Jump to content

Quadro 4000 Alterantive for 3DS Max 2013


Recommended Posts

Hello Everyone,

 

I have been reading the CG Architect forums for a while and love it, so much great information,

 

I would like everyones opinion on choosing a graphics card for a new rendering machine that will be using 3DS MAX 2013, The Quadro 4000 is the card I was qouted by a GPU computing company buy I dont see how it is required for my work and its also very expensive,

 

The Machine specs are:

i7 3930k 3.2Ghz 12MB Cache

36GB Kingston 1600mhz Ram

Asus P9X79

Samsung 256GB SSD

1TB Western Digital Re4

Thermaltake Toughpower XT 875W PS

 

I selected these components based on Andrew Lynn's list on amazon and all that is left is to choose the graphics card.

 

The machine will be used for very detailed renders,

 

Please feel free to make any sugestions about the system specs, I was hoping I might be able to use a card like the GTX 580

 

Thank You,

 

Matt

Link to comment
Share on other sites

Gtx 580 is good solution. But not enough!! Because other helpers (like photoshop, ae, premier pro etc..) use cuda (opengl) for speed and performance!! So i think u have to use minimum Quatro 4000!!!

I want to use gtx series 'couse these series is the performance series, powerfull series but their driver does not fully support for using cuda microprocessors SO!!!!!!!! you have to take a Quatro series like Quatro 4000

Link to comment
Share on other sites

I really think the Quadro cards are way overpriced. They have no extra value over a GTX card. I have a GTX 570 and love it. Maybe look at the GTX 670. You can get a 670 with 4GB of memory, 1006Mhz core clock, and 1344 CUDA cores. You can pick one up on Newegg for $500US. Compare the specs and price of the 4000 and its a no-brainer.

Link to comment
Share on other sites

woah 36gb of ram!

im no expert but i think a GTX 580 or similiar would be OK for you...Iv got one and am happy with performance on heavy 3d scenes.

depends on how detailed you are going to get i suppose!

 

this article compares them in detail and is worth a read

http://www.nyc3d.org/3dwp/PortfolioItems/geforce-gtx-580-vs-quadro-5000/

 

Great article! I personally would look at getting two GTX580's over one Quadro5000. The 5000 is just crazy priced! But again, it depends on what you need it for.

Link to comment
Share on other sites

I just built a new computer last week and it has an almost identical configuration. I'm using the GTX 580 with 3gb of memory (EVGA Classified) and it works nicely! The thing about the GTX 580 that you need to be aware of is that it takes 2 - 8 pin power connectors and one 6 pin, and the size of the thing just barely fit into the case I bought (Antec 300). There's much debate over Quadro versus GTX and I've found that the GTX cards work excellent for 3ds Max and my budget.

Link to comment
Share on other sites

I tend to disagree with Ahmet: for applications that are CUDA optimized, the raw power of the GTX 580 with its greater CUDA core count and faster core clock is unmatched by any of the current Quadros - even the 5000 and the 6000.

 

The issue exactly that most of the applications today - other than GPU renderers, some certain functions within Adobe Suite apps, Mari etc - do not rely heavily on CUDA or OpenCL.

 

If you are after purely viewport performance, Quadro cards perform better than GTX cards, but whether the difference is enough to justify the increased cost is situational and relevant to the exact scene complexity - as illustrated by the linked article a few posts above. Good value is also offered by the AMD FirePro cards, with the V5900 and V7900 having a very nice price/performance ratio. AMD Introduced the FirePro 8xxx series a month ago or so, so I would expect price drops for the previous generation FirePros. In programs that use OpenGL API for viewport acceleration (pretty much anything but 3DS Max), AMD cards usually have better performance in the mid range Pro cards, and Radeons have a definite advantage over GTX cards which in general are not optimized for OpenGL.

 

Be careful with your specs, two small notes:

Get a more modern PSU.

Wattage is (more than) enough, yet this Thermaltake is a bit outdated and has mediocre reviews. A 80+ Silver/Gold does make sense when working with energy hogs like the 3930K and powerful GTX/Radeon cards (especially the GTX 580) for long periods of time.

 

Most likely you want to say you will get 32GB of RAM, not 36, as the X79 platform utilizes quad channel Ram and makes sense using 4x or identical sticks - most likely 4x8GB -> 32.

I would get 1866 DDR3, which is a bit faster than 1600, yet not that much more expensive.

 

Edit: myself opted for a EVGA GTX SC670 4GB instead of a GTX 580 3GB. I could not wait for working drivers with VRay RT in order to go for the $300 Radeon 7950 3GB which in OpenCL simply embarrasses all nvidia offerings.

 

In my mind the vastly superior thermal and noise performance offered by Kepler 6xx cards, with the added bonus of some more VRam outweighs the 7-8% or less loss in performance for GPU rendering (real life results reported in Chaosgroup's forums want the 680 being 4% slower than the GTX 580 in current VRay RT / current drivers).

 

Viewport performance is not worse. The card is still big, yet a tad smaller, and if you really want fast GPU rendering you can add 3x 670s that will be faster, quieter and much easier to power with a single PSU than 2x 580s, while having that little bit of extra VRam to play with.

Edited by dtolios
Link to comment
Share on other sites

Very detailed Dimitris,

 

Thank you for your help, my setup is 32GB of ram (4x8) it was just a spelling mistake,

 

I ended up finding a Quadro 4000 for $400 and for the price and driver compatibilily I though it would be a good decision,

 

The shop did not have the Thermaltake Toughpower and I got the TR2 Bronze.

 

Once again thank you for the help, lets see how well the build goes,

Link to comment
Share on other sites

  • 4 months later...

Hi, I am new to this community. I realise this in a fairly old post but I am curious about something you said....

 

if you really want fast GPU rendering you can add 3x 670s that will be faster, quieter and much easier to power with a single PSU than 2x 580s, while having that little bit of extra VRam to play with.

 

I have a GTX 680 4GB, if I had one / two more of these, would it double/triple my rendering speeds or does it not work like that?

 

I use Vray, Fumefx and trying to learn Mental Ray slowly.

 

Would this also have any effect on my viewport speeds?

 

Thanks in advance

 

Sean

Link to comment
Share on other sites

VRay RT GPU / iRay / Octane and other activeshade-like GPU accelerated plugins do make use of more than one GPUs. I know fumefx has GPU accel, but I don't know if it will get better - I would guess yes. You don't have to have identical cards for the above programs to take adv. of them (i.e. SLI).

 

3DS viewport does not care for multiple cards, whether SLI linked or not.

 

Mental Ray renderer, which is exclusively CPU based as it comes with current 3DS won't care for your GPUs. iRay is mental ray's current unbiased GPU accel. engine.

Link to comment
Share on other sites

a few numbers...

 

http://www.tomshardware.co.uk/forum/382946-15-quadro-gtx680-hd7970-viewport-performance-fight

 

btw. i think it would be very interesting to get a 3dsmax viewport benchmark thread going, maybe using one small and one heavy scene to compare the performances of professional and gaming cards. There is still not much information available concerning this topic...

Link to comment
Share on other sites

Yeah, I did some tests myself using a few affordable Quadro cards, tested against my "own" GTX 670 4GB and the results were not that encouraging for the 670...but it depends on the application.

 

For sure Maya and 3DS are not horrible with the gaming cards, but even a Quadro 600 trumps them in most cases (yeah, castrated drivers, still...).

 

I did not do any "canned" 3DS testing, as the SPECapc 3DSmax module is not free.

I would be interested in setting up a standardized test...maybe an animated camera traveling around a high poly interior space model? Or exterior with trees etc. Then "Fraps" for counting frames?

Link to comment
Share on other sites

some visual comparisons quadro 600 vs gtx 560 ti (but no automated benchmark)

 

geforce 560ti:

 

quadro 600:

 

 

i just found this comparison chart of max 2013 using Nitrous:

Video+test+20Mln+tablica+BALL+1.jpg

Video+test+20Mln+tablica+BALL+2.jpg

 

from here: http://diasarnaat.blogspot.de/p/21.html

 

here it looks like the gtx6xx are not so bad for max 2013...

 

but the problem (as always) is that there are many different resolutions and processors used. The test should be limited to 1920x1080 (or 1200) and the table should be sorted by processor.

Edited by numerobis
Link to comment
Share on other sites

I honestly don't know how the thing operates...probably due to the K5000 kepler based Quadro now being offered as "the future", Autodesk optimized a bit for Kepler? Same for iRay which received no love before 2013 - i.e. "you have to have a Quadro or Tesla, and if there are none such Kepler cards yet, wait in line or get a "real" pro-card".

 

I am tempted to upload a single Specviewperf / Specapc Maya 2012 run in youtube, but I was "burned" running those for 4-5x cards the other day (Maya test was about 45m-1h each). Ofc I won't do 3-4x loops this time, still...it is painful (plus I could not find my fraps serial).

 

...all my tests were running on the same i7-3820 4.5GHz, but afaik the CPU plays very little role, unless we are talking something seriously outdated.

Link to comment
Share on other sites

i don't know how the frame counts in this chart are measured. I think by eye...

There are some entries marked with "Fraps". So i think some had used Fraps as you suggested. But it looks like Fraps can only count full frames when i compare the entries. So in this case it hink a smaller scene resulting in higher frame rates would be better to differentiate the results.

I don't know if there are any scripts available that can run a benchmark inside max (load a scene, switch to a defined single frame or quad frame, start the animation and display capture the rfame results).

Link to comment
Share on other sites

Yeah, fraps won't do you much good in scenes where you register "2.5" vs. "2.9" fps :p

Yes, I would be thinking of a medium complexity scene that spans some time, like 2 minutes, during which you would record the output (perhaps with fraps, displaying the actual fps embedded in the video), and then actually see the total count of frames the card was able to render those 120 seconds and build a small database.

 

The only standardized script for 3DS max I know of is the Specapc 2011 - that is before Nitrous, and it does cost $495 if you want to publish your results...there is a personal $20 license, that I would not pay for something in many ways outdated, that I could still not use in blogs or forums.

Link to comment
Share on other sites

yes, i've seen that 3dsmax is not included anymore in the free SPECviewperf suite... stupid.

 

I just saw that there is a thread about a viewport benchmark for particle animation.

 

http://forums.cgarchitect.com/72537-3ds-max-viewport-performance-test-9000-animate-particle.html

 

 

I found this benchmark script: http://www.scriptspot.com/bobo/mxs2/bmarker/

 

A bit aged, but maybe still useful...

Link to comment
Share on other sites

Sorry from google translates

Behavior in max Vray cards - does not fit into the accepted parameters for video cards.

Reasons - in the Max, in different settings, scenes Vray, in configurations of computers.

A determined with a video to work in max not, then what would not be surprised that fancy new card - works no better than the former in Max.

 

If it is impossible to choose the Max vidyahu by well-known test, the performance charts in the "clean",

you can try to test the "field" in those particular scene in Max-Vray.

 

Offered scene VideoTest20Mln

Scene sharpened universally fitted to the averages of real scenes.

 

Resolution in this scene is not important - it is one in the Max.

Effect of processors seen in the table.

 

The main difference - it's like he max2013 - valid card

and geometry - the number of

max 2012 - no this test

Link to comment
Share on other sites

  • 8 months later...

Yes. You can have both. The "Quadro driver" doesn't really do much for the GTX. The GTX will kick in when a cuda application pretty much asks it to.

I have tried it and it was working flawlessly with Quadro drivers for the quadro, and Geforce drivers for the GTX.

 

Have tried combinations of Quadro 4000, 2000, 600, K2000 and GTX 670, 660 and GTX Titan.

Also tried 4000 + 2x GTX cards, a 670 and a Titan.

If you've already have a driver that works with your version of VRay RT*, the combination works from the get-go with no glitches, and you can GPGPU progressive renderings with whichever combination you wish: just the GTX, just one of the many cards, all the cards together etc.

 

*sometimes the "latest" doesn't work with the version of VRay RT and there has to be some back and forth trial/error, but not due to a Quadro / GTX conflict. I would not work even if you had a single card if the driver version "breaks" something.

 

If I would get a GTX for compute, I would go for a 4GB 670 or 760 as the "cheapest"...going lower in CUDA cores starts being inneficient.

Edited by dtolios
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...