Podcast Archive - StorageReview.com - Podcast #129: Keeping Your Cool With Noctua

Episode Date: June 22, 2024

Brian goes for another live podcast this week featuring Jakob Dellinger, the “air cooling… The post Podcast #129: Keeping Your Cool With Noctua appeared first on StorageReview.com. ...

Transcript
Discussion (0)
Starting point is 00:00:00 All right. After several technical difficulties, we are trying to do this live again. Coordinating people across the globe seems like it should be an easy thing by now, but we keep on trying to push the boundaries of how we're doing this podcast. We really want the interaction with our audience, and so we've been trying to do some new things, and with that comes some new challenges, but hopefully by the end of the year we'll have it sorted out. Today we've got what I hope to be a pretty exciting conversation if you can tell by the stuff on my desk we're chatting with Noctua who are a I said I made not even an intentional joke when I said a fan favorite on social media teasing this live podcast. Everybody loves Noctua.
Starting point is 00:00:51 They love the design, the quietness. But there's all sorts of new stuff that happened at Computex last week. There's some new stuff with the Homelab 15 box. And we're going to tear into all of that. As a reminder, though, this is live. So be sure to submit your questions. If they're any good, we'll try to tear into all of that. As a reminder though, this is live. So be sure to submit your questions. If they're any good, we'll try to get them in. If you're just complaining about the fans being brown, I can't help with that. Although we'll talk about that too in this
Starting point is 00:01:15 conversation. So Jakob from Noctua is here to have this conversation, to participate with us. Thanks for doing this. I know it's evening time where you are but thanks for tuning in and participating how are you doing today my pleasure thank you very much for having me yeah looking forward to to answering your questions well let's just start why is it in your view that people love your products so much? That's a very nice question to kick things off. I would say they just appreciate the attention to detail and the obsession with quality and performance that we have here at Noctua.
Starting point is 00:02:02 I would say that's something you can feel when you pick up a Noctua product. Well, go back a little bit and tell us a little bit more about the origin story for Noctua, how we got here. And then I want to get into some of the new stuff you were showing at Computex, because where you started and where you are today are maybe side-by-side paths, but some of the stuff you guys are doing now is really cool.
Starting point is 00:02:26 Not that it hasn't always been, but tell us a little bit more about how the company started and how you got to this point. Sure, sure. So we're actually pretty close to our 20th anniversary. We started up with Noctua in 2005, but the parent company in Austria here, RASCOM, existed since 2001. And yeah, we started out as basically your average computer part distributor and reseller.
Starting point is 00:03:01 And eventually we focused more and more on quiet, high efficient, high quality cooling components, started doing OEM business, importing for other brands, and then realized that we had built up so much expertise in the field that we just wanted to get our own thing going. And that was how Noctua was born. Basically we started working with one of our Taiwanese partners that we've at this stage already been working for for a few years importing and doing OEM stuff for other brands and then just kicked off our own thing and channeled all the expertise we had built up into our own products rather than other people's products.
Starting point is 00:03:48 Well, what was the problem you saw with the fans available to you? There must have been at that time dozens of options. Back in these days, first of all, a lot of the cooling fans used in PCs were like 60 millimeter, sorry, 60 millimeter, 80 millimeter, relatively small diameter fans. I think the older people among us remember those like 60 millimeter, 7K RPM fans screaming inside their machines. And yeah, that was a gradual shift from those small diameter fans to larger diameter units like 120, 92 millimeter that enabled a significant reduction in noise levels. And then at the same time, the first brands who were doing quiet, basically what they did was just buying standard fans from Chinese or Taiwanese suppliers and reducing the rpm speed on those fans but
Starting point is 00:04:53 those fans have been designed for relatively high rpm ranges maybe 2000 3000 rpm something like that so the their efficiency when you drop the RPM was not stellar exactly. So I think we were the first to really design a fan, a 120 millimeter fan from scratch, optimized for those low speed regions and that enabled us to have a really significant performance boost to customers. What about the obsession with quietness? Where did that come from? Difficult to tell. I mean, I think this is something you just live with. For me personally, I just remember always tuning even the very first PCs I've built for being quieter because I just hated the noise. And yeah, I would say it was very similar for Roland, our CEO. I mean, I remember in those very early days when we built the rigs to run at
Starting point is 00:06:11 the boss's desk. Quietness was always the first thing that had to be taken care of. So yeah, I would say that we were probably just bored and hating noise. Well, that makes sense. I mean, it's funny where we find inspiration, right? If you just start out by hating noise, that makes sense. I thought maybe you had like a media and entertainment twist or something because we see that a lot. And in fact, when Vince, who does our editing, and you guys supplied a ton of fans to
Starting point is 00:06:47 help him make that super quiet rack. Yeah, that was perfect. Yeah, it's a good video, and it's pretty fun. So Vince essentially had a, we got an Eaton quiet rack, so an insulated rack with doors and such on it. Really nice, by the way. And then he took out all the fans he could out of his NAS and servers and JBODs and everything, replaced as many as possible, some which required a little more intervention than others with the Noctua product. And the silence for him was real in terms of business impact because he needs to be able to hear what he's editing. And then if he's recording, he doesn't want a bunch of ambient noise in the way because he's working in the same space as his
Starting point is 00:07:29 little server rack so for guys like that the silence is is tremendous but I think too until you get used to I don't even know how to say this until you get used to not hearing the noise from the fans, I think you just assume that all fans are like that. And you guys prove over and over again that that's just not the case. Yeah, that's a message we need to get out there to people. You don't need to live with the noise. It's not necessary. There are things you can do. Yeah, I said the beige or brown fans at the beginning.
Starting point is 00:08:06 I know you've got some of these products on the desk here are in fact black and you make black things. But the Noctua that most people associate with the instant visual is that brown and beige color scheme. Is there a history to that? And where did that come from? Yeah, we already spoke a little bit about those early days of the company and the brand. And those early 2000s were also like the first wave of PC modding
Starting point is 00:08:41 where you had that super flashy LED blinking in all colors aesthetics and quite a brass coloring on all the other components. So it was either like black and red, stuff like that. And yeah, we wanted to offer something different and to stand out not just from a quietness of operation perspective and a quality perspective, not just from a quietness of operation perspective and a quality perspective, but also visually. And then we were looking for a call scheme that on the one hand like breaks with that aggressive modding kind of aesthetics,
Starting point is 00:09:18 but on the other hand also reflects that quietness of operation, like by using earth tone colors and one more background to it is that we're coffee lovers here at Noctua and yeah the dark brown we use and but also the beige it's it's kind of reminiscent of Italian design Italian coffee coffee culture in some way. So that's like an extra little ring to it. But the main point was really finding a call scheme that reflects that quietness of operation
Starting point is 00:09:55 and the shift in focus compared to those visually driven modeling component. Okay, so I'm significantly older than probably most of our audience in here. And I grew up in the 80s when one of the single greatest inventions was the scratch and sniff sticker. So nobody born lately will have any idea what I'm talking about. But there was a period of time where the greatest bit of entertainment on the recessed schoolyard was to share your sticker book with friends
Starting point is 00:10:32 and the scratch and sniff stickers that were in there. So you would literally scratch it and it would release a scent. And then we'd be walking around smelling bubble gum. Or if you were super fortunate, you had the skunk what i'm thinking here is the next generation of the coffee colored fans as they heat up emit a pleasant espresso aroma can that be done i'll put it i'll put that on the list and uh i'll let you know tomorrow how darn do you guys react to that. They'll love it. Yeah, okay. We've got a comment in the chat about scratch and sniff stickers.
Starting point is 00:11:13 Apparently, it was a popular thing, at least enough in that generation where at least one person knows what I'm talking about. I think they even made ones. The modern equivalent is the Bean Boozled game, the jelly beans where they're colored the same, but one will have like a fruit taste and one will taste like gym socks or something. They had these stickers, it would show a bubble gum, but it would smell like the skunk or something.
Starting point is 00:11:37 But I digress as per usual. So we've talked a lot about where you guys came from and answered, I think, a lot of the popular questions about the company and the color scheme and why the commitment to quiet. Talk about what you guys did at Computex because you had a bunch of stuff going on there, power supply, new fans. Pick the thing that resonated most with the audience out there, but I want to tear into several things that was part of the Computex booth.
Starting point is 00:12:12 Yeah, we had a pretty broad mix this year, I would say. Like one side was updates to projects that we've been showing for, well, a couple of years actually. but those are also some of the most eagerly awaited by our customers. So that's the next gen 140 millimeter fan and the next gen D15 CPU cooler. So yeah, this year we gave the final updates
Starting point is 00:12:42 on these projects, told our visitors that basically mass production is in full swing. The coolers and the round frame fans are shipping now and we're targeting early July release for them. I can show you I brought one of the coolers here just freshly unboxed. They have arrived in Vienna. And yeah, at Computex, we told you it's going to be a late June release. We actually had some logistical delays. So it's going to be early July now. So we had to push back by another week or maybe two at the most.
Starting point is 00:13:23 But they are coming. They're going to come very soon and the square frame version is targeted for September now. So for anyone on the audio, you just held up a fan, radiator, fan, radiator, big chunk of equipment there. What is that rated for and what sockets are you supporting with that product? Socket support is AMD AM4 and AM5 as well as Intel LGA 1700, but you can also retrofit it to 2066, 2011, 775 with the free of charge mounting kits we offer. So we don't want to include those legacy socket mounting hardware in the box
Starting point is 00:14:08 because it would be just too wasteful to include all those parts. And then, I don't know, 0.5% of the customers actually use them. But yeah, all the current sockets from the two main players are supported. And so that's one that's been in demand for a while, but you also showed off something that surprised the heck out of me. You guys had, and I'm not sure if it's a shipping product yet, but was it a Grace Hopper cooler or Grace Blackwell? It was a gigantic thing.
Starting point is 00:14:43 Yeah, NVIDIA GH200. This is still a prototype. A GH200? That's insane. And that's an air cooler only? Yeah, yeah. I mean, the total power budget is 1,000 watts. And yeah, we can handle that
Starting point is 00:15:01 with two large air cooling heat sinks. There's this general misconception that air cooling wouldn't be able to handle, I don't know, north of 500 watt. But it totally depends on what chip you're putting it on. So, for example, we had Xeon coolers, socket 4677, running it up to 700 watt air. And yeah, that's just due to the enormous surface area and large die size of the latest Xeon CPUs. But if you put the same cooler on a, let's say M5 CPU, if it's a single CCD one, the same cooler will top out
Starting point is 00:15:43 at, I don't know, maybe 150, 160, 170 watts, something like that. So it really depends on the CPU. And with GH200, we're fortunate enough to have large heat spreaders where we can nicely dissipate a lot of thermal energy. Well, who's crazy enough to put that into a, it's a workstation form factor, right? I mean, maybe you can't talk about it yet, but you've got partners that are trying to squeeze these enterprise boards into workstations, which is... Absolutely, yeah. The use case we mainly see from the clients we're working with on this is people who want to train large language models locally due to latency concerns, but also due to data
Starting point is 00:16:27 safety and data privacy concerns. So if you need that computing power in a local environment, like an edge server or a lab environment, all of a sudden quietness of operation matters, and that's where we step in. Yeah. So you brought up something that I wanted to get into. I didn't expect to hit it just yet. But while we're here, let's talk about it.
Starting point is 00:16:50 At all the big infrastructure shows, Dell World, presumably HPE, Discover, which is going on right now. Liquid cooling is a hot topic. I mean, NVIDIA GTC showed off the NVL72 where everything's liquid cooled. Now your products aren't going after those 1U high density servers. And that's, I guess, where we get into the conversation about where air can continue playing and where liquid cooling has to play a role in some of these super dense configurations. Is there a place for liquid cooling in the Noctua product line someday? Or is that just too far afield from where your expertise resides?
Starting point is 00:17:37 Well, we actually showed a liquid cooler in the broader sense of the word at Computex. We showed a thermosiphon based cooling solution for PCs. So yeah, definitely. I mean, it's not like liquid cooling is something we're ruling out in the first place. The problem is just that many of the now so popular all-in-one liquid cooling solutions are really hard to square with our core values of quality reliability and quietness of operation because we see so many customers switching back to air cooling from all-in-one water coolers because they had like multiple units failing on them
Starting point is 00:18:20 pumps failing on them within a couple of years or just having significant problems with vibration or pump noise. And that's something that wouldn't work in an octo product. And that's also the background of the... Yeah, well, I guess it depends. Like if you're in a data center, you can generally handle a little more noise and infrastructure, right? And at that scale, at that AI, HPC, big, big scale, then the rules are different. But you're right. I mean, in the workstation or PC world,
Starting point is 00:18:56 if you can get more out of air in terms of the TDP support and headroom there, that's a pretty amazing story to be able to be that efficient in the design. How much of that relies on what you're able to do with these heat sinks and case fans? And how much of that relies on your system partners to design cases with airflow and air shrouds and other components inside, including fan placement. Do you guys get involved in helping partners figure that out so that they're as efficient as possible? Because I don't get the feeling that in these high TDP systems that you can just slap three fans in the front, one on your CPU and be okay there.
Starting point is 00:19:45 You may need to be a little more creative. Yeah, I mean, I would say that PC and workstation case design has come a long way. So unless you make a really bad choice, modern cases are well-ventilated and well-suited for cooling high-voltage components. So So generally if a competent system builder chooses a decent case, there's not so much you can do wrong honestly. Of course you can get that extra edge by using custom ducting and stuff like that, but as long as you follow the basic rules and pick big good components I would say you're mostly fine it's usually those projects
Starting point is 00:20:33 where we get more involved with the overall like system level thermal design that are a bit more customized so our fans go into dozens of different products. For example, we had a theater and filming equipment, lamps lately where we zoomed a little bit deeper into the thermal design of the units. And there you can make, I would say, much bigger impact by changing a few things around and improving airflow throughout the system
Starting point is 00:21:11 than with an average workstation or PC. Yeah, that's interesting. Your efficiency is... Not to say that you cannot gain a lot from optimizing a workstation or pc by just putting in the right fans and doing some tweaks here and there but yeah i would say the better pc in the workstation cases out there they offer you a good basis to start off from yeah it's it's interesting we're getting some comments on the YouTube feed about some custom solutions or creative solutions
Starting point is 00:21:51 that these guys will go through, including loading fans, it looks like, near the PCIe slots if they're unused and using Molex or SATA adapters to power these things. Probably not the number one way that people are using your fans but i guess you don't mind if the the end user gets creative with these as well right not not at all on the contrary i mean we love seeing people doing uh doing crazy stuff and uh because it i mean it's all feeding back to us, right? We get a lot of the ideas we just get from customers who are creative, finding better
Starting point is 00:22:33 solutions for their specific thermal problems. So yeah, we love seeing people doing crazy stuff. Absolutely. So speaking of thermal problems, one of the things you guys were talking about at Computex too, that again, I wasn't there, I'm getting this feedback from Jordan who swung by your booth, but he really highlighted the thermal study you guys did around CPU lids. I think we all assume or many assume that those are rectangle CPUs and the lid is flat. And if I'm applying my thermal paste to the CPU lid and putting my heat sink on there, that the heat is relatively evenly distributed across that lid.
Starting point is 00:23:20 And so the heat sink will pick up sort of equal weight of heat across the entire thing. It seems nothing could be further from the truth and you guys had a big study highlighting that. Can you tell me a little more about what you did there? I wish that was the case. It's actually a very complicated and challenging situation at the moment with the two key platforms, AMD AM5 and Intel LGA1700, differing so significantly in surface flatness and also hotspot location. So as a rule of thumb, AMD AM5 CPUs, they stay relatively flat. By relatively flat I mean they what I mean is that they still go slightly concave but only to around the 15 micron level whereas Intel LGA 1700 CPUs they deform heavily under the pressure that the internal loading
Starting point is 00:24:21 mechanism exerts on the on the heat. So with LGA1700, we've measured up to like 55 micron level concavity. And that means that if you take a relatively flat CPU cooler and put it on an Intel LGA1700 CPU, you will end up having contact basically only on the north and the south side of the heat spreader, and the entire middle part will be left uncovered. And of course, thermal paste is there to fill the gap, but you generally want your thermal paste layer to be as thin as possible for optimal thermal performance. So in a situation like the one I've described where you have a like a bad match between the cooler contact area and the CPU heat spreader
Starting point is 00:25:11 you can lose like five six degrees easily in cooling performance with all other things being equal. So yeah when we worked on this next generation D15 cooler, I've just shown you before, we've spent years like squeezing out two to three degrees through fan design, heatsink design, tailoring the heatsink to the fans. So we just couldn't accept losing a couple of degrees due to poor contact quality. And that's why we ran all those studies and in the end decided to offer the D15G2 in three different versions to optimize it for different CPU and flatness scenarios. Well how would that manifest itself? How would an end user know?
Starting point is 00:26:02 I mean it's hard to know, right? I guess you have some idea of what the CPU temp should be. But I mean, here, what we can do in our lab is run something with the shipping heat sink, and then change the heat sink and run it again, or put a cold plate on and run it again, and start to get some measurements of CPU temp and then performance, if we go to a max load, like a Y cruncher to really stress the CPU, it's really obvious in that scenario with two or three different heat sinks
Starting point is 00:26:35 or thermal paste or whatever in a controlled environment, running the same workload over and over, we can chart that out and see, oh, hey, this was the best, this was the worst. And the Delta in performance could be five, 6%. So those couple degrees can really make a big difference in performance too. But if you're buying a system or if you're building a system
Starting point is 00:26:59 and you're just doing this kind of on a one-off, it would be almost impossible to know that you're being robbed of some performance or exposing your system to more heat than you really need to be and even if you knew that then what would you do so i don't know i don't know what my question is there but this is a complicated problem yeah it is a complicated problem but um and i see where you're heading but the good news is that the deformation patterns or the general flatness is mostly platform specific. relatively flat and that it has the hotspot location slightly shifted towards the south side of the package because that's where the CCDs are located or the single CCD if you have a single CCD AM5 CPU. So that's a situation that's defined by the specific platform. So it's not like you
Starting point is 00:27:57 have to measure temperatures on your specific AM5 CPU. It's just that you have to be aware of the overall situation with the platform. And the situation is similar on the other side of the spectrum with Intel LGA1700. You can be confident that if you put full ILM pressure onto your CPU, it will get heavily deformed and then you're better off picking a higher base convexity heatsink. With Intel, it's a bit more complicated because there are means to reduce ILM pressure. So you can either use one of those contact frames or do what's been called a washer mod. You take one millimeter washers, put them underneath the internal loading mechanism and thereby reduce the pressure
Starting point is 00:28:47 the ILM exerts onto the CPU. And this way you can also reduce this deformation and improve contact quality with heatsinks that are not as convex. Now, where it gets really complicated now is that we have seen situations where Intel CPUs got permanently deformed from ILM pressure. So if you run your CPUs for months and
Starting point is 00:29:11 months with lots of thermal cycling, high heat loads, under full ILM pressure, it will not really go back to its original shape. So even if you add those washers underneath the ILM later on, you're still not going to end up with a flat CPU. But we try to give customers a relatively simple guidance. And that is, if you're unsure about this or if that all sounds like too much for you, just pick the standard version. The standard version, like all our multi-socket coolers, comes with a medium base convexity. And that means that you can run it on AMD AM5 with the included offset mounting. You will have perfect pressure over the CCDs, best possible temperatures on AM5.
Starting point is 00:29:59 And you can use the included 1mm shim washers on LGA1700. That's going to give you the best possible results on the fresh LGA 1700 CPU and even if your CPU has become deformed from from long-term use the results are still going to be very much okay it's only if you want to edge out in a degree or two on deformed LGA 1700 CPUs or if for some reason you don't want to mess with the socket or the ILM and run full ILM pressure, then you can get a little bit extra performance from using the high base convexity version. And on the other side of the spectrum, if you don't want to use the included offset mounting on AM5 for some reason,
Starting point is 00:30:45 you can get a little bit better temperatures with the low base convexity version. Same thing if you're still running, let's say, LGA 2066 or AM4, then the low base convexity version is going to give you a little bit better results. But it's not like we're forcing customers to make a choice between two extremes. You're going to have that standard solution that's going to work great in virtually all scenarios. And then you have those edge niche case solutions if you want to push things a little bit further in specific configurations. So we've got two questions. I'll try to combine them here from Jimmy and Jake that are asking about this issue specifically around LGA 1700. How much of this can be moderated with an aftermarket contact frame?
Starting point is 00:31:39 And the second part of that being, rather than making a variety of models to account for this, could you mediate some of that issue with your own contact frame with the universal fan design? So the contact frames have a relatively similar effect as the washer mod. They reduce the amount of pressure that goes into the CPU that presses it into the socket and thereby result in less deformation however in our so just to stick to the contact frame for a second if you have a contact frame already that's great and you're going to see very good results with the standard media based complexity. However, I must note that we actually have seen slightly better results with the one millimeter washer mod versus the contact frame. So from our perspective, the washer mod is actually the better solution.
Starting point is 00:32:42 That's why we include the washers rather than a contact frame, because it's just a much more simple part to include. We have not seen any issues with lost memory channels or something like that that some customers have reported with the contact frames. And yeah, that's why we think that the washers are are the better solution from our perspective because they gave us better results thermally and also i would say a simpler solution to the same thing interesting yeah i mean the just yeah for
Starting point is 00:33:23 anyone that didn't see it you should check out these these Noctua posters that show the heat on the lids. I mean, it's unbelievable how much variation there is. Those are actually just pressure scans, not scans. So to just show when you put the two surfaces together where you have the highest pressure concentration. Right. Yeah, my bad. But still really interesting data, right, to show where that variance lies. I mean, that's impressive. Also at Computex, you guys are showing off a power supply.
Starting point is 00:34:02 Tell me about that. Yes, that's one of my favorite projects of the year, actually. Yeah, we've teamed up with Seasonic and took their flagship model Prime TX, so titanium level efficiency, 1600 watt power supply and made it even quieter because it's a very quiet unit to begin with. So when we started working on it, honestly, I was a bit worried if we could provide a meaningful improvement there. And then when we got the first results, yeah, I was very happy to say the least because in the end we managed to knock off like an 8 to 10 decibel noise reduction across
Starting point is 00:34:48 the entire fan speed range and yeah it's not a simple fan swap it's not just we put in our fans and that gave us this huge reduction we also custom engineered a new fan grill that works in tandem with our fan that's been designed from ground up to reduce turbulence and improve acoustics. So with that 8 to 10 decibel improvement, the grill alone contributes around 2 decibels on a flow normalized perspective. Well, and you have to make it brown too. Don't diminish that, right? Absolutely.
Starting point is 00:35:28 There has to be some visual distinction. So it's going to have a brown accent on the cover and there's going to be some brown accents on the cables as well. What's the ETA for the arrival of that product? We hope to have it out before the end of the year. Okay. So you guys... It's a guys uh promises but i'm pretty confident that we should be able to to pull it off i mean it's somewhat brave though and um i'll admit that um i'm not the deepest on the history of the company but for you to show the products
Starting point is 00:36:02 and you talked about the gen 2 fans, so early in the development cycle is, I guess on one hand you could argue it keeps you accountable to deliver these things, but most companies won't show product that early for fear of missing deadlines or the social backlash or whatever. It's pretty brave. Yeah, I mean, our advantage is that we are a family-owned, family-run business, so it's
Starting point is 00:36:29 not like we're going to get pressure from shareholders if we don't meet a deadline. That's not to say that we don't want to meet deadlines, but with the G2 products, it definitely wasn't our intention to show them years before release. It's just that we ran into so many issues that significantly delayed the rollout. And that's how that whole situation of having them shown again and again actually occurred. Well, but you're almost there. So the light is very much at the end of the tunnel, right? Finally.
Starting point is 00:37:09 But that, quite frankly speaking, was not intentional from our side. And we've definitely learned the lesson from customers that they don't appreciate stuff being delayed again and again. So we're for sure a bit more cautious putting things on the roadmap nowadays. So with the Thermal Siphon project, for example, you can see that we've shown it at Computex, but it was clear that this is a development announcement and a long-term project we're working on. So we deliberately decided not to give an ETA and not to put it on the roadmap yet to avoid precisely this kind of situation. Fair enough. We got a couple questions. Harry's got
Starting point is 00:37:53 one from YouTube and I'll take his and add it on to mine. We've talked a lot about CPUs, a little bit about case fans, power supplies and those fans. We can't ignore the GPU though. I mean, that's the thermal beast in all of these systems. On the gaming side, we're all familiar with those big plastic housing units with two or three fans on those. But on the pro graphics side, where we spend most of our time with like RTX 6000, 8000, 5000, 4000, whatever, those are typically really slim with just a little teeny fan on there. How much performance is being left on the table by insufficient cooling on GPUs? And kind of what's your view on whatA can do there or maybe not do there to help on the GPU side?
Starting point is 00:38:49 Well, so far as far as GPUs are concerned, we've been entirely focused on the consumer and gaming space. So we've teamed up with Asus. I think the first time was a couple of years ago by now for the 3070 and then we did a 3080, 4080 and a 4080 Super with them where we co-engineered the heatsink and put on our fans. This enabled us to provide quite a significant performance to noise efficiency boost over the stock cards. From my perspective, it's not so much performance opportunities being missed
Starting point is 00:39:34 with non-optimized consumer level cards. It's more the fact that they run louder than they would need to with a heavily tuned cooling solution. So I think the key benefit of choosing one of those Noctua Edition Aces cards is that you're getting the same performance as other, let's say, 4080 Super cards would give you, but you're running a few decibels quite. Is the GPU category, it's largely the differentiation in the designs is largely the cosmetic bits. Is there an opportunity one day to have a more universal kit that could work on you know anyone's 4090 versus just you know one brand or another or the designs too tied together?
Starting point is 00:40:26 I would say that's pretty difficult to pull off. First of all, we have an exclusive agreement with Asus. We're really focused towards rolling out more GPUs with ACES. I think there's a reason for the fact that the aftermarket GPU cooler market has pretty much vanished. There have been these universal one-size-fits-all heat sinks, but I think if people go the extra mile of remove buying an expensive let's say 4090 level card removing the cooler then they probably do it because they want to put on a DIY water cooling and like go really crazy on the full cooling solution we don't see much of a market for
Starting point is 00:41:21 people who take buy such an expensive card and probably void their warranty by removing the heatsink and then just putting on an air cooler. That's maybe a little bit better, but perhaps not all that much. There's just no way around the fact that maybe a little bit due to the push we gave with our first doctor edition aces car um the the stock coolers of today's consumer cpu uh gpus have come a long way they are now like on a much higher level than let's say uh five two to eight years ago yeah no it makes sense um it's just something that always bugs me especially when we get some of the pre-configured workstations in from the big brands, is that it just feels like sometimes
Starting point is 00:42:12 the whole system's a little loud, and then, especially when they stack several GPUs in there, you would think, man, it just seems like there's a more efficient way to do this. And it sounds like there is. It's just a lot of work to get there in terms of designing those fan solutions for GPUs. It's a lot of work to get there because you definitely want to optimize the heatsink for the fans or vice versa. It's not like you just put on the fans and see the heatsink and you'll get great results.
Starting point is 00:42:42 And for the pro-level workstation, multi-GPU setups, of course, space is also a problem. The reason why a consumer car, or one of the contributing reasons why consumer-level cars have become so much better from a thermal engineering perspective is because the heat sinks have grown in size. It's just quite straightforward that the larger heat sinks and larger diameter fans help and if you want to squeeze in four GPUs there's just no way you can go for a let's say four and a half slot design. Yeah, yeah, yeah well we're seeing some like HPZ for instance it's doing a little spacing.
Starting point is 00:43:27 So it's like two and a half between the slots to try to get a little more gap in there for airflow. And they're needing to be creative to get there. I don't want to lose track of one other relatively new piece of news that you guys are involved with the 45 drives guys who we know really well slightly off cameras are HL 15 which is I peek through the front grill looks distinctly coffee colored now where it was not before tell me about that relationship on on their home lab HL 15 product how you got involved there and kind of what you guys are doing in collaboration together.
Starting point is 00:44:08 We've had the HL15 on the radar for a while, but it was really Jeff Geerling's video where he used our Ampere Ultra heatsinks in an HL15 case that got the whole thing spinning. Because he swapped the stock 2U Empire Ultra cooler for our coolers and then realized, great, CPU cooling is virtually inaudible now, but the case fans are too loud so um he uh he asked us for a like a upgrade set of fans to to put into the hl15 and then he he did a video where he tried to measure the the acoustic improvement and was basically unable to measure anything with the updated version because it was like below his noise floor in the studio and at that point we felt like okay it's time to get in touch with 45 drives and make that
Starting point is 00:45:14 at least an upgrade option in their um in their configuration system and uh yeah that's what we uh thankfully managed to achieve so now when you when you an HL15, you can have it stock fitted with Noctua fans to run much quieter. Or even if you already have an HL15, you can buy an upgrade kit through 45 Drives to do that Noctua upgrade later on. Right, and so the case fans are six in there, right? Three on the front that I can see on mine and then three more internal. So are you offering a CPU heatsink with that as well or just those six case fans? No, the kit contains the six case fans. It contains inlet spacers for the three front fans because they sit very close to the intake grilles.
Starting point is 00:46:11 So without those spacers, we would see a lot of turbulence from interaction of the fan blades and the grills. So yeah, we're offsetting the fans by five millimeter to give them a little bit more breathing space, improves airflow, reduces turbulence and thereby improves acoustics. So you get the six fans, three inlet spacers, you get a eight channel fan hub, a PWM fan hub and a PWM fan controller in case your system doesn't allow for automatic PWM based fan speed control you can manually set the fan speed to the design level. So talk to me too about fan design I know you've got the new ones coming we talked a lot about this already Jay has got a question on on YouTube and this is one that we're starting to see a little more of, especially on designs that might use a radiator with a closed loop to handle
Starting point is 00:47:11 cooling in a system. They're putting these, in some cases, really thick fans, I don't know, two or three times as thick as one of your standard 120 mil fans in front of the radiators. Can you talk at all about from a design standpoint, does a thicker fan get you anything in terms of performance per decibel or anything along those lines? Is there any value in that or do you still feel really comfortable with your current designs? I would say yes and no. It's not like going thicker automatically gives you a better fan, but it certainly gives you additional options in fan design that can be leveraged to enhance performance.
Starting point is 00:48:01 That's for sure. But it's not as simple as doubling the thickness and there you have a better fan. That's not going to work. It also depends on fan size a lot, of course. I would say that for a 120 millimeter diameter, 25 millimeter death or thickness is something like a sweet spot where you can get get very good efficiency levels and it also depends on the fan speed so there's a good reason that high speed server or industrial fans in 120 millimeter size often come in let's say the 32 or 38 millimeter because when you increase rpm significantly then i would say the the benefit of using a thicker profile is also increasing well i tell you what this has been an amazing conversation i'm a little surprised we've been doing an hour pretty much on oh wow
Starting point is 00:49:03 on on air cooling and and talking about fans but you you know, it's cool. There's so many nerdy things here that make a big difference in people's lives. And I don't want to overvalue the fan in terms of its impact on the world. But for a lot of people, it makes a huge difference in their workflow and how they get things done and uh on that note uh i mean your product has been great we've been using them a ton more in in our lab in our environment and it's not just the fans it's the thermal paste that we've used kevin's been enamored by even the wipes because we switch out uh uh views and heat sinks the wipes that's his favorite thing we go through those faster than anything else now we used to use a paper towel and until you actually have a a nicely pre-wetted
Starting point is 00:49:50 little wipe that's so much cleaner and easier to use there's a lot of little things that make life easier so noctua has been a big part of that we appreciate your support in the lab and continue to look forward to working with you guys. But this has been awesome. Again, thank you so much for your time and for all the people that watch live, appreciate your questions in the chat. We tried to get to as many of those as we could, but if anything else comes up, feel free to reach out to these guys at Noctua. I'm sure they'd be happy to field your questions and help you make sure you're making the right investments in your PC and whatever else you've got going on.
Starting point is 00:50:30 But Jakob, thanks again. I really appreciate it. Thank you very much for having me. It was a pleasure. Thank you.

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.