-
Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!
Stalled Winter One -- 15.6L SFF case, 3090 Support, 3-slot GPUs, dual 280mm radiators, CFD Optimized Design
- Thread starter WinterCharm
- Start date
With the release of these new 12-16L cases, how do we think winter one will stand up?
Keen to see comparisons Vs nr200, meshlicious, Lian li q58 and the Dan C4.
Also was a shorter air only version of winter one modelled without the space for rads and how did/would this affect airflow?
No itx case about that has 4x140mm fans from my understanding.
Keen to see comparisons Vs nr200, meshlicious, Lian li q58 and the Dan C4.
Also was a shorter air only version of winter one modelled without the space for rads and how did/would this affect airflow?
No itx case about that has 4x140mm fans from my understanding.
Whatever happened to this?Also, expect an update later today
From what I’ve seen of their designs none have the forced airflow characteristics that makes W1 such an absurdly good case. Meshlicious has the most going for it as it’s the least restrictive of those and allows components to “breathe”, but even better than letting your components breathe is sticking them in a chimney with an absurd amount of pressure forcing heat out.
There’s a reason this case outperforms open-air test benches
I ordered a Meshlicious myself just because the small surface are is important to me but if I was indeed looking for a case all about the best performance, I'd probably have just gone for a W1 too.
Whatever happened to this?
Give me another 45 minutes, and It'll be up!
Had some website issues with hosting a 300MB PDF, and had to jump through a surprising amount of hoops to reduce the size of the user manual down to just ~ 13MB without reducing image quality.
----------------
EDIT:
UPDATE #12 is Live!
Inside, there's a link to the Winter One user's manual!
Also, for the very first time, I'm going to show you all the true extent of just how far we took CFD-driven optimization in Winter One
If you love airflow and cooling, you do not want to miss this one
Last edited:
Somehow our weeks appear to have had the same trajectory, although we were fighting battles on different fronts—spent a good 30 hours troubleshooting HTML -> PDF generation, but in some stroke of luck avoided any bloating issues.
@WinterCharm Hmmmm. Something seems off in the recommended airflow configuration pages. There’s a configuration that is both recommended *and* not recommended.
O_O
AHHHHHHH. I have no idea how that got swapped in on the second page. .-.
I will go fix it right now.
EDIT:
Fixed!!!! (Link has been updated in the KS post, as well).
The corrected page is below:
Last edited:
No problem! Glad I read this, just made sure to set my pciex16 slot to 3.0.
Loving all the info, really comprehensive.
One observation on the manual, are you deliberately leaving the rear IO panel off the motherboard when assembling?
Edit: Also any flow pictures with the dual 280 rad setup, with tubing?
One observation on the manual, are you deliberately leaving the rear IO panel off the motherboard when assembling?
Edit: Also any flow pictures with the dual 280 rad setup, with tubing?
@WinterCharm
I noticed the low QC rejection rate in your update, which seems to be really good for a premium case like yours.
Did you consider to sell B-Grade cases to lower the amount of parts which would need to be scrapped due to e.g. cosmetic flaws or slight tolerance discrepancies?
I mean, it's additional work for you / the factory but even if it's just a few cases, would be a shame to scrap if they are good to use (also considering the environmental footprint which you pointed out is important for your project) ...
I noticed the low QC rejection rate in your update, which seems to be really good for a premium case like yours.
Did you consider to sell B-Grade cases to lower the amount of parts which would need to be scrapped due to e.g. cosmetic flaws or slight tolerance discrepancies?
I mean, it's additional work for you / the factory but even if it's just a few cases, would be a shame to scrap if they are good to use (also considering the environmental footprint which you pointed out is important for your project) ...
1. That's an image we intend to fix. Just haven't gotten around to making a CAD model for a backplate, because it's so low on the priority list compared to things like supply chain and QC. (Winter One will fit your I/O plates just fine, you'll see in the build video next week)
2. There are some flow pictures with 280s and tubing. I just haven't released them. There's a lot more we did that I'm trying to keep internal to the company just to protect some of the IP we developed. For now I think we'll keep those internal, but I can say dual 280's will give you unbelievable thermal performance.
Sometimes, it's just nice to know that you're not alone. ^-^ Glad it worked out for you!Somehow our weeks appear to have had the same trajectory, although we were fighting battles on different fronts—spent a good 30 hours troubleshooting HTML -> PDF generation, but in some stroke of luck avoided any bloating issues.
And that's exactly the reputation I was going for with Winter One ^-^ so thanks for those kind words. Relentless improvement on airflow and noise until the case is simply known for excellent cooling.
A shorter air-only version would perform okay, but would require extensive retooling of the airflow design. If you read Update 12 you'll get a fair idea of just how detailed the airflow simulations got for Winter One. I didn't go into tons of detail about how I crafted the airflow for the case, but there are a ton of things dependent on the extra gap between the fans and components (from unrestricted exhaust, to some of the flow separation barriers.
While it could be done, it's no simple task. It would likely outperform other shorter cases for air cooling, but we already plan to do that with the SPK for Winter One -- that extra space will be used for giant heatsinks
Exactly. Couldn't have said it better.From what I’ve seen of their designs none have the forced airflow characteristics that makes W1 such an absurdly good case. Meshlicious has the most going for it as it’s the least restrictive of those and allows components to “breathe”, but even better than letting your components breathe is sticking them in a chimney with an absurd amount of pressure forcing heat out.
There’s a reason this case outperforms open-air test benches
So I stuck a sensor behind the GPU in Winter One, and fired up an AI training workload (really memory intensive, so should cause a 3090's GDDR6X memory get *very* hot (it's about as memory intensive as mining, if not more). Now, for those of you who may not know, the 3090FE has an issue - Nvidia did *not* properly thermal pad the rear of the card, and the memory there tends to get incredibly hot. Around 110C and then begins to throttle. in most cases, if you throw memory intensive work at the card.
During an ML workload, air temperatures between the GPU backplate and motherboard backplate were warm, hovering around 50C. (this is what the riser is exposed to). I updated HWInfo64 and can now confirm GPU memory is hovering at 100C (10C lower than most other cases) in AI training workloads. Case fans were running at 1300rpm, and the 3090FE fans were also running at 1350 RPM, as set by the GPU itself (54% PWM judging by the readout in afterburner). My case is set up for all-exhaust, which is the less optimal of the two airflow setups for cooling the central spine area. (In Bottom >> top airflow, components in the central spine area get even better cooling, read Update 12 if you're curious).
------
I will be modding my 3090FE with proper thermal pads, probably later tonight, because I really do not like the idea of memory temps hovering at 100C. Despite it being 4-6C lower than some other cases at stock, silicon should *not* be running that hot. I'll livestream the modding / teardown tonight at 8:30pm EST here, if anyone wants to hang out. It'll include a teardown of Winter One and putting the heatsink back on my m.2 SSD
Last edited:
There are be *very* few B-grade cases. 2 QC steps are done early in the factory (CMM + Optical) If something fails CMM, it's actually recycled and the raw metal is used again (our metals are verified / lab tested, it would be a gigantic waste to not re-use what was confirmed high grade recycled aluminum). If something fails finishing, it's usually caught by the Optical QC and re-finished properly. (again, cutting back on wasted time / effort and having to send stuff back for failing QC).
By the time they reach me, everything is inspected once again (this time by yours truly), for cosmetic issues that might slip past automated stuff. There are occasional failures here, but very very few. It will be quite a while before we have enough "bad" parts to create a sizable drop of B-stock.
I will be modding my 3090FE with proper thermal pads, probably later tonight, because I really do not like the idea of memory temps hovering at 100C. Despite it being 4-6C lower than some other cases at stock, silicon should *not* be running that hot. I'll livestream the modding / teardown tonight at 8:30pm EST here, if anyone wants to hang out. It'll include a teardown of Winter One and putting the heatsink back on my m.2 SSD
Stream is running a bit late. We should be good to start by 9pm EST. Technical issues. :3
FYI for others on doing thermal pad swaps. If you have an X570, then consider replacing the thermal pad if you can hear the fan.I will be modding my 3090FE with proper thermal pads, probably later tonight, because I really do not like the idea of memory temps hovering at 100C. Despite it being 4-6C lower than some other cases at stock, silicon should *not* be running that hot. I'll livestream the modding / teardown tonight at 8:30pm EST here, if anyone wants to hang out. It'll include a teardown of Winter One and putting the heatsink back on my m.2 SSD
This pad replacement on the strix, for example, is the real deal. Swapping out the crappy OEM pad to fujiploy caused the chipset temp to drop from 73C to 60C and the chipset fan went from being the dominant noise to inaudible.
@WinterCharm In your build do you use the standard Corsair SFX cables? Are there cable lengths that you recommend for the case? Could be a nice resource to include in the manual.
I did use the standard Corsair SFX cables in my build.
this is a great idea for something else we can include in the manual! I’ll see what we can do. Generally speaking any cable kit for a sandwich case like the Dan A4 / Ghost S1 or FormdT1 should be in the ballpark of proper cable lengths
also, here’s a few bonus image
taken during the stream.
and for those of you who showed up, thanks for hanging out
internet crapped out 2 hours in but, everything is running really well now, so I didn’t destroy my GPU ^_^
Were you able to install the WD heatsink on the m.2 installed on the back of the motherboard? Do you have before and after temps?
Also, what is the clearance back there for a m.2 heatsink? Curious to know, so that I can search for aftermarket ones that will fit.
Semi-concerned about sticking a hot SN850 into the rear m.2 slot on the motherboard.
Thank you!
Also, what is the clearance back there for a m.2 heatsink? Curious to know, so that I can search for aftermarket ones that will fit.
Semi-concerned about sticking a hot SN850 into the rear m.2 slot on the motherboard.
Thank you!
It was great seeing the tear down of the case and the GPU, even if there were some especially bum twitching moments with those two connectors not wanting to come out!for those of you who showed up, thanks for hanging out
internet crapped out 2 hours in but, everything is running really well now, so I didn’t destroy my GPU ^_^
