thinking of a new PC Build - your thoughts?

[scostumatu]

Hi all

 

Firsty, forgive me as Im not too techy when it comes to computers - i like to think my talents lay in the creative field Im wanting to get a new PC built - i currently have a 4 year old (1st Generation) i7 950 @3.07ghz with 12 gig of ram and quadro 4000 graphics card (which i still love and think its a workhorse).

 

Im wanting to use it for 3DS and V-ray. So CPU power is important.

I filled out a form for a custom online build (see below link) and it come to around $2500 AUD

http://www.mln.com.au/product/desktop_customiser.php?item_id=4170

 

Here are the specs:

Processor: Intel Core i7-4930K Processor (Six Core, 12MB Cache, Unlocked for Overclocking)

Motherboard: ASRock X79 Extreme6 Motherboard

Memory: Corsair Dominator GT 16GB DDR3 Memory 2133MHz

Solid State Drive: Samsung 840 EVO 120GB SSD (RW : 540MB/s,410MB/s)

Hard Drive Storage: 1TB Hard Drive

Graphics Card: nVidia GeForce GT 630 2GB Graphics

Operating System: Excludes Operating System

Audio: Onboard HD Audio

Wireless Connectivity: 802.11n Wireless Receiver

Cooling System: Venom Stage 2 Liquid Cooling System

Power Supply: Seasonic 80 Plus Gold Rating 450W PSU

 

I know it may sound like a silly question, but I was wondering if I would notice a huge speed difference in terms of rendering with this new system compared to my existing?

Is this system a good choice or is it over/under done?

 

Ideally I would love to build myself a mini render farm but i wouldnt know where to start and all the local companies dont have much of an idea about it or what to do.

 

Any assistance is much appreciated - thanks guys.

[Zdravko Barisic]

Memory: Corsair Dominator GT 16GB DDR3 Memory 2133MHz

>>>> Can you go with 32Gb, at least?

 

 

Graphics Card: nVidia GeForce GT 630 2GB Graphics

>>>Old Quadro 4000 will perform better than GT 630

 

Power Supply: Seasonic 80 Plus Gold Rating 450W PSU

>>> not so much power for 4930K

[scostumatu]

Yes I was planning on getting 32 - is 64 MB RAM overkill?

re - graphics card, so is my quadro still considered a good graphics card?

 

re power supply - a friend of mine told me exactly the same thing

 

thanks for your response

 

Memory: Corsair Dominator GT 16GB DDR3 Memory 2133MHz

>>>> Can you go with 32Gb, at least?

 

 

Graphics Card: nVidia GeForce GT 630 2GB Graphics

>>>Old Quadro 4000 will perform better than GT 630

 

Power Supply: Seasonic 80 Plus Gold Rating 450W PSU

>>> not so much power for 4930K

[maymoonc]

yes you will find a huge difference especially it has 6 cores

definitely you need a better power supply & gpu as Zdravko Barisic said

 

 

If you don’t have more money just replace liquid cooling with air cooling

 

 

Edited July 16, 2014 by maymoonc

[Dimitris Tolios]

The GT 630 is horribly underpowered - its not just that the Q4000 works ok still, its the GT630 that is horrible!

If you are not getting something at least as fast as a GTX 750Ti, don't even think about replacing your Quadro.

 

450W is not that bad - but if you would plan to upgrade the GPU down the line, I would consider a 500-550W, to be on the safe side. I doubt that you will have issues with the 450W without overcloking the 4930 and adding a powerful GPU.

 

I have a 500W that can support overclocked 4770K @ 4.6GHz (nets more power draw than a stock 4930K easily) + used as a testbench has seen all shorts of GPUs on it - from GTX Titan (250W) to Quadro 4000/K4000 and W7000/V7900. Never complained even on long stress tests.

 

Dominator GTs are a waste of money - seriously. No performance advantage, its all about looks and the fancy heatsink.

2133 speeds don't yield measurable performance benefits. Going for a cheaper (per stick) 4x8GB solution, even 1866 speed, is much preferred.

[scostumatu]

Thanks for the response so far guys

 

Dimitris, Im happy to upgrade my power supply even to future proof it somewhat - its not that much of a price jump. So the i7 im looking at buying, is that a good choice in your opinion?

In terms of graphics card, if im to replace my q4000, what do you recommend? Happy to spend up to $800.

 

Thanks again

 

 

The GT 630 is horribly underpowered - its not just that the Q4000 works ok still, its the GT630 that is horrible!

If you are not getting something at least as fast as a GTX 750Ti, don't even think about replacing your Quadro.

 

450W is not that bad - but if you would plan to upgrade the GPU down the line, I would consider a 500-550W, to be on the safe side. I doubt that you will have issues with the 450W without overcloking the 4930 and adding a powerful GPU.

 

I have a 500W that can support overclocked 4770K @ 4.6GHz (nets more power draw than a stock 4930K easily) + used as a testbench has seen all shorts of GPUs on it - from GTX Titan (250W) to Quadro 4000/K4000 and W7000/V7900. Never complained even on long stress tests.

 

Dominator GTs are a waste of money - seriously. No performance advantage, its all about looks and the fancy heatsink.

2133 speeds don't yield measurable performance benefits. Going for a cheaper (per stick) 4x8GB solution, even 1866 speed, is much preferred.

[scostumatu]

PS - Dimitris, reading your article here and compiling a list of things to buy - your stuff is great!! keep up the good work!

http://pcfoo.com/2013/03/cg-workstation-the-pro-3930k/

[Dimitris Tolios]

Since you are reading the blog posts etc, I don't have much to add...

The 4930K choices appears to be a safe one. The upcoming Haswell-E based s2011-3 CPUs, will probably also be 6-core in the $600 and lower range, adding the expense of more expensive DDR4 and ofcourse "newer" X99 boards that will maintain a tad higher sale prices for the first few months.

 

DDR4 don't appear to be offering any groundbreaking advantages - Haswell, like Ivy and Sandy Bridge are not memory speed limited really, and especially rendering / modeling tasks don't really benefit from RAM speed bumps.

 

Supposingly those will launch by the end of Q3 2014, but I would not expect prices to "settle" before Xmas or early 2015 - might be wrong.

 

 

Just some thoughts on PSU + futureproofing:

 

Things get developed for lower consumption in a given performance, or in the worst case scenario stay the same.

Chips get designed to operate in a multitude of environments, and the strictest of those uses mandate thermal design: usually that's servers and small form machines.

 

Your i7-950, is notably slower, yet consumes more power by average vs. the i7-4930K. At least @ stock speeds.

The newer architectures utilize smaller lithographies => can keep higher clocks with lower voltage inputs and the efficiency of the lower power states gets better and better. Its not just laptops that get jumps in efficiency and nearly double the run times over those last 4-5 years.

 

Especially with combination with the GPU, with the 1st generation Fermi based Quadro 4000 being notoriously hot running and technically consuming more power by average in 3DS vs. any of the K5000 / K6000 / GTX 680 / 770 / 780 / Titan etc, any combination of those GPUs with a 4930K would yield lower average power draw vs. your current rig.

 

The ways you can possibly add to a rig to make it consume more:

 

* use completely different class of components - of similar architecture. Eg you had a 650Ti and you ad a 780. Both can do really low power states, both are Kepler architecture. Naturally the massively bigger chip on the 780 will consume more than that on the 650Ti, much like a 5.0L V8 consumes more gas idling than an economy 1.6L I4 of similar tech. Physics.

You had a Q4000 and you add a 780 At modeling environment the 780 will pull similar or lower average power with the Q4000, as the 780 is far more advanced in power saving features. Its not apples to apples anymore, thus "this" bigger engine (780) can actually idle and casually cruise around far more efficiently.

 

* do different tasks that stress your components more: the 780 in the example above, doesn't stress that much in 3DS due to the nature of the 3D engine. In fact getting any of GTX 760, 770 and 780 won't really do much different for your viewport performance: will be nearly identical. But GPGPU tasks, like accelerating Adobe apps, MARI etc or GPU rendering with Vray RT GPU/Octane etc, will start pushing those 200+ W GPUs to their rated TDP - much higher than what 3D viewports do.

 

* add more components: going for multiple GPUs - perhaps for gaming on the side or accelerating those GPGPU tasks even more, will of course increase the PSU requirements.

 

* overclock + any of the above: overclocking is adding power draw like nothing else: remember, higher clocks mean that current travels back and forth the CPU @ a higher rate. Add 20% more MHz = you add 20% more "back and forths" heating up the component that is subject to o/c.

Usually we increase Vcore ontop of increasing clocks, to ensure stability. Increased Vcore means higher current flowing, at an already higher back and forth rate = those 130W s2011 CPUs can easily go above 250W real power draw. Extreme overclocks easily pull above 300W with s2011, and s1155/1150 can easily reach 180-200W when pushed hard. Yes, each platform can nearly triple rated TDP.

 

So...after what appears to be another long - winded introduction:

 

How old is the system you are replacing now? 4-5 years old?

What are you keeping out of that rig to use on the new one? -> pretty much nothing after this long. RAM is compatible but slower than the average today, old HDDs cannot keep up with SSDs and newer/faster HDDs, perhaps the GPU is salvageable, the PSU is old... - and that's it.

 

Wait, what? Yes, PSUs don't like aging. Despite being nearly "solid state", capacitors inside PSUs get old and increase their possibilities of failing on you. If I had a 5yo PSU of really good quality, I would not retire it per se, but I would definitely think twice about using it in my new high-end workstation.

 

So, in my mind, unless you are leaving headroom for adding a second GPU or overclocking, a good 500W will do. You will rarely pull more than 200W with a 4930K rig in modeling / rendering tasks. Gaming on AAA titles? Maybe, still 500W is more than enough for getting 100% of both CPU and GPU simultaneously @ stock clocks.

 

By the time you will upgrade again, w/e higher capacity PSU will probably be "old".

Edited July 17, 2014 by dtolios

[scostumatu]

wow thanks so much for the extended info - much appreciated. I spoke with some local technicians today and this is what im most likely going to get:

 

FRACTAL DESIGN DEFINE R4 TITANIUM MIDI NO PSU (CASE

THRMLTKE 1275W TGHPWR XT GOLD PSU

ASUS P9X79-LE SKT2011/ATX/8xDDR3

KINGSTON 16GB KIT (2x8GB) 2133MHz x 2

INTEL I7-4930K SKT2011/NO FAN/12MB/6CORE

THERMALTAKE WATER 2.0 PRO LIQUID COOLER KIT

Samsung 256 SSD

LEADTEK QUADRO K4000 3GB

 

Sorry ive i copied and pasted (it was from the quotation). Im getting a little more watts in the power supply to accomodate for potentially more graphics cards as per your article ( GeForce GTX 770 SC 4GB) - might be a bit overkill but the price wasnt that much more.

 

 

 

Since you are reading the blog posts etc, I don't have much to add...

The 4930K choices appears to be a safe one. The upcoming Haswell-E based s2011-3 CPUs, will probably also be 6-core in the $600 and lower range, adding the expense of more expensive DDR4 and ofcourse "newer" X99 boards that will maintain a tad higher sale prices for the first few months.

 

DDR4 don't appear to be offering any groundbreaking advantages - Haswell, like Ivy and Sandy Bridge are not memory speed limited really, and especially rendering / modeling tasks don't really benefit from RAM speed bumps.

 

Supposingly those will launch by the end of Q3 2014, but I would not expect prices to "settle" before Xmas or early 2015 - might be wrong.

 

 

Just some thoughts on PSU + futureproofing:

 

Things get developed for lower consumption in a given performance, or in the worst case scenario stay the same.

Chips get designed to operate in a multitude of environments, and the strictest of those uses mandate thermal design: usually that's servers and small form machines.

 

Your i7-950, is notably slower, yet consumes more power by average vs. the i7-4930K. At least @ stock speeds.

The newer architectures utilize smaller lithographies => can keep higher clocks with lower voltage inputs and the efficiency of the lower power states gets better and better. Its not just laptops that get jumps in efficiency and nearly double the run times over those last 4-5 years.

 

Especially with combination with the GPU, with the 1st generation Fermi based Quadro 4000 being notoriously hot running and technically consuming more power by average in 3DS vs. any of the K5000 / K6000 / GTX 680 / 770 / 780 / Titan etc, any combination of those GPUs with a 4930K would yield lower average power draw vs. your current rig.

 

The ways you can possibly add to a rig to make it consume more:

 

* use completely different class of components - of similar architecture. Eg you had a 650Ti and you ad a 780. Both can do really low power states, both are Kepler architecture. Naturally the massively bigger chip on the 780 will consume more than that on the 650Ti, much like a 5.0L V8 consumes more gas idling than an economy 1.6L I4 of similar tech. Physics.

You had a Q4000 and you add a 780 At modeling environment the 780 will pull similar or lower average power with the Q4000, as the 780 is far more advanced in power saving features. Its not apples to apples anymore, thus "this" bigger engine (780) can actually idle and casually cruise around far more efficiently.

 

* do different tasks that stress your components more: the 780 in the example above, doesn't stress that much in 3DS due to the nature of the 3D engine. In fact getting any of GTX 760, 770 and 780 won't really do much different for your viewport performance: will be nearly identical. But GPGPU tasks, like accelerating Adobe apps, MARI etc or GPU rendering with Vray RT GPU/Octane etc, will start pushing those 200+ W GPUs to their rated TDP - much higher than what 3D viewports do.

 

* add more components: going for multiple GPUs - perhaps for gaming on the side or accelerating those GPGPU tasks even more, will of course increase the PSU requirements.

 

* overclock + any of the above: overclocking is adding power draw like nothing else: remember, higher clocks mean that current travels back and forth the CPU @ a higher rate. Add 20% more MHz = you add 20% more "back and forths" heating up the component that is subject to o/c.

Usually we increase Vcore ontop of increasing clocks, to ensure stability. Increased Vcore means higher current flowing, at an already higher back and forth rate = those 130W s2011 CPUs can easily go above 250W real power draw. Extreme overclocks easily pull above 300W with s2011, and s1155/1150 can easily reach 180-200W when pushed hard. Yes, each platform can nearly triple rated TDP.

 

So...after what appears to be another long - winded introduction:

 

How old is the system you are replacing now? 4-5 years old?

What are you keeping out of that rig to use on the new one? -> pretty much nothing after this long. RAM is compatible but slower than the average today, old HDDs cannot keep up with SSDs and newer/faster HDDs, perhaps the GPU is salvageable, the PSU is old... - and that's it.

 

Wait, what? Yes, PSUs don't like aging. Despite being nearly "solid state", capacitors inside PSUs get old and increase their possibilities of failing on you. If I had a 5yo PSU of really good quality, I would not retire it per se, but I would definitely think twice about using it in my new high-end workstation.

 

So, in my mind, unless you are leaving headroom for adding a second GPU or overclocking, a good 500W will do. You will rarely pull more than 200W with a 4930K rig in modeling / rendering tasks. Gaming on AAA titles? Maybe, still 500W is more than enough for getting 100% of both CPU and GPU simultaneously @ stock clocks.

 

By the time you will upgrade again, w/e higher capacity PSU will probably be "old".

[Zdravko Barisic]

If you are doung archiviz, and I guess you are, than 32GB or RAM, and vanilla GTX660-760-770 will make more for you, than 16GB of RAM and Quadro K4000.

The Quadro age is gone forever in 3DSMAX, for last few years, unless you pick head model like K6000, which is equivalent on GTX680.

So, go for more ram and gaming GTX model.

I am talking from personal view, not reading the net. Quadro for 3dsmax, have no use anymore. Nitrous display driver made that.

...

With 32GB of RAM, you can use multiple instance of MAX, and use some of these magic scripts.

http://www.scriptspot.com/search/apachesolr_search/copy%20paste

...and have open some image and cad editor, like a charm.

 

...

GTX 760/770 4 GB is best by for 3ds max, these days, and 2GB edition for common tasks, like interior and studio setups.

Good luck!

[scostumatu]

Thanks Zdravko - I just came across this youtube link and im pretty impressed. Im thinking 1 and maybe 2 gtx 780s would be awesome.

or here

 

so just to confirm - i if i did get a gtx 780, viewport performance would be as good as the current quadro 4000 i have?

 

cheers

 

 

If you are doung archiviz, and I guess you are, than 32GB or RAM, and vanilla GTX660-760-770 will make more for you, than 16GB of RAM and Quadro K4000.

The Quadro age is gone forever in 3DSMAX, for last few years, unless you pick head model like K6000, which is equivalent on GTX680.

So, go for more ram and gaming GTX model.

I am talking from personal view, not reading the net. Quadro for 3dsmax, have no use anymore. Nitrous display driver made that.

...

With 32GB of RAM, you can use multiple instance of MAX, and use some of these magic scripts.

http://www.scriptspot.com/search/apachesolr_search/copy%20paste

...and have open some image and cad editor, like a charm.

 

...

GTX 760/770 4 GB is best by for 3ds max, these days, and 2GB edition for common tasks, like interior and studio setups.

Good luck!

Edited July 18, 2014 by scostumatu

[Zdravko Barisic]

No need for 2 GPUs, unless you are using graphice renderers in active production, in everyday work. And I am pretty shore you dont.

So, just grab some 760 4Gb, and thats it.

750/770/780 are almost same in viewport manipulation, and they are not slower than Q4000, for shore, can only be better.

[scostumatu]

I do use VRAY RT, but usally just for checking lighting and materials. If having 2 speeds that up significantly, then its worth it in my opinion. WIll having 2 speed that up dramatically?

So as far as my current quadro 4000 goes, just confirming that i can use it alongside 1 or 2 760/770s?

 

Cheers again

 

No need for 2 GPUs, unless you are using graphice renderers in active production, in everyday work. And I am pretty shore you dont.

So, just grab some 760 4Gb, and thats it.

750/770/780 are almost same in viewport manipulation, and they are not slower than Q4000, for shore, can only be better.

[Zdravko Barisic]

Depends on mobo, how many PCI express slots do you have, but 2 GPUs you can use for shore, using 3 GPUs needs some extra PSU like 850W.

If you use VRT, 780 will be muche better than 760/770.

Buying 2 780 will be best buy, so you can sold you Q4000, as used GPU for some money.

This is good CUDA benchmark tool:

http://www.randomcontrol.com/arionbench

[scostumatu]

the power supply i intend to get has something like 1000W so that should be fine. In terms of express slots, ill have enough to cater atleast 3 or 4 from memory of what the technician said.

Ill look into the 780s now - i might start with 2 to begin - thanks for your help dude.

 

Depends on mobo, how many PCI express slots do you have, but 2 GPUs you can use for shore, using 3 GPUs needs some extra PSU like 850W.

If you use VRT, 780 will be muche better than 760/770.

Buying 2 780 will be best buy, so you can sold you Q4000, as used GPU for some money.

This is good CUDA benchmark tool:

http://www.randomcontrol.com/arionbench

[paneli]

Thanks Zdravko - I just came across this youtube link and im pretty impressed. Im thinking 1 and maybe 2 gtx 780s would be awesome.

or here

 

so just to confirm - i if i did get a gtx 780, viewport performance would be as good as the current quadro 4000 i have?

 

cheers

 

NO. GTX780 is NO match to Quadro 4000. The Quadro is an overpriced piece of crap.

 

This is my long story told short:

I had a huge problem about a year ago with my previous workstation that forced me to buy a new nearly identical workstation and send the broken one to the lab.

 

I was in a lot of pressure and short of time, I couldn't disappoint my clients so I just did things as fast as possible. I didn't want any surprises and wanted to go back to work as fast as possible so both machines ended up the same.

 

The only difference is the GTX 780 instead of Quadro 4000.

 

Since day 1 the new workstation didn't want to work with Windows 7 (I got a BSOD when doing nothing. Even pluging a USB could cause BSOD. It was that bad), so I tried to install Windows 8.1 Preview.

Regarding Windows 8.1 - Its magic. This windows literally revived my new machine. All BSODs were gone for good.

 

Anyway. I installed all my software again and continued working.

After 10 days my previous workstation came back from repair (I was told they started memtest86 and went for a coffee. When they came back there were like 40K errors. They showed me a picture of it, unfortunately I don't have it today) and only after I launched the same project on it I realized what I have been missing with this stupid Quadro.

 

I recreated the same conditions in both machines (both machines are based on same hardware so I just mirrored the system image) and still the machine with the GTX behaved much better in viewports.

For test I made a scene full of GeoSpheres. The Quadro at 25 million poly was already a piss-off. The GTX was great at 25 million poly and at 70 million poly I stopped since I have proved that point to myself.

 

So - the age of Quadro for 3ds max is over.

[Dimitris Tolios]

Just make it clear that VRay RT GPU is NOT viewport - it is actual preview ala ActiveShade engine for Vray materials / lights, or for some production rendering engine - given you can work around some shortfalls / unsupported features - not that hard for most simple scenes.

 

3DS Max's viewport engine is using Direct3D, so the OpenGL optimized drivers that are the Quadro niche that remains, won't give you much of an advantage. Especially when you are comparing an old piece of hardware that retails in the sub $250 used, with a new $450~$500 GTX.

 

That said, only one nVidia card can be used with 3DS max. Having more than one - of any type - will only benefit VRay RT GPU. Just the GPU accelerated mode. Regular Vray Advanced - any version - won't care about you having 1, 2 or 20 GPUs.

 

Also, despite VRay RT GPU being able to squeeze almost 100% of the performance of as many GTX 780/780Ti/Titan/Titan Z etc you can throw at it, communicating with them almost directly with CUDA/OpenCL, viewport engines don't: the CPU has to pre-render / pre-"chew" portions of the scene as you are navigating through it (orbiting, zooming, panning etc) for the GPU to finish projecting in the screen space. This "preparation" adds lots of overhead exclusively to the CPU, so the GPU is rarely the only bottleneck in the display performance - at least not after a certain level.

 

You won't be able to reap 100% of the performance of a 780 or Titan with 3DS Max's engine, as there is no CPU that can really keep up. A GTX 760 or 770 will give you almost identical performance.

 

Investing straight away in 2x $500+ GPUs that can be utilized to 100% almost exclusively for GPU rendering, without being absolutely sure it suits your workflow, is not wise.

I would start with one, and getting a second once I am convinced that I can work within the limitations.

[scostumatu]

Thanks for the response guys. I have been tucked away doing extensive research and ive read Dimitris post at least 10 times (like i said in my initialy post im not too tech savvy - so i need to read it over haha) - having said that, its making a lot of sense.

I dont normally use VRAY RT with my quadro because its really bad - having said that, i really like the idea of tweaking lights/materials during an activeshade preview - it will help speed things up a lot i imagine instead of re-rendering and re-rendering - besides that its super cool to look at

 

Most of my scenes come to (at the most) 6million polygons - so i dont think im going to have an issue with it.

 

Eli - thats strange that you had such a hard time with your quadro - i bought mine 2nd hand on ebay and its been fantastic - not one thing bad i can say about it (oh actually, RT sucks) but for me it was all about viewport performance and it handles it very well.

 

Dimitris, there are 2 cards im considering (available locally) and if after reading your post there isnt much difference between the below?

Gigabyte 3GB GTX 780 for $555

EVGA 2G GTX 770 for $405

 

If not, then ill get the cheaper on obviously.

 

This is my final configeration - Dimitris, thanks so much as your post helped me a lot. Please let me know if anything is standing out and is better swapped with something else.

Power supply - Coolermaster Vanguard 1000watts 80+ Gold Full-Modular Power Supply Unit

motherboard - ASUS P9X79

RAM - Kingston HyperX Predator

CPU - i7 4930

Cooler - Noctua NH-D14

SSD - samsung pro 250gig

HDD barracude 2tb

CASE - so far theyve recommended a Corsair CC750D. Theyre expensive at $197 but they have recommended it. It has good reviews. I would prefer a case with a hard drive hot swap but having trouble finding the right one.

 

 

 

 

Just make it clear that VRay RT GPU is NOT viewport - it is actual preview ala ActiveShade engine for Vray materials / lights, or for some production rendering engine - given you can work around some shortfalls / unsupported features - not that hard for most simple scenes.

 

3DS Max's viewport engine is using Direct3D, so the OpenGL optimized drivers that are the Quadro niche that remains, won't give you much of an advantage. Especially when you are comparing an old piece of hardware that retails in the sub $250 used, with a new $450~$500 GTX.

 

That said, only one nVidia card can be used with 3DS max. Having more than one - of any type - will only benefit VRay RT GPU. Just the GPU accelerated mode. Regular Vray Advanced - any version - won't care about you having 1, 2 or 20 GPUs.

 

Also, despite VRay RT GPU being able to squeeze almost 100% of the performance of as many GTX 780/780Ti/Titan/Titan Z etc you can throw at it, communicating with them almost directly with CUDA/OpenCL, viewport engines don't: the CPU has to pre-render / pre-"chew" portions of the scene as you are navigating through it (orbiting, zooming, panning etc) for the GPU to finish projecting in the screen space. This "preparation" adds lots of overhead exclusively to the CPU, so the GPU is rarely the only bottleneck in the display performance - at least not after a certain level.

 

You won't be able to reap 100% of the performance of a 780 or Titan with 3DS Max's engine, as there is no CPU that can really keep up. A GTX 760 or 770 will give you almost identical performance.

 

Investing straight away in 2x $500+ GPUs that can be utilized to 100% almost exclusively for GPU rendering, without being absolutely sure it suits your workflow, is not wise.

I would start with one, and getting a second once I am convinced that I can work within the limitations.

[Dimitris Tolios]

I don't believe either the 770 or the 780 will give you much of an advantage in viewport, so if it wasn't for Vray RT, I would say even a GTX 760 2GB should be sufficient, giving you 95-100% of the performance the 780 would.

 

It boils down to VRay RT and GPU accelerated compute (or games) to see real differences in performance by investing more money.

In this case, I would suggest you get the biggest Vram buffer available within your budget: that's 4 GB for the 760/770, or 6GB for the 780.

 

Most likely the 3GB with the "basic" 780 or even 2GB for the smaller 7xx GPUs will be enough for scenes with just 6-10M polys like those you report, but "just in case" you would like to try out more complex stuff, you could play it safe. Again, a 760 4GB would be a safe all-around bet, and should you need more speed, you could always add a second - either identical, or even faster. Nothing stops VRay RT from working with a 780 and a 760 or any combination and number of compatible cards - as long as the scene fits in each card's buffer independently.

 

So you could start either with a 780 3GB or a 760 4GB if you want to stay on the cheaper side for now, try out Vray RT and if it really suits your style (and you can "fit in RT's style") you can build upon with either identical or w/e cards are available at the time & budget when time comes.

[scostumatu]

awesome awesome thanks again - im re-assured now.

 

legend.

 

ok off i go to spend to money

 

I don't believe either the 770 or the 780 will give you much of an advantage in viewport, so if it wasn't for Vray RT, I would say even a GTX 760 2GB should be sufficient, giving you 95-100% of the performance the 780 would.

 

It boils down to VRay RT and GPU accelerated compute (or games) to see real differences in performance by investing more money.

In this case, I would suggest you get the biggest Vram buffer available within your budget: that's 4 GB for the 760/770, or 6GB for the 780.

 

Most likely the 3GB with the "basic" 780 or even 2GB for the smaller 7xx GPUs will be enough for scenes with just 6-10M polys like those you report, but "just in case" you would like to try out more complex stuff, you could play it safe. Again, a 760 4GB would be a safe all-around bet, and should you need more speed, you could always add a second - either identical, or even faster. Nothing stops VRay RT from working with a 780 and a 760 or any combination and number of compatible cards - as long as the scene fits in each card's buffer independently.

 

So you could start either with a 780 3GB or a 760 4GB if you want to stay on the cheaper side for now, try out Vray RT and if it really suits your style (and you can "fit in RT's style") you can build upon with either identical or w/e cards are available at the time & budget when time comes.

[Stephen Thomas]

I have a Quadro 2000 + Tesla C2075 running in Nvidia Maximus configuration on one machine, and a GTX 770 on the other. Performance is about the same for RT so I would definitely say go for the GeForce card. Especially since you could get multiple for the same price as the Quadro! I went for the EVGA OC Superclocked 4GB 770 and it is lovely, I have to say

[scostumatu]

nice one - so ive never had multiple cards on one machine. If I have my original quadro 4000 + a gtx 780, there shouldnt be any problems? Hows does the PC/software know which card to use for say viewport performance and vray RT?

 

I have a Quadro 2000 + Tesla C2075 running in Nvidia Maximus configuration on one machine, and a GTX 770 on the other. Performance is about the same for RT so I would definitely say go for the GeForce card. Especially since you could get multiple for the same price as the Quadro! I went for the EVGA OC Superclocked 4GB 770 and it is lovely, I have to say

[Stephen Thomas]

I think viewport performance is simply down to which card you plug your display into, which should be the Quadro. Vray RT will detect all cuda cores available, without the need for any SLI-type setup. Using the Quadro as your main display card will also give you the option of using higher colour depth monitors if you wish, which the gaming cards don't support.

[scostumatu]

awesome thanks a lot. Cant wait to get this put together! Ill get my hands on teh quadro k4000 when i can afford it, but for now the original quadro 4000 will do. Its a pretty good performer.

 

I think viewport performance is simply down to which card you plug your display into, which should be the Quadro. Vray RT will detect all cuda cores available, without the need for any SLI-type setup. Using the Quadro as your main display card will also give you the option of using higher colour depth monitors if you wish, which the gaming cards don't support.

[paneli]

In your place I would just take a GeForce and ditch the Quadro idea.

 

I don't understand why waisting 1000$ (in my country) on a useless Quadro when you can get a decent GeForce for half the price, if not more.