NVIDIA Unleashes Quadro 6000 and 5000 Series GPUs

Article Index:   

Performance Summary:  Before we go over the numbers, we would like to note the critical impact of drivers in all of this. A day before the launch of this review, ATI released a new workstation driver which promised to significantly improve performance in several key applications, specifically those used in the new SPECviewperf 11 testing suite. Although we already finished up all the testing, created graphs, and came to our conclusions based on that data, we felt it would be worth the effort to re-test the FirePro cards to see how much of a difference the new driver would make while providing you with the most up to date information possible. In short, it made a significant improvement for the FirePro cards in both Lightwave 3D and SolidWorks. It still wasn't enough to overtake the Quadros, but definitely closed the gap.

The point is to take these comparison numbers with a grain of salt, especially when comparing scores from different camps. As time goes on, expect to see increased performance in certain software programs due to driver optimization. ATI's FirePro cards have been out for months so we aren't surprised to see more mature drivers from them. And since Fermi-based hardware is new to the professional market, we fully expect to see a boost in performance as new drivers are released.

Note: After posting our review on launch day, we looked over the results and determined our Quadro scores were lower than expected in several tests. After some troubleshooting, we tracked down the issue. Although SPECviewperf 11 doesn't offer the option to toggle Vsync on or off, it was enabled in the NVIDIA control panel and we needed to disable it for the new benchmark. We re-tested the cards and have updated the article to reflect the new Quadro scores. 

With that said, let's analyze the data we have in front of us. Looking at SPECviewperf 11 results and the real world applications it makes use of, the Quadro 6000 videocard demonstrated the highest performance in all eight benchmarks that make up SPECviewperf 11. On average, we found it to be 103% faster than its ATI's flagship model, the FirePro V8800. At the same time, the Quadro 6000 came in at 26% faster than the Quadro 5000, a much closer comparison. Without a doubt, the Quadro 5000 is also a very strong high end card. Throughout SPEC testing, it averaged a remarkable 74% performance increase over the Quadro FX 4800, and a 61% boost over the V8800. Here's a scary thought. Both of these cards are extremely fast and will only improve with every new driver release.

Let's talk about pricing. We've seen our share of $2,000 workstation graphics cards across these pages through the years, but this is definitely an eye opener. Although we mentioned it on the first and second pages of this article, its worth noting again here. The Quadro 6000 retails for $4,999. That's a whole lot of cheddar, even for a workstation card. Amazingly, the 5000 model can be had for less than half of that, at $2,249. From our performance results, we can't say the 6000 model is worth twice as much as the 5000 card, but there is a considerable performance upgrade to be had. We saw an average difference of 26% separate the two cards throughout SPEC testing. Granted, we only used a single 30" monitor and could not have come close to maxing out the card's 6GB of on-board memory. And of course, those who desire a product with no compromises have the option of purchasing the 6000 model, while relying on future driver updates to provide performance enhancements. But if top end performance is the goal, it might be worth getting two Quadro 5000's and using them in SLI, at roughly the same cost of one Quadro 6000.

In conclusion, the Quadro 6000 and 5000 graphics cards represent two of the most powerful workstation products currently on the market. Both are substantially faster than ATI's top card in most of the real world applications we tested, but they have their drawbacks. The Quadros support a maximum of two displays per card, while most of the FirePro models provide ATI's triple monitor EyeFinity feature. The V8800 actually offers four video display outputs. Also, both Quadro cards produce significant heat and consume a considerable amount of power, irrefutable characteristics of the GF100 GPU. And of course, you must be willing to pay in order to make use of their features and performance. But in the world of workstation graphics, the cost of hardware can be quickly offset by the benefits and production throughput these products provide. If you can get past these limitations, what you're left with is a powerful set of videocards capable of increasing the productivity of users across many different applications. That's the bottom line folks. In the real world, the bottom line is usually the deciding factor. If you're looking for the highest level of performance possible, check out the new Quadro 6000 and 5000 workstation graphics cards from NVIDIA.  

  • Untouchable performance
  • Quiet operation
  • PhysX and CUDA support
  • Very expensive
  • Only two monitor support per card
  • Power hungry
  • Runs hot under full load


Image gallery

Related content


AKwyn 4 years ago

I'd never thought another Quadro card would see the light of day, I'm amazed at the fact that it failed some tests but I'm sure future drivers will help improve that score.

Chainzsaw 4 years ago

Not too bad of video cards, but they still draw more power and heat up much more than the ATI cards. They both seem to trade blows on diferent programs, but the conclusion states they are substantially faster than ATI's cards. I wouldn't go as far as to say that though. They are a big improvement over previous generation cards, but ATI is already what, 6 months ahead of Nvidia?

If there was a power (as in electricity) to peroformance ratio, the ATI cards would most likely win that.

AKwyn 4 years ago

I sorta disagree. While ATI is ahead of the curb, NVIDIA seems to have the best cards performance-wise. If I were Pixar then I would want to use the most powerful rendering solutions known to man.

But you are right on the fact that they take a lot of power, that seems to be the biggest flaw of the GF100 chip.

animatortom 4 years ago

I was going to say they finally have an answer to the FirePro's? Yet not being able to use three monitors? Most of us would like to have at least two plus our Wacoms. I guess the next question would be, does it support more of you have a second card or does the porting still limit you to two?

I understand the Cinebench marks look better for the ATI. Yet when it comes to frame rate testing, in the view port anything around a 40 is good enough for smooth work flow. Unless you really need to see scientific fluid dynamics live on screen? The CATIA and EnSight levels prove this point very well! Now I know why NVidia has a better reputation for being better integrated to many DCC apps!

"ATI's FirePro cards have been out for months so we aren't surprised to see more mature drivers from them."

I am not so sure about that! I have had the V8800 for a while and they have not released a new driver since the last and only update on 4/25/10?! They even told me how lucky I am to have good driver support, yet didn't answer any concerns about driver updates. I would now say that when it comes to CGI, Nvidia has better performance when it comes to rendering. An animated scene with high resolution textures, motion blur and Photometric lighting is better handled from a company that has been at the forefront of that industry for the past twenty years.

I agree that yes...the cost really offsets productivity! Nvidia has a better reputation for workstation support and knowledgeable integration for 3D graphics. Now I am finding that out the hard way! If you have an awesome high performance system to begin with then you wont notice many troubles when switching to a higher workstation card, except within the veiwports. Yet like they do in these reviews, you will have to start off with a fresh build when switching to something like the Quadros. I have always supported ATI, and now I know why it is that I am not as productive as I want to be!! In studio, I usually have an Nvidia solution, and ATI at home. Since I am usually using scenes already modeled, lite and textured, I have never really noticed much of a difference until rendertime. That, I always caulked up the difference to not having a render farm. Now I know better!!

The PhysX and Cuda features are something that should drive professionals to these cards. It would also be interesting to see how these get integrated with a Tesla c1060? I guess the best way to sum the two companies up?! ATI develops for the end entertainment, down towards the development of content... only as a side note! Nvidia develops for the developer, knowing the end result is to make the best product for entertainment!

So if Y'all are going to just toss this test system into the test-Bin toybox, maybe pitch it my way :) Or I could trade you a FirePro V8800 for one of these Quadros? I would gladly switch the four monitor support for smoother integrated performance:)

RA1D 4 years ago

The new Quadros support two monitors per card (just like the GTX 4-series)

You can download the new FirePro driver 8.762 from ATI right now.  Choose 'FirePro Beta' from the drop down menu.


What problems are you having with the V8800?


animatortom 4 years ago

Thanks Raid, Ill give it a try.

The 8800 has problems in Max on stability in rendering issues. It also doesn't like when you load the Direct X driver. It only seems stable if I use Open GL. At times the wireframe gets about four pixels larger, like it is being drawn in crayon. When I load the X driver it has two of the viewports going blank.  There is still the globals marker for any selected object yet nothing else. I can still switch any of the working veiwports to any view and they still work, just two are MIA.

In Maya the blend shapes sometimes work and most other times don't. It gets frustrating when you move the slider and you think the Blend doesn't actually work! Also any sub-object selection usually does not update when you are moving them or making any kind of adjustments. In both programs the material mapping tends to not update when changing UV's. That only seems to work half of the time. I end up having to zoom in and out decides to refresh the view, then it catches up and adjusts the map.


TDeYoung 4 years ago

To animatortom

At SIGGRAPH I saw many demos by Autodesk of both Maya and 3ds Max 2011 in the AMD booth using Eyefinity w/ 3 displays. They demo’ed every day with no issues. Also as Max really only supports DX (they quit adding features to OpenGL code 4 or 5 years ago), so it was all DX and worked flawlessly for several days of demos.

So wonder if you were using an old driver?

animatortom 4 years ago

"I saw many demos ....several days of demos."

That is the key word,....DEMO!

It is one thing to set up a demo system with demo files. When run through properly set up files it can be made to look like it works well. In practical application, it is a whole other matter. If they had a completely tricked out system with twin Xenons and 128GB of RAM with dual 8800's then yes it would look like it rocks. They haven't had to do much to OpenGl because the development has been complete and for the basics of 3D is working well after years of development.

DX is a new direction for DCC.  The thing with Max, is that it requires a ton of RAM and a powerful CPU. If you have a six-core with around 64GB or RAM then with just the Firepro 4800, it would rock! the others are overkill unless you have a strong system to begin with. Maya relies more on the GPU so it benefits as long as it has more than 1GB on the GPU.  With a crappy system, the 8800 is fairly useless in Max. In Maya it is almost three times as fast.

So if they are demoing it, then I can guarantee, they would have it running on a completely tricked out system inside a simple looking case just for show.

gescom 4 years ago

@ animatortom

I would suggest you to try Linux - Fedora or Debian for example. There aren't any problems with ATI OpenGL Maya/Linux drivers. Cheers, g.

GrapeApe 4 years ago
[quote]"I would now say that when it comes to CGI, Nvidia has better performance when it comes to rendering...is better handled from a company that has been at the forefront of that industry for the past twenty years."[/quote]

nV hasn't been at the Forefront that long, they filled the gaps Matrox and 3DLabs left behind 10 years ago. If anything ATi's been in it longer, nV just did a better job of capitalizing on the failure of others (including 3Dfx).

[quote]The PhysX and Cuda features are something that should drive professionals to these cards.[/quote]

What exactly is the benefit of PhysX here... or anywhere? Confused

With OpenCL and Direct compute both of those features will start to become marginalized. Proprietary is fine if you can develop a compelling reason, but even Adobe has said they are going to OpenCL, not CUDA for the next build to get the more global CPU & GPU benefits of OpenCL.

CUDA like Brook was a nice bridge solution, but their time has come and gone, just like Cg on which Cuda was based.

[quote]"I guess the best way to sum the two companies up?! ATI develops for the end entertainment, down towards the development of content... only as a side note! Nvidia develops for the developer, knowing the end result is to make the best product for entertainment!"[/quote]

That would be a way to sum it up, but not the 'best way', it's a little Myopic. If anything this review shows that ATi makes better hardware, but they are held back by their much MUCH weaker software/driver/dev_relations teams. AMD was supposed to improve that but they have done little other than a tiny improvement in Linux and only able to at best equal nV.

* Dang, hate the quoting tool here. Even after editing out remnants remain.*

acarzt 4 years ago


Click HTML and Remove the <p> </p> to get rid of the gray stuff lol

acarzt 4 years ago

64-bit is obviously a requirement for these cards... they got absolutely destroyed in all of the 32-bit tests.

They won quite a few of the 64-bit tests.

(32-bit can only address 4GBs of memory total, including system and video memory, the 6000 has 6GBs of memory)

InfinityzeN 4 years ago

This release is all well and good, but if you want one of these cards and live in the US, you better buy it *NOW*.  The FTC has placed a ban on the import of all Nvidia chips (including those used in motherboards) and is considering an injunction that will prevent Nvidia from selling any of their current stock in the US.

All of this thanks to those friendly trolls at RAMBUS.

acarzt 4 years ago

Well, Nvidia claims to have a loop hole, and it not exactly clear which of Nvidia's products violate the patent. If these do not... they can continue to sell them.

vexar 4 years ago

Again like my comment on Evga Water Cooled 480, TEST duel monitors, most artists use duel monitors, one for the main 3D display while modeling or animating, and the second for all our windows like the outliner in Maya or the Track View in Unreal 3, please use duel 30's if that's what you can. Also Max 2010 has HW rendering that gives you ambient occlusion, soft shadows, and a lot of nice hardware specific features that aren't included in Maya yet. This would be great to see tested. Take a real-time rendered cinematic from an Unreal 3 and play it back, all this stuff would make me happy like simple jack to see.


Thank you, it would be great, and I have hope!

GrapeApe 4 years ago

Considering that the DP performance of the GTX480 has been crippled it looks like they Quadros are equally crippled based on the results in Sandra GPGPU.

A little sad to see for such expensive cards.

For the workstation Apps, once again it comes down tothe driver optimizations more than the hardware ones, when looking at the relative performances in the Quadro vs FirePro and Geforce vs Radeon performance, even in those apps that the Firepro performed poorly, the GTX480 usually was unusable way way behind a single HD5870 which more often than not was well ahead of the HD5970 too sue to lack of multi-GPU support.

Thanks as always for the first crack/look at the Pro Cards.

SGurram 4 years ago

I am wondering, How the tests been verified on 2560x resolution ?, where the spec is showing maximum of 1920X1200 resolution in benchmark tests...

even the submitted results in SPEC are showing 1920x1200 and 1920x1080 resolution...

i don't think it's a right way of comparison...

animatortom 4 years ago

You would probably have to have two running in SLI mode.  Those tests are usually for your standard working resolutions.

If you are considering it for rendering out 2560 or even up to 4K then the render time wont have much to do with what resolution you use onscreen. If you are going to render at high resolutions then you just have to figure that the time will be doubled.

The card can handle Dual 3840 x 2400 @ 24Hz, or 2560 at 60Hz. So the resolution is there! Although if you are planning on using something like a 30" At 2560, or even two of those, then you will probably need to get a kvmp switch to boost the power.

Antizealot 4 years ago

I must say, I was impressed with the way the Ati 5870 performed. I have one myself and from the benches it performs better than I knew.

TLong 3 years ago

Quadro 6000 videocard demonstrated the highest performance in all eight benchmarks that make up SPECviewperf 11. On average, we found it to be 103% faster than its ATI's flagship model, the FirePro V8800.

Not exactly trading blows as one guy wrote. This is why Nvidia owns the pro market with mid 80 per cent.

JMann1 2 years ago

GrapeApe a few points you seem to have overlooked.

1. I've delt with quite a few design houses in and around where I live. And all of them after using both camps, say no questions asked, they prefer Nvidia Quadros over everything else. Why you might be asking, driver support.

2. The price you pay for a Quadro isn't for the card, its for the driver support that comes with it. The drivers have built in profiles, just like gaming profiles but for professional apps, to better maximize the performance.

3. Driver support is second to none. You use Auto Cad, get an error, report it, you have a customized fixed driver within 24h. Not days, weeks or months, WITH IN 24 HOURS! And thats the same for all professional apps.

Post a Comment
or Register to comment