Stalled Winter One -- 15.6L SFF case, 3090 Support, 3-slot GPUs, dual 280mm radiators, CFD Optimized Design

[WinterCharm]

I am wondering if perhaps it's possible to adjust/reduce the mobo standoffs by 2- 3 mm to allow the popular Noctua L12S to fit without having to bend it. Who knows how long the Ghost S1 edition L12 will be around. That and the L12S with it's 120mm fan is demonstrably better than the Ghost edition.

Standoffs cannot really be shorter, and SPK will solve the air-cooled heatsink problem.

 

[G23 Mr Gimp]

I've been following the progress on this project for some time and have been really impressed with the attention to detail and flow testing.
Looking forward to the KickStarter as you'll fully have my support !

 

[WinterCharm]

I've been following the progress on this project for some time and have been really impressed with the attention to detail and flow testing.
Looking forward to the KickStarter as you'll fully have my support !

Thank you so much. that means a lot.

Definitely check out the website and sign up to get an email notification when the Kickstarter launches

 

[Alloy Craft]

None of the CFD backs this up, so No. Airflow in this case was meticulously designed with CFD (see the link), and having these holes lower is better balancing the intake flow in all-exhaust situations, where you want to maximize radiator access to clean air, and minimize restriction. Components near the holes would severely restrict practical flow in those situations, so having a few extra rows above / below makes for better intake. I added 2 more rows of holes, over the original design, despite the increase in cost, because the benefits were clear to airflow, without affecting the airflow delivered to components in all-exhaust.

There were also clear benefits to bottom >> top airflow configurations with the slightly expanded holes, where the airflow creates a jacket delivering fresh airflow into the intake of any fans that are pulling air in through perforated panels, while also disrupting any potential of recirculated hot air from a top-down cooler being sucked back into the intake of a part like NH-L12 or a GPU cooler. If you don't do this, especially for GPU's that are taking air in from the side and blowing it top / bottom, you get a lot of recirculation. This is one of the major flaws of sandwich cases that have weak or nonexistent airflow, and why just one or two fans amounts to a huge drop in temperatures.

Why are all the simulations you show missing the chassis components? The simulations seem to be just airflow velocity in an empty case?

 

[WinterCharm]

Why are all the simulations you show missing the chassis components? The simulations seem to be just airflow velocity in an empty case?

Great observation! Yes, the simulations I posted *do not* include CPU and GPU coolers. The Tl;Dr is: they don't affect *global airflow* in the case in a meaningful way. However, if you want the full explanation, please read on! ?

----------------​

Initially, I set up my models with the assumption that these parts *did* affect global flow. Because of the variation in product geometry, I chose the *least* aerodynamic shape for CPU / GPU heatsinks, for the early modeling of this case... A Rectangular "Brick" (think: Radeon VII) was used to model the GPU cooler, and a similar, down firing (but Square) Brick was used to model the CPU cooler (think: Noctua L9x65).

Now, the reason these heatsinks all have fans strapped to them, is that fin density is high and drag is therefore very high, as the more surfaces you have the more flow loses kinetic energy due to boundary layer interactions. The Fan imparts KE and P to the fluid, and the fluid expends ∆P & KE to move through the heatsink.

In all my testing, I found that with most CPU and GPU coolers, the kinetic energy / pressure imparted by the fan *mostly* expended when getting air through the heatsink, with velocities approaching the low 15-20cm/s at the heatsink "exhaust". So, in modeling these fans, while they act as an "intake" with a wide cone, and performance of the heatsinks was affected by that intake air being a clean, steady supply, their affect on airflow within the case was mostly confined to their local area, rather than any meaningful global effect, except for diverting flow, based on the volume they occupied. Additionally, the warm air that was exhausted followed the "global" flow of air, and was carried out with the global flow in the case.

What mattered for the performance of these components was 2 main things: (1) -- access to a consistent supply of COOL AIR, and (2) avoiding recirculation of HOT air. Remember my discussion about the dangers of all-intake creating recirculation in SFF cases? That problem is amplified when your system becomes a heat sink with a fan on it, that's only a few cm in length! This is why some heat sink designs use a shorter pathway with 2 separate fan+fin stacks (NH-D15, for example), with each having higher exhaust velocity, or if they use a single fan they keep the pathways short to ensure enough exhaust velocity (NH-U14S, for example), and if they want a longer pathway for more "compact" cooling performance, they add an outtake fan (NH-U12A, for example) to push exhaust away, and create a larger pressure gradient across the long pathway.

To address these problems, I designed Winter One to ensure that (1) The Airflow in the case provides cool air to any intake fans, and (2) airflow in the case was moving quickly enough that it carries away the "slow and hot" flow coming from these heatsinks. So the solution is this: IF you have a solid side panel, make sure there is enough clearance (1cm, recommended) between that intake fan, and the side of the case... IF you want to use thicker components (say, a 3-slot GPU, or an intake fan very close to a panel), Go with the Perforated Side Panel Option. They were designed so the airflow actually bows out of the case a bit (while still maintaining good flow within the case itself, about a 10cm/s *drop* in flow (from 80 to 70) cm/s but that flow moving over the outside of the panels *consistently* provides a clean intake for something like a triple slot air-cooled GPU... Meanwhile flow inside the case overcomes the "local" behavior of the slow exhaust air coming from those heatsinks, and carries it to the top exhaust. In the case of a blower card, It simply acts like a "sink" pulling in some flow from all directions, but again, very much a local effect, extending In a few CM in every direction.... it doesn't really interfere with, or interrupt the global flow in the case.

The best way I can explain local vs global flow, is you blowing at some smoke / steam coming from boiling pasta... Your blowing overcomes the tendency of the steam to rise in a straight line... so you are pushing the water vapor to a different location on one axis, but *even as you do, the steam continues to rise* -- these behaviors are exhibited together, but the flow with the highest velocity becomes the predominant behavior. This is how flow is summed up on exhaust air at low velocities coming from a CPU / GPU heatsink...

So, after a few iterations of this, and realizing that these systems had nothing to do with the global flow in the case, I excluded them from *most* simulations, and occasionally did a few sanity check simulations after major redesigns, to ensure that this behavior hadn't changed.

---------------​

I will not be posting any of the CFD results that include components inside, to try and protect some of the IP around how I managed airflow around those components, and optimized the various configurations and layouts to account for varied local flow based on what parts were installed.

At the same time, the manual will include suggested airflow setups for best performance, so that users building in the case can benefit from this work. ?

 

[dotnetapp]

None of the CFD backs this up, so No. Airflow in this case was meticulously designed with CFD (see the link), and having these holes lower is better balancing the intake flow in all-exhaust situations, where you want to maximize radiator access to clean air, and minimize restriction. Components near the holes would severely restrict practical flow in those situations, so having a few extra rows above / below makes for better intake. I added 2 more rows of holes, over the original design, despite the increase in cost, because the benefits were clear to airflow, without affecting the airflow delivered to components in all-exhaust.

There were also clear benefits to bottom >> top airflow configurations with the slightly expanded holes, where the airflow creates a jacket delivering fresh airflow into the intake of any fans that are pulling air in through perforated panels, while also disrupting any potential of recirculated hot air from a top-down cooler being sucked back into the intake of a part like NH-L12 or a GPU cooler. If you don't do this, especially for GPU's that are taking air in from the side and blowing it top / bottom, you get a lot of recirculation. This is one of the major flaws of sandwich cases that have weak or nonexistent airflow, and why just one or two fans amounts to a huge drop in temperatures.

Hmm did you maybe simulate if it makes a difference if you make a push or pull config with radiators?
Or is it completly neglectable ?

EDIT: Just saw your last post i hopeto read than what the differences are

 

[Ilyu]

have you thought about enhancing the radiator mounting plate to be able to shift the 140mm or 280 rads like with the 120 holes? I think something like the corsair xr5 280 (310x140x30) would fit quiet nice I guess? here

 

[WinterCharm]

Hmm did you maybe simulate if it makes a difference if you make a push or pull config with radiators?
Or is it completly neglectable ?

Pull on radiators is typically *marginally* better (consistently by around 1.6%) vs Push. Typically you net around 1ºC or 0.6ºC better temperatures as a result...

have you thought about enhancing the radiator mounting plate to be able to shift the 140mm or 280 rads like with the 120 holes? I think something like the corsair xr5 280 (310x140x30) would fit quiet nice I guess? here

This was actually requested by @Qzrx who is in the Beta Program and asked for the exact feature, for the exact radiator. I can confirm it's being changed. ^-^

 

[REVOCCASES]

Great observation! Yes, the simulations I posted *do not* include CPU and GPU coolers. The Tl;Dr is: they don't affect *global airflow* in the case in a meaningful way. However, if you want the full explanation, please read on! ?

----------------​

Initially, I set up my models with the assumption that these parts *did* affect global flow. Because of the variation in product geometry, I chose the *least* aerodynamic shape for CPU / GPU heatsinks, for the early modeling of this case... A Rectangular "Brick" (think: Radeon VII) was used to model the GPU cooler, and a similar, down firing (but Square) Brick was used to model the CPU cooler (think: Noctua L9x65).

Now, the reason these heatsinks all have fans strapped to them, is that fin density is high and drag is therefore very high, as the more surfaces you have the more flow loses kinetic energy due to boundary layer interactions. The Fan imparts KE and P to the fluid, and the fluid expends ∆P & KE to move through the heatsink.

In all my testing, I found that with most CPU and GPU coolers, the kinetic energy / pressure imparted by the fan *mostly* expended when getting air through the heatsink, with velocities approaching the low 15-20cm/s at the heatsink "exhaust". So, in modeling these fans, while they act as an "intake" with a wide cone, and performance of the heatsinks was affected by that intake air being a clean, steady supply, their affect on airflow within the case was mostly confined to their local area, rather than any meaningful global effect, except for diverting flow, based on the volume they occupied. Additionally, the warm air that was exhausted followed the "global" flow of air, and was carried out with the global flow in the case.

What mattered for the performance of these components was 2 main things: (1) -- access to a consistent supply of COOL AIR, and (2) avoiding recirculation of HOT air. Remember my discussion about the dangers of all-intake creating recirculation in SFF cases? That problem is amplified when your system becomes a heat sink with a fan on it, that's only a few cm in length! This is why some heat sink designs use a shorter pathway with 2 separate fan+fin stacks (NH-D15, for example), with each having higher exhaust velocity, or if they use a single fan they keep the pathways short to ensure enough exhaust velocity (NH-U14S, for example), and if they want a longer pathway for more "compact" cooling performance, they add an outtake fan (NH-U12A, for example) to push exhaust away, and create a larger pressure gradient across the long pathway.

To address these problems, I designed Winter One to ensure that (1) The Airflow in the case provides cool air to any intake fans, and (2) airflow in the case was moving quickly enough that it carries away the "slow and hot" flow coming from these heatsinks. So the solution is this: IF you have a solid side panel, make sure there is enough clearance (1cm, recommended) between that intake fan, and the side of the case... IF you want to use thicker components (say, a 3-slot GPU, or an intake fan very close to a panel), Go with the Perforated Side Panel Option. They were designed so the airflow actually bows out of the case a bit (while still maintaining good flow within the case itself, about a 10cm/s *drop* in flow (from 80 to 70) cm/s but that flow moving over the outside of the panels *consistently* provides a clean intake for something like a triple slot air-cooled GPU... Meanwhile flow inside the case overcomes the "local" behavior of the slow exhaust air coming from those heatsinks, and carries it to the top exhaust. In the case of a blower card, It simply acts like a "sink" pulling in some flow from all directions, but again, very much a local effect, extending In a few CM in every direction.... it doesn't really interfere with, or interrupt the global flow in the case.

The best way I can explain local vs global flow, is you blowing at some smoke / steam coming from boiling pasta... Your blowing overcomes the tendency of the steam to rise in a straight line... so you are pushing the water vapor to a different location on one axis, but *even as you do, the steam continues to rise* -- these behaviors are exhibited together, but the flow with the highest velocity becomes the predominant behavior. This is how flow is summed up on exhaust air at low velocities coming from a CPU / GPU heatsink...

So, after a few iterations of this, and realizing that these systems had nothing to do with the global flow in the case, I excluded them from *most* simulations, and occasionally did a few sanity check simulations after major redesigns, to ensure that this behavior hadn't changed.

---------------​

I will not be posting any of the CFD results that include components inside, to try and protect some of the IP around how I managed airflow around those components, and optimized the various configurations and layouts to account for varied local flow based on what parts were installed.

At the same time, the manual will include suggested airflow setups for best performance, so that users building in the case can benefit from this work. ?

When do you expect we will see actual thermal test results? Especially interested in the air cooled setup with closed side panels and different fan setups (e.g. 2 exhaust vs 2 intake vs 2 in plus 2 ex).

 

[WinterCharm]

When do you expect we will see actual thermal test results? Especially interested in the air cooled setup with closed side panels and different fan setups (e.g. 2 exhaust vs 2 intake vs 2 in plus 2 ex).

Beta Testers are currently building in the case, and trying out a variety of hardware configurations. They'll be posting their first impressions and thermal results in the coming week.

In the meantime, here are some preliminary results from my own testing:

DUST

So this is a quick teardown of a look at dust in the system after running for 1 month continuously (24/7, doing F@H when not being used) in pull-only (negative pressure) setup with 2 x 140mm fans. This confirms what I was saying earlier -- if Airflow is good, you won't get tons of dust trapped in a case, even one that's operating in the "worst case" scenario of a fully negative pressure build.

This was sitting on the floor, by the way, not even on the desk.


EARLY THERMALS

Here's some Early Thermal Testing with FurMark CPU + GPU stress tests. The system configuration is as follows. Keep in mind this is a relatively low spec system by modern standards, simply because I tend to do most work on the go on my laptop.

* i7 4790K (Stock Clocks)
* Cryorig C7G + Noctua 92mm x 15mm fan
* Gigabyte H97 ITX motherboard
* 16GB DDR3 2666Mhz RAM
* Samsung 840 Pro SSD (512GB)
* Radeon RX 570 (Sapphire Pulse)
* 2 x 140mm fans set to EXHAUST (top) (motherboard only has CPU + 1 fan header)
* Corsair SF450 power supply

-------------

Case Fans were set on a gentle fan curve between 850 >> 1200 RPM based on system temperatures. CPU Fan Curve was set to ramp up slowly, and then max out when the CPU touches 75ºC.

Here's what the ramp up, sustained loads, and cool down look like, including plots of the power draw and fan speeds of all components. Feel free to open up the images and zoom in, I took the shots at full resolution.

Temperatures settle in at an incredibly reasonable 70ºC on the GPU and 81ºC on the CPU, with case fans running at just 1060 RPM... CPU fan is kicking into high gear at 2500RPM, and I could probably reasonably slow that down without much impact on CPU thermals, but I wanted to be safe with the first round of thermal testing.

This is a pretty good look at how quickly the system is able to cool down after being at steady state. One of the most important things this tells us is how quickly cool air is being supplied and hot air is being removed. The rapid decline in temperature is a good indicator that everything has fresh air supply.

Keep in mind FurMark is an ABSOLUTE WORST CASE scenario for thermals. Here's how the system did with a more "realistic" workload in Geekbench 5:

Winter One has incredibly tame thermals for an SFF system, and is holding its own against well-ventilated ATX desktops. My system Idles at 38ºC for the CPU and 30ºC for the GPU. Most Notably, the CPU results are 6ºC cooler than an open air test bench, and the Sapphire RX 570 thermals were 4ºC below an open air test bench

If anyone wants to see thermals under gaming workloads, let me know, and I'll fire up BFV, KSP, or Mordhau and record that information, too

The system stays remarkably and pleasantly quiet even under load. The case fans are inaudible, and the only thing you can hear is the CPU fan when it jumps up to 2500 RPM in FurMark. Otherwise, it's nice and quiet. Keep in mind this is with Bottom >> Top Airflow and just 2 fans (so not even push/pull) and represents the worst case scenario of cooling mid range system from a few years ago, in Winter One.

I also plan to do some testing with OCCT in the next few days

 

[CluckyTaco]

DUST

So this is a quick teardown of a look at dust in the system after running for 2 weeks continuously (24/7, doing F@H when not being used) in pull-only (negative pressure) setup with 2 x 140mm fans. This confirms what I was saying earlier -- if Airflow is good, you won't get tons of dust trapped in a case, even one that's operating in the "worst case" scenario of a fully negative pressure build.

This was sitting on the floor, by the way, not even on the desk.

I know you mentioned that there isn't much dust accumulation from your initial testing. Would the case allow for installing filters like DEMCiflex magnetic ones? I'm asking as it would help with easy maintenance for us dusty folks. Also appreciate your thermal results for synthetic workloads. I signed up on the website for notifications for case updates. Kickstarter here I come

 

[Ayfid]

In order to keep complexity down, we're going to have to start with just anodized black. However, if the KS campaign passes 300 units, I'll unlock a silver option as an additional perk, and everyone who placed an order can choose their color via survey around when the campaign ends

This prototype is 2-toned because I was testing out both colors and vetting different factories for their parts quality.

I have to say... I really want it in "prototype 2-tone". ?

This was actually requested by @Qzrx who is in the Beta Program and asked for the exact feature, for the exact radiator. I can confirm it's being changed. ^-^

There are a few cases I have been eyeing for my next build, but if the Winter One can fit 2x 280mm radiators then it really doesn't have any contenders for me.

 

[DmanX]

Really liking the gap between the back of the GPU and the back of the PSU. Looks like >20mm. Perfect for the RTX 3080.

 

[rayuma]

Do you think custom cables for a Ghost S1 would be able to be used here as well? The distance between the power supply and motherboard look similar.

 

[WinterCharm]

Do you think custom cables for a Ghost S1 would be able to be used here as well? The distance between the power supply and motherboard look similar.

It is still a sandwich layout (with a tiny bit of variation) so I don't see why not. The cable runs are quite short

 

[REVOCCASES]

Beta Testers are currently building in the case, and trying out a variety of hardware configurations. They'll be posting their first impressions and thermal results in the coming week.

In the meantime, here are some preliminary results from my own testing:

DUST

So this is a quick teardown of a look at dust in the system after running for 2 weeks continuously (24/7, doing F@H when not being used) in pull-only (negative pressure) setup with 2 x 140mm fans. This confirms what I was saying earlier -- if Airflow is good, you won't get tons of dust trapped in a case, even one that's operating in the "worst case" scenario of a fully negative pressure build.

This was sitting on the floor, by the way, not even on the desk.


EARLY THERMALS

Here's some Early Thermal Testing with FurMark CPU + GPU stress tests. The system configuration is as follows. Keep in mind this is a relatively low spec system by modern standards, simply because I tend to do most work on the go on my laptop.

* i7 4790K (Stock Clocks)
* Cryorig C7G + Noctua 92mm x 15mm fan
* Gigabyte H97 ITX motherboard
* 16GB DDR3 2666Mhz RAM
* Samsung 840 Pro SSD (512GB)
* Radeon RX 570 (Sapphire Pulse)
* 2 x 140mm fans set to EXHAUST (top) (motherboard only has CPU + 1 fan header)
* Corsair SF450 power supply

-------------

Case Fans were set on a gentle fan curve between 850 >> 1200 RPM based on system temperatures. CPU Fan Curve was set to ramp up slowly, and then max out when the CPU touches 75ºC.

Here's what the ramp up, sustained loads, and cool down look like, including plots of the power draw and fan speeds of all components. Feel free to open up the images and zoom in, I took the shots at full resolution.

Temperatures settle in at an incredibly reasonable 70ºC on the GPU and 81ºC on the CPU, with case fans running at just 1060 RPM... CPU fan is kicking into high gear at 2500RPM, and I could probably reasonably slow that down without much impact on CPU thermals, but I wanted to be safe with the first round of thermal testing.

This is a pretty good look at how quickly the system is able to cool down after being at steady state. One of the most important things this tells us is how quickly cool air is being supplied and hot air is being removed. The rapid decline in temperature is a good indicator that everything has fresh air supply.

Keep in mind FurMark is an ABSOLUTE WORST CASE scenario for thermals. Here's how the system did with a more "realistic" workload in Geekbench 5:

Winter One has incredibly tame thermals for an SFF system, and is holding its own against well-ventilated ATX desktops. My system Idles at 38ºC for the CPU and 30ºC for the GPU. Most Notably, the CPU results are 6ºC cooler than an open air test bench, and the Sapphire RX 570 thermals were 4ºC below an open air test bench

If anyone wants to see thermals under gaming workloads, let me know, and I'll fire up BFV, KSP, or Mordhau and record that information, too

The system stays remarkably and pleasantly quiet even under load. The case fans are inaudible, and the only thing you can hear is the CPU fan when it jumps up to 2500 RPM in FurMark. Otherwise, it's nice and quiet. Keep in mind this is with Bottom >> Top Airflow and just 2 fans (so not even push/pull) and represents the worst case scenario of cooling mid range system from a few years ago, in Winter One.

I also plan to do some testing with OCCT in the next few days

Looks great, thank you.

This totally confirms my findings I made during development of the RCC-BIG1 and MID1 cases: a optimized wind-tunnel-like air flow concept does work very well and you do not need many fans or holes all over the case to achieve good thermals.

Now I am pretty curious if adding more fans would really improve thermals significantly. From my experience you somewhere reach a point where more airflow does not improve thermals any more because the GPU and CPU coolers (heatsinks) become the limiting factor.

Just for reference, here are my stress-test thermals with the BIG1 which only uses one slim 120mm exhaust fan:

• Ryzen R7 3800X (@105W) Tmax = 80°C
• RTX2080 SUPER OC, (@250W) Tmax= 80°C

Considering you are using 2 x 140mm fans and having similar temps like me, the CPU cooler seems to be the limiting factor here as Winter One has definitely enough air flow.

 

[WinterCharm]

I know you mentioned that there isn't much dust accumulation from your initial testing. Would the case allow for installing filters like DEMCiflex magnetic ones? I'm asking as it would help with easy maintenance for us dusty folks. Also appreciate your thermal results for synthetic workloads. I signed up on the website for notifications for case updates. Kickstarter here I come

Yes, between the fan/radiator plate and the top / bottom plates, you can cut a filter to shape and place it there, and it will just clamp into place. For the perforated side panels, you'll be able to add some magnetic tape and attach DEMCiflex filters in that area. So if you want to filter airflow through the case, it is certainly possible, and handy if you have lots of pets that shed or something

Looks great, thank you.

Now I am pretty curious if adding more fans would really improve thermals significantly. From my experience you somewhere reach a point where more airflow does not improve thermals any more because the GPU and CPU coolers (heatsinks) become the limiting factor.

Just for reference, here are my stress-test thermals with the BIG1 which only uses one slim 120mm exhaust fan:

• Ryzen R7 3800X (@105W) Tmax = 80°C
• RTX2080 SUPER OC, (@250W) Tmax= 80°C

Considering you are using 2 x 140mm fans and having similar temps like me, the CPU cooler seems to be the limiting factor here as Winter One has definitely enough air flow.

My particular CPU cooler is the definitely the limiting factor here. However, larger coolers such as the L12S and SPK will make use of the airflow in this case, and same goes for radiators.

 

[WinterCharm]

Some Quick Updates

1. I'll be working on the Winter One website today, so you may see it go down / etc. I'll make a post with changes once it's back up
2. Beta testers have already started to build in the case We're getting ready to collect some thermal data soon.
3. I've already confirmed that some dimensional changes will be made to support the RTX 3090. I'll be finalizing and announcing the exact changes soon.
4. I'm working with the Beta Testers to build out a Parts Compatibility Tables, and made sure things can fit.
5. There were a few minor issues with some of the Beta Test parts, and some design tweaks are being made so this can be addressed. These are mostly QoL and 'niceness' updates, which some Beta Testers have asked for.

I thought I'd go ahead and take a moment to introduce you to the beta testers for Winter One.

@Qzrx
@gelbetron
@JDragon

While they're under NDA, and can't tell you everything, I'll be relaxing the NDAs regularly (so they can disclose more and more) leading up to the KS so you can all talk to them about what it's like building in Winter One Most of them are taking some time to build out their system this weekend, and it's going quite well so far, so if you have questions about the build process, feel free to ask them.

Currently, Beta Testers are allowed to talk about:

1. Building in Winter One the overall Build Process
2. Post Build Photos of the case itself, and the system they're making
3. Issues they've encountered and fixes coming in the final version.

The selection process for the case had multiple layers, and there were a lot of applicants, which is why it took me a while to select Beta Testers. If you were unable to beta test the case this time, don't feel bad. Almost all the applications were very good, and it's clear that lots of effort was put in. Thank you all for being so passionate about helping me test a new design.

 

[Demthios]

Awesome news now I just wanna see build pics! Come-on you beta testers ?

 

[Qzrx]

A little Sunday-night sneak preview of some things to come... ?

3900X
2080ti
SF750
Dual 240mm radiator custom loop
Aquaero6