ASUS and Noctua Announce ASUS GeForce RTX 3070 Noctua Edition Graphics Card

[qubit]

I asked ASUS for a sample, Noctua made it clear in their press release that they can't provide samples. We'll see if anyone really cares about putting these claims to the test, they'll sell the whole production run either way

I'll bet that's why they can't be bothered with samples, as they believe there's no advantage for them.

 

[Valantar]

For those that have the physical space to accommodate a wide card, I'd love to see more manufacturers offer GPUs with a heatsink that's 240x120 and with a board that has two PWM 4-pin headers.

Conceivably, a standard shroud and heatsink dimension could do two unbelievably awesome things;

1. Reduce build costs and increase profits by consolidating parts across multiple models. I think naively I would have the slimmest hope that these savings can be passed onto the consumer, but I'm a realist and this is 2021 after all...

2. Provide a no-nonsense cooling solution that doesn't rely on RGBLED vomit and dumb plastic bolt-on rubbish to sell a GPU.

That would be really cool. Even if they didn't go that far, just making deshrouding and attaching case fans easier would be a huge boon.

 

[Flyordie]

One ugly GPU.

But repairable. Something you can buy without jumping through hoops.

 

[Dammeron]

I would expect this to rival most water cooled GPUs in thermals and noise though - and it certainly isn't bigger than if you strapped a 240mm rad to your GPU

And You would be very much wrong... The difference in both temps and noise, compared to a fullcover, is huge.

 

[Valantar]

And You would be very much wrong... The difference in both temps and noise, compared to a fullcover, is huge.

Have you tested this specific GPU? Have you seen the type of thermals people report when deshrouding and adding high performance 25mm thick fans to their GPUs? The difference is definitely not huge. It might be if you have one of those utterly overblown water cooling setups with 3+ large radiators and a dozen+ fans, but that isn't what I was comparing to. My 6900 XT Liquid Devil Ultimate sits at around 60-65 with a 280mm rad and Arctic P14s at full speed (~1600rpm). That's in a pretty small case (Meshlicious) and with a single rad for both it and the CPU, so it's clearly an edge case, but it's not that extreme - here are Igor's lab's results when benching the same GPU in its own loop (a 240mm rad from what I can tell). It seems to even out in the mid-50s, with no word on pump or fan speeds. I would frankly be surprised if this 3070 couldn't stay in the low 50s (edge temps) with the fans still running quite slowly. Those fans are, after all, fantastic, and it has a pretty beefy heatsink to work with.

 

[Dammeron]

I would frankly be surprised if this 3070 couldn't stay in the low 50s (edge temps) with the fans still running quite slowly. Those fans are, after all, fantastic, and it has a pretty beefy heatsink to work with.

And I would. There is a major difference between a cooler mounted on a card itself, heating up the card's PCB and air inside the PC case, compared to a LC loop, that moves all the heat away from the source and outside the enclosed space.

 

[Valantar]

And I would. There is a major difference between a cooler mounted on a card itself, heating up the card's PCB and air inside the PC case, compared to a LC loop, that moves all the heat away from the source and outside the enclosed space.

The most common mounting location for radiators is as a front intake, so not really. There's also the difference between having active airflow across the PCB and not, which complicates this further. Even warm air off a heatsink flowing across the PCB will serve to cool it as long as the air is noticeably cooler than the PCB and components on it, after all, and lack of airflow across the PCB is a known drawback of (even full cover) water cooling. And I assume we're starting from the assumption of a case with functional airflow, right? That being said, most deshroud mods I've seen recently are in SFF cases with 25mm case fans mounted on or very near the case side panel, whether as intake or exhaust, which obviously helps things. No matter your method of cooling, you need access to cool air at the site where heat is to be exchanged. Given how thick this GPU is, there will no doubt be situations in which its access to cool air is limited to such a degree that its performance is noticeably reduced. But that's the same with a radiator - if mounted in a non-optimal place (say, behind a restrictive panel), it will underperform. You seem to be starting from some implicit assumptions of worse operating conditions for this card compared to your theoretical water loop, which would of course make it a biased comparison to begin with.

 

[Chrispy_]

I actually have a 3060Ti TUF on water at the moment and prior to the block arriving it was doing ridiculous sub-50C temperatures at +200MHz overclock with the fans at ~1200rpm which was below my local noise floor.

I'm not going to weigh in how well air vs water would work when pushed to the limit but at a 200W TDP the air cooler was actually so good I'd have been happy to leave it as an air-cooled card if I hadn't already ordered the waterblock. To my knowledge, VRM temperatures only affect the lifespan, not performance. VRMs at 105C will perform as well as VRMs at 35C and the usual spec is 5 years continous at 105C. So, whilst there may be advantages to card lifespan by keeping VRMs cooler, the point at which it matters is likely to be irrelevant for the useful lifespan of the card, so long as the cooling is adequate to keep it within spec in the first place.

I don't have the FLIR camera at home, nor am I inclined to replace the block with the aircooler again for science, but a quick google says that it has VRM temps in the 60-70C range which is cool enough that you can basically just forget about VRM cooling altogether. It's a non-issue among non-issues.

credit: HW Unboxed.

 

[las]

So, will they produce more than 100 units? Great time to launch a GPU like this!

 

[CrAsHnBuRnXp]

If people dont want to use their ugly ass fans in their cases, what makes them think we are going to want to put this ugly ass card in our computer? Only time it will work is if someone builds a Noctua themed computer.

 

[awesomesauce]

If people dont want to use their ugly ass fans in their cases, what makes them think we are going to want to put this ugly ass card in our computer? Only time it will work is if someone builds a Noctua themed computer.

back in the day, asus TUF mobo had a color scheme that could fit with this card but i think nova day TUF use black..

here Z270

 

[CrAsHnBuRnXp]

That is the ugliest motherboard Ive ever seen. Id rather have a green pcb motherboard over that.

TBH though, Asus missed an opportunity not making this a STRIX card. STRIX cards are represented by an owl and so is Noctua.

 

[Valantar]

If people dont want to use their ugly ass fans in their cases, what makes them think we are going to want to put this ugly ass card in our computer? Only time it will work is if someone builds a Noctua themed computer.

Given that Noctua sells plenty of their very expensive brown fans I don't think "people don't want to use [them]" is an accurate description. Sure, the colors are polarizing, but the detractors being loud doesn't mean there aren't plenty of fans or people who don't mind out there.

 

[ThrashZone]

That is the ugliest motherboard Ive ever seen. Id rather have a green pcb motherboard over that.

TBH though, Asus missed an opportunity not making this a STRIX card. STRIX cards are represented by an owl and so is Noctua.

Hi,
x99 looks a bit better if you use the fan instead of the cover

Taking up 5 slots is a massive waste that would interrupt with the second pci-e slot wouldn't it ?

People use sound cards/ m.2 cards/... on boards too.

 

[thesmokingman]

Have you tested this specific GPU? Have you seen the type of thermals people report when deshrouding and adding high performance 25mm thick fans to their GPUs? The difference is definitely not huge. It might be if you have one of those utterly overblown water cooling setups with 3+ large radiators and a dozen+ fans, but that isn't what I was comparing to. My 6900 XT Liquid Devil Ultimate sits at around 60-65 with a 280mm rad and Arctic P14s at full speed (~1600rpm). That's in a pretty small case (Meshlicious) and with a single rad for both it and the CPU, so it's clearly an edge case, but it's not that extreme - here are Igor's lab's results when benching the same GPU in its own loop (a 240mm rad from what I can tell). It seems to even out in the mid-50s, with no word on pump or fan speeds. I would frankly be surprised if this 3070 couldn't stay in the low 50s (edge temps) with the fans still running quite slowly. Those fans are, after all, fantastic, and it has a pretty beefy heatsink to work with.

Dude, it's your choice to have a loop that is barely or questionably better than aircooling.

 

[Valantar]

Dude, it's your choice to have a loop that is barely or questionably better than aircooling.

Yes, I say as much in the post you quoted, in case you didn't notice. But then I also cited a source testing the same card with a much less restrictive setup, which still only delivered slightly better results. So... congratulations on a highly selective reading of my post?

Also "barely or questionably better than air cooling"? Lol. Good luck making a build even remotely as quiet as mine with air cooling and a 500W heat output in a <15l case.

Hi,
x99 looks a bit better if you use the fan instead of the cover

Taking up 5 slots is a massive waste that would interrupt with the second pci-e slot wouldn't it ?

People use sound cards/ m.2 cards/... on boards too.

This will make installing/removing m.2s a but more of a hassle, sure, but how often do you do that? And the vast majority of PC builds with a dGPU never see another AIC through the entire lifetime of the build. If you're in the minority using an internal sound card (or NIC, or capture card, or other AIC), you're likely to know and thus avoid this card. For everyone else, it won't matter.

 

[thesmokingman]

Yes, I say as much in the post you quoted, in case you didn't notice. But then I also cited a source testing the same card with a much less restrictive setup, which still only delivered slightly better results. So... congratulations on a highly selective reading of my post?

Also "barely or questionably better than air cooling"? Lol. Good luck making a build even remotely as quiet as mine with air cooling and a 500W heat output in a <15l case.


This will make installing/removing m.2s a but more of a hassle, sure, but how often do you do that? And the vast majority of PC builds with a dGPU never see another AIC through the entire lifetime of the build. If you're in the minority using an internal sound card (or NIC, or capture card, or other AIC), you're likely to know and thus avoid this card. For everyone else, it won't matter.

You're ignoring the details of your own link to make a really incorrect assumption. The cooling capability of any loop is a choice one makes. The test from Igor's is limited by its cooling, DUH. This is just like your average at best loop, it's your choice too make a poor loop just like Igor's is not really a test of the loop in general but of the limited cooling that he choice to use.

 

[Chrispy_]

Without getting too tied up in the arguments here regarding air vs water, you have to remember that this is a 3070 with an MRSP of $499 and a street scalper price of $1100 if you are impatient and don't bother waiting for $699 stocks to come back in.

Sure, you can push a 3070 to the absolute limits if you drop a $200 waterblock onto it (invalidating the warranty) and drop that into a custom loop that probably already cost $500. Or you could just spend that money on getting a 3080 instead. Which will be much faster. With a warranty. With more VRAM. With higher resale value.

I love my custom loop but I do it for quietness reasons. If I was after more performance I'd just get a faster card that will effortlessly provide more performance. Extreme watercooling discussions with expensive loops are complete stupidity if you could have just bought a faster card with the money in the first place. Why spend $100K on aftermarket mods for your Honda Civic when you can just buy a complete Corvette Z06 for $83K. No Civic will do 0-60 in 3s nor will it reach 200mph and you're going to shorten the lifespan and resale value of the Civic attempting to even get close.

 

[thesmokingman]

Without getting too tied up in the arguments here regarding air vs water, you have to remember that this is a 3070 with an MRSP of $499 and a street scalper price of $1100 if you are impatient and don't bother waiting for $699 stocks to come back in.

Sure, you can push a 3070 to the absolute limits if you drop a $200 waterblock onto it (invalidating the warranty) and drop that into a custom loop that probably already cost $500.

Or you could just spend that money on getting a 3080 instead. Which will be much faster. With a warranty. With more VRAM. With higher resale value.

I love my custom loop but I do it for quietness reasons. If I was after more performance I'd just get a faster card FFS. Extreme watercooling discussions with expensive loops are complete stupidity if you could have just bought a faster card with the money in the first place. Consider it like spending $100K on aftermarket mods for your Honda Civic when you could have just bought a much faster, nicer car for less.

That's also a false justification. We're not talking anything extreme here. You put into a full loop what you want out of it. The bulk of what you put in is carried over to the next build, you don't lose it like you are suggesting, like wtf man? You lose out of the cost of the block for that gpu but the rads, tubes, fittings, pumps, etc etc carry over. Lmao, spend 100K on a civic, can you create a more moronic example, really?

 

[Chrispy_]

That's also a false justification. We're not talking anything extreme here. You put into a full loop what you want out of it. The bulk of what you put in is carried over to the next build, you don't lose it like you are suggesting, like wtf man? You lose out of the cost of the block for that gpu but the rads, tubes, fittings, pumps, etc etc carry over. Lmao, spend 100K on a civic, can you create a more moronic example, really?

You're strawmanning this. The argument in question is that dropping a waterblock onto a GPU of this calibre will have such a dramatic improvement in temps and performance that it makes it worth it from a performance/$ perspective, which simply isn't true.

Even if you ignore the cost of the rest of the loop, the cost of the full-cover block by itself is barely enough to justify the performance or noise gains. I went from a silent aircooled TUF with +200MHz on the core to a watercooled TUF and I very much doubt even pushing my dual-radiator loop to its limits would manage more than another 100MHz. The block was €190 and if you add that to the cost of the TUF in the first place I could easily have just bought a faster card.

 

[thesmokingman]

You're strawmanning this. The argument in question is that dropping a waterblock onto a GPU of this calibre will have such a dramatic improvement in temps and performance that it makes it worth it from a performance/$ perspective, which simply isn't true.

Even if you ignore the cost of the rest of the loop, the cost of the full-cover block by itself is barely enough to justify the performance or noise gains. I went from a silent aircooled TUF with +200MHz on the core to a watercooled TUF and I very much doubt even pushing my dual-radiator loop to its limits would manage more than another 100MHz. The block was €190 and if you add that to the cost of the TUF in the first place I could easily have just bought a faster card.

There was no argument but suddenly a couple posters including you are creating one. No one said watercooling is free or doesn't have a cost. WTF man??!! You're creating your own argument right now. Watercooling loops are luxuries make no mistake about that. If you cannot handle the cost, its a game you should not play. If you have to make these stupid arguments, it's not for you. On the flip side stop making these inane justifications because for some of us it is important enough to spend the money on it.

 

[Chrispy_]

The argument started with this ridiculous comment

The difference in noise and temps is marginal at best, and more likely irrelevant, unless discussing extreme overclocks at mad voltages and vastly inflated TDPs. A loop designed to keep 500W cool will do better than an air-cooler for sure, but we're talking about hardware with a 220W power limit here. If you want faster stuff, go and clog up an overclocking thread with that argument - this thread is about a very quiet air-cooled card that has a 220W default TDP and most likely is locked to 110% power limit by ASUS.

 

[thesmokingman]

The argument started with this ridiculous comment

The difference in noise and temps is marginal at best, and more likely irrelevant, unless discussing extreme overclocks at mad voltages and vastly inflated TDPs. A loop designed to keep 500W cool will do better than an air-cooler for sure, but we're talking about hardware with a 220W power limit here. If you want faster stuff, go and clog up an overclocking thread with that argument - this thread is about a very quiet air-cooled card that has a 220W default TDP and most likely is locked to 110% power limit by ASUS.

Facepalm. And you didn't find the quoted post equally ridiculous? What the F is going on here?

 

[Chrispy_]

And you didn't find the quoted post equally ridiculous?

What, that a TUF heatsink upgraded with two NF A12x25 fans can likely rival most watercooling setups in temps and noise?
I've physical proof that the default TUF can do exactly that and you don't need to take my word for it, there are plenty of reviews that show the TUF cooler to be exceptionally good.

I'd be willing to bet that this Noctua version of the TUF is as good as most watercooled variants. Don't forget, the upper power limit on this is 240W from the default of 220W. A heatsink of this size and performance is going to have absolutely no problems whatsoever keeping this thing at very low temperatures under load and the press release itself says that this Noctua card can manage 100% load (so presumably 220W) at 12.6 dB(A). Now, I'm fairly sure my D5 pump makes more noise than that.

Now if (and ONLY if) you take Valantars post (#3) out of the context of this card, then yes. But clearly he's refering to this card with it's puny 240W upper power draw limit. The already overkill air cooler that's doing a damn fine job on a 3080 was ripped straight off a 320W 3080TUF and put on a 3070 with a 220W TDP and it's even more overkill. Now we're making it take up five slots and throwing Noctua at the double-overkill and I think it's fair to say this thing is going to be both quiet enough and cool enough to rival full cover waterblocks even in a decent loop.

If someone doesn't agree with that then this is a discussion forum. State your hypothesis and put your arguments to the test instead of just putting someone down incorrectly and with no supporting evidence or reasoning.

 

[Valantar]

You're ignoring the details of your own link to make a really incorrect assumption. The cooling capability of any loop is a choice one makes. The test from Igor's is limited by its cooling, DUH. This is just like your average at best loop, it's your choice too make a poor loop just like Igor's is not really a test of the loop in general but of the limited cooling that he choice to use.

@Chrispy_ sums it up excellently above here, but just to reiterate: you're completely missing the point. What I said is that this card - the 3070 with the TUF heatsink with two NF-A12x25s on it - can likely perform on par with "most water cooled GPUs". Just for reference, most water cooled GPUs are AIO cooled, and not in a custom loop. And AIO-equipped GPUs don't cool that well, and certainly not at low noise levels. Look at the Asus 3080 Ti Strix LC - with a dedicated 240mm rad it runs at about 50°C, but at more than 41dBA - way too noisy for me, and indeed among the noisiest on TPU's charts. The quiet BIOS is better, but still not particularly impressive. I would definitely expect this 3070 to be quieter than that GPU, and comparable in terms of thermals.

I'd also appreciate if you at least had the courtesy to keep civil and not drag us into some quasi-personal discussion of whether my loop is "poor" - you clearly have a pretty poor basis for making a judgement, and the only reason to take the discussion in that direction is if you want to derail this by making things personal. So, please don't? Also, since you clearly missed it (I'd suggest you go back and re-read some posts tbh): that I specifically wasn't talking about huge builds stuffed full of radiators was the first point I addressed:

It might be if you have one of those utterly overblown water cooling setups with 3+ large radiators and a dozen+ fans, but that isn't what I was comparing to.

Most water cooled GPUs aren't in massive showpiece builds. Most water cooled GPUs have, at best, a single 240 or 360mm rad to themselves. Many have a 120mm, though they've been moving to 240mm rads for most GPU AIOs in recent years. The point is: whatever it is you're comparing to, it isn't what I was comparing to, and the further you try to insist that the point I made was based on your examples, the more you make it clear that you're not here to have an actually interesting discussion.