PDA

View Full Version : Need technical guides and suggestions.


AllanGuigan
07-23-2012, 02:44 PM
Hello fellow artists.
I am about 2 years old in the 3D world.
I think I've learned a lot yet it seems I've learned nothing!
I am planning to buy a System and would like to know what all things should one consider while buying the components.

Primary Softwares to be used:
3ds Max, Maya, Fusion, Machmover, Realflow and the usual.
(RayFire and FumeFX by November)

About Graphic card,
I want to know whether they take part in any kind of simulation or calculation.
Quadro or GTX?
Do they effect rendertime? Or they are just limited to the viewport?
Whats the deal with Tesla cards? How do I use them? Would they be helpful?

About Processor,
I can either get an i7 2600k or Xeon X6550 x 2 (LGA1567) or Intel Xeon E7-4830 (LGA 1567 ES)
Which would be more benificial?
For Xeon processor, can someone please recommend a motherboard for it as I can not find one with pci-e slot.
And in the Dual socket Xeon board, can I use it with just one processor or do I need to
install 2 physical processors for it to work?

About HDD,
I already have a spare 2Tb Hdd, so dont need one but would like to know
whether I'll need a SSD. Is it really beneficial?

About Powersupply,
Got a spare 600W Coolermaster so i guess don't need to worry about it as most modern card are compatible with it. (not sure about TESLA)

About RAM,
I have 2x4GB 1333 RAM and when I check the RAM usage its hardly up 3.5gb. Do I need more RAM?
Would also like to know what difference does it make if I get the something like corsair Vengence?

Looking forward for some great answers.
Thank you for reading.

olson
08-02-2012, 05:08 PM
About Graphic card,
I want to know whether they take part in any kind of simulation or calculation.
Quadro or GTX?
Do they effect rendertime? Or they are just limited to the viewport?
Whats the deal with Tesla cards? How do I use them? Would they be helpful?


Graphics cards are primarily used to render 3D viewports. Faster card = smoother viewport. Some renderers and simulators use the graphics card to accelerate calculations with OpenCL or CUDA but those are fairly niche. Gaming cards and workstation cards are very similar but have firmware and driver tweaks to improve various tasks. Gaming cards can be used in professional 3D applications and vice versa but there are some caveats (too many to cover here). Tesla and FireStream cards have similar hardware to their workstation and gaming cousins but are for processing only and have no display outputs.


About Processor,
I can either get an i7 2600k or Xeon X6550 x 2 (LGA1567) or Intel Xeon E7-4830 (LGA 1567 ES)
Which would be more benificial?
For Xeon processor, can someone please recommend a motherboard for it as I can not find one with pci-e slot.
And in the Dual socket Xeon board, can I use it with just one processor or do I need to
install 2 physical processors for it to work?


It all depends on how well the software being used takes advantage of multiple processor cores. Software that takes advantage of many cores (like ray tracing rendering and a lot of simulations) will benefit from workstation hardware with more cores (like the Xeon configurations). Software that doesn't use multiple processor cores or uses them poorly (like Adobe products and most of Maya) will benefit from the higher clock speed of a single socket machine.


About HDD,
I already have a spare 2Tb Hdd, so dont need one but would like to know
whether I'll need a SSD. Is it really beneficial?


Middle and higher end SSD are faster than HDD but come at a much higher cost and lower capacity compared to HDD. They offer some benefits but they are not worth it at the expense of other components like the processor or memory. So if you can spend an extra $400 on a SSD go for it, if you would have to cut back on other components to afford it you're better off with HDD.


About Powersupply,
Got a spare 600W Coolermaster so i guess don't need to worry about it as most modern card are compatible with it. (not sure about TESLA)


That's enough for almost all single socket machines out there. Dual socket systems might need more power depending on the power requirements of the specific components.


About RAM,
I have 2x4GB 1333 RAM and when I check the RAM usage its hardly up 3.5gb. Do I need more RAM?
Would also like to know what difference does it make if I get the something like corsair Vengence?


Right now 16GB of memory is less than $100. Even if you don't need it (yet) I'd get it anyway. The "Vengence" and other stupidly named products like "Ripjaws" are targeted towards gamers and is 99% marketing. They have slightly better performance but its not worth the premium and also some require higher operating voltage which can cause some issues on its own.

AllanGuigan
08-05-2012, 08:09 AM
Finally a reply.
Thanks a lot buddy, I really appreciate it.
I need just one more favor to ask, please suggest me a motherboard for Intel Xeon X6550,
which has a pci express (mostly going for 560 Ti) slot and 2x slot for cpu.
Once again,
Thanks a lot.

vlad
08-05-2012, 03:47 PM
But why build an expensive system around an already outdated platform? If I may ask, how much are you paying for those Xeons?

AllanGuigan
08-05-2012, 04:09 PM
Lets say a friend is selling me the xeons at the same price I'd get a single Intel 3960X.
With Xeon I'd get 16 core - 32 threads (as I'll get 2 physical CPU's) vs 6 core 12 threads of 3960.
According to basic math, I think Xeon is a better option (atleast I hope so!)
And one more thing, I'm also confused between GTX 560 Ti and GTX 580. The GTX 6xx series is too expensive for me. I decided not to go for Quadro.
I am not a gamer though, but I do occasionally play game which do not need dedicated graphic card! (You guessed it, SOLITAIRE!)
Thanks so much guys for the interest and efforts!
Really appreciate it.

vlad
08-05-2012, 06:43 PM
Perhaps, but the X6550 is (gasp) 2.0 ghz, with a max OC capability of 2.4ghz. So 16 cores would total 38ghz (with HT taken out of the equation for simplicity's sake). Consider a 3930K OC'ed to 4.8ghz would give you 29ghz for half the price. BUT it would be twice as fast as the Xeon system is single threaded applications, which represents the vast majority of the tasks you'll likely be doing. Now if you factor in the additional cost of a dual socket mobo, fatter psu, e-atx case (or even hptx), you could build a barebone render slave equipped with the same 3930 equally OC'ed to 4.8ghz for probably not that much more $$ in total for the 2 boxes than the single 2 socket system.
The twin box system would total 58ghz (20 more than the dual Xeon one) for rendering while being twice as fast for pretty much everything else. Plus you have a backup if something goes wrong with the main one.

I've always been a strong proponent of dual socket builds in the past, but now, IMO they only make sense if money is not a factor and you go all for the fastest, most expensive octocores and max out ram slots with the fastest ram available, multiple Quadro 6000 galore for GPU rendering, PCI SSD drives and the whole shebang.

And like I said, socket 1567 is already over 3 years old...

AllanGuigan
08-05-2012, 07:35 PM
Hmm...
You make a very interesting point my friend.
I'm not too familiar with OC'ing stuff.
Lets say I drop the consideration of the xeons,
which other processors would you suggest in comparison to the almighty 3930K.
(Including i7 990x and even AMD's)

A very innocent question:
What matters more, number of cores and threads or the processor speed (ie, GHz)
Fore example,
Lets say CPU A has 4 cores and 8 threads clocked at 2 GHz
and CPU B has 2 cores and 4 threads clocked at 4 GHz
(Not considering to OC anything)
Would both of them perform almost the same? If yes, then why the need to divide cpu's with cores and threads? And what would the difference between Xeon and I7 be?
I think I know the answer but want to make sure.
I hope I'm not annoying people with my questions, but let me tell you guys, I've been looking for answers for so long, but here (where I'm at) no one seems to know much about the hardware.
Thanks once again.

Srek
08-05-2012, 08:27 PM
Perhaps, but the X6550 is (gasp) 2.0 ghz, with a max OC capability of 2.4ghz. So 16 cores would total 38ghz (with HT taken out of the equation for simplicity's sake). Consider a 3930K OC'ed to 4.8ghz would give you 29ghz for half the price. BUT it would be twice as fast as the Xeon system is single threaded applications, which represents the vast majority of the tasks you'll likely be doing....
Just one more nail for the X6550 coffin, it's 45nm based and slower per GHz than the 3930
Cheers
Björn

tswalk
08-05-2012, 11:53 PM
I would agree with Vlad.. the xeon's vs i7 price/performance for a workstation is not that good.

however, it would make an ideal render server to offload things to.

AllanGuigan
08-19-2012, 01:23 AM
Thanks a lot guys.
Your feedback has been very helpful.
As of now, I'm gonna go with the 3930K.
Still a bit confused regarding the graphic card {(GTX 560ti & GTX 580) suggest if you can!}
Can anyone of you please tell me about the cooling system for the 3930?
I have never used anything apart from the cooler fan supplied with the processor.
Thanks again guys.
You all have been very helpful

vlad
08-19-2012, 02:32 PM
Thanks a lot guys.
Your feedback has been very helpful.
As of now, I'm gonna go with the 3930K.
Still a bit confused regarding the graphic card {(GTX 560ti & GTX 580) suggest if you can!}
Can anyone of you please tell me about the cooling system for the 3930?
I have never used anything apart from the cooler fan supplied with the processor.
Thanks again guys.
You all have been very helpful

I have both cards in different systems and the 580 is a bit faster obviously. If you can afford it, go for it. If you cant, the 560 isnt a bad performer either.

I have the Noctua NH-D14 on my 3930. It's considered by many the best air cooler out there. I was hesitating between this and the Corsair H100 water cooling system. Benchmarks put them pretty much on par so I went with the Noctua, mainly for peace of mind. Plus it's quieter. It's one huge chunk of hardware though, so make sure you have room for it in your case and use low profile ram sticks to avoid clearance issues. It gives me around 75C under full load on a 4.5ghz overclock.

coldside-digital
08-19-2012, 05:01 PM
About RAM,
I have 2x4GB 1333 RAM and when I check the RAM usage its hardly up 3.5gb. Do I need more RAM?


Sounds to me like you are using a 32-bit operating system and/or 32-bit software if you only use upto 3.5gb of RAM out of 8gb while doing heavy work.

Make sure to factor upgrades to 64-bit software in your pricing (if you will need to pay for upgrades to 64-bit versions, depending on the software)

AllanGuigan
08-20-2012, 02:09 PM
No buddy,
I use everything in 64 bit.
Its just that I don't use all the softwares at the same time!
And Vlad, that was quick man!
Please suggest a motherboard as well cause there are so many to choose from.
I'm not gona sli so don't need multiple express slots!
And how good is GTX 460 2win?
It says it outperforms the 580.
Any thoughts on that.
And please suggest a good motherboard.
Thanks again guys.

sentry66
08-20-2012, 05:04 PM
my asus sabertooth x79 works really well at 4.7ghz for the 3930k. I'm sure there's others that work just as well. Just don't go cheap if you're doing big overclocking. You'll need those chipset heatsinks and extra MOSFETS

AllanGuigan
09-01-2012, 03:41 PM
Thank you all for your replies, they were really helpful.
Will get my pc anytime now.
I still haven't ordered the graphic card yet.
Im confused with AMD's FirePro and FireGl and nVidia GTX and Quadro seriese.
Options:
nVidia Quadro 4000/2000, GTX 580, even considering GTX 460 2win (says its faster then 580)
AMD FirePro 7900
FireGl 8650
Please give me your suggestion and if possible even the reason for choosing.

Thanks again guys!

coldside-digital
09-01-2012, 10:02 PM
Although two GTX 460's may be faster than a 580 for gaming, a 3d application can only utilize one graphics card so there is not benefit in running SLI for 3d applications.

I have the Palit GTX 580 3GB model, and it performs fantastically in 3ds Max and Houdini.

tswalk
09-02-2012, 04:55 AM
i'ld avoid purchasing a quadro right now until after December when the K5000 and other kepler based models are released.

AllanGuigan
09-02-2012, 12:40 PM
Hey, I'm not talking about sli,
I'm talking about the twin gpu 460, a single card. Is it any good?
The 580 3gb is beyond my reach so wont be considering it.
What about the Firepro and FireGl cards?
Anyone with any experience with them?
AMD's are a bit cheaper but don't know about their professional cards.
Thanks again friends.

vlad
09-02-2012, 03:27 PM
Never heard of that twin 460. If it exists, it's probably the same as all other twin cards : 2 gpus in SLI in a single encasing. So again, useless for 3D. And if the 580 is out of your reach, any Quadro or Firepro priced within your reach will be a bottom end item and would perform much worse than a mid-range geforce, especially in Max.

AllanGuigan
09-02-2012, 05:53 PM
GTX 460 2win
http://www.evga.com/articles/00613/

I can get either the GTX 580 1.5gb or the 460 2win.

vlad
09-02-2012, 06:30 PM
Ah well, like I said, those 2 gpus inside are connected via SLI. So what you get in 3D apps is only one 460 gpu working + half the ram of whatever's inside that (since graphic ram is not cumulative between both gpus) and another 460 gpu doing nothing except provide more heat in an already hot environment (plus the other half of ram wasted). A single 460 will perform the same, cost less, draw less power and be cooler. Get the 1.5gb 580, no brainer. ;)

AllanGuigan
09-02-2012, 07:38 PM
Seems 580 1.5 gb is the option to go with.
One last thing, should i try to increase my budget a bit and go for the 660 ti?
or thers not much difference between the two?
The 580 is a 384 bit where as the 660 ti is 192 bit.
Does it matter?
Let me ask you in a different way, what aspects should be considered keeping in mind its gona be used for 3d applications,
the Memory bandwidth GB/s, number of cuda cores etc?
Cause in some places 580 shows higher numbers as compared to 660ti.
Don't know why do I keep asking soo many questions, I guess I just like to know the differences.

Srek
09-03-2012, 07:47 AM
Memory Bandwidth makes only a very small contribution to over all render or editor speed, i never bothered to check at all.
Cuda cores are only important if you have a specific use for them and the application you use utilizes them sufficently. Most general DCC apps don't do much if anything with them.
All the numbers you quoted are imo mostly marketing text, aimed to make people think they need the latest and greatest. This might be true for game benchmarks etc. but in real life the differences are often not noticable and sometimes even not measurable.
Get a current i7 with matching mainboard, a current single GPU graphics card in the upper middle range, two or three fast and big disks (it does not have to be SSD to be able to work fast and efficent), 16GB RAM and be done with it.
Just my 2 cents of course
Cheers
Björn

CGTalk Moderation
09-03-2012, 07:47 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.