quadro 6000 and quadro 2000

[aristocratic3d]

HI friends,

I feel like i need to upgrade my graphic card. what should i buy? quadro 4000 or GTX770 4GB? or Guadro2000/ GTX770 4GB?

I need to have great viewport performance. I dont do GPU rendering. I can manage rendering time buy multiple CPU. SO I need the card for better view port performance only.

How big is the difference between quadro 6000 and quadro 2000 in terms of viewport performance.?

 

Thanks!

Abdullah

[Dimitris Tolios]

Depends on the model complexity, software viewport engine, and lately version of the software you are using.

What GPU do you have now? What's the rest of your system roughly consisted of? What's the complexity of your work, in which program etc?

[aristocratic3d]

I am using a "GTX580 1.5GB" now. 16GB RAM. i7 3770.

I use 3ds max 2014. I often need to deal with heavy environment. I use vray, multiscatter and adobe ps cs6.

My slave nodes are all i7 and have 12GB RAM with a built in graphic card that came along with a MOBO. ( I think increasing ram for slave nodes should be fine. no?)

 

I am increasing my ram to 32GB. buying four new HDD. (have 850W PSU, dont know if i need to increase it as well.)

 

Dimitris I went through some other thread to see if I find what I need. I noticed you have put some valuable information there as well.

I am glad that you commented on my thread as well.

 

Depends on the model complexity, software viewport engine, and lately version of the software you are using.

What GPU do you have now? What's the rest of your system roughly consisted of? What's the complexity of your work, in which program etc?

[aristocratic3d]

WOW. what a coincidence! Yesterday I watched two videos in youtube and visited PCFoo.com as well. that video was comparing in between two cards. from where I came to a decision that I need quadro instead of GTX. I was wondering how a 1GB quadroo 2000 was beating a GTX 4GB card though.

Its my great pleasure to get your suggestion here.

[Dimitris Tolios]

You won't be needing a PSU for HDDs...each HDD is 10W give or take, less when not under load.

 

You as for viewport performance, so how much RAM your nodes have is irrelevant.

Unfortunately, the GTX 580 is a very powerful card...the unfortunate portions goes to the fact that you won't be able to replace it with something radically faster without spending a lot of money.

 

There is a chance of you being kind of VRam limited with 1.5GB if you are using lots of heavy textures, but if this is a case, you will need a 384bit card to really help access more than 2GB fast. 4GB versions of 256bit cards like the 760/670/680-770 (slower->faster) won't help you much in viewport. That leaves you with the GTX 780 and Titan range of cards as far as nvidia gaming cards go.

 

Anything less than a Quadro 4000 will probably just struggle to match and not surpass the 580 in Nitrous.

All non-K Quadros are kinda outdated tho, and even if their performance is not bad, I would not consider getting a Quadro 4000/5000/6000, unless it was priced very low, something you can achieve through eBay if you are patient - at least for the 4000 and rarely for the 5000. The 6000 is a niche product and very few get on the used market, furthermore in a competitive price.

 

Quadro and Firepro/FireGL cards beat GTX cards in OpenGL applications, as usually the drivers for such are purposly neglected for the gaming cards, or the developer of the viewport engine didn't care to support gaming cards enough in order to take the steps needed for them to work properly.

New versions of software, including Maya and Solidworks (past 2013) appear to be working decently with GTX and Radeon cards.

 

Also, even if it is not obvious to all, it is wrong to compare cards solely based on VRam onboard.

There are multiple 2 and even 4GB cards out there that are "horrible", usually based on 64bit bus DDR3 that makes them no-better than "on-board" GPUs (IGP).

Memory bitrate (multiples of 64bit, 64-128-384-512 etc), memory speed and memory type itself (these days almost exclusively DDR3 & DDR5) is very important also.

Edited February 10, 2014 by dtolios

[aristocratic3d]

I thought if I want to have more powerful graphic card I will need more PSU support. I did not mean that for HDD.

In this video-

It shows that the quadro 2000 is 2x faster than GTX 760. as I know that the GTX760 is much faster than a 1.5GB GTX580.

So I was thinking that a quadro 2000 will be at least 3x faster than my current GTX580 1.5GB.

 

Let me know if disagree with that.

 

Thanks!

[Dimitris Tolios]

I thought if I want to have more powerful graphic card I will need more PSU support. I did not mean that for HDD.

In this video-

It shows that the quadro 2000 is 2x faster than GTX 760. as I know that the GTX760 is much faster than a 1.5GB GTX580.

So I was thinking that a quadro 2000 will be at least 3x faster than my current GTX580 1.5GB.

 

Let me know if disagree with that.

 

Thanks!

 

 

Why is the 760 2x faster than a 580?

Are you counting CUDA cores here and getting this verdict?

 

The 760 is indeed faster in games, but that is again an optimization issue + the fact that the 580 is already much older.

You won't see any significant gain in 3DS Max with the 760, and if you do, there should be an issue with your driver installation or a problematic 580. There are quite a few fields, including compute, where a 580 surpasses or holds its own even vs. a 770.

 

In the video you've linked, I was using Maya 2012 in OpenGL.

Even my Quadro 600 would beat not only the 670, but even my GTX Titan in OpenGL benchmarks, again, only because drivers for OpenGL are crippled for GeForce cards, and developers weren't paying attention to them when making the original OpenGL engines for Maya, Solidworks etc.

Newer versions of those software suites, mainly past 2013, are a bit different and work decently with OpenGL, while very good with D3D (like the ones used in 3DS Max, Viewport 2.0 mode in Maya 2013 and on etc).

[aristocratic3d]

So, so you are saying that I wont get 2x benefit if I use Quadroo 2000 instead of GTX580 in 3ds max 2014?

BTW, I thought 760 is faster because it is 4GB!?

 

Why is the 760 2x faster than a 580?

Are you counting CUDA cores here and getting this verdict?

 

The 760 is indeed faster in games, but that is again an optimization issue + the fact that the 580 is already much older.

You won't see any significant gain in 3DS Max with the 760, and if you do, there should be an issue with your driver installation or a problematic 580. There are quite a few fields, including compute, where a 580 surpasses or holds its own even vs. a 770.

 

In the video you've linked, I was using Maya 2012 in OpenGL.

Even my Quadro 600 would beat not only the 670, but even my GTX Titan in OpenGL benchmarks, again, only because drivers for OpenGL are crippled for GeForce cards, and developers weren't paying attention to them when making the original OpenGL engines for Maya, Solidworks etc.

Newer versions of those software suites, mainly past 2013, are a bit different and work decently with OpenGL, while very good with D3D (like the ones used in 3DS Max, Viewport 2.0 mode in Maya 2013 and on etc).

[Dimitris Tolios]

So, so you are saying that I wont get 2x benefit if I use Quadroo 2000 instead of GTX580 in 3ds max 2014?

BTW, I thought 760 is faster because it is 4GB!?

 

Memory bandwidth:

GTX 580: 384bit * 4008MHz = 192.384 GB/s

GTX 760: 256bit * 6008Mhz = 192.256 GB/s

Quadro 2000: 192bit * 2500MHz = 41.6 GB/s

GT 610: 64bit * 1800MHz = 14.4GB/s

 

Memory size is irrelevant to how fast it exchanges data back and forth the GPU.

The 610 was included, as this card is actually available with 2GB DDR3 memory configurations. There is a multitude of cards with less than 2GB that would be far faster than this card, even today.

 

The game is not about how much memory you can put in, but how fast the memory needs to be, which capacity can we get with the current lithographic processes, at which price, so that we can do what we have to do within reasonable time and cost.

 

DDR3 used in low-end GPUs is the same kind used as main system memory. There is an abundance of large capacity chips, but speeds simply don't cut it. Thus, high-end GPUs are forced to use GDDR5, that is far more scarce and expensive, and in many ways it is GDDR5 availability that guides what really happens in the GPU world.

 

When the 580 came out, the new they needed roughly 190GB/s bandwidth, but the memory chip suppliers could not work with packages with more than 128MB / 1GHz (effectively *4 the speed, as this is how GDDR5 is working) / 32bit per BGA package that would be cheap enough for nVidia to make the profit margins they wanted. So, using 12 of them, there you got it: 32bit chips * 12 = 384bit and 128MB * 12 = 1,536MB, all pumping together the required bandwidth. The 3GB version uses the same 384bit bus, but splits it onto double the chips (24). There was no "magical" increase in bandwidth with more memory. Still the same bus, still the same clocks.

 

When the 6xx series and the 7xx series came along, memory suppliers had achieved 1500MHz RAM speeds, and managed to cram "two" 128MB GDDR5 modules in one BGA package, effectively 2 * 32 = 64bit and 256MB for each chip. Put 8 of those together, and you have 64bit * 8 = 256bit and 256MB * 8 = 2048MB (2GB). Since frequency was 50% faster, they would achieve a similar theoretical bandwidth despite the lower bus width.

The 4GB version uses the same 256bit bus, but splits it onto double the chips (16). There was no "magical" increase in bandwidth with more memory. Still the same bus, still the same clocks.

 

Since the Kepler architecture (GTX 6xx & 7xx, Quadro Kxxxx) uses much more and simpler cores than the Fermi arch did (GTX 4xx & 5xx), those 256bit are spread to far more "players". It is really effective for small packages, but in general the GK104 GPUs (GTX 660,670,680 and the 760 & 770) are memory stagnated. This doesn't make them slow, just not as fast as those could be.

But the dealio is for nVidia (and AMD) to make money, and not about you getting the best - and this involves trying to use as little silicon as they can to reduce production costs.

Edited February 10, 2014 by dtolios

[Francisco Penaloza]

With a Quadro card you'll get a better performance in Max (drivers related), now this is more noticeable on ArchViz, because if you are creating game assets that GTX will do just fine. We archViz artist are the few who put crazy amounts of polys and objects on screen

Now as Dimitris mentioned compared to your 580 this increase won't be that dramatically, unless you go very high on the Quadro series such Quadros k5000 or k6000 but even that the performance is kind of logarithmic.

Other advantage of Quadro are lower power consuption, less heat and consistent performance between other apps, such Photoshop, Premiere, After Effects.

 

Other thing to have in mind 3DsMax is not the king of viewport performance, Maya has better viewport performance and other app too.

[aristocratic3d]

So investing $600 into a refurbished quadro 4000 does not worth the money for me?

With a Quadro card you'll get a better performance in Max (drivers related), now this is more noticeable on ArchViz, because if you are creating game assets that GTX will do just fine. We archViz artist are the few who put crazy amounts of polys and objects on screen

Now as Dimitris mentioned compared to your 580 this increase won't be that dramatically, unless you go very high on the Quadro series such Quadros k5000 or k6000 but even that the performance is kind of logarithmic.

Other advantage of Quadro are lower power consuption, less heat and consistent performance between other apps, such Photoshop, Premiere, After Effects.

 

Other thing to have in mind 3DsMax is not the king of viewport performance, Maya has better viewport performance and other app too.

Edited February 11, 2014 by aristocratic3d

[Dimitris Tolios]

So investing $600 into a refurbished quadro 4000 does not worth the money for me?

 

I don't know how expensive stuff are in your part of the World, but $600 are a lot of money for a used 4000.

How much is it new, and more importantly, how much is a K4000?

 

What me and Francisco are trying to say to you, is that there is not definitive answer. It might, it might not.

There are a lot of factors into play.

 

What is definitive is that you won't be seeing "2x" performance increase etc across the board. In some applications, maybe more than that, but not in 3DS Max, so don't get your expectations too high.

[Tommy L]

I think you should wait for the 'next big thing'. The GTX580 is still a good card and upgrading will probably be a disappointment. I use a gtx580 (3gb) and no issues. I use some pretty intense scenes, (high poly, dynamics, particles etc) in Max 2014 and I dont feel the need to upgrade.

[Zdravko Barisic]

So, you are trying to say that 7xx series is not worth of money, comparing to 6xx?

Egz. 660 is almost like 760 when talking about geometry viewport work?

...

And maybe 580 is best of all for viewport working?

[Francisco Penaloza]

So, you are trying to say that 7xx series is not worth of money, comparing to 6xx?

Egz. 660 is almost like 760 when talking about geometry viewport work?

...

And maybe 580 is best of all for viewport working?

 

there is not definitive answer. It might, it might not.

There are a lot of factors into play.

 

What is definitive is that you won't be seeing "2x" performance increase etc across the board. In some applications, maybe more than that, but not in 3DS Max, so don't get your expectations too high.

 

 

I know it is hard to believe or even understand, (to me too for sure) because all the publicity and "performance test" are design for you to buy the new best thing.

We as 3D artist have a very specific niche of hardware, so for a Gamer will be the major leap in performance for us is not or just the same, or even slower (Intel CPU for example). What is constant is watts performance ratio is always better for Quadro series, and now a day heat has to take in consideration too.

If you are not doing GPU rendering there is not big improvement in performance on video card in the past generation, maybe two, that's where all the fuss and interest from NVidia has been, CUDA processing.

 

At the office and at Home I have Quadro 4000, with latest Autodesk products and Adobe products, it work just fine for me, I know its limits and software limits so I work around, but it work just fine.

[Zdravko Barisic]

I understand pretty much what you are talking about, as I have been using 7900GTO@Quadro 1500 512MB, 2006-2012 /!!!!!!!/.

With no lag at all...but it was time to switch for some more powerful.

Thanx!

[Dimitris Tolios]

I understand pretty much what you are talking about, as I have been using 7900GTO@Quadro 1500 512MB, 2006-2012 /!!!!!!!/.

With no lag at all...but it was time to switch for some more powerful.

Thanx!

 

Well, this is the deal we are talking about: depending on what you do, but also your personal tolerances/patience, the fastest card might appear disappointing despite the $1000s invested, or a mid-range-ol-card might be ok.

 

Thus it is a bad spot to be in, when someone asks you "will it be better".

We can give quantitative answers when it comes to a standardized test: the card A did 5, the card B did 6, thus A

The question "will it be 2x faster", is not something you can prudently answer without knowing the exact context.

 

It is also hard to give qualitative answers when we don't share the same workflow, same models, same experience or even the same tolerance levels etc, but at least you can claim that a fast Quadro like the 4000, despite its age, does pretty good with 3DS, Maya etc, in general more fluidly than older GTXs.