High End 3D Animation (Maya), Compositing, and Editing Build

[jcastane]

Hi Fellas, I´m putting together a 3D animation workstation, mainly for use for Maya. I´ll also be using Adobe CS5 Suite as well (Photoshop, After Effects, and Premiere) and Foundry Nuke. I´m considering putting Autodesk Flame on this guy as well.

Approximate Purchase Date: this week

Budget Range: $4,000 usd

System Usage from Most to Least Important: 3D Animation, Compositing, Editing, Unity3d Development.

Parts Not Required: keyboard, mouse, monitor, speakers, OS, HD drives, tablet

Preferred Website(s) for Parts: newegg.com

Country: Colombia

Parts Preferences:

Nvidia Quadro workstation card, thinking about quadro 5000:
http://www.newegg.com/Product/Product.aspx?Item=N82E16814133325

Still debating whether to go with dual opteron 6274´s for 32 cores -- something like: http://www.newegg.com/Product/Product.aspx?Item=N82E16819113038

or with a single, 6 core i7:
http://www.newegg.com/Product/Product.aspx?Item=N82E16819116491

actually, while posting, just saw this other one, which looks almost exactly the same, except $400 cheaper:
http://www.newegg.com/Product/Product.aspx?Item=N82E16819116492

if i went with i7, the rest of the build would be as follows:

ASUS Rampage IV Extreme LGA 2011 Board:
http://www.newegg.com/Product/Product.aspx?Item=N82E16813131802

G.SKILL Ripjaws Z Series 32 GB (8x4GB) DDR3 2133 Ram
http://www.newegg.com/Product/Product.aspx?Item=N82E16820231504

2x Intel 120GB SSD -- (one for OS drive, one for "active projects" drive) (I have close to 2TB of available storage in other drives for "passive" files)
http://www.newegg.com/Product/Product.aspx?Item=N82E16820167050

Overclocking: Maybe, if not too complicated or risky.

SLI or Crossfire: No

Monitor Resolution: Highest Possible

Additional Comments: No idea as to what power source I should be looking at, or whether additional cooling will be necessary. I´ve checked that the parts (in case of the i7 build) are compatible with each other, but there could be things that I´ve missed. Any thoughts/opinions are welcome. I think I have a good price range and want to make sure I get the best workstation possible.


Best Regards,
-Jac

 

[blazorthon]

Dual Opterons would probably be the fastest setup here, but that only works if your software supports 32 threads. I'm not sure if your software goes beyond 16 or 24 threads. The i7-3960X is a waste because despite the $400 price difference between it and the i7-3930K, the two are almost identical. It's kinda like the i7-960 versus the higher Nehalem i7s, except even closer in performance and even farther in price. There is a huge price difference, but the performance difference is negligible. Basically, by getting the 3960X, you pay for it to have an X in it's model number and for it to be called an extreme edition, not for significantly more performance (the only difference is that the 3960X has slightly more cache and a 4% higher frequency, so it's about 5% faster for 70% more money)

Make sure that your software supports 32 threads and can utilize that many threads well. If it does, then the dual Opteron 6274 should be the best option. A 6274 is almost as fast as an i7-3930K (the i7 is probably about 10%-15% faster, maybe a little less), so two of them will be almost twice as fast as the i7 for about double the cost. Considering that if it were a similarly performing Xeon it would probably cost a lot more, some of the Opterons have some serious server performance for the money. The 6274 also uses less power than the i7-3930K does, so you should make some/all of the price difference back over about one year (at least, according to my math if this system is on almost 24/7) if you use a single 6274. Using two 6274s will use more power than a single one or even a single i7 of any kind, but it will still be more energy efficient and worth the extra power.

I can't recommend a professional graphics card because I'm not experienced with them.

However, if your software does not make good use of 32 threads, then the i7-3930K would be your best option for less than $1500 or so from Intel because nothing from Intel cheaper than that actually beats it.

Sorry, but not many people with true experience with server/workstation class systems scour the forums, so you might have some trouble getting an expert on these machines to talk to here. Perhaps you could mention something in one of the articles that involves such machines? That way, you're more likely to recieve help from someone who can help more than I can.

 

[Draven35]

It depends what yo're doing. If you;re doing a lot of modelling and opengl work then no, the Opterons won't be faster, they would only be faster come render time. A lot of Maya and MAx and Lightwave and other 3d app work is very very single threaded so it comes down to the per-core FPU performance of your processor.

 

[serialkiller]

hello op

i know what the best you want.


the first thing the part where the system is tested is when you are rendering.

basically rendering is only done by your cpu.

the gpu just shows out the output the cpu just rendered.

the gpu is only maximumly used if you do real time raytracing. and for that a gpu must be a powerful tessler especiallty if its nvidia it really good.



so mainly concentrate on your cpus and not the gpu.

and no program which is consumerablly available uses more than 16 cores.

so i would suggest


intel xeon E-5 2640 x2
evga srx motherboard
corsair gt dominator 4x4gb 1600mhz x2
pny quadro 4000
seasonic x560 w 80+ gold


for the storage

intel 520 120 gb ssd for os + programs

wd carvier black 1tb x2 in raid 1 for system files and projects

seagate baracuda 3tb hdd for storage


case

lanli armorsuit or corsair 800D or what you like and fits that mammoth srx which is of hptx size.

also case is personal perference

for cooling and overclocking go with noctua nh d14 but xeons run a loot cooler then even cm hyper 212 evo would be enough to get the cpu past 3.5ghz easily.

 

[Harm Millaard]

Dual Opterons would probably be the fastest setup here, but that only works if your software supports 32 threads.

Let me correct this with regards to Adobe CS5.x. Dual Opterons are way slower than even an outdated single i7-920. AMD lacks support for SSE 4.1+ extensions making them extremely slow in comparison with Intel. A dual Opteron 6134 with 32 GB RAM, a Quadro 4000 and a multitude of SSD's and an LSI hardware raid controller is nearly 5 times slower than an overclocked i7-920.

In our benchmark for Premiere Pro, the best Opteron comes out at rank 444. out of more than 800 systems.

For these kind of applications AMD is absolutely out.

 

[blazorthon]

Let me correct this with regards to Adobe CS5.x. Dual Opterons are way slower than even an outdated single i7-920. AMD lacks support for SSE 4.1+ extensions making them extremely slow in comparison with Intel. A dual Opteron 6134 with 32 GB RAM, a Quadro 4000 and a multitude of SSD's and an LSI hardware raid controller is nearly 5 times slower than an overclocked i7-920.

In our benchmark for Premiere Pro, the best Opteron comes out at rank 444. out of more than 800 systems.

For these kind of applications AMD is absolutely out.

Opteron 61xx are OLD AND OUTDATED. Opteron 62xx are not. The 62xxs have proper SSE 4.1+ support and are very good for it (although their dual 128 bit FPUs working together for 256 bit FP math seems to still be slightly inferior to the Sandy Bridge FPU, it is better than the Nehalem FPUs). Considering the fact that I only mentioned the Interlagos Opterons (the 62xxs) as viable for this, this should have been obvious to you. I really dislike it when someone who *corrects* me is THAT obviously wrong. As more of a desktop specialist, even I know this much.

 

[blazorthon]

If I'm going to be corrected when I'm wrong, then please correct me, not be even more wrong than I was (I wasn't wrong, so I'm already ahead).

All Bulldozer based CPUs have SSE compatibility up to 4.1+ and 4.2. The Opteron 6200 series is a purely Bulldozer series. Go to AMD's site, Wiki, or a huge number of other places to see this for yourself if you must. They also have AVX and more.

 

[Harm Millaard]

Opteron 61xx are OLD AND OUTDATED. Opteron 62xx are not. The 62xxs have proper SSE 4.1+ support and are very good for it (although their dual 128 bit FPUs working together for 256 bit FP math seems to still be slightly inferior to the Sandy Bridge FPU, it is better than the Nehalem FPUs). Considering the fact that I only mentioned the Interlagos Opterons (the 62xxs) as viable for this, this should have been obvious to you. God do I hate it when someone who *corrects* me is THAT obviously wrong. As more of a desktop specialist, even I know this much.

So your Opteron 62xx are at least 5 times faster than the old 61xx??? I wonder if that is accurate.

 

[jcastane]

Hi Fellas,
Thanks for the suggestions, I really appreciate it. I'm actually a little less certain of what to get now from all the feedback heheh. I have the following questions:

Blazor, Harm, and Draven:

Thanks for the insights on the processor, all good points.
Draven got me thinking; In my particular case, the priority is having a responsive, powerful workstaion. (fast viewport action in Maya even in heavy scenes, good ram caching in Adobe applications, closest to realtime I can get when building networks in Nuke, or working in complex comps in After Effects). In any case, I would like to count with some serious rendering firepower as well. Still, the priority is being able to do the moment to moment work efficiently.


Thanks Blazor, i7-3960X is definitely off the list. Now what worries me about the AMD Opteron 6274 Interlagos, despite seeing positive benchmarks (still a little behind i7s, but close) is that they are 2.2GHz per core. I've also read that dual processor setups don't actually mean exactly 2x the power, but closer to 1.5x -- is that correct? Would this mean that in single-core (or even dual-core) applications they would be a lot slower than the i7-3930K (3.2GHz/ 3.8 Turbo)? If push comes to shove, i'd rather have a smooth work experience than hardcore rendering power (although it'd be nice to have both).

Any thoughts on the 6274, Harm?


Serialkiller, I checked out your proposed build. I see that the intel xeon E-5 2640s are factory clocked at 2.5GHz. By overclocking them to 3.5ghz it sounds like i'd be nearing the performance of the i7-3930K. My question is, is there really a big difference between one 6 core hyperthreaded processor, and two six core non-hyperthreaded processors? Is the question here one 6 physical+6 logical cores vs 12 physical cores, or am I thinking about something else entirely? What would be the benefit of the dual xeons vs the single i7?


I'm also still leaning towards the Quadro 5000 -- although rendering is handled on the cpu, I still want the best Mercury Engine, Maya, and Nuke viewport experiences I can get. Do you think the leap from 4000 to 5000 is unnecessary?


Thanks again, I really appreciate you guys helping me out.

Best regards,
-Jac

 

[BuBzXXL]

I'm not sure if you can overclock server processors... I don't know anything about them

 

[blazorthon]

I'm not sure if you can overclock server processors... I don't know anything about them

Nehalem based Xeons can be overclocked (well, at least some of them can). Sandy Bridge based Xeons can not be overclocked very much. I'm not sure about Opterons, but my guess is that they can't be overclocked.

 

[blazorthon]

Hi Fellas,
Thanks for the suggestions, I really appreciate it. I'm actually a little less certain of what to get now from all the feedback heheh. I have the following questions:

Blazor, Harm, and Draven:

Thanks for the insights on the processor, all good points.
Draven got me thinking; In my particular case, the priority is having a responsive, powerful workstaion. (fast viewport action in Maya even in heavy scenes, good ram caching in Adobe applications, closest to realtime I can get when building networks in Nuke, or working in complex comps in After Effects). In any case, I would like to count with some serious rendering firepower as well. Still, the priority is being able to do the moment to moment work efficiently.


Thanks Blazor, i7-3960X is definitely off the list. Now what worries me about the AMD Opteron 6274 Interlagos, despite seeing positive benchmarks (still a little behind i7s, but close) is that they are 2.2GHz per core. I've also read that dual processor setups don't actually mean exactly 2x the power, but closer to 1.5x -- is that correct? Would this mean that in single-core (or even dual-core) applications they would be a lot slower than the i7-3930K (3.2GHz/ 3.8 Turbo)? If push comes to shove, i'd rather have a smooth work experience than hardcore rendering power (although it'd be nice to have both).

Any thoughts on the 6274, Harm?


Serialkiller, I checked out your proposed build. I see that the intel xeon E-5 2640s are factory clocked at 2.5GHz. By overclocking them to 3.5ghz it sounds like i'd be nearing the performance of the i7-3930K. My question is, is there really a big difference between one 6 core hyperthreaded processor, and two six core non-hyperthreaded processors? Is the question here one 6 physical+6 logical cores vs 12 physical cores, or am I thinking about something else entirely? What would be the benefit of the dual xeons vs the single i7?


I'm also still leaning towards the Quadro 5000 -- although rendering is handled on the cpu, I still want the best Mercury Engine, Maya, and Nuke viewport experiences I can get. Do you think the leap from 4000 to 5000 is unnecessary?


Thanks again, I really appreciate you guys helping me out.

Best regards,
-Jac

For single threaded work, yes, the Opterons would be much slower than the i7s and it's not just because of the lower clock frequency (which is just due to them being server chips and not really workstation chips), it's also because Sandy Bridge has about 50% more IPC (instructions per clock, aka performance per Hz) than Bulldozer. Looking back on this, Opterons would be less than half the speed of the i7s in one to six threaded software and a lot slower in anything less than 16 threaded.

Unless you are doing purely 32 threaded software, then the i7 is the better choice because it has a much greater balance of highly and lightly threaded performance than the Opterons would. Yeah, dual Opteron would not be exactly twice as fast as a single one. However, I'm not sure of exactly how much faster it would be for 32 threaded software... Although I'm pretty sure it will be closer to twice as fast than only 50% faster. If it were only 50% faster, then that would practically ruin the point of the dual and quad G34 socket servers because the scaling efficiency would be too horrible to make them worth while.

Go for the i7-3930K. If you could get two of them then I would recommend that, but alas, Intel made sure of that not being an option...

Like I said before, I'm not experienced with professional graphics cards so I can't really comment on that.

The highest overclock that I've found to be possible on the E-5 2640 is from 2.5GHz to 2.8GHz, so 3.5GHz isn't even an option. Intel intentionally killed the overclocking potential of the Sandy Bridge based Xeons.

 

[Harm Millaard]

I just updated the http://ppbm5.com/DB-PPBM5-2.php with the latest submissions and now have 827 system in the benchmark. Notice this is purely oriented towards Premiere Pro Cs5.x, but since AE has similar hardware requirements, it is also indicative of AE performance. It does not help with 3D application like Maya or Vue.

For CS5.x and CS6, no AMD chip or combination of chips can match Intel because of their architectural limitations. If you click on the column header 'Brand CPU' the complete table is sorted on the brand and thus will show all AMD systems first. The very best AMD is a dual Opteron 2427 at rank #357, nearly four times slower than the fastest Intel system. Unfortunately we don't have any 6274 systems in the benchmark, but when the 6134 is already nearly 5 times slower than the best Intel system, I have very, very few hopes that the 6274 will perform 5 times better than the 6134 to just make up the difference with Intel, let alone pass it. I mean performance gains from generation to generation are usually measured in 10's of percentages, from 100 to maybe 130 or 140, but from 100 to 500 would be a huge and unprecedented move, that would have stunned the whole world.

 

[blazorthon]

I just updated the http://ppbm5.com/DB-PPBM5-2.php with the latest submissions and now have 827 system in the benchmark. Notice this is purely oriented towards Premiere Pro Cs5.x, but since AE has similar hardware requirements, it is also indicative of AE performance. It does not help with 3D application like Maya or Vue.

For CS5.x and CS6, no AMD chip or combination of chips can match Intel because of their architectural limitations. If you click on the column header 'Brand CPU' the complete table is sorted on the brand and thus will show all AMD systems first. The very best AMD is a dual Opteron 2427 at rank #357, nearly four times slower than the fastest Intel system. Unfortunately we don't have any 6274 systems in the benchmark, but when the 6134 is already nearly 5 times slower than the best Intel system, I have very, very few hopes that the 6274 will perform 5 times better than the 6134 to just make up the difference with Intel, let alone pass it. I mean performance gains from generation to generation are usually measured in 10's of percentages, from 100 to maybe 130 or 140, but from 100 to 500 would be a huge and unprecedented move, that would have stunned the whole world.

If you doubt it so much, then why don't you look up the Tom's article that compares the new Bulldozer to the old CPUs? Bulldozer is several times faster than Magny-Cours and the others because it has the SSE 4.x compatibility that you yourself mentioned and vastly superior FPUs. Bulldozer compromises this huge advantage with being slightly slower per Hz than the older chips (lower IPC), but then again, it has more cores in than it's predecessors and generally higher clock frequencies that more than make up for it. The 16 core Interlagos CPUs are a lot faster than the 12 core Magny-Cours even in integer work, although the difference isn't close to the floating point performance difference.

Believe what you want, Toms, Anand, and many others disagree with you.

 

[BuBzXXL]

Hey, Jcastane how long did it take for you to get used to using all the editing programs. I wanna get into this kind of stuff just can't afford it yet lol...

 

[Harm Millaard]

If you doubt it so much, then why don't you look up the Tom's article that compares the new Bulldozer to the old CPUs? Bulldozer is several times faster than Magny-Cours and the others because it has the SSE 4.x compatibility that you yourself mentioned and vastly superior FPUs. Bulldozer compromises this huge advantage with being slightly slower per Hz than the older chips (lower IPC), but then again, it has more cores in than it's predecessors and generally higher clock frequencies that more than make up for it. The 16 core Interlagos CPUs are a lot faster than the 12 core Magny-Cours even in integer work, although the difference isn't close to the floating point performance difference.

Believe what you want, Toms, Anand, and many others disagree with you.

Care to give me a link to the article you mean?

 

[serialkiller]

ops

i did found out that the sandy bridge xeons are not overclocking frendly but

but their are few models (i dont rember now where i see it) is havind sufix x means can be oced.

see its not about cores and clock speed but its about performance per clock and performance per core and also about how much cache do you have.


xeons are engineered to work better with workstation programs and not games.

they are basically

main stream and high performance or enthusists cpu

underclocked , increased cache , low power consumption and tdp , more cores and support for multi cpu linking or connecting 4-8 cpu in series on a motherboard and working simultenously.


the last thing is what i7 and xeons differ from each other.



so the best bet is go with i7 3930k and spend on quadro and ssd.


also as i said cpu is more than important while rendering than gpu.


quadro 5000 is not that worth it in my view as 4000 does it a second slower.

so a 1000 bucks for 1~10 sec decrease is pointless.


thats why i pointed out the xeons.

also xeons are the ones that have more life time and better stability in cost of price and clock speed.

but as i said earlier clock speed is not that important.

and they try to rectify that loss by increasing the core count and cache.


so i would suggest



i7 3930k

asus p9x79 or sabertooth x79

4x8 gb 1600mhz ddr3 ram of least cas and lattency and of corsair , gskill , crucial , kingston , and few more

seasonic psu are the best in the world but you want to lower the cost but what that kind of quality then

get corsair ax , hx , tx series psu

xfx , and pc power and cooling silencer series psu as they are made by seasonic.

but the best choice will be seasonic x series psu.



and pny quadro 5000 or to lower the cost but not on performance then 4000


intel , crucial and samsung ssd for the most reliable and trouble free ssds.

wd and seagate for the true performance and quality for the buck hdd.

try to make a seperate ssd/hdd for the work files / important files and keep as any raid except raid 0 but raid 1 is better as it will keep the data backed up if the primary storage craches.



also check for the case as it support e atx as ws motherboard are really large.


and for the problem free operation few tips

download all the latest and stable drivers of motherboard bios , ssd , and gpu in a pendrive and update the drivers after the os installation.

also update the windows 7 pro to the point when it says no updates available.

by this the computer will be bug free , crash free and a performer.



hope this help and best of luck for your build.

 

[blazorthon]

Care to give me a link to the article you mean?

AMD supports 4 instructions from the SSE4 instruction set, but have also added four new SSE instructions, naming the group SSE4a. These instructions are not found in Intel's processors supporting SSE4.1 and AMD processors only started supporting Intel's SSE4.1 and SSE4.2 in the Bulldozer-based FX processors. Support was added for SSE4a for unaligned SSE load-operation instructions (which formerly required 16-byte alignment).[3]

Pulled from the Wiki for SSE4

http://en.wikipedia.org/wiki/SSE4

AMD FX processors will support the latest instruction set extensions, SSE 4.1 and 4.2; AVX, -256-bit YMM registers, -Non-destructive source operand capability, -AES subset, -FMAC subset (AMD 4-operand form); XSAVE state space management; XOP Instructions. The FX series of processors will also include Light Weight Profiling with Low-overhead user-level profiling; they use XSAVE state space; and will store records for configured events.

Pulled from a Legit reviews article

The new instruction sets will most likely help out AMD in some of the benchmarks tests that we will run today and down the road when Windows 8 comes out. For example Microsoft Visual Studio 2010 SP1 supports the new instructions in the AMD FX processors such as XOP/FMA4/AVX/SSE 4.x.

Pulled from the same article

http://www.legitreviews.com/article/1741/2/

Look at this Anand article that shows the difference in performance for the Opterons and Xeons specifically:

http://www.anandtech.com/show/5058/amds-opteron-interlagos-6200/13

Notice how the 6276 is nearly four times faster than the 6174 system. There you go. Do I need to also look for the Tom's article, or is this enough? The Tom's shows roughly the same thing, just with desktop FX instead of server class Opterons.

EDIT: As I said in my last comment or two, ignore my Opteron recommendation and go for the i7-3930K. The Opterons are more server class than workstation class. This post was more in reply to the doubts of Bulldozer fixing AMD's poor floating point performance of their former generations.

 

[BuBzXXL]

I saw a link for Windows Server OS it was like $600 is that correct?!??!

 

[blazorthon]

I saw a link for Windows Server OS it was like $600 is that correct?!??!

Yeah, the price is up there. That seems like a cheaper version if it's $600 or a pretty good deal. Some of the Server 2008r2 versions cost more than $1000. Do you need a server OS? You can get Windows Server OSs for free if you are a college student or know a college student if you use Microsoft Dreamspark.

 

[jcastane]

Hi Guys,
Thanks a lot for the replies, I really appreciate it. I´ll be posting the final build I went with in a bit.

BuBzXXL -- For editing, I only really work with Adobe Premiere, which is fantastic. I studied Computer Animation in school, so mainly I focus on 3D Animation and Compositing Applications. I think in most cases, understanding the principles behind the work allows you to pick up any program and run in a pretty short amount of time.

Editing, specifically, isn´t as complex in technical terms as the other things. The programs are fairly straightforward, and could likely be picked up in a week or so, if you're a computer person. The real thing to think about is having the understanding of film/design/story principles to be able to get a sequence to read.

Again, fellas, I really appreciate all the help. Sorry about the slow response.


Best Regards,
-Jac