Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Jeff's Dragon Machine... Let's do it Jeff... ;)

The quick'n'easy (yet not so cheap) solution is a fully decked out HP z820 and a K6000 card in it. Wroooom!
 
what kind of price points would you be shooting for? i am potentially interested, but some scalability to provide a tiered system with a lower budget, and then more lucrative options would probably suit the most users is my 2 cents. or a system where you can start out with a bare bones and add things to it...
 
The w9100 proved a better choice... But we didn't tested on the HP z820.
At present day I'd choose a CUDA card over an OpenCL - in 6 months I'm not so sure. No doubt in the next year or two we'll see more traction towards OpenCL but for best power & versatility I'd go with CUDA.
 
I am still on a Supermicro X8 MB with dual 6-core CPUs. I was going to go to E5-V2 last month but other than upgradeitis, I can wait for E5-V3.

Hehe. You're doing fine to wait. Biggest thing you would gain by an upgrade right now is the faster RAM. You can get more cores on a CPU, but at a lower clock speed. Biggest advantage to waiting for the V3 chips is the PCIe v4. No real-world gains from that in the immediate future, but it will future-proof the system to some extent. At least in a couple years, you'll be able to support a current GPU, rather than trying to make due with a GTX 980 or whatever in a PCIe v3 slot that can't handle its full bandwidth potential. ;)

I'll be building a dual Xeon box with V3 chips I think. Because Broadwell core Xeons will probably be 18+ months away at the time the V3 Haswell Xeons ship.

That might be a way to round out this discussion later this year, or more likely early next year. In the near future, discuss various options and configurations and then when the new motherboards and Xeon CPUs show up I can put together what I would expect to be the best real-world performer at that time and see how it does. ;)

I'll start putting together some performance numbers with the hardware I'm currently using. Still waiting for word on my Dragon's return so I can get some current clips. I know Ketch and some of you other guys can send some over, but I'd rather just start shooting and testing on my own. I'll have to do a few days of just testing and playing once Dragon lands anyway!
 
Notes for Jeff

Notes for Jeff

IMO the focus of this thread is building a machine that eats R3Ds of any flavor for breakfast, sans RR-X, without spending more coin than necessary.

CPU(s): AFAIK clock speed is more critical to real time monitoring/operations than # of cores using current builds of key packages. This seems like a good place to save money, though rendering/processing on properly multithreaded tools would be slower. Your knowledge of CPU options is a huge plus in recommending the "sweet spot" of cost/capability for this use case. The tricky part is that even Jeff can't accurately predict when key tools will be coded for better multithreading support. Moreover, Ketch stipulated the system be tuned to the capabilities of today's tools, so in hopes of narrowing things down...

Mobo: I would think X79 or L2011. Plenty of x16 PCIe3 slots and SATAIII (6Gb/s) support to start.

PSU: Brands to avoid? Any issues besides expected service life we need to consider?

Case/Cooling: This seems like a critical issue to me. Recently specced a CoolMaster case for a build that used larger/slower fans to keep noise levels down - been great so far, downside is that it's quite large.

GPU(s): This is where it gets interesting. Theoretically, a $999 Titan Black with 6GB VRAM has enough muscle. There are plenty of other options that might provide good bang for the buck depending on your needs. The FirePro W9100 w/16GB VRAM at $3,900 is very attractive if you have the budget - is it safe to assume that big buffer would negate most stuttering and dropped frame issues?

RAM: 128GB of 1866MHz DDR3 would be the minimum I'd spec for imaging work as it seems I'm always cacheing for play back and wishing it could pre-load more frames. Would love some perspective on how worthwhile that strategy is since 64GBs would be less expensive.

Boot drive: 500GB Samsung 840EVO SSD

Storage: So many options. My dream machine would have two internal 4x SSD RAIDS using 1TB Samsung 840EVOs (at $450/each that's close to $4K plus RAID controller(s) so it's $$$$).

Cheers - #19
 
Biggest advantage to waiting for the V3 chips is the PCIe v4
Well, that and isn't DDR4 coming along for the ride too? I'll be interested to see if any Haswells support NvLink.
 
X99 + Haswell-E + DDR4 in one month. Thunderbolt2, more PCI lanes, faster RAM, 8 cores unlocked for overclocking. Not worth spending now with only four weeks to wait. Throw in a Titan Black SLI, and you have a pretty awesome setup. Maingear, Origin, Digital Storm and other high-end PC manufacturers will do a top notch build and overclock the lot for you. Hackintosh will be problematic for some time, but Windows will sing from day one. This is absolutely the direction I will go.
 
IMO the focus of this thread is building a machine that eats R3Ds of any flavor for breakfast, sans RR-X, without spending more coin than necessary.

CPU(s): AFAIK clock speed is more critical to real time monitoring/operations than # of cores using current builds of key packages. This seems like a good place to save money, though rendering/processing on properly multithreaded tools would be slower. Your knowledge of CPU options is a huge plus in recommending the "sweet spot" of cost/capability for this use case. The tricky part is that even Jeff can't accurately predict when key tools will be coded for better multithreading support. Moreover, Ketch stipulated the system be tuned to the capabilities of today's tools, so in hopes of narrowing things down...

Mobo: I would think X79 or L2011. Plenty of x16 PCIe3 slots and SATAIII (6Gb/s) support to start.

PSU: Brands to avoid? Any issues besides expected service life we need to consider?

Case/Cooling: This seems like a critical issue to me. Recently specced a CoolMaster case for a build that used larger/slower fans to keep noise levels down - been great so far, downside is that it's quite large.

GPU(s): This is where it gets interesting. Theoretically, a $999 Titan Black with 6GB VRAM has enough muscle. There are plenty of other options that might provide good bang for the buck depending on your needs. The FirePro W9100 w/16GB VRAM at $3,900 is very attractive if you have the budget - is it safe to assume that big buffer would negate most stuttering and dropped frame issues?

RAM: 128GB of 1866MHz DDR3 would be the minimum I'd spec for imaging work as it seems I'm always cacheing for play back and wishing it could pre-load more frames. Would love some perspective on how worthwhile that strategy is since 64GBs would be less expensive.

Boot drive: 500GB Samsung 840EVO SSD

Storage: So many options. My dream machine would have two internal 4x SSD RAIDS using 1TB Samsung 840EVOs (at $450/each that's close to $4K plus RAID controller(s) so it's $$$$).

Cheers - #19


BINGO!!
 
Well, that and isn't DDR4 coming along for the ride too? I'll be interested to see if any Haswells support NvLink.

DDR4 as well, but I haven't seen any benchmarks or other info yet to show the CPUs will be able to make good use of it. Currently the quad-channel DDR3 at 1866MHz is more than capable of handling the 8GT/s and associated bandwidth the CPUs are capable of. The bus I/O is not increasing with the v3 update, or at least not according to anything I've seen. They're just pushing out DDR4 and PCIe v4 support to coincide with the launch of those into the desktop market. Think of it as Intel's new marketing plan to combat the ever-slumping Xeon sales. On one hand the PCIe v4 support is nice, but in reality is it going to make a difference? When that system is 3 years old and there are actually GPUs out there that absolutely need that PCIe bandwidth, are you really going to be upgrading that PC's GPU or just buying a whole new box? Same with the RAM…. Are you going to spend for DDR4 ECC RAM, which is like ultra-super-duper-expensive right now. Especially if the CPU can't truly use it? I mean, we can put a pig in a tutu, but that doesn't make it dance.

Will have to see how the benchmarks and specs look as these updated components start shipping. May make a lot more sense to pick up E5-V2 parts and systems on clearance and wait for Broadwell Xeon in 2016.
 
You just perfectly described why I think Google's Ara project is stillborn. Shame about DDR4 though. It could really really help 3D rendering shader calls if they gave it room to breath.
 
This is where Broadwell will be a significant CPU update, probably the biggest to come along for some time. We will see lower power consumption and faster core speeds along with the ability to actually utilize DDR4 memory throughput.

Generally speaking, DDR4 support on the Haswell core chips isn't going to offer increased throughput or abilities. However, one thing that will help is the DDR4 will offer much lower latency in that we should get faster transaction initiation timings. So to say there is absolutely no benefit is not true. We won't gain the ability to move more data from point A to point B any faster with the Haswell-EP update. But the task of moving that data will start a few clicks faster every time. That's huge medicine for database servers and 3D render nodes. So there is that to factor in.
 
Can anyone help me with building a Hackintosh Maya and After Effects beast with a Supermicro GPU SuperWorkstation SYS-7047GR-TRF Dual LGA2011 Xeon 1620W 4U Rackmount/Tower Workstation Barebone System (Dark Grey) with a Motherboard: Super X9DRG-QF.
THanks,
Will
 
Hay Ketch
- The new Mac Pro Fully Loaded 2.7GHz 12 Core, 12G ATI D700 , 64G Ram
- Red Rocket-X
- Any thunderBoult 2 Expansion but not Sonnet the Thunder boult 2 its still Un-compatible with the new MAC PRO.
- SSD Raid Hard Drives Thunderbolt 2
- AJA Kona 3G Or Black Magic Cards for Monitoring.
- Final CutX (offline)
- Adobe Premiere
- Assimilate Scratch V8.
- Smoke.

it works fine with me and its Super Fast ;).





 
Is this thread not what this company was made for?

http://devilanddemon.com


Battistella

Well, actually I worked with them on providing Dragon files to test and see what was needed to make the machine church those R3D's away, but then I never heard back form them, so a dead end there, but fact remains that they are simply TOO expensive for most users, and this is why I invited, well kind of put Jeff on the spot to look to see what we would need to build a Dragon Machine... :)


Hay Ketch
- The new Mac Pro Fully Loaded 2.7GHz 12 Core, 12G ATI D700 , 64G Ram
- Red Rocket-X
- Any thunderBoult 2 Expansion but not Sonnet the Thunder boult 2 its still Un-compatible with the new MAC PRO.
- SSD Raid Hard Drives Thunderbolt 2
- AJA Kona 3G Or Black Magic Cards for Monitoring.
- Final CutX (offline)
- Adobe Premiere
- Assimilate Scratch V8.
- Smoke.

it works fine with me and its Super Fast ;).

Yeah that is a nice system, still pricey for some, but what are you getting out of 6K Dragon R3D's?

A análisis of how each Program is handling those files would be nice, since you have them installed.

RCX-Pro
FC-X
APP-CC
AS-V8
Smoke

Thanks buddy!!
 
Back
Top