Mind if I ask what background you have on your laptop in that picture or what you'd normally have? I run 3 displays and a laptop like that as well and usually just take the highest res and let display fusion stretch and crop across all 4. I'm running Accretion that way right now and it looks good but obviously not quite as intended. I'm just curious what you do since it looks like you've got the same situation.

Wow is that a beautiful setup. I'm jealous, even if I don't have a project that would remotely challenge that beast.

I used to think that 2 screens were better than 1, but now I realise that 3 screens are better than 2. That's one seriously awesome rig you have there. Enjoy.

I was lucky enough to pick up some new hardware as well. The Samsung Galaxy S6 phone has a 1440 x 2560 display, any chance of getting renders for that resolution/orientation?

Great 3D effect!

I commend your decision on having 3 systems - one going to each monitor. Use Synergy, Multiplicity Pro similar to allow for seamless keyboard/mouse movement between systems. Don't forget to also do a physical KVM (just do keyboard/mouse) so you can control each computer directly when booting, troubleshooting, etc. I've been using a setup like this for years. Enjoy!

Congrats Ryan on your late/early Christmas. When I saw your note I thought: why does Ryan need to buy new computers to render? Isn't there distributed rendering yet? I imagined all DB fans installing the software that Seti@home used to generate new renders on hundreds or thousands of lower power computers and sending the info back to Ryan (or even the cloud). After a quick google, most top hits are research papers, hrm. Not there yet I guess...

Display Fusion is also worth looking at in addition to Mouse Without Borders and Synergy. I personally use Mouse Without Borders. :)

Since you'll have 3 computers side to side, I'd recommend using Mouse Without Borders or Synergy to control them all with one mouse/keyboard. What do you think?

So now that Bucephalus has a twin, I'm thinking in a few more years when you finally pull the plug, you'll have to replace them with computers named Skinfaxi and Hrímfaxi. One for day renders and one for night scenes, to keep with the mythology!

Ryan,



It's a little late, I know, but that Seagate 4TB disk sticks out to me as the Achilles Heel of your new project. Seagate's drives consistently rate near the bottom for reliability (failures per disk/year), and you've only put in one of them so when it fails, as all disks eventually do, you're looking at either a very long restore from backup or loss of data. I'm sure that server motherboard has on-board RAID capability; you should seriously consider getting 2 drives in RAID-1 or 3 in RAID-5 so that one disk failure can't take you down in the middle of a render.

I can completely identify with the idea of wanting to get the noise and heat out of your office :o) I have all my computers/machines in a "server room" downstairs and my office is directly above, therefore peace, quiet and most of all lovely and cool compared to how it used to be, especially in the summer.



For me the key to getting this right was finding a decent KVM that satisfied the requirements of allowing relatively high cable runs (compared to a normal desktop KVM) and no impact to the display signal.



I've been through a number of brands and models, mainly as I switched from VGA to DVI to DisplayPort over the years, currently I'm using Geffen gear which has been the most reliable of all my choices over the years, I'd highly recommend them, that being said, you'll need multiscreen ideally and super high resolution support, so that's going to be tough, I've not kept a close eye on their range, but that's where I'd go looking first.



Congrats on the achievements so far, the basement move will really allow you to focus, speaking as someone who's already walked that path. Best of luck±

@Ryan R: That price at Puget is a total ripoff.



We just bought two servers from Dell, same CPUs, twice the RAM (16GB sticks), no GPU/storage, for $20k total. If you don't put these in your office you could consider getting a server from Dell, next business day on site part swaps if something fails is pretty nice for 3 years.

I played around with configuring one of their water-cooled gaming PCs with specs almost identical to your proposed Shadowfax build, and it came in right about $15k (not including the monitors).



http://puget.systems/?u=142162



There are some advantages to having them build it, like full burn-in, cooling tuning, benchmarking, lifetime labor and support, and really cool thermal images of the machine during burn-in. :)

We have a local company in Seattle that does high quality work and builds PCs for very reasonable prices compared to building your own. Just playing around, I did some configuration of a 32-core 3.3 GHz system (4 CPUs) with similar specs to Shadowfax otherwise. You'd have to talk to them about a second GPU in SLI. It uses last-gen DDR3-1600 RAM but it's a beast for $20k. I think they are at least worth a look, if for nothing else than to read their blog on hardware reliability statistics. http://puget.systems/?u=142161

I think that's best. It strikes me as a bit underhanded to keep raising the goal to bring in donations from the people that will donate when they see that we're "almost there". It's a stunt to elicit more money, and I don't feel like it should be pulled on people who are honoring you for your talent, skill, and hard work.

Besides, I think it's more impressive to see how much people will knowingly give beyond what you need for the new workstation, because you know they are giving that directly to your family. We're sponsoring you. :)

You used to mention wanting to use Maya as well. Have you moved away from that?

When I donated, the goal was $15k and like 80% there. Now I see it's $20k and 80% there again. What gives?

I don't see a progress bar on how close you are to your goal? It might cause in influx of detonation once your fairly close.

Particularly if you look at the turbo boost speeds, which the chips are more likely to run at, the 12-core 2690 falls short by only 100 MHz compared to the 2687W when all cores are active, and is actually faster when only a few cores are. For the same price, 20% more almost equally fast cores may be the better option, although I remember you saying when choosing parts for Bucephalus that your renders are clockspeed sensitive, although only a 100 MHz deficit doesn't seem like a lot. Anyway, exciting stuff, good luck!

If you are not strictly frequency bound, then you might want to look at the Xeon E5-2680 v3. It's 12x 2.5 GHz (30 ghz-cores) vs 10x 3.1 (31 ghz-cores). So it is a tad slower (~3%) but they run about a grand less for the pair.



Likewise, the e5-2690 v3 (12x 2.6) would be basically the same as the 2687 (31.2 ghz-cores), so it should be on your short list as well (in case that one happens to be cheaper than the other).

I get that you want to be able to serve the highest-end users you can - but I would bet the overwhelming majority of users running multi-monitor setups are doing so at 1920x1080.



Wouldn't it make sense in the short term to just render multiscreens at 2x and 3x 1920x1080? it seems like satisfying 90% of multi-screen users would be better than nothing?

Ryan,



If you're going to be dropping this much cash on a new workstation, why not get someone like James Walter or B-Negative at overclock.net, who professionally build systems and that could make a good cooling option possible. Additionally, Hukkel at OCN built a 4P machine inside a modified SuperMicro case that was entirely water cooled for Folding@Home. I imagine folding would be just as computer intensive as 4K or 8K digital renders.



Jameswalt1: http://www.overclock.net/t/1529883/sponsored-monochrome/0_40

B-Negative: http://www.overclock.net/t/1531403/sponsored-hexgear-idoru/0_40

Hukkel: http://www.overclock.net/t/1501011/sponsored-opteron-prime-watercooled-supermicro-4-cpu-folding-home-crazyness-folding-numbers/0_40

Ryan-

Once you have a completely final scene, how many multiscreen renders do you do? IE do you do a single render at the highest resolution and just resize/crop it down, or do you do multiple of them then resize/crop?



The reason I ask is because although a render farm probably doesn't efficiently split up a single frame, you can spin up one VM per render that you need done, and do them in parallel, which is a nice, and efficient. And if you figure you get $X dollars per day per multiscreen image (across all revenue sources leading to that image), then there is probably an argument to made for rendering in parallel, and getting images to your customers faster.

Hey Ryan, would you not benefit from getting a workstation gpu like Nvidia's Quadro series instead of using two gaming focused gpu's in SLI?

I'm sure you've done your homework, but a couple more places I found:



http://www.bit-tech.net/hardware/cpus/2011/11/14/intel-sandy-bridge-e-review/4 (comparing 12-core)



http://www.cbscores.com/index.php?sort=cores&order=desc



might be worth downloading and running that benchmark, see how you do, give you an idea of what those newer types of hardware can do.

Ryan,



Curious why you're not going with a 4x16 core opteron based machine. I'd have to go digging for benchmarks, but I'd imagine that 64 real cores would render faster/wider than 20 real cores + 20 HT cores.



If you don't mind going 1 generation older (and slightly slower), you can get 16 core CPU's on ebay for <$150 each.

Ryan-

Read that article you linked with a huge grain of salt, they are a company that is selling a competing service and trying to get you to use their product. In specific, it is full of FUD around performance and pricing. They claim 30% or worse performance drop from bare metal, this is nonsense - and has been nonsense for years see: https://major.io/2014/06/22/performance-benchmarks-kvm-vs-xen/



They also claim bad pricing. I just checked and you can get an instance with a Xeon E5-2666 Haswell with 10 cores (20 threads) at 2.9ghz with turbo up to 3.5ghz, 60G of RAM, for $1.856/hour for Linux ($3.184 for Windows Server). If you need more RAM they have other instances that trade some compute power for way more RAM.



In any case, I'd encourage you to at least consider it, and model out how much your hardware costs for a new machine, the electricity to power the machine locally, and the air conditioning to cool it, etc. Not to mention the time it takes for you to get these images into your paid customers hands, which I of course don't have the data on, but I have to assume that a significant fraction of you users only want multiscreen.



-David

Thanks, Ryan, for giving us the heads up. I wondered why you moved on after one render of some projects (such as Fire Below Ice, Moonshadow, and Cloud Canyon that didn't even make it to the gallery). I also like the idea of a "render beast" tip jar for those who want to donate expressly to the new equipment. Perhaps a few hints about the setup might bring donations.

I too was concerned about the lack of multi screen renders. Thank you for letting us Dual and Triple screen users know what is up.

How about setting up a donation page for those of us who are already lifetime members? I would like to help ya upgrade to a new render beast but don't really need to buy another membership or go as a monthly commitment?

Ryan-

Why not spin up EC2 VMs and render in parallel? You could probably do a test render or two there and easily figure out what your actual cost is to render an image and then project that out over time.

I guess my only annoyance is that the new 5k renders (a relatively small userbase) are taking precedence over any of the multi-screen renders (a huge userbase). I'd expect your priority pipeline to be the other way around... but I'm clearly biased.