Refractive Software® is discontinuing Cubix Xpander support

[Chris_TC]

So the Config without the GPU can be estimated at 5585 Euros.

It's really just another example where you pay over €1,500 extra just for the fact that it's pre-assembled and targeted at businesses.

€580 EVGA SR-2 dual socket
€2500 2x Xeon 5670
€150 2x 1TB hard drive
€320 1500W PSU
€300 Case, Keyboard, Mouse, DVD burner, Case fans
€150 Having it all assembled

Total = €4,000

[Carl S.]

@ROUBAL Some more benefits I see with the external option is the fact that the gpu's don't have to be powered on 100% of the time with the computer, can be easily moved from one workstation to another if needed, a dedicated cooling system, and the ability to add more gpu's. In the case of the motherboard you mentioned with 7 single wide pci slots you can theoretically add 5 xpander boxes and one double wide gpu totaling 11 cards. If you make the all 11 cards the double gtx480's you will have a supercomputer.

[Carl S.]

Anyway, Carl, the EVGA Classified Super Record 2 (SR-2) Motherboard is also 4 way SLI and supports 3x nVidia® EVGA® GeForce® GTX 480 in PCIe 2.0.

With that motherboard you could then have 6 gtx480's with xpander boxes.

I also think the only real benefit of the PCIe 2.0 is the speed at which the scene is loaded onto the card, once the scene is on the card it renders the scene the same speed. The cards to not render slower over PCIe 1 AFIK.

[ROUBAL]

@Chris_TC : you forgot at least the 24 GB DDR3 memory that costs in average 700 Euros...

Added to that, compared to prices at my main providers around here, your 4000 Euros estimation is closer to the price without VAT... and Carri Systems solution is provided with 3 years on site warranty, so I think that they don't charge too much for assembly.

If I could afford that solution, thinking about the time I'd had to spend to build the machine, risks of doings things wrong and voiding warranty on some sensitive parts, or having to send back defective parts if something doesn't run, I really prefer to buy a ready to use machine, already tested even if it costs a bit more.

[bob30]

Well, i don't know where you see for workstation building but i found yesterday evening on the net a respectable company which proposes a basic workstation configuration with 3 GTX 480 for $3000, PCI 2.0, and with good materials otherwise. In a dekstop-like computer box, not as a big cube. Still too expensive for me, but quite more cheaper and efficient than Carl's configuration (not wanting to offend you by the way).

Xpander boxes display the information for all cards by one PCI 1.0 adapter card, so, it can not be as faster than if the cards were all in the system in separate PCI 1.0 slots. Now i f you add PCI 2.0 slots and *16, it is surely far superior to external boxes.

And considering the prices of external boxes, as you paid it, there is no "photo", as one said.

[Jaberwocky]

Going back to the GTX580 Idea.Of Course 4 x 580's would in all probability consume 1300-1400 watts at full tilt.I would therefore suggest a Coolermaster HAF X Case and 2 x 750watt Coolermaster GX power supplies.(one to be mounted externally of course).Bearing in mind that the probable idle power consumption for each card when not in use would likley be no more than 30 watts.So Consumption for the beast when not rendering would not be to extream.

See link: http://www.coolermaster.co.uk/product.p ... ct_id=6637

for details

[Jaberwocky]

Ok. I think i may have the answer everybody

check this link out

http://www.magma.com/expressbox7x8.asp

[bob30]

Ok. I think i may have the answer everybody

check this link out

http://www.magma.com/expressbox7x8.asp

If there is nothing indicated, i think it should be PCI gen 1.0, so it remains very expensive as an external solution, to my view. For now, according to what i saw here and there, external boxes are not cheap solutions for GPU rendering, and better is to build specific machines for that, with full integration directly on the motherboard and PCI 2.0 slots.

[Carl S.]

Well, i don't know where you see for workstation building but i found yesterday evening on the net a respectable company which proposes a basic workstation configuration with 3 GTX 480 for $3000, PCI 2.0, and with good materials otherwise. In a dekstop-like computer box, not as a big cube. Still too expensive for me, but quite more cheaper and efficient than Carl's configuration (not wanting to offend you by the way).

Xpander boxes display the information for all cards by one PCI 1.0 adapter card, so, it can not be as faster than if the cards were all in the system in separate PCI 1.0 slots. Now i f you add PCI 2.0 slots and *16, it is surely far superior to external boxes.

And considering the prices of external boxes, as you paid it, there is no "photo", as one said.

Not offended at all.

The cards render just as fast over PCIe 1.0 as they do with PCIe 2.0. The only difference is the scene loads a few seconds faster over 2.0 onto the video cards. With small/medium sized scenes it is barley noticeable.
Once on the card it uses the full power of the card sending only little bits of info back. People are not being ripped off and PCIe 2.0 is not far superior in this case.
Here is a post to clear it up for you.

Yeah the load lag would be roughly the same if it were physically installed in your system. This is what happens when you load the model into the gpu. You really only see a difference when using the pcie x1 bus speed version as that is 8 times slower than the x8 version and 16 times slower than the x16 version. Between the x8 and x16 the load speed difference is pretty negligible. You lose a little with the cable length, but once its on the gpu all reandering should be the same. I forgot you said you would water cool the units. They do make a four slot xpander. I think that's the xpander pro 2 if memory servers. You should be able to fit all three in if the watercooling doesn't go beyond the single wide slot configuration.
I am hoping that cubix release the special version they made for Siggraph as it looks awesome. We'll get some photos of it u and the rest of the booth in a day or so.

Not sure what you mean by no "photo" at all.

[bob30]

Carl, it is physically impossible that PCI 1.0 is as fast as PCI 2.0 even onto medium scenes. Roubal said on another post:

If you consider that the tested scene weights only 290MB, on a PCIe x 16 (even if it is a 1.0) the loading time(so, the transfer speed of the PCIe link) seems to be a very critical factor !

I am currenlty doing a test on an animation of 35 frames of a scene requiring 290 MB of VRAM, on a GTX 260 mounted in a Cubix-Xpander which is connectet to a PCIe 16x 1.0 : For a rendering time of 4 seconds (at 32 samples per pixel), the loading time of each frame is 124 seconds ! More than 2 minutes per frame .

At this speed, the whole 35 frames animation requires 72 minutes of loading for 140 seconds of rendering !

For what I have tested so far, my fears are confirmed and the current loading time for animation is really beyond what I consider as acceptable.

and also:

this necessary loading time is a so huge waste of power...

But Carl, perhaps, do you mean that you have made some benchmark tests for illustrating what you advanced ?


On the other hand, we're speaking about PCI 2.0 16x and 8x into a system, directly linked by the motherboard to the CPU(s).