Tuesday, July 27th 2010

New NVIDIA Fermi-Class Quadro Launches the Era of Computational Visualization

NVIDIA today launched the era of the 'computational visualization workstation' for designers, engineers, researchers and animators by introducing its Quadro graphics processing units (GPUs) based on NVIDIA Fermi architecture, and by also introducing the new NVIDIA 3D Vision Pro solution. The new Quadro GPUs deliver performance that is up to five times faster for 3D applications and up to eight times faster for computational simulation, shattering previous benchmarks.

The NVIDIA Quadro Plex 7000 array, and Quadro 6000, Quadro 5000 and Quadro 4000 GPUs feature the new NVIDIA Scalable Geometry Engines and leverage NVIDIA Application Acceleration Engines (AXE) to enable the world's fastest performance across a broad range of CAD, DCC and visualization applications. Rated at an unheard of 1.3 billion triangles per second in raw performance, the Quadro 6000 enables users to interactively work with models and scenes that are five times more complex than ever before.
"For over a decade Quadro has been the choice of professionals around the world. We've also built Engines like AXE to enable the creation of next-gen applications," said Jeff Brown, general manager, Professional Solutions Group, NVIDIA. "When you couple these technologies with our Fermi architecture, the result is a new Quadro family that's exponentially better than anything the market has ever seen."

Redesigned from the Ground Up to Accelerate Entire Workflows
Combining high performance computing capabilities with advanced visualization, the new Quadro GPU is the world's first professional graphics solution with Error Correction Codes (ECC) memory and fast, IEEE double precision floating point performance. These are intended for applications demanding the highest accuracy, such as medical imaging, finite element analysis and computational fluid dynamics.

"In high-end visual effects development, fast iteration is essential," said Olivier Maury, research and development engineer, Industrial Light & Magic (ILM). "By using NVIDIA Quadro GPUs, we are seeing up to eight iterations each day of complex fire, dust and air simulations, representing speed improvements of ten to fifteen times. NVIDIA CUDA and Quadro GPUs have entirely changed the way we solve complex visual effects challenges."

The new Quadro GPUs are built on industry standards, including OpenGL 4.1, DirectX 11, DirectCompute and OpenCL. In addition, Quadro leverages technologies that are unique to NVIDIA -- such as the company's portfolio of Application Acceleration Engines (AXE) and NVIDIA CUDA parallel processing architecture. The result is that software developers are able to create and deliver the next-generation of professional applications that incorporate compute-intensive tasks. Among these tasks are ray tracing, physics simulation, computational fluid dynamics and real-time video effects processing.

"NVIDIA isn't just working on building better graphics hardware, but the software tools needed to drive the needed advancements that are now changing the industry," said Rob Enderle, principal analyst, Enderle Group. "NVIDIA was committed to building tool after tool this past decade in order to drive the market where they needed it to go, and their efforts are paying off."

The new Quadro professional solutions introduced today include:

-Scalable Visualization Systems:
  • Quadro Plex 7000, with 12 GB (total) of memory and 896 CUDA cores
-Board and Desktop Workstation Solutions:
  • Quadro 6000, with 6 GB of GDDR5 memory and 448 CUDA cores
  • Quadro 5000, with 2.5 GB of GDDR5 memory and 352 CUDA cores
  • Quadro 4000, with 2 GB of GDDR5 memory and 256 CUDA cores
-Mobile Workstations:
  • Quadro 5000M, mobile workstation GPU with 2 GB of GDDR5 memory and 320 CUDA cores
Additionally, all Quadro products are compatible with the new NVIDIA 3D Vision Pro active shutter-glasses solution, also announced today, for the highest-quality stereoscopic 3D immersive experience.
Widespread Adoption
Workstation OEMs and System Integrators, including Dell, HP, Lenovo, BOXX Technologies and NextComputing, are among those that are offering the newest Quadro solutions. Quadro products are also available from value added resellers through master distributors PNY Technologies in North America and Europe, Leadtek in Asia Pacific, and ELSA in Japan.

"As the first professional-class GPUs to integrate high performance computing with advanced visualization, NVIDIA Quadro combined with Dell Precision workstations are poised to transform workflows," said Greg Weir, senior manager, Dell Precision Workstations Product and ISV Marketing. "With this technology, we are enabling our design, research, animation and film customers to deliver higher quality results in less time."

"NVIDIA continues to push the envelope on delivering cutting-edge graphics solutions,"said Rob Herman, director of product marketing, worldwide ThinkStation Business Unit, Lenovo. We're actively incorporating these latest graphics into our newly upgraded ThinkStation C20 and D20 workstations to deliver innovative, 'visual supercomputers' to our customers."

Availability and Pricing
The Quadro 4000 ($1,199 MSRP, USD) and Quadro 5000 ($2,249 MSRP, USD) are available immediately through all channels. The Quadro 6000 ($4,999 MSRP, USD) and Quadro Plex 7000 ($14,500 MSRP, USD) will be available this fall. Mobile workstations based on the Quadro 5000M will be available in the third quarter of 2010 from HP and Dell.
Add your own comment

51 Comments on New NVIDIA Fermi-Class Quadro Launches the Era of Computational Visualization

#1
(FIH) The Don
we could ask teh mighty W1zz, he should be able to get a card to test from nvidia
Posted on Reply
#2
Unregistered
DanishDevil said:
Just read a couple lines of this article near the bottom:

http://www.xbitlabs.com/news/video/display/20070502085641.html

Besides, these are NOT consumer graphics cards. They are meant for workstations and have specialized drivers that make them good at what they are intended to do, and crap at gaming.
I remember see those prices too :shadedshu what a shame, but no i meant way back, that was still in the GW Bush era when things were starting to crumble, talking back in the good old days like >04. (before DX10)
Posted on Edit | Reply
#3
Pickles24
If you had an EVGA Classified SR-2 with twin Xeon 5600's, 48g of memory and one of these cards.. You could set it up with Autodesk Smoke, Flame, Flint, Inferno.. it even comes with it's own Red Hat distro. You could argue that you have one of the badest composting machines on the planet. That's big boy effect houses..
Posted on Reply
#4
mdsx1950
LAN_deRf_HA said:
Could someone please benchmark a quadro against it's desktop equivalent in gaming with the appropriate drivers (performance I think?). It's been far too long since we've seen a real comparison like, but there's been evidence here and there that the gaming performance gap between the two has closed quite a bit. Probably to make it easier on developers wanting to work and test quickly on the same rig. I figure it should be done both at stock, and then with the consumer card clocked down to match the workstation card to see the clock for clock difference.
+1
I would love to see a benchmark like that. :rockout:
Posted on Reply
#5
btarunr
Editor & Senior Moderator
Taskforce said:
Really, i don't remember Quadros costing 14k, nor do i remember high-end graphics card costing over 500 bucks on release.
That's because there was no Quadro Plex back then. Quadro Plex is not a graphics card, it's a system (graphics server). There were plenty of graphics cards (even back in the early '80s) that went for over $500 on release. Maybe you're new to the computer hardware scene.
Posted on Reply
#7
SluggyJ
Hothardware have benched and reviewed the 5000 and 6000 and can be found here... I would love to see how the Plex 7000 does
Posted on Reply
#8
mdsx1950
SluggyJ said:
Hothardware have benched and reviewed the 5000 and 6000 and can be found here... I would love to see how the Plex 7000 does
I wish they put some game benchmarks. But anyways thanks for the link. :toast:
Posted on Reply
#9
Cheeseball
These things can probably fold like a bat out of hell.
Posted on Reply
#10
btarunr
Editor & Senior Moderator
Cheeseball said:
These things can probably fold like a bat out of hell.
F@H CUDA is not video memory intensive. So these cards will in turn be slower than GeForce cards (since they're clocked lower).
Posted on Reply
#11
KainXS
that hothardware review is the old one before amd updated their firepro drivers right
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 difference in both Lightwave 3D and SolidWorks. Previously, the Quadro cards showed the best performance in those two tests, but after the ATI driver update, the FirePro models leapfrogged their scores.
based on the shader counts these quadro's are based on the old gpu's, not the new superscalar gpu's right.
Posted on Reply
#12
mdsx1950
After reading the HotHardware review, i wasn't so impressed with the Quadros. I don't think it deserves such a big price tag.
Posted on Reply
#13
Unregistered
btarunr said:
That's because there was no Quadro Plex back then. Quadro Plex is not a graphics card, it's a system (graphics server). There were plenty of graphics cards (even back in the early '80s) that went for over $500 on release. Maybe you're new to the computer hardware scene.
Yes i know Quadro isn't aimed at your average Consumers and yes they had something similar back then with the same purpose obviously not called Quadro Plex.
"early 80's" seriously? back then you wouldn't call people who bought graphics cards average consumers, graphics card segment really took off when DOOM came about in 1993. However lets not time travel to the 80's, what I'm saying there was a time when we had a "sweet spot" if you will, when computer hardware were reasonably priced and not overpriced as they're now.

No I've been dealing with computers since i was a toddler.
Posted on Edit | Reply
#14
newtekie1
Semi-Retired Folder
Taskforce said:
Really, i don't remember Quadros costing 14k, nor do i remember high-end graphics card costing over 500 bucks on release.
The original Quadro Plex started at $10,000.

And if you go way back to the early days of the video card reviews here you'll find:

X850XT@$415

7900GTX@$500

7950GX2@$550

Going further back, I paid $500 for my 9800Pro, and $400 for my 9700Pro before that.

The high end graphics market has always been in that $400-600 range, and overpriced. And even today there are wonderful sweet spots that reasonably price.
Posted on Reply
#15
mdsx1950
newtekie1 said:

7950GX2@$550
.
143W power draw is nothing to take lightly
Thats funny considering nowaday midrange GPUs take that much and are considered as non-power hungry cards. :laugh:
Posted on Reply
#16
1freedude
part of the price tag includes superior customer support, as in talk to a nVidia engineer support.

most of the people using these cards aren't computer wizards, they draw and edit "stuff."
Posted on Reply
#17
Bjorn_Of_Iceland
Quadro Plex 7000, with 12 GB (total) of memory and 896 CUDA cores...

so basically a dual 470?
Posted on Reply
#19
newtekie1
Semi-Retired Folder
Bjorn_Of_Iceland said:
Quadro Plex 7000, with 12 GB (total) of memory and 896 CUDA cores...

so basically a dual 470?
Dual Quadro 6000 actually, which does use the same core as the 470 w/ 448 "CUDA Cores", but it has 6GB of memory.

Also, I seem to remember seeing it mentioned that these cards have hardware ECC enabled, while the desktop parts do not. This slows them down even further, but makes them more accurate and reliable for data processing and rendering, it also means a certain amount of the memory on the board is actually used for the parity. So if it is 6GB usable, I think each card might actually have something like 7.5GB physical memory on the board.
Posted on Reply
#20
AccessCode
How many digits of Pi per second can this thing calculate???
Posted on Reply
#21
Laurijan
Is the GFX memory duplicated in the RAM leaving you with up to 6Gb (or 12Gb) less RAM for apps?
Posted on Reply
#22
kaneda
Laurijan said:
Is the GFX memory duplicated in the RAM leaving you with up to 6Gb (or 12Gb) less RAM for apps?
These are not for DX9 appz.

People need to stop asking for game reviews too, its boring and happens every time workstation cards are released(or news on them is released).

Workstation cards are based on the same core as commercial graphics cards, you pay a premium for specialized drivers and stability.The drivers are tailored to meet the requirements of professional applications to offer considerably better performance. Just think of regular graphics driver updates, you often get slight improvements in certain games? same thing, software(drivers) is updated and designed to work with these professional applications rather than games. They also tend to be more stable, choice components and the such to ensure people using these cards get the stability their work requires.

These cards are not for gaming, if you buy one for gaming you are a fool.


As for this PLEX. BLAH.
Unless you can turn it into a render node(s). that would be kinda cool, having X amount of dedicated GPU's for accelerated rendering/encoding.
Posted on Reply
#23
ToTTenTranz
That Quadro 5000M is intriguing, to say the least.

320 isn't a multiple of 48, so this card definitely has a GF100.

Why use the GF100, if they already have the more power-eficient GF104 with them?
L2 Cache size? Memory bandwidth?
Posted on Reply
#24
kaneda
ToTTenTranz said:
That Quadro 5000M is intriguing, to say the least.

320 isn't a multiple of 48, so this card definitely has a GF100.

Why use the GF100, if they already have the more power-eficient GF104 with them?
L2 Cache size? Memory bandwidth?
Probably getting rid of stock/was already in production. etc.

something like that would be my guess.


companies do enjoy milking the market :/
Posted on Reply
#25
Jizzler
From Xbitlabs.com:



Can read the rest of the review here. Doesn't have additional game benchmarks, however, is interesting if you want to see how a GTX 470 performs in Maya, 3ds max, and others.
Posted on Reply
Add your own comment