Stalled Hutzy XS — Ultra Compact Gaming Case (<4L)

[hardcore_gamer]

ASRock X99E-ITX/ac MB
Intel i7-5820K Haswell-E 6-core 3.3GHz CPU
Dynatron T318 copper / vapor chamber CPU cooler (27mm height)
Cryorig XT140 140mm slim PWM fan (13mm height)
G.Skill NT Series 16GB (2 x 8GB DIMMs) SDRAM DDR4 2133 CAS 15
Intel 600p 512GB M.2 / PCIe 3.0 x4 NVMe SSD
Gigabyte GeForce GTX 1070 mITX OC GPU

Maximum power draw (according to PCPartsPicker) would be about 364 watts…

6 cores / 12 threads, 16GB RAM, a 512GB NVMe system drive AND a GTX 1070 in a sub-4 liter package…?!?

Yes, please…!

So, thoughts towards overall chassis temps & such…?

Thermal Grizzly Kryonaut TIM to mount the CPU cooler & to replace the stock TIM on the GPU…

(…zip ties to secure 140mm fan to the CPU heat sink, jut like in the A4-SFX X99 build-up…)

If it's a gaming build, that CPU is overkill for a GTX 1070. Moreover, it has a TDP of 140W. Considering the size of this case, airflow options, and the rated TDP (135W) of Dynatron T318, the CPU is likely to run very hot without any overclocking headroom left to take advantage of this unlocked K-series processor. Also, noise output of Flex ATX PSUs typically increase exponentially with respect to power draw.

I'd personally select a 65W Core i7-6700 (non-K) for lower temperatures and lower noise.

 

[Boil]

If it's a gaming build, that CPU is overkill for a GTX 1070. Moreover, it has a TDP of 140W. Considering the size of this case, airflow options, and the rated TDP (135W) of Dynatron T318, the CPU is likely to run very hot without any overclocking headroom left to take advantage of this unlocked K-series processor. Also, noise output of Flex ATX PSUs typically increase exponentially with respect to power draw.

I'd personally select a 65W Core i7-6700 (non-K) for lower temperatures and lower noise.

Yeah, I was just busy playing a game of "What if…?"

If one could keep the temps low enough, and if any manufacturer could manage to make a ITX sized 1080; this chassis would make one heck of a super compact enthusiast rig…!

 

[Gautam]

Looks great, I expect designs like this to become more popular. Mini ITX GTX 1070's make full-size cards almost unnecessary. It's time for more designs to take advantage of the space savings. I saw you got some flack for using flex ATX, but its the most sensible choice. Pico PSU's can be bordering on too weak, and SFX unnecessarily large in most cases.

 

[Hahutzy]

Thanks. Yea the combination of mini cards and FlexATX pave a logical route to sizing down chassis.
The unknown to be explored is how to eject heat in a small volume, which your build put great emphasis on.

There's been a lot of doubt in using FlexATX in builds even before my time as a case designer, but it is seeing a trend of gaining acceptance as time passes, with more models that offer respectable wattages and online retailers making them easy to purchase for individuals.
Good eye on the Amazon version of the FSP400 by the way, it's the only offering I've seen that comes stock with a PCIE 8-pin connector.

 

[BirdofPrey]

I think it may be a bit soon to think of flex ATX as a route to compactness.
It's gotten powerful enough to run systems, yes, and the average power draw has declined recently, so they are viable, of course, but the market still isn't there, so there aren't very many options for flex leaving it very much a specialist market for SFF enthusiasts.
Of course, this is where SFX was just a few years ago, and now we've seen ATX sized systems with SFX PSU mountings at computex, while actual small(er) systems have stopped trying to make do with a full ATX PSU

 

[Kmpkt]

I agree with Birdofprey and at this point would probably even go one step further to say that I don't think that given my experience with the form factor that Flex ATX is going to be the future of small form factor building at any point.

I think the fundamental problem with Flex going forward is there isn't even the possibility of a good cooling solution given the constraints of the form factor. Without exception, adequate 40mm fans howl like a banshee at relatively low loads. Perhaps a second fan added in a push/pull config could reduce this problem, but I can't fathom where manufacturers are going to find the room to do this in contemporary Flex PSUs. The only other reasonable solution I can think of is the implementation of a Flex-Tall sort of configuration where you stack an 80mm fan like they do in some TFX units. As soon as you take this approach or alternatively duct in a larger fan to push air through the front, the size of the unit goes up substantially while the usability goes down, basically eliminating the power density advantage of the unit which is the whole point of using Flex in the first place. Probably the best idea I've played with for solving this problem is the idea of developing an externally mountable 120mm fan and duct mounted on the back of the PSU to pull air, possibly using a rear USB port as a power source (ie http://www.frozencpu.com/products/16995/ can run a fan on 5V power).

While I really love the idea of Flex-ATX in theory (enough that I bought quite a few of them myself including the FSP-500 Platinum, FSP-400 Gold and Seasonic 300 SUG), in practicality I just don't feel that it's going to get the level of adoption from consumers that SFX did for the aforementioned reasons. In my mind for Flex ATX to see any kind of mainstream adoption is going to take the following:

1.) Proper modularity (not sloppy harnesses a la Seasonic), at the very least equating that offered on the HDPlex 250. Expecting all but the most advanced user to be able to do custom harness work themselves is a pipe dream. Non-modular units are just crappy because you need to either deal with a tremedous amount of cable bulk or void your warranty clipping all the cables (I believe). That of course is completely trumped by the fact it's just a pain in the ass to do.

2.) Shrinking the height of internals enough or relocating taller elements elsewhere on the PCB so that you can top mount an 80mm fan for quieter cooling or as I mentioned earlier get a second 40mm involved to do a push/pull. Assuming that each fan could operate at 50% of max load in this situation (and therefore a fraction of the noise), I see this being the best possible fan solution.

3.) Efficiency needs to get up to Titanium to reduce cooling needs even further, thus bumping up the point at which the fan is disruptively loud. If/when GaN power supplies mature, Flex ATX will probably be much too large a form factor as the AC-DC component of the PSU would likely shrink by about 75%.

Not to be a Debbie Downer to those doing what are intended to be production builds based around the Flex ATX, but after spending a lot of time playing with them and being frustrated each time by either cable bulk, noise, or the inconvenience of recabling stock units I just feel that the form factor is too unfriendly for mass adoption. The case I am presently designing will have support for Flex, but only because it's so easy to implement without really changing the case (ie. not a classical placement attached the the rear IO plate).

 

[iFreilicht]

Very true. FlexATX is the niche right now that SFX was years ago. If SFX replaces ATX as the most popular kind of PSU, then we can talk about Flex becoming more popular.

 

[BirdofPrey]

What I'd REALLY like to see is some advancement of the PicoPSU and HDPlex type units. Both are great of small and low wattage systems, since they can provide decent power density, and separation of components gives better placement options in tight spaces, but, as it stands, they don't deal well with GPUs.

A Pico is more than enough to power a motherboard and slew of peripherals and fits in the motherboard area, but struggles with a GPU, and, at the moment, there's no good way to get a 12V DC input and split it amongst the two. HDPlex does a bit better since you can squeeze a bit more wattage out of them, but I think both lose out by not having very many options for internal AC-DC converters; while bricks are useful, not having to use one would be appreciated, and, again, being able to easily take a high wattage AC to 12V DC power supply, and delivering it to both the GPU and a DC-DC PSU for the motherboard and peripherals would be most helpful.

 

[jeshikat]

If FinSix can really reduce the size of the AC-DC half of the equation like @ilovelampshade is saying, then I think HDPLEX and PicoPSU type units will become extremely compelling.

 

[remixedMind]

Excellent case, i like it and i see my self buying one for my next build

 

[Kmpkt]

The FinSix solution is actually why I've chosen to wait on producing my case and leave Flex ATX as a solution. They're either going to have a high wattage ACDC brick or an internal by 2Q 2017 and if it does what they say it should, it'll be a game changer for SFF.

 

[Kmpkt]

What I'd REALLY like to see is some advancement of the PicoPSU and HDPlex type units. Both are great of small and low wattage systems, since they can provide decent power density, and separation of components gives better placement options in tight spaces, but, as it stands, they don't deal well with GPUs.

A Pico is more than enough to power a motherboard and slew of peripherals and fits in the motherboard area, but struggles with a GPU, and, at the moment, there's no good way to get a 12V DC input and split it amongst the two. HDPlex does a bit better since you can squeeze a bit more wattage out of them, but I think both lose out by not having very many options for internal AC-DC converters; while bricks are useful, not having to use one would be appreciated, and, again, being able to easily take a high wattage AC to 12V DC power supply, and delivering it to both the GPU and a DC-DC PSU for the motherboard and peripherals would be most helpful.

What I'd really like to do if anyone is interested is look at getting one of the Pico manufacturers or HD Plex to do a unit to the community's spec. I don't know if anyone here has the knowledge to help engineer something like this, but I think if a Pico or Pico/board hybrid were done specifically for this application (8 Pin GPU, 8 Pin EPS, SATA Power) it could be pretty awesome. What I'd like to see is something like the Dual DC solution Josh used in one of his build videos over on NFC's youtube channel. In his build, the Pico I believe was used to carry the 24 pin and 8 pin EPS from an HDPlex like board and then the GPU and HDD could be plugged into the DC-DC board. Would eliminate the clutter of the 24 pin cables without adding additional stress to the Pico of running the GPU and additional modules.

 

[Hahutzy]

Yes, investing the time and resources to develop on an unproven standard always comes with a risk of it not taking off. Even with technology such as the FinSix Dart hinting at the rise of HDPLEX / PicoPSU, we could be having the same conversation comparing it to the next big (small?) thing.

In particular though, I don't like the fact that Dart is built upon patented technology and effectively has a monopoly until someone comes up with a different way to achieve the same result. So for now, you end up with a system that has DC-ATX inside (HDPLEX / PicoPSU) and either a standard AC-DC laptop brick (which is an available strategy today), or you buy into the success of FinSix and only FinSix, as opposed to other PSU form factors that are open and supported by multiple companies.

Now don't get me wrong, I don't take any of your doubts lightly; you're all experienced people in the subject, and I'm sure @ilovelampshade has had experience with FlexATX before I even knew about the standard. But there are several circumstances that only recently arose that gave me some confidence in the standard:

1) The general trend with Intel is to value efficiency over significant performance jumps (reported here, for example).
In contrast to Haswell's 84W standard with "low-wattage" chips at 65W. Skylake comes out with 65W as the standard TDP. We will continue to see TDP drop.

2) The trend of GPUs getting smaller and using less power while gaining performance boosts. This is advantageous for SFF as a whole, until GPUs reach a point that they are integrated onto the motherboard -- at which point NUCs will take over SFF.

3) The combination of the above makes for a situation where we stand now -- the FlexATX standard is able to supply much more wattage than what a build needs. For example, a i5-6400 + GTX 1060 build uses 240W, just a bit under 50% of a FSP500 unit. Or with a FSP400, that is 60%. As we know, fan curves are exponential, so the lower % load tend to let the fan spin much slower (if at all). This means that for the first time, we may be able to see the FlexATX run a respectable system without needing to stress the 40mm fan (and our ears for that matter).

4) The FlexATX unit that is both consumer-pc-ready with an 8-pin PEG, and is accessible to consumers happened. It is the FSP400 on Amazon that @Gautam presented to us in his build log. This is both insignificant and significant, depending on how you look at it. Yes, it is only 1 model, by 1 company, sold at 1 outlet. But it happened, and definitely for a reason. Is this the start of a trend? Perhaps, if there's demand. And for all intents and purpose, I want to do my part in trying to make that trend happen.

On a lighter note, Hutzy XS supports both FlexATX and HDPlex mounting, so I will be okay

 

[Gautam]

4) The FlexATX unit that is both consumer-pc-ready with an 8-pin PEG, and is accessible to consumers happened. It is the FSP400 on Amazon that @Gautam presented to us in his build log. This is both insignificant and significant, depending on how you look at it. Yes, it is only 1 model, by 1 company, sold at 1 outlet. But it happened, and definitely for a reason. Is this the start of a trend? Perhaps, if there's demand. And for all intents and purpose, I want to do my part in trying to make that trend happen.

Thank you for thanking me for finding it, but it was just sitting there in plain sight, on the world's biggest e-tailer, by one of the biggest PSU OEMs. (It was shipped directly by FSP as far as I remember) I was surprised that such a user-friendly variant existed out of the box. In fact, I was just about to pull the trigger on one on eBay with some industrial spec when I figured I should give Amazon a second look. Glad I did. I was expecting that I'd have to solder a PEG connector myself if not even EPS... luckily I needed nothing else.

The SFX and FlexATX specs are both close to 20 years old at this point. I have no evidence for this at all, but my gut says that FlexATX is about as common as SFX in consumer desktops. Namely stuff like Dell Optiplexes and the like, so its not at all an unproven standard. In the enthusiast community, it looks like SFX is overwhelmingly more popular.

I agree with both you guys about the pros and cons of flex. I wouldn't be able to tolerate it in my build with the stock fan. Instead, I placed a 92mm fan by the blowhole, which is pretty ghetto but happens to work fine. The route I was originally planning was to crack it open entirely and put a fan on the top, or cut the housing open. All these options are very user-unfriendly.

But in the end, I think you have it right. The progression of technology implies that if you're heading in the direction of smaller components, with less power draw and less heat output, you're on the right side of history. It seems like @iFreilicht is encountering good noise levels with the FSP500. Not too long ago, you had to have a full-size ATX PSU if you wanted to power a high-end rig quietly and efficiently. Now SFX PSUs are readily available with huge power output, and concurrently, components are drawing far less. Time is a flat circle.

I guess that also tells us that HDPlex type solutions aren't going anywhere either...

 

[EdZ]

In particular though, I don't like the fact that Dart is built upon patented technology and effectively has a monopoly until someone comes up with a different way to achieve the same result.

FinSix are using Gallium Nitride based ICs to increase power density. GaN is new for consumer power electronics, but it's been used in high-power amplifiers for radio equipment for easily a decade (and even longer for military RADAR applications). The main barrier to more widespread use has been the large additional expense over regular Silicon MOSFETs for what is only a modest increase in capability.

 

[CC Ricers]

So it's more of a question of whether or not the profit margins are worth the development costs using GaN for power electronics, and most companies have decided it's not yet worth looking into for getting into that market.

 

[Hahutzy]

What about their so called "high frequency power technology"?

http://finsix.com/power-platform/

Is it just marketing bull?

 

[hardcore_gamer]

What about their so called "high frequency power technology"?

http://finsix.com/power-platform/

Is it just marketing bull?

They use higher switching frequency so that size of transformers and filtering elements are reduced. To reduce switching losses, they're probably using AlGaN/GaN HEMTs which are more efficient than silicon MOSFETs at high switching frequencies.

 

[iFreilicht]

They use higher switching frequency so that size of transformers and filtering elements are reduced. To reduce switching losses, they're probably using AlGaN/GaN HEMTs which are more efficient than silicon MOSFETs at high switching frequencies.

Why isn't everyone doing this then?

 

[hardcore_gamer]

Why isn't everyone doing this then?

Because III-V group semiconductor devices are expensive compared to their Silicon counterparts.