Prototype Project Orthrus - Smallest SLI mATX case (V2)

[papierschiff]

oh man, I'm reading this thread now for multiple years and I look into it every week... I want to have this sooooo much. I don't care how much it will cost... but please, make this case reality.
(and yes, I don't need 2 gpus, but with only one gpu there is extra space for other things...)

 

[ThePositiveWay]

oh man, I'm reading this thread now for multiple years and I look into it every week... I want to have this sooooo much. I don't care how much it will cost... but please, make this case reality.
(and yes, I don't need 2 gpus, but with only one gpu there is extra space for other things...)

Most people don't need the 2 gpus. I'm watching this case because it looks like the HTC Vive wireless adapter will require a pcie slot that I don't have available in my NCase. I'm hoping some release estimate to be announced by end of year

 

[Phuncz]

I was hoping for VR to actually make good use of two cards with each rendering one eye, but apparently that's still not a thing developers want to look into.

 

[Ellimist]

VR SLI does work way better though. you get 100% increase due to each GPU rendering one eye each. Support would be good. I was thinking though in this case you could maybe use the 2.5" SSD slot for a PCIE card with some mods if you use M.2 drives only

 

[VegetableStu]

on the two cards for two eyes bent for a while: if you have two cards and two displays and in Display settings you tell them to show the same thing, would they display in sync? o_o

 

[EdZ]

VR multi-GPU is a LOT of extra work (basically you need to take all your engine optimisation work, throw it straight in the trash, then start again from scratch) for a lot less than 100% scaling. You only gain performance in operations that are unique per-eye (shared operations do not gain any performance) and there are so many existing techniques like Single Pass Stereo that obviate most of the per-eye gains without the need to re-optimise that you end up with a doubling of developer workload to optimise for a use-case that may gain ~20% performance and is used by a single-digit percent (or less) of the install base. And if you get it wrong, you end up with a massive increase in latency due to the inter-card link making the game unusable.

Unlike for 'flat-screen' gaming, there is no implicit multi-GPU option for VR. You need to explicitly implement it, or it will simply perform worse than a single GPU.

 

[Player 1]

I’m here for the updates. Need this as my setup!! Top job

 

[Deleted member 3325]

As of right now, what is the estimate on completion date? Soon, or still going to be a while (months or years)? I know you get asked this a lot but I've just been super excited for this case ever since I saw LTT video on it a year ago, so any information you can give will be greatly appreciated. Amazing work here.

 

[FAQBytes]

As of right now, what is the estimate on completion date? Soon, or still going to be a while (months or years)? I know you get asked this a lot but I've just been super excited for this case ever since I saw LTT video on it a year ago, so any information you can give will be greatly appreciated. Amazing work here.

Based on what I'm seeing, late Q4 at the earliest unless something drastically changed between the last update and now.

More realistically I'd say late Q1-Q2 next year.

 

[xSanchez78]

@LukeD Apparently the GTX 1180 PCB was leaked - it has a new SLI finger... does that mean these cards straight up can't be used in this case?

 

[vikinggoat]

The image links seem to all be borked, were they ever made available elsewhere?

So i modded the FSP 700 Watt Platinum PSU ... here's what I've come up with:

• Painted black with high temp paint
• Desoldered all old mustard cables and replaced with Silverstone extensions

The plugs are as follows:

• 2 x 6+2PIN + 6PIN connectors (for GPUs)
• 1 x 4PIN Molex
• 1 x 8PIN (for CPU)
• 2 x SATA
• 1 x 24PIN ATX

 

[FAQBytes]

@LukeD Apparently the GTX 1180 PCB was leaked - it has a new SLI finger... does that mean these cards straight up can't be used in this case?

That's definitely a Quadro card unless they're moving other graphics cards to the NVLink system.

Edit: At second glance, perhaps not. Seems to be reverse ordered from previous NVLink fingers.

Seems to be very similar, so maybe the roumors of much higher performance cards is true. That and/or perhaps an overhaul of the SLI system for a unified core and true extended memory similar to NVLink.

Also, after watching Buildzoid's video (which I had meant to do prior), it only raises more questions on what NVidia plans on accomplishing with this generation.

Edit 2: I think I know what those 2 buck converters on the left hand of the card is. The new VR port is capable of supplying 27 watts, hence you need something to be able to supply that power. I think that all but confirms that is a VR Port.

I think with the complexity of this board + other factors, we are looking at an increase of around $75-100 GoG. $50 or less is seeming less likely, despite others thinking so, especially if NVidia moves to a dual fan system (Potentially implying a hotter chip, but I think the Titan Z proves it can be done with a reference style cooler.). Not having a founders card will be disappointing to me, purely for this case.

 

[xSanchez78]

Well at this point, pretty much the entire lineup of the new cards has been revealed and there are detailed specs for them and everything... including the fact that the RTX 2070, RTX 2080, and RTX 2080 Ti are basically stripped down quadro cards, and the fact that they use NVLink for SLI now. That's great and all because performance scaling might be drastically better than it used to with two cards, but the main issue is the connector.

As far as I've seen, the NVLink bridge is only available in a solid bridge much like the latest "High-Bandwidth" bridges. Luckily for the Pascal generation, soft bridges still worked and ended up giving the same performance as the HB ones as long as you used two, but what about now? Is there some way around this or is it going to simply be that these new generation cards simply won't work in this case?

 

[FAQBytes]

Well at this point, pretty much the entire lineup of the new cards has been revealed and there are detailed specs for them and everything... including the fact that the RTX 2070, RTX 2080, and RTX 2080 Ti are basically stripped down quadro cards, and the fact that they use NVLink for SLI now. That's great and all because performance scaling might be drastically better than it used to with two cards, but the main issue is the connector.

As far as I've seen, the NVLink bridge is only available in a solid bridge much like the latest "High-Bandwidth" bridges. Luckily for the Pascal generation, soft bridges still worked and ended up giving the same performance as the HB ones as long as you used two, but what about now? Is there some way around this or is it going to simply be that these new generation cards simply won't work in this case?

It's not that difficult to make custom PCBs. Worst comes to worst, I could whip a couple up to test them, ideally flexible ones.
What concerns me is that, up to now, Nvidia has been pretty secretive about all of it. What's the frequency the communication runs at? How tolerable is it to interference?
You have to know that and a little more for me to do signal integrity analysis on any PCB I would make.
I also don't quite have the capital to buy two cards immediately. I'm also sure there's people out there far more qualified to take your money for that sort of thing. I've really only developed low frequency embedded electronics. High frequency stuff is slightly out of my area of expertise.

 

[xSanchez78]

oh wow so you're saying we've got a community on our hands willing to make a custom NVLink bridge specifically for this case? Jesus man I appreciate whoever is willing to spend their own money and time to make that for us. I'll be willing to pay good money for that because I want some RTX 2070's in this case really really bad lol.

 

[FAQBytes]

Considering the roumors or the non-blower reference coolers being confirmed, what does the cooling performance of standard coolers look like? I'm concerned this may have just taken this case off the table for me unless some design changes happen to allow for choice to have porting for the new style coolers on the acrylic side panel if temperatures are much higher.

 

[xSanchez78]

Well I'm pretty sure you can still purchase a blower design from a vendor like Asus, because an open air card like that will destroy the cooling performance in a case like this. But for me, I've planned to do water cooling since day 1, and especially now, it seems like a necessity. I'm probably going to put the 2070's in their own custom loop and use a Cryorig C7 Cu for the Ryzen 5 2600. If you're able to, I suggest doing something similar.

 

[Ellimist]

At this stage I'd settle for just knowing when we will be able to order the case. Can always customise or build new side panels when they are acrylic.

 

[LukeD]

UPDATE:

Guys let me first start off with an apology, I've been a bit quiet over the past few months.
Life hit me hard and between work pressures, a completely unexpected divorce and health complications I had to go back to basics and start taking care of my physical and mental health. I didn't want to do a half-ass job and rush this to market so the project was delayed on my own accord only.

Where things are at, at the moment:
- I need to provide feedback back to manufacturing around the most recent changes / optimisations.
- I need to order PCIe riser cables once the optimisations are done
Not sure how we're going to tackle the new NVLink connector. I'm hoping LiHeat can pull a magic trick out of their hat and make a shielded flexible NVLink connector.

 

[DSmithBYU]

Take all the time you need. Deep breaths, proceed slowly and steadily.

You can do it.