Exactly. The rigidity of the case in UNCOVERED large expanded and extra-large expanded mode is what I'm curious about. The primary appeal for the is the case is the expandability and contraction of the enclosure according to the build. Going with the uncovered options maintain the maximum flexibility. @diceboii13
Stalled DICE CASES - FLOW : expandable SFF case for watercooling enthusiast (built in distroplate)
- Thread starter diceboii13
- Start date
Exactly. The rigidity of the case in UNCOVERED large expanded and extra-large expanded mode is what I'm curious about. The primary appeal for the is the case is the expandability and contraction of the enclosure according to the build. Going with the uncovered options maintain the maximum flexibility. @diceboii13
Ah I see the point. I just talked about the covered ones. I remember when I did the testing with the first prototype it was an uncovered one, and there was no covered option. It held well, definitely not as strong and rigid like the covered ones but if you not kicking or throwing the case it can stay nice and stiff in the desk. It's like a bit of a compromise, but the owner can change his expand any time. I'm gonna make some tests with them tomorrow and share it here.
Talking about the large vs extra large expands.. I saw the extra large a bit overkill, considering there is not much extra cooling potential in 45mm rads. A top end PC today can run well with 1 or 2 280*30 rads. Thicker rads maybe can add some headroom for overclocks but I didn't see the point where that extra 100mhz can make any benefit. And the numbers talk, the uncovered and the extra large expands are the smallest amount in the orders.
For me the expandability is the core value proposition of Dicecase. Since I plan to experiment with different setups. Dicecase would allow me to experiment with different cooling solutions while keeping all the parts contained within the case. Without the expandability, there really little functional differentiator from other 12~15L cases.
There are a couple cases upcoming like the WinterOne that are targeting the 2x240 market, but if you want 2x240 you could just get an Ncase M1. I see this case as the best case for 280mm radiators, because the SM580 has no room for a pump or res, and the Cerberus/X is too big.
indeed it's interesting but is it a big enough market? would most buyers prefer better construction and/or price at the expense of flexibility? I don't know the answer tbh.
Aircooling cases too much. Not many want one more. One rad SSF cases also present. But two 240/280 rad SFF case are all waiting for.
Buy an Open stand. Like Openbenchtable and test all what you want. Case is not for testing setups. I can say without loosing your time - two 280 rads > two 240 rads > one 280 > one 240 > 4 fans > 2 fans.
So what you want to test?
Exclude expandability and make case in one size will free resources and cost of this case to make this case better.
tbh that is what i wanted but with RDNA2 being much lower energy use, i may not need dual rad after all?
Nobody know how RDNA2 will be in power use. Aspesialy after 30x0 nvidia cards and their price.
And two 280 it is in any way more cool and less noise. You can just use lower fan speed.
If you need just one rad - choose another case.
And two 280 it is in any way more cool and less noise. You can just use lower fan speed.
If you need just one rad - choose another case.
Last edited:
im being patient for now. still wanna follow what happens with this case anyway
If anyone is interested in the 3080 and 3090 FE cards, they seem to share PCB, or at least the important stuff is on the same place on the 3080 and 3090 FE, and it is AMAZING how small it is, with such high power density and GDDR instead of HBM. The VRM of those cards is way better than the reference PCB of others. They are pretty much selling custom PCB stuff at MSRP.
Which means, 3090 FE for me boy
Which means, 3090 FE for me boy
Last edited:
Where exactly did you see this information? Because from where I'm sitting, they DEFINITELY DO NOT SHARE THE SAME PCB design, the 3090 have a MUCH taller PCB that expans far far beyond the PCI-e slot (https://cdn.videocardz.com/1/2020/09/NVIDIA-GeForce-RTX-3090-PCB-3-850x563.jpg) and includes SLI/NVLink fingers, whereas the 3080 PCB ends exactly where the PCIe bracket retention flange is, and do not include any SLI/NVLink fingers.
I think the distinction really needs to be clear here, the 3090 amd 3080 Founders Edition cards DO NOT HAVE REFERENCE PCBs, they are custom Nvidia board design, and is different from previous generations where the Founders Editition used reference board design.
Please see this article directly from EKWB where they specifically spelled this out: https://www.ekwb.com/blog/did-you-k...ers-edition-layout-is-not-a-reference-design/
Expandability will stay for sure because that's one of my selling point, but it will be more rigid and easier to build.
I don't need to see the article. I perfectly know the reference AIB PCB is not the same as the FE design. The top part of the 3090 FE PCB is completely blank, for the exception of the NVLINK fingers. The rest of the PCB components that are important to cool are in the same place, (CORE, VRM and RAM).
NVIDIA-Ge-Force-RTX-3090-PCB-3-850x563 hosted at ImgBB
Image NVIDIA-Ge-Force-RTX-3090-PCB-3-850x563 hosted in ImgBB
ibb.co
The terminal for that waterblock is placed on the red square, thile the green VRM are the ones that will be missing in the 3080 FE design. Like I said on my post...
I wrote post saying the Reference AIB PCB is different, and probably worse. Nvidia uses SMD capacitors, tantalumns and Al Polymer capacitors, while reference PCB are using cylindrical ones.
It is probably not the pretiest waterblock out there, but I don't want to have a card that can't fit even in some Mid tower cases... if the terminal is at the end with the 140mm wide, you can't put it in many cases, as it will be 160mm at least. Lian Li O11 is out of the race, and that case is hell popular right now. And as well... being a black PCB, I don't care if some is exposed, as I always buy the POM Acetal blocks anyway, or full nickel.
I think selling point of this case it is dual 240/280 rad support, waterplate. And size less then Cerberus or other A brands.
If you remove expandability, you can free some case cost and resources to do it better in one size.
I don't need to see the article. I perfectly know the reference AIB PCB is not the same as the FE design. The top part of the 3090 FE PCB is completely blank, for the exception of the NVLINK fingers. The rest of the PCB components that are important to cool are in the same place, (CORE, VRM and RAM).
NVIDIA-Ge-Force-RTX-3090-PCB-3-850x563 hosted at ImgBB
Image NVIDIA-Ge-Force-RTX-3090-PCB-3-850x563 hosted in ImgBB
The terminal for that waterblock is placed on the red square, thile the green VRM are the ones that will be missing in the 3080 FE design. Like I said on my post...
I wrote post saying the Reference AIB PCB is different, and probably worse. Nvidia uses SMD capacitors, tantalumns and Al Polymer capacitors, while reference PCB are using cylindrical ones.
It is probably not the pretiest waterblock out there, but I don't want to have a card that can't fit even in some Mid tower cases... if the terminal is at the end with the 140mm wide, you can't put it in many cases, as it will be 160mm at least. Lian Li O11 is out of the race, and that case is hell popular right now. And as well... being a black PCB, I don't care if some is exposed, as I always buy the POM Acetal blocks anyway, or full nickel.
FWIW, it does look like that Bykski 3090FE block is designed to fit 3080FE card, judging by the 12-pin connector cutout. However, at this point we don't know exactly what the reference PCBs will look like yet (unless you have leaked materials, in which case please do share). So I find your assertion that the PCB is probably worse is mainly speculation at this point (if you do have leaked materials regarding the reference PCBs, do share them, I find it quite interesting). Based on the 'reference design' EK blocks that were announced earlier, it doesn't look like reference PCBs will be much longer anyway (EK reference blocks are 230mm long based on their product page).
Regarding capacitors, you can get Tantalums and Alu-polmer caps in either SMD or cylindrical cans, and at this point we don't know what the implementation is going to be for reference cards (not to mention that it's only a reference, AIB partners are free to customise/substitute components if they feel like). Sure, surface mount devices usually perform a little better than through-hole, but do you have any hard data on the actual differences that manifests in terms of actual performance?
There are at least 2 reference PCB, where capacitors and inductors are changed, and according to Gigabyte (it was on their web page), the reference design is 14 + 4, and gigabyte is using 15 + 4, which is still lower than 16 + 4 (and probably no noticeable). But SMD FLAT are better than cylindrical capacitors, even if cylindrical ones are SMD (which exist), as the first one has 3 contacts to the PCB, with 2 grounds, and being shorter, have better ESL (even if they have only 2 contats). It is not a coincidence that all GPU that overclock better than the median use SMD capacitors (even if it is only 50 Mhz), as well as ITX motherboards, like the asus strix x-570, have SMD with 3 contact paths, and probably one of the best capacitors, at 470 uF capacitor at 3 milliohms, which means that it has one of the best output filtering. Adding that to the fact that you can mount the SMD literally on the back of the socket, and on a intel motherbord, in the middle of it, means that the voltage ripple is going to be way better than what any cylindrical capacitor can dream of.
Here is how the reference pcb looks like. But apparently there are some that use different inductors, with pcb reference n-rtx3090-x and n-rtx3090h-x for Bykski, but I can't find the webpage now. I will continue searhing for it, as I landed there yesterday.
ZOyNqjS.png
That's an interesting PCB shot. Based on the pad sizing it looks like that reference PCB can accomodate SMT caps instead of those cans too (probably at OEM's discretions). Regarding, actual performance impact of THT vs SMT filtering caps, I don't have a big enough data set to make a proper conclusion, but I think binning makes far more impact on overclocking than the type of packaging of the caps on the board, but I'll assume you're right on that count (I'm a mechanical engineer with only fairly basic understanding of PWM power electronics). Anecdotally, Asus and MSI still uses cans for their higher-end cards (Matrix, Strix OC, Lightning Z), I'd be pretty interested to see if anyone has a decent sample size between those cards and their SMT-equipped peers and see if there's an actually measurable trend.
Just out of interest, you mentioned the Strix ITX board with their SMT caps, do you happen to have some scope measurements vs comparable boards with cans? Would be nice to see some actual hard data (I saw buildzoid mentioned trying to jury rig a 3900X with measure points for scope output, but don't recall seeing any recent videos on that)
