Installation & Licensing
Welcome to Autodesk’s Installation and Licensing Forums. Share your knowledge, ask questions, and explore popular Download, Installation, and Licensing topics.
cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Reply
Message 1 of 14
jub0raj
1970 Views, 13 Replies

Need help about GPU

hellow everyone i will be working with 2K resolutions and i will be buying a new computer but before i  have a question. i need a good GPU to do my work. my work consists Visual Effects, 3D modelling and Animations. im looking for some expert suggestions should i go for Quadro or Firepro? and which GPU should i go for? i have a restricted budget. i can spend only USD 1900 for GPU. which one do u suggest for me? thanks 🙂

Tags (1)
13 REPLIES 13
Message 2 of 14
eodeo
in reply to: jub0raj

If you dont plan on GPU rendering, get the best gaming GPU you can afford. If, however you do plan to use GPU based rendering like "iray" or "VRay RT" you should go for a quadro card as only they support CUDA based rendering. The only reason to go for a quadro over "gaming" GPU is video RAM size. While slower in every instance than their gaming cousians, quadro cards do come with larger video RAM that is perfect for this situation. CUDA based renderers need to place your entire scene in your video RAM as system RAM is ignored.

 

If you don't plan to use CUDA based rendering solutions, quadros are not your best option- not that they will work poorly @ 2k$ price point.

 

You should get a firepro only if you know exactly why you want it as its use is even more restricted than that of a quadro.

Message 3 of 14
darawork
in reply to: eodeo

GTX cards do support iRay and vRayRT. They have loads of CUDA cores, loads of them... and that's what you want. That and memory size, bandwidth and interface bit depth are purchasing factors;

 

Quadro K6000: 2880 CUDA Cores

GTX 780: 2304 CUDA Cores

 

Quadro K6000 Memory bandwidth: 288 G/Ps

GTX 780 Memory bandwidth: 288.4 G/Ps

 

Both have a 384bit memory interface.

 

Quadro K6000: $1600

GTX 780: $500

 

Hrmmmm.....

Darawork
AutoDesk User
Windows 10/11, 3DS Max 2022/24, Revit 2022, AutoCad 2024, Dell Precision 5810/20, ASUS DIY, nVidia Quadro P5000/RTX 5000/GTX760

Message 4 of 14
eodeo
in reply to: darawork

You could compare k6000 to its actual gaming counterpart 780 TI that is a different and noticably faster then the vanila gtx 780.

 

Quadro vs GTX

 

Core count:

2880 vs 2880

 

GPU clock speed:

900 vs 928mhz

 

Memory speed:

6000 vs 7000mhz

 

Memory size:

12 vs 3gb

 

Price:

6000$ (with 1000$ discount acording to newegg) vs 700$

http://www.newegg.com/Product/Product.aspx?Item=N82E16814133494

http://www.newegg.com/Product/Product.aspx?Item=N82E16814487002

 

Cards are virtually identicall in every other aspect. For 3d perfomance, 780ti is undesputed king. With that said, if you have a large 3d scene, and you need to more then 3gb of ram and (most important of all) plan to use GPU based rendering solutions like iray or Vray RT, quadro is the only way to go.

Message 5 of 14
drendir
in reply to: eodeo

even if i combine 4 TITAN GTX'es i wouldn't have half the price of a high end quadro...

 

Still wouldn't do for cuda based rendering?...

3dstudio max 2016/2017
Windows 7 & 10
Message 6 of 14
eodeo
in reply to: drendir

Titan is very fine for CUDA based rendering systems like iray and vray rt. With 6gb of ram and 2688 cuda cores, it's slightly slower than gtx 780ti, but has twice the amount of RAM, making it more desirable in most cases. The difference in core count is next to negligible, while ram size is doubled.

Message 7 of 14
drendir
in reply to: eodeo

hence my question: WHY GO QUADRO for the love of god i'm so lost in this marketing skeeme:

 

(1x) NVidia Quadro K6000 /// 6500€

memory: 12Gb

cuda cores: 2880

usage: 250Watt

 

(1x) GTX TITAN /// 961€

memory: 6Gb

cuda cores: 2688

usage: 250Watt

 

I mean... When u load your models' geometry into the GPU-ram, does this get spread over 2 GPU's-inner ram when using SLI-GPU-rendering? or will it bombard one and chocke it up?

 

In any case, it doesn't take a math genius to see where the money lies here? If i buy 2 TITANS i'm whiping the floor with a Quadro K6000 at a 33% cost of the K6000 and 200% Watt usage, 200% Cuda Core # and the same amount of GPU RAM

 

Am i seeing pink elephants or it Quadro a "professional" hype?

 

In another thread Steve Curley pointed out that Quadros are more reliable and on the long run more stable. I'm assuming a GTX will run fine for 2 years after which most pple will start thinking about updating their harware anyway looking at todays changing rate in hardware

 

So i'd "guess" gaming cards are seriously catching up with professional 3D cards to the point they're simply better then Quadros if you think of it as a whole rendering-package and in what you wanna invest your hard earned dollars...

I'm working on a GTX 760 now, coming from a Quadro 4000 and i have to say: GTX ❤️

 

regards,

 

Ethan

3dstudio max 2016/2017
Windows 7 & 10
Message 8 of 14
eodeo
in reply to: drendir

First to answer your question- getting more of the same GPUs do not increase final RAM size- only computational power is added, as each card has access to its own RAM amount- whatever it may be.

 

It should be pointed out that you can mix and match different cards, like a titan, a gtx 780, a quadro and a tesla card. All can and will be effectively used together for the GPU rendering in iray and vray rt. While the speed is nearly linearly added (as in 2+3=5 math) for core speed- all the cards will work in RAM amount of the smallest one. This is important as all the possible "extra" RAM of your quado, tesla and titan will idle.

 

With that said, it is paramount to understand what actually RAM does. It is used only for loading your scene. It provides no other benefits.

 

Take this hypothetical scene for example- you have an arch scene that takes up 4gb of ram to fully load. Gtx 780 is incapable of loading it and thus fails to render it. Titan loads it fine, with 2gb of ram idling. Quadro k6000 loads it fine with 8 gb of ram idling. If you have all 3 cards selected to render in the same machine, all will fail as the weakest of the bunch is unable to load it. You will need to manually uncheck gtx 780 and proceed with rendering.

 

3ds Max is telling you what amount of system RAM you're using right now- when you render in the special rendering window that pops up when you start rendering. This is roughly the same amount of RAM your GPU will need to complete rendering process too.

 

I'd like to point out that most of my scenes fit nicely into a single GB of ram. You should check your own scenes and see how much on average do you use. If your scene takes more than 12gb of ram to render, not even the highest-end quadro will fit you. However, for most arch/car scenes 3gb of ram is plenty and gtx 780 ti is the way to go- for fastest possible GPU rendering speed.

 

I'm pointing to arch/car scenes specifically as these are the clear examples where GPU rendering shines. I'm not sure how good GPU based rendering is for animation.

 

Now for reliability, it should be pointed out that all GPUs come with at least 3 year warranty. This means that should one accidentally fail, you'll get a replacement. After 3 years, you'll more likely than not want to get new GPUs as these nearly double in speed every 2 years or so.

 

It's only in a big enterprise environment where you'd want stability and longevity of a quadro as they will likely use these cards longer than a small studio would. For small studios, physical space trumps extra speed provided by old cards.

Message 9 of 14
drendir
in reply to: eodeo

yea u nailed my question pretty good.

 

So our "urban" scenes easily go up to 8Gb RAM usage when rendering everything out.

 

> Does this mean the same amount of ram will be needed from the GPU-ram? or are these two ram-modes (pure computer ram vs gpu-ram) differently "measured"?

 

- And i was wondering why so many "cpu-rendering"-example movies are based on shoes and glasses and diamonds and all these 5 by 5 meter objects that have generaly no real scale challange to em.

 

I seriously would love to hand them a 6million poly model for an urban environment that we need to put out in animations... sigh 😛

 

So yeah... all in all i think i'll stay withing the cheap GTX'es for now, they kick the ceiling out of photoshop and other graphical applications - as to the only reason for a high-end quadro is the memory size when rendering GPU-animations... i think i'll skip that for the next 2 years and see what better technologies come out. I'm sure there's some "urban-lover-guru-rendering-mastermind" out there who will think of something.***

 

Or the day GPU's can stack memory and the "weakest ram- link" won't be the standard anymore. that one is sneaky.

3dstudio max 2016/2017
Windows 7 & 10
Message 10 of 14
rehan_chaudhri
in reply to: eodeo

whic gpu is best for 3ds max vray rt rendering?

quadro 5000 or gtx 780ti?

gtx 780 ti supprot vray or iray rendering?

 

Message 11 of 14
darawork
in reply to: rehan_chaudhri

Hi,

 

Please read the above posts, and make your decision.

Considering that the K5000 is lesser spec than the K6000,

I think that the answer to both questions should be easy to work out.

 

Robot wink

Darawork
AutoDesk User
Windows 10/11, 3DS Max 2022/24, Revit 2022, AutoCad 2024, Dell Precision 5810/20, ASUS DIY, nVidia Quadro P5000/RTX 5000/GTX760

Message 12 of 14
eodeo
in reply to: rehan_chaudhri

"whic gpu is best for 3ds max vray rt rendering? quadro 5000 or gtx 780ti? gtx 780 ti supprot vray or iray rendering?" All post 2006 Nvidia GPUs support CUDA and thus vray RT, including the fairly modern, if a bit obsolete gtx 780 ti. It's important to note that GTX 780ti (2880 cores) is Keppler based GPU and as such it has more CUDA cores compared to the newer (and current as of march 2016) Maxwell based GPUs like the GTX 970 (1664 cores), but is in fact slower for most things, including GPU based rendering. On top of that, GTX 970 is a lot cheaper and has more base vram 4gb, of which 3.5gb is usable, which is still more than the 3gb base that 780ti has. As for quadro, things get slightly more complicated as Nvidia has chosen to keep thousand numbers scheme for a while now, opting to differentiate their Quadro line with a single prefix letter. So the Quadro 5000 is the oldest model, the Quadro k5000 is the Keppler based one, and the Quadro m5000 is the most current, Maxwell based one. Of them all, only the latest quadro, m5000 is actually better than either GTX 780ti/970 and noticeably so with 2048 Maxwell CUDA cores (effectively equivalent of gtx 980 non ti version. Older quadro versions are several times slower than either of the GTX models. If you're buying a new GPU and you can wait, it would actually be wise to wait 3 months for the next GPU- Pascal - to ship and than a couple of months more for it to be supported by the vray rt.
Message 13 of 14
darawork
in reply to: eodeo

Personally I always wait for an item of hardware to become 'last year' tech
before buying. After the tidal 'swell' of hype and cost goes down.

780ti is a nice card. 😃

Darawork
AutoDesk User
Windows 10/11, 3DS Max 2022/24, Revit 2022, AutoCad 2024, Dell Precision 5810/20, ASUS DIY, nVidia Quadro P5000/RTX 5000/GTX760

Message 14 of 14
darawork
in reply to: jub0raj

980 user. 😃

Darawork
AutoDesk User
Windows 10/11, 3DS Max 2022/24, Revit 2022, AutoCad 2024, Dell Precision 5810/20, ASUS DIY, nVidia Quadro P5000/RTX 5000/GTX760

Can't find what you're looking for? Ask the community or share your knowledge.

Post to forums  

Administrator Productivity


Autodesk Design & Make Report