Instead of maintaining 60C with fans off, the three included 92mm fans could be spun at 300-400RPM (completely silently!!!), and the idle temperature would be 40C!

I agree with this 100%
The "zero rpm" is a marketing gimmick. Of course hardware will last longer if cooled properly, of course 400 rpm is inaudible. We should have the choice of a low rpm mode.

 

They used to print fan noise curves on the boxes fans came in. This seems to no longer be done then. Fans start to produce a lot more noise as the tips of their blades approach the speed of sound. Below that they’re often very quiet. This varies quite naturally by blade length. For a 120mm fan it’s generally around 800rpm, though tricks can be pulled to get it a bit higher. One of them is curving the blades, but there are others as well. Different fans will have different sound curves but plus or minus 10% of that number is generally hard to achieve. I could see 400 for a smaller fan though. My personal solution is to buy fans with a low max speed so it isn’t a problem. I also don’t tend to build extremely hot systems though so this isn’t something that will work for everyone. If fan design is your thing blade design math was more or less worked out by the end of world war 2 (because fighter planes used em) there was some stuff in the 80’s regarding bearings and some more stuff regarding stiffening systems for blades and other materials science stuff, but the algorithms were worked out in the 40’s.

 

it depends on what temps the vbios think it's ok before actually turning on the fans, i think the threshold is around 60 but that's wayy to high

 

I ditched my 6900xt due to the zero rpm mode and the driver not being able to keep my zero rpm disabled setting persist on a reboot.
I would rather run the fan at low speed all the time vs letting the card heat soak at 60c before the fans turn on.

 

Damn, I didn't even think of this!
Even if I stick in one of those Noctua low noise fan adapters (basically an 80 ohm resistor I think) the driver won't remember!
I know it won't keep my custom clocks and voltages which is annoying AF

 

Degradation is caused by a combination of both temperature and current draw, simply letting a chip idle up to 60C without drawing a large current will do it negligible harm, the worst thing with Zero RPM that can happen is it will wear down the fan bearings faster due to constantly stop-starting - the GPU die/substrate experiences far worse damage during transient loads like cold booting a PC, or starting a game from idle.

 

But since temperature differential is required for heat transfer, wouldn't giving a 60C chip with a 60C heatsink 170 watts out of nowhere cause a higher temperature jump than a 40C chip with a 40C heatsink?

edit: -I agree that a constant 60C isn't so much a problem for silicon, but it is for capacitors. With capacitors, degradation is pretty much entirely tied to temperature, not what they've stored and released

 

0 RPM mode is actually very useful if you want quiet system, of course if other components are quiet. I think I have Asus 960? I think in my sons PC that have been on 0 rpm mode since 10 years or so only because his CPU is on passive cooler and the whole system he silent so he can do either homework or wherever at night, it only kicks in I believe when temp gets to 60C so Its quite nice feature. My Strix 3090 runs on 0RPM since I got it 2 years ago and once temps hits 55C fans are spinning for 15-20 second, I to care about silents so I think its good feature.

 

This is important for reviewers. They need to show minimum rpm. Some manufactures still set way higher idle or minimum rpm, which makes no sense.

Some cards tolerate lowering minimum rpm with mpt tool. Depends on fan capabilities.

 

This is important for reviewers. They need to show minimum rpm. Some manufactures still set way higher idle or minimum rpm, which makes no sense.

It is very important! Minimum RPM of 1600-1800 RPM and three 92mm fans? That's a lot of noise

 

View attachment 2632878

This is the descriptor beside the Zero RPM setting in the Fan Tuning section of AMD's "AMD Software: Adrenaline Edition" in the general "Performance" tab (top), then the Tuning sub-tab (near top at left)

Sure, this sounds good "save power and reduce noise by idling your fan while your GPU is within the Zero RPM temperature range"
And it sounds especially good when you realize that when you try to customize the fan curve, that the lowest speed you can set the PWM to is 30%, and that 30% is actually:
View attachment 2632881 View attachment 2632882

Three 92mm fans running at 1600RPM is not exactly friendly to the ears when you're in a quiet room.

Since View attachment 2632883 at idle, and the card is double height with three 92mm fans, all of them spinning at 1600RPM all of the time is obviously excessive, and lends credence to AMD's claim of "saving power" when Zero RPM is enabled.

The GPU's 3 fans running at 1600RPM can bring the the heatsink/card component temperatures down REALLY fast. 16W? 3 x 92mm fans? About 5 watts per 92mm fan? Cools down right quick.

In fact, if the fans are prevented from running and the card is left operating (showing the windows desktop) in a case with normal airflow (light, nothing special), it will rise to about 60 degrees C, and remain there indefinitely. This system's previous card was a Gigabyte GT 1030, passive edition:
View attachment 2632886
The board consumes 12.6W at the desktop. Depending on ambient and case temperature, the temperature varied a bit, but with the PC not working too hard and the normal room temperature of 21 to 23 deg C, it ran around 50 degrees. Yes it's much smaller, but the fins are more exposed to free flowing air in the case because there's no fan or plastic around the heatsink.

The GT1030's 50C running temperature never bothered me much because the heatsink wasn't covering or in direct proximity to the capacitors.

In my Gigabyte RX 6600 XT Gaming OC 8G (GV-R66XTGAMING OC-8GD):

View attachment 2632890
(above is accurate relative scale (per pixel count of PCIe slot width) to GT1030)

View attachment 2632892 View attachment 2632893
(these are (obviously) not to scale to anything (but each other lol) ! )

You can see the shroud all the way around the 6600XT.
It's metal, and is, itself, a heatsink of sorts. I believe there are VRMs under it. The capacitors must also be under it (or are on the other side under a portion of the heatsink that is raised to make room for them).

These capacitors, being enclosed by metal and air that is at least the outside temperature of the card (which I've measured to be 55C when the card reports 60C) need to be at least as hot as the material encasing them. So at minimum they're 55C, plus they make heat of their own, plus the VRMs are close proximity and radiate heat through the air and through the board to the caps. Let's call it like it is and say the caps are sitting at 60C when the card reports its temperature as 60C...

I don't think this is OK. (to use Zero RPM fan tuning...)! It's a well documented fact that the higher the temperature of a capacitor, the shorter its lifespan. I believe that the rule is approximately - for every 10 degrees C added, lifetime in hours is halved. So that goes the other way, too. 10 degrees lower? Doubled. Capacitor failures are definitely not unheard of on cards (and not just mining cards..)

Instead of maintaining 60C with fans off, the three included 92mm fans could be spun at 300-400RPM (completely silently!!!), and the idle temperature would be 40C! This is good not only for the capacitors, but all included components. Something important to consider, too, is that although the 20C difference for solid state parts might not seem relevant in the 40-60C temperature range, not all components are manufactured entirely perfectly, and so will fail sooner then average. Depending on the manufacturing fault and its severity, this could be the difference between the card lasting as long as you want, or failing a year early.

Why, then does Gigabyte (and likely other manufacturers) make this card unable to spin its fans any slower than 1600RPM? Their maximum speed is almost 4000RPM, which not even 3000 is needed to dissipate heat with +20% TDP enabled... The cooling solution is good, but it seems set up in a way to sabotage the card.

This is something I've thought about before, but now that my 6600XT has been on for as long as it has of by that you mean that if the , my concern has returned...
Does anyone else have the same concern? Experiences with cards with this Zero RPM fan tuning (that you used which kept the card at 60C?)?
What about solutions to lower the 1600RPM @ 30% PWM? Should I find the header and somehow work a Noctua LNA (low noise adapter) in the way?

If by that you mean a tuning where if something isn’t hot enough to need cooling the fan doesn’t spin then yes. Judging from the sheer length of your statement though (didn’t read it) you are probably making a special case though. My understanding is that even in special cases there is so much slop built in to things and so many safeties that even if you found a way to beat it you still wouldn’t. Heat is just too pernicious. It conducts everywhere.

 

I'll re-read (it's been a while), summarize, and clarify what I mean for you (it's gonna take me quite a bit longer to write than read so pls read lolol):

Basically, with "Zero RPM" mode, the fans don't turn on until the GPU die itself hits 60C, and when it does, three 92mm fans spin at 1600RPM (LOUD!!!) until the temperature is brought to under 50C, at which point they turn off. Once heat builds up to 60C again, the cycle repeats. And repeats and repeats...

For every 10C increase in temperature, the life of things like capacitors halves. Say a capacitor on the GPU is rated for 3000 hours at 105C (common) - that's 6000 hours at 95C, 12000 hours at 85C, 24000 hours at 75C and 48000 hours at 65C, and 96000 hours at 55C. If the card reports 60C, the entire outside metal covering of it measures 60C with my IR thermometer (the temp sensor is on the when the GPU itself). I guess the reason the GPU's temp matches the external temp is because the VRM section is also radiating a significant amount of heat due to how low its efficiency at 10 watts (vs. its design: to be efficient between about 70 and 165W).

So the caps are under the heatsinks which measure 60C, so they must be at least 60C - meaning their rated life is about 72,000 hours, or about 8 years in a PC that's on 24/7. Of course, that is the average lifespan - MTBF: mean time between failure - so you're going components failing before and after that (ie. 4 years, 12 years). Problem with that is it just takes one failing, and often times there are a bunch of them on the card - sometimes a couple dozen especially important ones.

I have the "Gigabyte Gaming OC" version of two cards, the 6600XT and 3080 (original). I didn't plan it, they were just what was available and on sale lol. Anyway, if I recall correctly, the 3080 manual fan mode starts spinning the fans at 38%, and the speed is roughly 360-390RPM. Of course the 6600XT takes less power, but it itself is also smaller, as well as its entire cooling apparatus - it's basically the same thing as the 3080 but scaled down for 170W max. instead of 375W max. The 3080 does take disproportionally more power at idle though - 36-40W vs 10W. The 3080 with its fans spinning at just 380RPM keeps the entire card cool (~35-40C) while remaining quiet. Stark contrast is the 6600XT's minimum fan speed is upwards of 1600RPM which does also keep the card [maybe even a bit but not a lot more] cool at 34-37C.

If the 3080 and 6600XT use the same VRM design and components, with the 3080 just having more phases to account for its higher power requirement, and one 3080 has a custom fan profile to always have its fans spinning at ~400RPM when idle, compared to a 3080 in "Zero RPM" mode, it should last at least twice as long, with the average user who maybe uses their card heavily for 10% of the time it's powered up, 4x

Unfortunately the 6600XT's 1600RPM minimum makes lengthening its life an annoyance. A bunch of noise just to keep a noisy card around longer! No thank you! lol

 

Yeah as they get closer (they never actually get to it but they push) the thing starts to spend energy making sound instead of speed. So the noise curve is sort of hook shaped. If you stay inside that hooked bit it’s generally really quiet. I had the thing backward I think. The longer the blade the lower the rpm before you hit it. It’s one of those square vs non-square things though so bigger fans will push more air for a given noise level. The will also have more space between. The blades though so static pressure goes down. So a 140 will be quieter than a 120 but have less static pressure. You need static pressure to push through a radiator so 280s tend to be thinner than 240s so even though they are “bigger” they cool about the same.

 

Cool!

Oh, you forgot my one question - what speed do you think the 800RPM 5V fan I took pictures of is spinning at with ~18V?

With 5V, the display I took a picture of above reports 0.7something watts. With 17.76V you can see the power is MUCH higher, like 14 TIMES higher, at 10.86 watts.

So yeah, 800RPM with, say 0.75W, how many RPM with 11W? (assuming all the extra power goes to moving air, and other losses (like electrical) are small enough to disregard). If you don't know, what would you estimate?

 

No idea. There are tools to measure it but I don’t remember what it’s called. Back in the day there were only 12v fans. The 5v thing came along with rgb so I’ve avoided it. 18v is hard to arrange though and the thing might not be able to handle it for very long. Design envelope and all that.gonna be a lot faster than 800 but it don’t know how much. I could make assumptions based on it being a noctua fan so an older company with a large design and mold library) and them maybe just using their old 12v stuff and then downrating it to 5v, but I have no idea whether it’s true or not. The motor could be totally different. One way to measure fan speed is put a marker on one blade and take a video of it. (Known framerate on the video) depending on the marker though it can throw the balance off though so there are issues. Also that fan is moving at a good clip so there are potential multiplication issues (was that .2 revolutions or 1.2 or 4.2?) so you may need a high speed camera.