AMD Radeon RX 6900 XT Liquid Cooled Edition Now On Sale in Europe

[Valantar]

View attachment 227204View attachment 227205

Do you see there is that fan on the GPU shroud? they are to cool the VRAM and VRM. Those AIO only cool the GPU, that means only ~200W is dumped into the loop (with 250W TGP)
Whereas the full 350W is dumped into the loop with the 6900XT LC, just like FuryX.

Yep, and so far we've yet to see any conclusive proof that the Fury X had widespread pump failure issues (nor any evidence to indicate that if it had such an issue, it was down to thermals and not something like a poorly designed pump). It definitely had pump whine issues as you brought up - that was widely reported even in the initial reviews - but that is entirely unrelated to thermally induced pump failure. You're also right that the off-the-shelf AIOs do lower the thermal input to the loop, though at least some of those integrate the VRAM into the loop in various ways (either through enlarged cold plates or some kind of intermediate VRAM plate that makes contact with the cold plate), so that calculation isn't quite as cut-and-dried - particularly when these models were very directly geared towards overclocking.

There are shockingly few reviews from this era with GPU power measurements, but 1080 Tis exceeding 250W and nearing 300W from the factory was relatively common, as seen in TPU's reviews (1, 2, 3). Manual OC's would push them above 300W relatively easily - and those cards were widely known to be easy to OC. And yes, some of that heat would be exhausted by the VRM fan, but a well configured VRM is >90% efficient, so the majority of power will be going through the AIO. In the end, the thermal difference between an AIO-cooled 1080 Ti and a Fury X is quite small, in the low tens of watts. Also, you're neglecting the fact that the Fury X has a much thicker radiator than both of those pictured cards - and likely a much more powerful fan (as EVGA's fans have been known to perform rather poorly on radiators, and Gentle Typhoons are still pretty high up performance charts.

At 60C and above liquid temp, these Asetek AIO are gonna fail rather quick, as mentioned by Asetek engineer (at 6:50)

Hm, that's interesting. Makes me wonder what changes around that level - whether it passes some threshold for permeation, whether that's where you start seeing the pump's "ambient" temperature (casing, block) run too high for it to be able to keep itself cool. Still, he confirms what I said in the previous post though: "We know 60°C is good, we've done extensive testing to ensure that 60°C is what will create a lasting unit." So from what they're saying, passing 60°C is bad, but running at 60°C should be no problem at all.

To be clear, I would definitely not run a water loop that hot, nor do I consider that okay for an AIO, but he's pretty clear on that being "safe" for a "lasting unit" in the interview. That doesn't change the fact that the 63°C seen in that 6900 XT LC review is way too hot (whether it's an Asetek design or not), but it does say something about how little AIOs are affected by fluid temperature.

I still think it looks like the 6900 XT LC is drastically misconfigured in how slow it runs its fan, and I really hope that's not how they're supposed to run - even a moderate speed increase would be a significant improvement there. It's no wonder the thing is struggling to stay cool when it's trying to cool nearly 350W with a single 120mm fan running at 1300rpm.

Seems like I know more about watercooling than you do

Man, thanks for the laugh. I mean, jesus, you can't help yourself, can you? You were making dumb assumptions about the specific reasons why I run an UV profile on my GPU. You were completely and utterly wrong, and now have egg on your face. Responding to that with "Seems like I know more about watercooling than you do" ... wow. That's some grade school level defensiveness and deflection. Rather than make yourself look even worse, I would really suggest toning it down a few notches. Seriously. Can we at least try to have an adult discussion here? You've been pelting me with personal attacks for a while now. It only undermines what you're trying to argue. Please stop.

 

[nguyen]

Yep, and so far we've yet to see any conclusive proof that the Fury X had widespread pump failure issues (nor any evidence to indicate that if it had such an issue, it was down to thermals and not something like a poorly designed pump). It definitely had pump whine issues as you brought up - that was widely reported even in the initial reviews - but that is entirely unrelated to thermally induced pump failure. You're also right that the off-the-shelf AIOs do lower the thermal input to the loop, though at least some of those integrate the VRAM into the loop in various ways (either through enlarged cold plates or some kind of intermediate VRAM plate that makes contact with the cold plate), so that calculation isn't quite as cut-and-dried - particularly when these models were very directly geared towards overclocking.

There are shockingly few reviews from this era with GPU power measurements, but 1080 Tis exceeding 250W and nearing 300W from the factory was relatively common, as seen in TPU's reviews (1, 2, 3). Manual OC's would push them above 300W relatively easily - and those cards were widely known to be easy to OC. And yes, some of that heat would be exhausted by the VRM fan, but a well configured VRM is >90% efficient, so the majority of power will be going through the AIO. In the end, the thermal difference between an AIO-cooled 1080 Ti and a Fury X is quite small, in the low tens of watts. Also, you're neglecting the fact that the Fury X has a much thicker radiator than both of those pictured cards - and likely a much more powerful fan (as EVGA's fans have been known to perform rather poorly on radiators, and Gentle Typhoons are still pretty high up performance charts.

Hm, that's interesting. Makes me wonder what changes around that level - whether it passes some threshold for permeation, whether that's where you start seeing the pump's "ambient" temperature (casing, block) run too high for it to be able to keep itself cool. Still, he confirms what I said in the previous post though: "We know 60°C is good, we've done extensive testing to ensure that 60°C is what will create a lasting unit." So from what they're saying, passing 60°C is bad, but running at 60°C should be no problem at all.

To be clear, I would definitely not run a water loop that hot, nor do I consider that okay for an AIO, but he's pretty clear on that being "safe" for a "lasting unit" in the interview. That doesn't change the fact that the 63°C seen in that 6900 XT LC review is way too hot (whether it's an Asetek design or not), but it does say something about how little AIOs are affected by fluid temperature.

I still think it looks like the 6900 XT LC is drastically misconfigured in how slow it runs its fan, and I really hope that's not how they're supposed to run - even a moderate speed increase would be a significant improvement there. It's no wonder the thing is struggling to stay cool when it's trying to cool nearly 350W with a single 120mm fan running at 1300rpm.


Man, thanks for the laugh. I mean, jesus, you can't help yourself, can you? You were making dumb assumptions about the specific reasons why I run an UV profile on my GPU. You were completely and utterly wrong, and now have egg on your face. Responding to that with "Seems like I know more about watercooling than you do" ... wow. That's some grade school level defensiveness and deflection. Rather than make yourself look even worse, I would really suggest toning it down a few notches. Seriously. Can we at least try to have an adult discussion here? You've been pelting me with personal attacks for a while now. It only undermines what you're trying to argue. Please stop.

Hah, I could see from a mile away that cooling 350W with a 120mm AIO was a bad idea, yet you kept asking for proof even though I already gave the review link on my #1 post in this thread. Like previous commenter mentioned, you are the one clogging the thread demanding proof that I already gave out.

Now imagine some chumps who bought that 6900XT LC and give it another 15% more power (400W), I would love to see those AIO fail spectacularly in the near future.

 

[Valantar]

Hah, I could see from a mile away that cooling 350W with a 120mm AIO was a bad idea, yet you kept asking for proof even though I already gave the review link on my #1 post in this thread. Like previous commenter mentioned, you are the one clogging the thread demanding proof that I already gave out.

But that's the thing - we've only seen conclusive proof that this is true while running the fan at a ludicrously low speed - I wouldn't expect a 120mm AIO at 1300rpm to cool a 125W CPU, let alone a 350W GPU. So while we could of course argue whether a badly configured fan curve is an inherent part of the cooler or not (I'd be squarely on the side of not), there's little doubt that this current configuration - assuming that GPU tested was working as intended - is unsuited for the application at hand. What it doesn't prove is that "cooling 350W with a 120mm AIO is a bad idea". It proves that "cooling 350W with a 120mm AIO at 1300rpm is a bad idea". That extra caveat changes things quite dramatically after all, and this fan curve is a rather drastic change from how AMD has configured their previous GPU AIOs. I would love to see PCgameshardware re-test the card with a more sensible fan curve, to see what this could do if configured by someone with a functioning brain. If it does 63°C liquid temperature and 80°C core at 1300rpm, I'd expect both of those to drop quite significantly if the fan speed was raised to something like 2000rpm. Liquid temperatures would definitely drop well below 60°C. Of course then you might say it's too loud, but then we'd be discussing tastes and preferences, not absolute cooler capacity.

As for "seeing this a mile away" and linking the review in your first comment here: it's useful to note that you don't actually point to that review with any part of your take on this (other than saying it exists). If you had, for example, pointed to the liquid temperatures from the review back then (which, given how late they were referenced here, it seems you hadn't read the review in much detail either?) we could have gotten to a far more interesting point in this discussion far earlier, instead of useless back-and-forth general claims of "can a 120mm AIO cool X W? y/n". Still, without someone doing a more in-depth review with a custom fan curve (and, ideally contacting AMD to ask wtf they were thinking with that configuration), we don't really have the data to tell whether this cooler is sufficient or not - though the low fan speed does suggest a lot of headroom, but how much better it would perform is an open question.

Now imagine some chumps who bought that 6900XT LC and give it another 15% more power (400W), I would love to see those AIO fail spectacularly in the near future.

Well, it would likely thermal throttle and essentially not change at all, unless they also boosted the fan speed, so ... meh. My guess is it would stay the same.

 

[Mussels]

Look valantar, you defended the theory. That's fine.


But it's been discovered that in practise, it's poop. Logic and planning don't always match up to reality, and it seemed damn weird from the start this had a single 120mm rad, without a hybrid design to remove some of the heat

This is what happend to an asetek AIO (thermaltake) after the user lost the fan controller and ran pump only, before asking me to fix it. It was not fixable.

If this card edges thermal throttling brand new in perfect situations, we know, we KNOW it's going to be malfunctioning, fast.

 

[Valantar]

Look valantar, you defended the theory. That's fine.


But it's been discovered that in practise, it's poop. Logic and planning don't always match up to reality, and it seemed damn weird from the start this had a single 120mm rad, without a hybrid design to remove some of the heat

Haven't said anything else, have I?

This is what happend to an asetek AIO (thermaltake) after the user lost the fan controller and ran pump only, before asking me to fix it. It was not fixable.
View attachment 227254

If this card edges thermal throttling brand new in perfect situations, we know, we KNOW it's going to be malfunctioning, fast.

Yeah, as I said, that fan profile is ludicrous. Why it doesn't exceed 1300rpm in normal operation is ... well, not something that there can be an even remotely reasonable explanation for, and takes this from "likely fine, if a bit warm" to "unacceptable".

 

[T1beriu]

Why should anyone overpay almost twice the price from suggested price from Nvidia and AMD?

Because that someone wants one and they can afford it? What's up these questions?!

 

[nguyen]

well another update, my friend latest EK DDC 3.2 pump also bite de_dust, seems like high liquid temp (~50C) really kills these pump rather quickly (2 years for Bitspower S1 and 1 year for EK DDC 3.2).

His new pump is the Bistpower DDC 4.2 (the spec say max temp is 70C), he also remove the top, front and backside panels of the Lian Li 011D in hope to reduce the temp and make the pump last longer this time.

My trusty S1 still runs fine without any airflow nor heatsink, spending a little more for 6x NF-A12x25 fans really paid off in the long term

 

[Deleted member 24505]

well another update, my friend latest EK DDC 3.2 pump also bite de_dust, seems like high liquid temp (~50C) really kills these pump rather quickly (2 years for Bitspower S1 and 1 year for EK DDC 3.2).

His new pump is the Bistpower DDC 4.2 (the spec say max temp is 70C), he also remove the top, front and backside panels of the Lian Li 011D in hope to reduce the temp and make the pump last longer this time.

My trusty S1 still runs fine without any airflow nor heatsink, spending a little more for 6x NF-A12x25 fans really paid off in the long term

On EK the DDC 4.2 still says 60c max

 

[Bilo991]

Idk if this forum thread is still a live but ok... I have this gpu, here in europe it was the same price like sapphire nitro rx 6900xt. And some one you are right. Fan speed is to low for this kind of cooler. It run at max 1500rpm, then you have problem, liquid inside aio is aroud 60c etc. Fan is 3200 rpm delta, i put it at +- 2500rpm when gaming and liquid temperature wont go over 45c, even when i oc whit max power draw liquid temperature wont go over 51c, fan at +-2500rpm, and gpu core clock is 2800mhz, hot spot temperature at 83c max, gpu power draw is 285w at stock and 330w at oc. And cooler is made by cooler master not asetek because i take shroud of, kind da. If you know what are you doing at have some know how, you just need to setup custom fan curve and you are golden. I have rad at the top, exhaust in my o11 dynamic whit 6 intake fans, 3 of them are in 360mm aio for cpu.

 

[Deleted member 24505]

I have a fan blowing on my pump now.