I asked myself: What is ASUS up to? What's going on in their minds introducing shunt resistors only on the Astral? Why did they implement them in the first place? This isn’t an easy question to answer, I fear.
Here is my analysis:
1. Shunt resistors - Why in the first place?
The shunt resistors (of course) don’t solve any of the underlying problems, not stopping anything from overheating (without further features). They simply provide a software readout on the pins - probably not even error-corrected, just factory-calibrated. So far, they serve no real value in protecting the hardware. Let's look further.
2. The readout and load handling
From the picture posted by Roman (showing 9.7 amps on one of the pins of the Astral), we see that the software did nothing to stop the GPU at that load. That's 0.2 amps over spec—not a big deal in practice. Let’s assume a) error correction isn’t needed because reading errors at this load level are negligible (I’m not an electrician, so I could be wrong). b) The additional heat from feeding 0.2-0.5 amps per pin should also be negligible on the connector, pin surroundings, cable materials, sleeve texture etc. c) most 'normal' readouts can produce 9.6 or 9.7A on a single pin from time to time. If this slight overload is the new normal, we’re dealing with a very fragile standard. Because it's over the spec. Is the spec on point or is it defensive (like most specs their is a unofficial headroom over the official headroom - but this is an unknown territory with many variables like ambient temperature, age of cables, no. of seatings etc. etc.).
But here’s the real question: at what point will the ASUS software flag an issue? Can we assume that 60 amps total (or 10 amps per pin fully balanced) will be acceptable to Asus? Being over the spec seems to be fine for them, but how far over spec are they willing to go? How far is not ok? We have no information about this.
3. Why no automatic shutdown?
From what I’ve read, Asus detects a load issue on one or more pins but doesn’t shut the GPU down. Why, in the name of all that is holy would you do that?
Asus! Would you really leave this in the hands of the user? What if they’re away from keyboard when the message pops up? Why won’t the Software shut the GPU down immediately when it detects a dangerous load? They could - the Astral has direct readouts on the pins, or immediately downclock it. What is the mission statement for the shunts?
4. An msgbox as a failsafe?
So, instead of a proper failsafe, Asus simply displays a message box to the user. That feels wildly inadequate. What if the message goes unseen until it's too late? Even if the user isn't AFK, does it overlay properly ingame? Will Asus provide more details on how this alert system actually works?
5. Why only the Astral?
I know shunt resistors add some cost, but we're talking peanuts in the grand scheme of things. So why only the fastest card?
- Do lower-tier cards not suffer from the same problem (assuming there is one serving the reason for implementing the shunts)?
- Are they too cheap to include this feature? Is the feature too expensive?
- Are the amp loads on the pins just an interesting KPI rather than a critical safety feature?
Did Asus think that adding this only to their flagship card would make it seem like just another high-end gimmick, avoiding any questions about the real reason for their inclusion? Because, let’s be honest, these resistors didn’t just appear out of nowhere.
My Take:
Asus Knows (yeah, I know it sounds like a conspiracy theory
)
Asus 100% saw the Nvidia connector issue at high loads during lab testing. That's why we have a readout on the Astral now. Let’s do the math:
- At 50 amps, each of the six pins should handle 8.33A (50/6).
- At 55 amps, it’s slightly over 9.1 per pin.
- 9.5A is the absolute max when you take Nvidia seriously
- Something over 9.5A is the absolute max, if you take Asus seriously
I bet Asus regularly saw single pins exceeding 9.5 amps, maybe hitting 10 amps on the edge pins (5 and 6) in some scenarios and didn't know what to do. In theory the GPU is defect, if you take the spec sheet and look up what's allowed and what's not. Maybe they asked Nvidia. Maybe Nvidia told them to ignore it. Maybe they didn’t ask at all. Who knows?
Instead, Asus quietly implemented a soft solution that won't be too in your face: Shunt resistors on the GPU (in a bridged circuit). A software dashboard. A message box alert. No shutdown. No hard limits. No transparency. Now, even Astral 5090 owners are asking Asus for more details on this in the Asus forum. I bet they won't get an answer so soon.
But anyway. WTF, Asus? Did you think you could slip this hint to us without upsetting Nvidia? Drop a hint about the problem, then pretend everything’s fine?
Asus, blink twice if Nvidia is breathing down your neck.
Sorry if I am being too hard on Asus. It's just a means to an end. The real Problem is of course Nvidia and their 12VHPWR standard.
The more I read, the more it’s clear to me - the issue lies with Nvidia and their fully-loaded (no headroom) 12VHPWR standard. Any small problem in the power chain PSU ⟷ Cable(s) ⟷ GPU like foreign debris, slight cable defects, contact issues, issues from multiple reseatings or even simple material aging and suddenly, you're in a high risk territory. Thank you for that, just what my house needs.
IMHO we need a complete redesign of 12VHPWR - or an entirely new standard. Nvidia is acutally ghosting the community, as usual. Maybe they wake up next week. But you can bet they’re fully aware of the problem by now. I hope they will come up with a statement at least. Personally, I don’t want an update to the connector in the next gen or more efficient GPUs with smaller power draws (albeit, this is a good thing to focus on independently). I want something beefier than 12VHPWR, thicker, with more headroom. I don’t care if that means buying a new PSU or pricier cables. This toy connector should have been a proprietary solution for SFF enthusiasts and mid tier GPUs in the first place - not the de facto flagship connector for 575W/600W GPUs.
And before anyone brings up the Galax HOF 4090 again - No, I don’t want to connect 2 or 3 of those toyish connectors on my GPU and multiple 12V-2x6 cables in my rig. That’s a cheapo band-aid, not a fix.
