection will be identical.
Now let's analyse the dissipation, it's quite simple because every pin-wire-pin combination is 1mΩ-6mΩ-15mΩ. At a total 50A load, this gives us 0.07W-0.4W-1W. Put 6 of those on the end of the cable and we've got 6W in the half of connector (and 6.4W total), which is probably enough to start melting it, especially given that the GPU will be hot from drawing 50A so we can't dissipate much into the PCB. Or maybe it's not and we instead need 10mΩ and 100mΩ instead of 1mΩ and 15mΩ, but all the same logic and results still apply for those numbers.
Note again that our current balance would be perfect here but our dissipation is not.
Now let's consider what happens when we add ASUS's bridge. We effectively split our network into two series networks, so let's look at each side separately:
The first has the wire and the PSU side connector, resulting in six lots of 15mΩ-6mΩ and 6 lots of 1mΩ-6mΩ. I'm not going to write out all the maths here, but that gives us 0.26W-0.1W and 0.15W-1W, so not nearly as much dissipation in the connector (2.5 W total and not concentrated too much on one side) and 1W in each wire for half the wires, over a 45cm wire that's nothing.
Note here that the current balance is now really bad, but the resulting dissipation is lower, both in total and in the highest dissipation pins.
The second resistor network consists of the GPU connector pins in parallel: 3 lots of 1mΩ and 3 lots of 15mΩ, or 0.24W and 0.01W. This gives us only 1.5W in the GPU connector. It is all concentrated on one side, but it's 4 times less than the 6 W we have on one side without the bridge.
You might think, "But wait, that's 15A through some of the pins, isn't that really bad?". The answer is no because the low resistance results in very little thermal dissipation. If the resistance increases, less current flows through that pin because the current distributes to the other pins, keeping thermal dissipation low.
As you can see, the bridge is a very good solution here despite worsening the current balance. Current balance is ultimately just a symptom, one that is easily measured, but it is not what we actually care about for preventing a fire.
kwirky88@reddit
Can’t this all just be avoided using the previous generation power connectors? Can anybody explain?
__some__guy@reddit
It would be too embarrassing for Nvidia to admit their 12VHPWR is trash, so they just ignore the issue.
der8auer@reddit
It's a very interesting theory. I want to make clear that when the uneven current theory is not proven to be (the only) cause of melting connectors. What I can say is that our first WireView Pro also had such a bridge like the Equalizer and we have about 30-40 cases with melted connectors However, with the new WireView Pro II we have no case of a melted connector yet. I guess time will tell 😁
cp5184@reddit
I haven't done a proper analysis and I probably won't because I'm not planning on using this type of cable...
But isn't a big contributing factor of this that there's a single power plane on the GPU side?
So there's already a bridge on the GPU side...
And that bridge was one of the problems...
What this does... Is... if... say... there's a 3090 or something that still has multiple power planes on the GPU sides, creates with this cable bridge the problem the single power plane on the GPU creates...
So say a 3090 has 3 or 4 power planes, the gpu vrm power stages on each individual power plane can detect the voltage, infer high cable resistance, the vrm can detect a low voltage power plane and lower the draw from it, making this hypothetical 3090 with multiple power planes safer and less fire prone...
If you put an asus equalizer on this hypothetical 3090 aren't you introducing the single power plane problem to a card that didn't have it already?
Fatigue-Error@reddit
Hey. Love your videos!
A question for you. Would it be good for a 4070ti Super to get a Wireview Pro II as well? Or is the current draw low enough that melting isn’t a big concern?
IntoxicatedHippo@reddit (OP)
I wonder how much of that comes down to having a thermistor in there now, it would be interesting to see what triggers most often if those are statistics you collect.
der8auer@reddit
We already had thermistors in the previous WireView Pro and it was often not triggered :(
In general we think the imbalance can be the highest risk. P=I\^2 * R. So for example a cable with 15A is more relevant than the resistance for the power generated.
Canadian_Border_Czar@reddit
Ill help test the theory, just send me a WireView Pro II cough and a 5090 cough and I'll get right on that.
Sorry it appears something is stuck in my throat. Excuse me.
KARMAAACS@reddit
Not to mention don't the NVIDIA adapters also bridge the pins? Haven't those melted too?
The bridge seems like an irrelevant 'feature' really, if the pins can't take 17A, but the wire can take 17A, the pins will melt before the wire will, correct?
Jaislight@reddit
You work for Asus huh?
zopiac@reddit
Resistor networks aren't really my area of expertise making this hard for my brain to digest, so I figured I'd play with the circuitJS example a bit out of curiosity by swapping out the "perfect" 0Ω links on the bridge with 1mΩ, 0.1mΩ, and 0.01mΩ links just to see if there's any problem there (no, even if the contact resistance of each wire's connection within the bridge itself is so high as 1mΩ). I got the following dissipations for each combined link in the chain (PSU side, bridge, and GPU side):
|PSU|Bridge|GPU|Total -|-|-|-|- No bridge |3.333W|---|3.333W|6.666W Perfect bridge |1.248W|---|0.780W|2.028W 1mΩ bridge |1.777W|0.792W|1.155W|3.724W 0.1mΩ bridge |1.324W|0.209W|0.795W|2.328W 0.01mΩ bridge|1.255W|0.024W|0.782W|2.061W
It's interesting to see where the heat comes from in each example though, as adding even 0.01mΩ resistance shifts the load around slightly. But no matter, overall dissipation drops significantly and even the highest single-point dissipation halves with the terrible 1mΩ bridge.
Deshke@reddit
... yes.. but no? the problem being the resistance between the wire and the leaf connector and the connector on the consumer. The tolerantes in the spec are to tight to account for sh** material quality and poor contact. 9A per pin is on the upper end of the cable spec, yes a heatsink may allow you to push 15A for a short amount of time, but that only works until the heatsink is saturated and then you still end up with melted plastics
Rippthrough@reddit
Those ratings on connectors and pins are primarily heat and heat dissipation related anyway. It's why you can often get pins for the same footprint that drop in and are rated for 13 or 16 amps in the same connector that normally only comes with pins rated for 9a or similar.
You get pins that are single or 2 piece phosphor bronze instead of brass, that allow higher contact pressures for better heat dissipation and lower contact resistance.
It's the same for cables - the same pin crimped to a 12g cable has very different current ratings to that pin crimped to a 22g cable - because the copper cable provides a massive cooling path for the contacts.
Proglamer@reddit
People discussing 3rd party solutions to prevent their $4000 hardware from melting down. Riiiight 👍
pmjm@reddit
Found Tim Asus.
For real though, thanks for the well thought out post. It's nice to see a contrarian opinion that's fueled by logic and not emotions on /r/hardware.
Joezev98@reddit
Shouldn't your simulation take into account the contact resistance from the wire, to the bridge, to another wire? I wonder how much of an impact that makes.
buildzoid@reddit
One issue. The Nvidia 4x 8pin to 12VHPWR adapter bridges ALL the pins together right at the connector and still melts.
IntoxicatedHippo@reddit (OP)
You can of course still have a connection that's poor enough to melt even with the minimum possible dissipation, but bridging the pins will fix at least some cases.
nittanyofthings@reddit
I don't like your setup at all. Make it simple: set all pins to 1m. Have one bad gpu side pin at 50m. And one unusually good gpu side pin at 0.75m. Before the bar the good contact pin is 9.9A 75mW and after the bar it is 12.5A 116mW. An increase of 50% thermal. And the bad pin went from 9mW to 2mW, 75% decrease.
The manufacturers rate their pins for a current, not a dissipation.
IntoxicatedHippo@reddit (OP)
Dissipation is where the current rating comes from, or more precisely it comes from the temperature rise limit for the given connector due to thermal dissipation. If we have one particularly low resistance pin, then yes, the current and dissipation of that pin will be higher, but realistically not so much that it will melt the connector. The resistance will also go up a bit with temperature which will help balance it out. Note that in the scenario I posted the resistance has to go up a massive amount for the dissipation to start going back down.
egnegn1@reddit
There is a thread discussing this. Why not discussing there?
EloquentPinguin@reddit
It actually maths out.
Thanks.
I didn't notice the mistake in the methodology when watching the Der8auer video, but it can actually be true that in the standard cables the "worst pins" receive similar current while in the Asus bridged cable the difference between best to worst pin is measured but the interpretation would need to be the other way around.
Sparkycivic@reddit
This guy starts house fires