GTX or Quadro?

[dsp_418]

Hi there,

 

I'm trying to figure out what could be the best video card for my next computer.

I basically have a list of components for the other parts of the new machine, but I'm struggling with the video card.

I've tried to get info around and the thing is someone say I should go with Quadro, someone else told me to just go with GTX.

Now, in terms of budget I spotted two cards on almost the same range, Quadro k4000 and GTX 780. About the Quadro someone told me to forget about it, it's just slow even just with 3ds Max UI, whereas someone else told me Quadro performs much better than GTX, can handle lot of geo and particles.

And then they mentioned things like double precision floating point, but still is unclear to me what that means and what kind of benefits I could get from that. Also, when I look about DPFP I found out that the new GTX (780 too) seems to have that.

 

So, I'm lost.

Apps I'm gonna use: After Effects CS6, Nuke (6 or 7), 3ds Max probably 2014 (Vray), Softimage probably 2014 (MR, Vray), Blender (Cycles)

 

Any help/info will be highly appreciated

Thanks.

Edited July 8, 2013 by dsp_418

[Francisco Penaloza]

You can use your same thread title as a search in this forum and you'll see several post regarding the same, most of them well explained already pros and cons

[dsp_418]

I did it, I didn't find anything specific about k4000 vs GTX 780 though.

[numerobis]

You can use your same thread title as a search in this forum and you'll see several post regarding the same, most of them well explained already pros and cons

 

i think these older threads are pretty useless concerning the current hardware and software. Especially with max after the switch to directx/nitrous and with the GTX6xx series.

We have to wait for a proper review testing the performance of the latest cards in max 2014.

[pmaggia]

I have a Quadro 4000 that performs outstanding en real time for modeling. It´s not to use Iray or Vray RT, because it haven´t a good amount of cores. But in the other hand, it performs for hours and hours without problem. If you use a GTX take care of that, maybe a liquid cooling may help. If you are thinking to use Iray GTX are an option only if you take care of the card teperature. Too, more cards, less time to render, less temprature to deal of. Sorry my english.

[Francisco Penaloza]

i think these older threads are pretty useless concerning the current hardware and software. Especially with max after the switch to directx/nitrous and with the GTX6xx series.

We have to wait for a proper review testing the performance of the latest cards in max 2014.

 

I am sorry but I do not agree with you, some of the same thread are just a few weeks old and yes you can tell me new hardware is develop almost every week, also is true that the performance difference is not that big, as matter of fact new Quadros are less powerful than previews series, but this is not the point.

If you read carefully "old thread" you'll see that the answer is always the same, if you do progressive rendering such, Octane, VRay RT, IRay or others GTX will be a better option Money performance wise. But Quadro or teslas are better at energy consumption performance wise. Most of us are not that concern about this because non of us will set up a render farm with these cards, like 30 or 50 cards. shoot with 2 GTX you really fly.

But for instance I do not do any progressive rendering, but I do work a lot with CAD applications so Quadro is a better option. If you have a scene in Max with millions of lines and wire frame, a quadro will perform better if you do shaded mode GTX will perform a little better until the memory click then you'll see strange mesh, flick polygons and what no, a Quadro stay better a little longer, until the memory click again so.

as always the decision is how much money you want to spend, I bough a Quadro 4000 a few years ago after that I have not worried about video cards anymore, I do use all video productions Adobe software plus Fusion, Modo, Cinema4d, Max and when any of these application that CUDA enable you'll see a performance increase, such Premiere, or After Effects.

Now 3dsMax is not the faster Viewport performance software out there,if you use Softimage you know what I talk about .

[dsp_418]

Francisco,

 

that's why I was asking in a new thread. Quadro 4000 is not Quadro k4000, about which I didn't find anything. On top of that, my thread was specific about comparing the k4000 with a GTX 780. You mentioned is a matter of how much money one wants to spend, and that's another reason why I was posting, as k4000 and GTX 780 are on the same price range, just a difference of $80-$100.

 

So, is really not a matter of money at this point. I'm hoping to get feedback from users who are using or have the chance to see how k4000 and GTX 780 perform. Or from users who have knowledge on both sides.

As you can see is not easy to pick the right video card as they cost almost the same, but also new Quadro K are different from old Quadros.

[Dimitris Tolios]

Quadro Ks are different, true.

I have no experience with the K4000, but I do with a K2000 and it appears to be up to older 4000 performance - other than compute where it "sucks" even vs. the older 2000.

 

The 780, much like the Titan will be un-impressive for the price range vs. other 6xx/7xx (same architecture) GTX cards as far as viewport acceleration goes.

It is a driver issue, and the raw compute power of the GK110 over the GK104 based cards cannot overcome that: in viewports where the 780 will do great, so would a 670/680/770 GK104 flagship. In viewports where GTX cards do mediocre, so will the 780.

[Francisco Penaloza]

Yes I know Quadro 4000 is not K4000, I was just pointing out my experience with Quadros cards. Now as mentioned by Dimitris, Drivers are a big issue here, and IMO is the only way NVidia can control what type of card you can use for different situations. If price is same and you do not do progressive renderings I'll go with Quadro K4000 no doubts, mostly for the compatibility of Pro software such, NUKE After effects and so.

[dsp_418]

So Quadro probably is not impressive but better solution for heavy works, is that correct?

 

Btw, yes I'll do progressive renderings with Blender (Cycles) and Vray RT for preview purpose.

[Dimitris Tolios]

So Quadro probably is not impressive but better solution for heavy works, is that correct?

 

Btw, yes I'll do progressive renderings with Blender (Cycles) and Vray RT for preview purpose.

In some instances the difference might be impressive, but if you were pushed hard to find the money for one and you are expecting "magic", you might be disappointed.

 

Same goes for GPU accelerated renderings: after the 2nd card, you still get faster, but the overall % gains are diminishing.

Surely a fast GTX (or more than one) will be better for cycles and VRay RT GPU than most Quadros. The K5000 is "ok" fast, being practically a 680 4GB, but when you can get 3x-4x fast GK104 GTX cards (770/680 or 670) for the same money as one K5000, each one being just as fast or faster (due to higher clocks), its hard to justify the K5000 as a good value for GPGPU.

 

Depending on how complicated you want your setup, you can go for more than 1 cards. I'm using PCs with such setups now and then, with no issues. Others complain that they could not get Quadros to work with GTX cards in tandem. I have tried with Quadro 600/2000/4000/K2000 and all worked flawlessly with my 670. Quadro gets "workstation" drivers and the monitors, GTX gets usually recognized as GTX without re-installing "desktop" drivers and seats "headless". Works fine for CUDA/Open CL apps, with the ability to select either or both cards for compute, while the Quadro monopolizes the viewport acceleration. Most programs don't care about "nVidia maximus" support and fansy named stuff.

If one or more CUDA/Open CL enabled devices are present, most programs just use it.

 

If you want an "ok" all around solution, a single 680/770/670 4GB will work fine.

If you are serious about GPGPU and you think 3GB for the 780 are not enough, you could wait in case they come up with a 6GB version, or just go Titan. For the most part and previews, 3GBs should be more than enough. 2GB might be pushing it in case you have a single card accelerating both viewport and a GPU accel. activeshade window, thus I'm recommending the 4GB GK104s. Better safe than sorry. Still 2GBs are plenty for the most part.

 

[hankash]

Good day,

Thank you Dimitris for your explicit information.

I have a Z800, with Dual Xeons, my workflow is 80% 3ds Max, CPU Rendering, with very little Vray RT. and 20% adobe suite and other 3d packages.

I am considering a K4000, or an equivalent GTX, what do you advise? The Quadro or the GTX?

One more, Would it be better getting 2 K2000 (4GB), or a single K4000, slight difference of almost a 100$.

I have 2 Eizo Monitors and 850 W Power unit.

 

Would highly appreciate any advice

Big Thanks

[Dimitris Tolios]

Good day,

Thank you Dimitris for your explicit information.

I have a Z800, with Dual Xeons, my workflow is 80% 3ds Max, CPU Rendering, with very little Vray RT. and 20% adobe suite and other 3d packages.

I am considering a K4000, or an equivalent GTX, what do you advise? The Quadro or the GTX?

One more, Would it be better getting 2 K2000 (4GB), or a single K4000, slight difference of almost a 100$.

I have 2 Eizo Monitors and 850 W Power unit.

 

Would highly appreciate any advice

Big Thanks

 

First of all, there is no equivalent GTX...K2000 and K4000 are low end cores...would be

It's all in the drivers.

 

2x "any" quadro is not needed, and even if it would be, then 2x K2000 is not 4GB. Is 2GB each.

Much like with GPGPU/GPU Rendering/OpenCL acceleration for Adobe suite etc, each card needs to fit 100% of the "stuff" in its own memory. Vram is not shared between the cards, even when you do SLI/CF for games (which is not supported by Quadro K2000 anyways). "Each card for itself".

HSA is not here yet. "Soon™".

 

I don't know how much you care about VRay RT, but K2000 is patheric, and K4000 won't be that much better for it.

Kepler cards are gimped for GPGPU - period. Even the Titan is not that "special". Yes it is the faster nVidia card, BUT, for the core count and the mm2 the die has, it is not special. No idea what Maxwell (the next nVidia architecture codename) will bring, but nVidia has "game" in GPGPU right now only because of CUDA.

AMD's GCN architecture is so much faster, it is not even funny, as a 7970 being 1.5 year older than the Titan, single handely outperforms it in OpenCL. Hope is with Adobe switching to OpenCL, things will mature faster for it throughout the industry.

 

At any rate, if GPU acceleration is what you want, K2000/K4000 is not the card for you.

In viewports both do great. I am impressed with my K2000 as it gives me Quadro 4000 performance for far less noise/heat (others had issues after buying it tho and I hope they don't blame me). Its compute performance is 2/3s of the older 2000 tho, and nothing even close to the 4000. A K2000 can drive my 1440p 27" + 1080p 22" just fine. You don't need a seperate card per monitor, this is not Windows 2000 era.

[hankash]

Hi Dimitris,

Thank you again for your elaborate explanation. Big Help.

I guess I will follow your lead, with a K2000, as All I need from It is a good viewport performance, (I do not care about Vray RT in my production, in my opinion it is still underdevelopped) and enjoy the cost difference in my stay next month in Rhodes, any recommendations for some good places there, just kidding.

Just out of curiosity, Can I install a K2000 and a Good GTX on a Z800, would it need any special additional equipment, maybe then I will enjoy both performances.

 

Thank you Dimitris.

[Dimitris Tolios]

In theory you can install as many cards as PCIe slots your board has, and if there is enough juice from the PSU to power them, you will do fine.

 

I don't know any place in Rhodes personally, as I've never been there, but should you really want to, I have a couple of friends living there year round to ask (no kidding)...

[thiyagu arjunan]

Hi elie,

 

if u want a gd viewport performence go for old quadro cards instead of new k series..old quadro 2000 is so much better than new k2000 .i tried with max only and i ll suggest this oly if u looking for 3DS Max..even 4000 is grt..k2000 is not that much worth ..sry to say this ..i got and faced some performence laggings even a small views...my beloved dimitris know my prob;-) since he helped me a lot.. thanks bro....finally i changed my card and ordered gtx770..hope it ll help me ;-)

[hankash]

Hi thiyagu,

 

Thank you for your info.

Dimitris, More then helpful and thank you for the offer.

I am seriously pretty lost. a 2000 (1GB) performing better then "a new model" K2000 (2GB).

How mislead are we?

 

Would you trust this benchmark, is it representative to 3d work, as in 3ds max, and viewport performance?

 

http://www.videocardbenchmark.net/high_end_gpus.html

 

Cheers

Edited July 25, 2013 by hankash

[Dimitris Tolios]

Passmark: I would not trust that for neither games or 3D CAD...makes no sense having a 560ti being faster than a tesla c2075 (both fermi architecture cards, with the tesla being much faster / more cores, equal to a 480 that it is actually faster than both according to passmark), a c2050 faster than a c2075 etc...sounds like a driver screw-up, and most of the results make no sense.

 

It clearly its a mash-up of a lot of factors, it cannot be compute, it cannot be pure D3D or OpenGL performance, but w/e it is, doesn't match patterns I've seen in either modern games, nor SPEC benchmarks.

 

I strongly believe thiyagu arjunan was unlucky with his K2000...something was wrong with the card.

I have tested a K2000 and a 2000, and the K2000 dominates literally in anything but compute. It is a very clear improvement over the 2000.

Edited July 29, 2013 by dtolios

[thiyagu arjunan]

I strongly believe thiyagu arjunan was unlucky with his K2000...something was wrong with the card.

 

hmm well dimitris...am unlucky or don kno anything than that ...except my card selection all ur suggestion worked well and grt for me to build my pc..much happy abt it..and this 770 doing much grt with my 3ds max..so here after i wont worry abt quadro..i just told my experience with quadro..tats it..other than tat every one kno u r grt in it...so keep on sharing ur knowledge and guide all buddy;-)

 

And i strongly believe tat benchmark chart in my case...i particularly saying it coz i noticed improvements in max;-)) tats wat i need at last..

[junglee]

In some instances the difference might be impressive, but if you were pushed hard to find the money for one and you are expecting "magic", you might be disappointed.

 

Same goes for GPU accelerated renderings: after the 2nd card, you still get faster, but the overall % gains are diminishing.

Surely a fast GTX (or more than one) will be better for cycles and VRay RT GPU than most Quadros. The K5000 is "ok" fast, being practically a 680 4GB, but when you can get 3x-4x fast GK104 GTX cards (770/680 or 670) for the same money as one K5000, each one being just as fast or faster (due to higher clocks), its hard to justify the K5000 as a good value for GPGPU.

 

Depending on how complicated you want your setup, you can go for more than 1 cards. I'm using PCs with such setups now and then, with no issues. Others complain that they could not get Quadros to work with GTX cards in tandem. I have tried with Quadro 600/2000/4000/K2000 and all worked flawlessly with my 670. Quadro gets "workstation" drivers and the monitors, GTX gets usually recognized as GTX without re-installing "desktop" drivers and seats "headless". Works fine for CUDA/Open CL apps, with the ability to select either or both cards for compute, while the Quadro monopolizes the viewport acceleration. Most programs don't care about "nVidia maximus" support and fansy named stuff.

If one or more CUDA/Open CL enabled devices are present, most programs just use it.

 

If you want an "ok" all around solution, a single 680/770/670 4GB will work fine.

If you are serious about GPGPU and you think 3GB for the 780 are not enough, you could wait in case they come up with a 6GB version, or just go Titan. For the most part and previews, 3GBs should be more than enough. 2GB might be pushing it in case you have a single card accelerating both viewport and a GPU accel. activeshade window, thus I'm recommending the 4GB GK104s. Better safe than sorry. Still 2GBs are plenty for the most part.

 

 

 

 

Dimitris,

 

I'm confused about GPU selection.

In some post, you recommened K2000 over 2000 but your blog is recommending Quadro 2000 for rendering pro machine spec.

And what exactly is "compute" power in GPU world? - Is it ability to do progrossive renderings, eg Vray RT?

 

I only use 3dmax 2012 & 2014 with Vray advance and photoshop/illustrator/indesign with my PC and my main concern is to get better viewport response in 3dmax. I don't intend to use Vray RT/Iray anytime soon so I do not need any GPU rendering power. And I currently use Nitrous based viewport set up.

 

I read alot of people saying with Nitrous viewport, GTX works faster than Quadros on later version 3dmax and Quadro lost its edge in these environment over GTX.

Is this true?

 

With which GPU would I be expecting a better performance in my case? My budget is less than $600 for GPU.

 

My current computer set up is:

3930k

16gb RAM

256 SSD

2TB HDD

GTX 660

 

Thanks in advance.

 

Jung

[Dimitris Tolios]

Things change. When I was writing about it in my blog I had no experience of a K2000, as that was only available to OEM builders.

Took quite some time for me to get my hands on one (yes, I did not want to pay full price for it just to test it).

 

So keep an eye on the date the piece was written, and get a feeling of context: no 7xx GTX was around, no Kxxx Quadros other than the 5000 etc.

 

GPU Compute = yes, in the CG world that translates to accelerating progressive renderings, real time material editors/painters and video or photo processes that utilize CUDA/OpenCL acceleration. And yes, in this field, the 2000 remains faster than the K2000, yet both much much slower than a 660 or whichever 4xx/5xx/6xx/7xx GTX for that matter...

 

I don't know about 2014. Downloaded the package but never got to update it in order to test it =(

[junglee]

Things change. When I was writing about it in my blog I had no experience of a K2000, as that was only available to OEM builders.

Took quite some time for me to get my hands on one (yes, I did not want to pay full price for it just to test it).

 

So keep an eye on the date the piece was written, and get a feeling of context: no 7xx GTX was around, no Kxxx Quadros other than the 5000 etc.

 

GPU Compute = yes, in the CG world that translates to accelerating progressive renderings, real time material editors/painters and video or photo processes that utilize CUDA/OpenCL acceleration. And yes, in this field, the 2000 remains faster than the K2000, yet both much much slower than a 660 or whichever 4xx/5xx/6xx/7xx GTX for that matter...

 

I don't know about 2014. Downloaded the package but never got to update it in order to test it =(

 

Oh, sorry I didn't check your post date on those

 

You are making a very interesting point here.

I've been having massive delays on material editor recently and I never knew that it was due to GPU computation force.

And this delay really puts me on a serious lag as I constantly open and close the editor and it takes about 10 sec each time to load all materials. And just trying to open material library takes at least 30sec even to just load material images.

So if I get quadro, I would still have this lag on materials editors but will improve in viewport response.

 

Is advantage in viewport response with Quadros noticeably better than GTX in your opinion?

In the end, I don't want to get 100% better on one area and ditch the other, rather I would be happier if both areas can improve for 50%.

 

And if I get GTX 770 4gb, would I be seeing substantial increase in performace in viewport/material editor over GTX 660?

 

Thanks,

[Dimitris Tolios]

Material Editor: lag is a major complain for many users. I believe is totally CPU bount = no GPU has much to do with it.

Even 2P systems have issues with the material editor, I don't know even if the process of rendering the little thumbnail previews is multithreaded or not...

 

Viewport: If it wasn't for the 2014 changes that I have no experience with, I would say to you fair and straight that I'd doubt that you will see much of an improvement outside compute getting any GTX over your 660. 770s, 780s and Titans included.

 

Even with Quadros, it totally depends on model and scene complexity but was mostly a driver thing, as Quadros never had "magic" hardware...the contrary, most of them were slower than equally priced GTX (consumer cards) by a large margin.

 

I don't know whether "fixing" the viewport for GTX cards "broke" Quadros with 2014, and at the same time allowed similar performance tiering than that you would see with games etc, where a 780 is indeed much faster than a 770, and a 770 notably faster than a 660.

I also don't know what kind of models you have and if 4GB will make any difference over the 2GB. In my case, having a 670 4GB, I have never seen memory allocation bigger than 1.4GB for viewports. I have pushed pass that only with modded games and when rendering with VRay RT GPU.

And that is memory allocation, usually a generous amount "above" what the application really needs (think of it Photoshop scratch disk or RAM limit - it gets X amount allocated for exclusive use if it needs to, doesn't mean it uses it every time you launch the app and edit small images).

 

Your case might be different, and super large textures or w/e situation might benefit you going 4GB...I don't have the resources or time to actually test all those scenarios and come up with a more educated opinion at the moment, but I would not expect even close to 50% performance increase going 770 4GB over your current 660. I would probably suggest you sticking with the 660 and accept the nature of the "beast" and its limitations.

[junglee]

Thanks for your quick response.

My computer arrived as we speak so I'll see if this faster CPU takes care of material editor problem and see how this GTX660 work out with my current projects.

 

I appreciated your input.

 

Jung

[thiyagu arjunan]

Dimitris,

 

Do u hav any idea abt the max driver for quadros in autodesk site??they have performance driver for 2012 version oly..but most of peoples and offices using max 2013 and 2014... really it doesn't matter for them? they keep on updating software and recommending cards in tat list..but not a performance driver ??....am i wrong?