ASUS Showcases PSUs with Unique VGA Voltage Stabilizer and GaN MOSFETs

[btarunr]

ASUS at the 2024 Computex showed off the latest revisions of its ROG Strix Platinum and ROG Thor Titanium III power supplies that pack a slew of cutting edge innovations for high-end gaming PCs. In their latest revisions, these PSUs meet the ATX 3.1 specification, and offer updated 12V-2x6 (H++) graphics card power connectors, but there's a key bit of innovation here.

Notice the purple 2-pin connector on the PSU? It is a set of sense pins that lead to the 12V-2x6 connectors, and lets the PSU get active feedback from the graphics card (rather than the motherboard) on how the voltage stability is, so it can actively hold the +12 V domain within a tight tolerance band. This mechanism lets the PSU observe voltage drops and spikes beyond the excursion tolerances of the ATX 3.1 spec. ASUS holds the patent for this innovation. Another major innovation with both these PSU lines is the use of modern GaN MOSFETs for improved switching efficiency, and compact dimensions, which vastly improves the layout of the PSU's mainboard, and serves up more room for heatsinks or airflow. ASUS is backing both these PSU lines with 10-year warranties.



View at TechPowerUp Main Site

 

[maxfly]

Hmm, maybe Asus is onto something here. If it actually works that is and if people are willing to overlook their woeful CS shortcomings.

 

[Dr. Dro]

Hmm, maybe Asus is onto something here. If it actually works that is and if people are willing to overlook their woeful CS shortcomings.

Their R&D and hardware divisions were always top notch. It's really customer service that leaves much to be desired especially as ASUS, as a premium hardware company, steps into the vacuum left by EVGA (eternal CS king) in a somewhat natural manner.

 

[Ferrum Master]

You need compensation line for a standard, closed system, with standard cables and lengths, temperature ranges... You can test it till you puke on your limited load line scenarios.

People... That thing is an utter bullshit that is not needed and solves nothing. i2s based, slow, compromisable, prone to outside factors. There is a thing called tolerances. I would not argue, it would make sense on 3.3V rail, that nothing uses these days, but 12V? Give me a break and keep the innovation outside the PSU devices. Medical and space does not need such things, but here hey, let's put something for the PR drum. Actually internally that has been there for ages, while you debug and calibrate the platform. It is a bandaid for poor performance design.

They would improve on EMI patterns instead of this.

 

[Dr. Dro]

You need compensation line for a standard, closed system, with standard cables and lengths, temperature ranges... You can test it till you puke on your limited load line scenarios.

People... That thing is an utter bullshit that is not needed and solves nothing. i2s based, slow, compromisable, prone to outside factors. There is a thing called tolerances. I would not argue, it would make sense on 3.3V rail, that nothing uses these days, but 12V? Give me a break and keep the innovation outside the PSU devices. Medical and space does not need such things, but here hey, let's put something for the PR drum. Actually internally that has been there for ages, while you debug and calibrate the platform. It is a bandaid for poor performance design.

They would improve on EMI patterns instead of this.

Valid, although I don't think the medical or aerospace industry will be using a ROG gaming PSU in any of their systems so there's that.

The monitoring feature can be novel but Corsair had been doing it with the AXi on a per cable basis since the 1200i >10 years ago.

 

[Ferrum Master]

Valid, although I don't think the medical or aerospace industry will be using a ROG gaming PSU in any of their systems so there's that.

The monitoring feature can be novel but Corsair had been doing it with the AXi on a per cable basis since the 1200i >10 years ago.

That's not monitoring. That's a feedback loop. I am all in for more sensors, independent sources of data, especially not on mainboard, not causing additional latency, like ASUS dreaded EC hub does, but adjusting that should not be needed to adjust in the first place, is beyond weird in this area of field. It is comparable as a heart stimulant if you have it weak or failing.

ROG in medical... well... it is a deathwish for sure. But the platform they use can be used thou. Delta, Seasonics etc do medical grade PSU's also, but contrary they are as dumb and bulletproof they can be. Not sure what the OEM is for this one really.

Less noise, more efficient, less EMI, less ripple, less size.... not more size, more loud, more cables, more problems.

 

[Vayra86]

That's not monitoring. That's a feedback loop. I am all in for more sensors, independent sources of data, especially not on mainboard, not causing additional latency, like ASUS dreaded EC hub does, but adjusting that should not be needed to adjust in the first place, is beyond weird in this area of field. It is comparable as a heart stimulant if you have it weak or failing.

ROG in medical... well... it is a deathwish for sure. But the platform they use can be used thou. Delta, Seasonics etc do medical grade PSU's also, but contrary they are as dumb and bulletproof they can be. Not sure what the OEM is for this one really.

Less noise, more efficient, less EMI, less ripple, less size.... not more size, more loud, more cables, more problems.

Less is more. This really does apply to anything PC. The more you add, the more problems you can have.

 

[ExcuseMeWtf]

But will they send it back from warranty disassembled?

 

[sLowEnd]

Mmm. I'll stay away from GaN products of this type for at least the first few iterations. USB power bricks are one thing, but high wattage PSUs are another.

 

[Dr. Dro]

Mmm. I'll stay away from GaN products of this type for at least the first few iterations. USB power bricks are one thing, but high wattage PSUs are another.

High-performance PSUs with GaN FETs are not new anymore, but due to cost, it's not widely employed. It's the secret behind the AX1600i's supreme performance. There's even an Aris review on TPU

Corsair AXi Series 1600 W Review

Simply put, the Corsair AX1600i is the best PSU money can buy today. Because of its huge capacity and price tag, it only addresses a small portion of users; those only satisfied with having the best of the best.

www.techpowerup.com

Unsurpassed since 2018, just like the AX1500i remained unsurpassed since its launch all those years ago and is still one of the top 5 performers even today.

 

[Wirko]

It is a bandaid for poor performance design.

Or thin wires ... or thin CCA wires.

I'm bothered by the fact that a patent can be granted for plain voltage sensing. That would have been an ingenious invention and advanced use of technology ... in 1924.

But why did you say I2S based? Isn't it analog?

 

[sLowEnd]

High-performance PSUs with GaN FETs are not new anymore, but due to cost, it's not widely employed. It's the secret behind the AX1600i's supreme performance. There's even an Aris review on TPU

Corsair AXi Series 1600 W Review

Simply put, the Corsair AX1600i is the best PSU money can buy today. Because of its huge capacity and price tag, it only addresses a small portion of users; those only satisfied with having the best of the best.

www.techpowerup.com

Unsurpassed since 2018, just like the AX1500i remained unsurpassed since its launch all those years ago and is still one of the top 5 performers even today.

The performance is not in question. Most products using GaN I've seen perform quite well. I'm quite happy with my GaN USB chargers. It's just that high wattage products are still very expensive (that Corsair unit you linked is hundreds of dollars) and they're (edit: GaN FET components) still improving at a pretty rapid rate.

 

[billEST]

Their R&D and hardware divisions were always top notch. It's really customer service that leaves much to be desired especially as ASUS, as a premium hardware company, steps into the vacuum left by EVGA (eternal CS king) in a somewhat natural manner.

asus is only R&D and hardware divisions , no factory

 

[maxfly]

Their R&D and hardware divisions were always top notch. It's really customer service that leaves much to be desired especially as ASUS, as a premium hardware company, steps into the vacuum left by EVGA (eternal CS king) in a somewhat natural manner.

Sure but when your rep is in the toilet that doesn't do them much good.

They competed on the same hardware field for many moons. Yes. Yet Asus has never approached EVGAs level of quality CS. In fact it seems they have done the exact opposite of filling any customer support/service or hardware voids. After the GN story people are a lot less likely to take a chance with Asus from what I'm seeing and hearing. There are always other premium options.

 

[Dr. Dro]

Sure but when your rep is in the toilet that doesn't do them much good.

They competed on the same hardware field for many moons. Yes. Yet Asus has never approached EVGAs level of quality CS. In fact it seems they have done the exact opposite of filling any customer support/service or hardware voids. After the GN story people are a lot less likely to take a chance with Asus from what I'm seeing and hearing. There are always other premium options.

Yeah, that's what I meant. ASUS CS has always left a lot to be desired vs. EVGA. Hopefully people give Aorus a fair chance too. The Gigabyte fellows have always made reliable hardware, at least I've had great luck with it.

asus is only R&D and hardware divisions , no factory

ASUS does have some great factory-level improvements, for example, that Auto-Extreme thing which is robotic assembly of the hardware, it's really something else. What was being questioned is their customer service quality, which is almost the extreme opposite of their hardware's quality. It's... bad, to say the least. You may never need it, but God forbid if you do.

The performance is not in question. Most products using GaN I've seen perform quite well. I'm quite happy with my GaN USB chargers. It's just that high wattage products are still very expensive (that Corsair unit you linked is hundreds of dollars) and they're (edit: GaN FET components) still improving at a pretty rapid rate.

Fair, but I don't think cost on this type of component will come down any time soon. I share Aris' sentiment on the last sentence of his review, I personally don't think it's overpriced considered how long it's meant to last, but alas.

 

[W1zzard]

i2s based

its not, it gives you an option to move the sense wire from the mobo 24 pin to the gpu

 

[Ferrum Master]

its not, it gives you an option to move the sense wire from the mobo 24 pin to the gpu

Please explain in details, what it is then? PS-ON? IVS btw is trademarked name, not sure what are they trying to achieve there then.

But why did you say I2S based? Isn't it analog?

Because there actually are programmable power supplies, with the ivs naming.

 

[W1zzard]

Please explain in details, what it is then?

Just stolen from Stackoverflow:

Cables have some resistance, and when you pass current through them, they have a voltage drop. This causes the load to "Receive" less voltage than what is actually set on the power supply.

Sense wires measure the voltage at the load (sense wires pass very little current so they don't suffer that much voltage drop), and the PSU corrects the voltage based on the feedback they get.

 

[Ferrum Master]

Just stolen from Stackoverflow:

Cables have some resistance, and when you pass current through them, they have a voltage drop. This causes the load to "Receive" less voltage than what is actually set on the power supply.

Sense wires measure the voltage at the load (sense wires pass very little current so they don't suffer that much voltage drop), and the PSU corrects the voltage based on the feedback they get.

So the mess. The historical 3.3V compensation pin11. And ATX12VO has a optional 12V Sense line in similar fashion.

Then PCIe sense0 and sense1 wires that historically are used are just terminator/jumper pins that matters for the end device it feeds, the AIC board. Has both pins on ground, we can use full power, has only one, please half a pint of beer, no pins, means the PCIe connector is not populated. In general the 12V-2x6 does the same, but using the small pins, making it less logic even more considering the current and return path needs, if it ain't needed, then we omit it, if you need it, then additional ground return path serves just as sense pins, that they are not...the ASUS connecor resides besides the PCIe supply pin, making it more weird and misleading.

Okay, now the real crap. In benchmark PSU world there sense pin has different meaning, and it is not a binary presence indicator as described before. It is connected to simple error opamp. Like we don't have better names, but some monkeys did mix two things up. 4T method should be used for normal results, but not sure about here, it is a simple one wire in the original 3.3V sense, made on the budget(it works slogan). On +12V it doesn't make a sense for cable lengths for about a foot or two. It is not needed. Within specs especially for power supplies at such power capabilities, those sense pins were for historically anemic power supplies.
If you mess up, you can induce wrong readings with this being analog... or that's the idea... adding a pot and rising voltages to all rails? People had made bench supplies in old days in similar fashion.

Third option is like in laptop world, where middle pin is also called sense and is usually a voltage divider in order for laptop to distinct adapter model and power, to not to burn the sucker. But Dell uses one Wire protocol in some models and the sense pin is digital. Same goes for modern PD capable chargers now.

So what is this thing really? It does not make SENSE lol. Where it really connects on the other end, it should not be exposed in reality. So I supposed it communicates with ASUS EC, they like to do like all the time with their products.

If it is just a random idea of additional external compensation outside ATX specs, it should die, it is a hazard. They even named it in common PSU series name, known for digital adjustment and they can get sued for that.

 

[N3utro]

It's a cool new feature but i doubt this would make any significant change as how a regular graphic card would work with or without it. Perhaps useful to extreme overclockers?