Nvidia's gaming graphics cards have been slowly moving away from the old-faithful G80 chip. Now, in fact, Nvidia's latest workstation offerings are outfitted with derivatives of Nvidia's GT200 GPUs, which many have come to know through their inclusion in the cheaper mainstream GeForce GTX 260 and 280 cards.
The first entrants in this product line are the Quadro FX 5800 at the extreme high-end, with its 4 GB frame buffer, and the more "reasonably priced" Quadro FX 4800, with its 1.5 GB of graphics memory. That latter board found its way into our test labs, so that we could put it through its paces. In a few weeks, we expect to get our hands on other models in this series as well.
The FX 4800 commands a price premium of over $300 compared to the equivalent FirePro V8700 from AMD's ATI division, which online resellers retail for about $1,250. Can buyers expect a boost in features and performance for this extra outlay? First, let's take a look at the comparable cards and their speeds and feeds from both companies in the tables that follow, then we'll get around to answering this burning question.
| Workstation Graphics Cards and their Mainstream Equivalents | ||||||
|---|---|---|---|---|---|---|
| Workstation Model | Derivative GPU | Fab | Mainstream-Equivalent | Graphics RAM | 3-Pin Stereo | DisplayPort |
| Nvidia Quadro FX 5600 | G80 | 90 nm | GeForce 8800 | 1,536 MB GDDR3 | yes | no |
| Nvidia Quadro FX 4800 | GT200 | 65 nm | GeForce GTX 260 (280) | 1,536 MB GDDR3 | yes | yes |
| Nvidia Quadro FX 4600 | G80 | 90 nm | GeForce 8800 | 768 MB GDDR3 | yes | no |
| Nvidia Quadro FX 1700 | G84 | 80 nm | GeForce 8600 | 512 MB DDR2 | yes | no |
| ATI FirePro V8700 | RV770 | 55 nm | Radeon HD 4870 | 1,024 MB GDDR5 | yes | yes |
| ATI FireGL V7700 | RV670 | 55 nm | Radeon HD 3850 | 512 MB GDDR4 | yes | yes |
| ATI FireGL V5600 | RV630 | 65 nm | Radeon HD 2600 XT | 512 MB GDDR4 | no | no |
| ATI FireGL V3600 | RV630 | 65 nm | Radeon HD 2600 Pro | 256 MB DDR2 | no | no |
| Workstation Model | Memory (RAM) Bandwidth | DirectX | OpenGL | Shader Model | Core Clock | Memory Clock | Pixel & Vertex Processing |
| Nvidia Quadro FX 5600 | 76.8 GB/s | 10 | 2.1 | 4.0 | 600 MHz | 800 MHz | 112 SPUs |
| Nvidia Quadro FX 4800 | 76.8 GB/s | 10 | 3.0 | 4.0 | 600 MHz | 800 MHz | 192 SPUs |
| Nvidia Quadro FX 4600 | 67.2 GB/s | 10 | 2.1 | 4.0 | 500 MHz | 700 MHz | 112 SPUs |
| Nvidia Quadro FX 1700 | 12.8 GB/s | 10 | 2.1 | 4.0 | 460 MHz | 400 MHz | 32 SPUs |
| ATI FirePro V8700 | 115.2 GB/s | 10.1 | 2.1 | 4.0 | 750 MHz | 900 MHz | 800 SPUs |
| ATI FireGL V7700 | 72.0 GB/s | 10.1 | 2.1 | 4.0 | 775 MHz | 1,125 MHz | 320 SPUs |
| ATI FireGL V5600 | 35.1 GB/s | 10 | 2.1 | 4.0 | 800 MHz | 1,100 MHz | 120 SPUs |
| ATI FireGL V3600 | 15.8 GB/s | 10 | 2.1 | 4.0 | 600 MHz | 500 MHz | 120 SPUs |
- Introduction
- Quadro FX 4800 Hardware Details
- Software Driver Features
- Test Configuration
- Benchmark Results: Maya
- Benchmark Results: 3ds Max
- Benchmark Results: Solidworks
- Benchmark Results: Viewperf I
- Benchmark Results: Viewperf II
- Performance Gaming Vs. Workstation: GeForce GTX 280 And Quadro FX 4800
- Conclusion: Editor's Recommendation
More and more applications are using DirectX as their main graphics API, yes Quadros will out preform GeForce cards using OpenGL (that is what they have been histiruically designbed to do) but DirectX performnce is another story. For 3DS Max, Autodesk Inventor, AutoCAD users to name a few, there is no compelling reason to spend big bucks on a Quadro
More and more applications are using DirectX as their main graphics API, yes Quadros will out preform GeForce cards using OpenGL (that is what they have been histiruically designbed to do) but DirectX performnce is another story. For 3DS Max, Autodesk Inventor, AutoCAD users to name a few, there is no compelling reason to spend big bucks on a Quadro
The entire review sounded like it was paid and edited for by Nvidia.
I work with Maya, Modo, Zbrush and Mudbox on a daily basis as part of my work and i also have trial versions at home for testing, and there is no practicle difference between the gaming card and the workstation. And most notably you dont need the Quadro driver hack, using normal Geforce drivers are plenty quick enough.
There are occasionally the odd driver issue when using games orientated drivers but these are normaly fixed with newer versions or going back a driver version.
It would be nice, when reviewing workstation graphics cards if you could post images of whats being rendered. As these cards handle totaly different depending on circumstance and type of render irrelevant to whether they are gaming or workstation cards.
You can run still them on non open_gl optimised drivers but they are incredibly flaky and performance is a joke.
3Ds Max while a powerful 3d design tool is closer to being a consumer level tool when compared against some of the bigger Product Lifecycle Management suites used in those industries.
Like the reviewer mentioned the drivers are written and optimised specifically around these applications and the application writers specify which version of the driver is supported and ratified stable.
This gives companies the peace of mind that their expensive CAD users are not going to lose 3 days work on a huge assembly because of a driver crashing.
Some of the models used in these apps are nothing short of huge. In the company I work for it can take over an 30 mins to check out and load a complex assembly and that's with best of breed PLM solution, great network architecture and extremely fast twin xeon workstations.
Decent review I thought.
Ed.
I totally agree that Workstation class cards are the only option for some high end CAD/PLM products that *still* run exclusively under OpenGL. After all you would not like to think the Airbus you just strapped yourelf into, was designed by some Catia monkey using a £200 GeForce card now would you?
No the only GPU acclerated rendering system at the moment is nVidias 'Gelato' which is frankly pants. I'm still waiting for Autodesk to get their head out of thei rears and launch a GPU rendering solution.
How much of this is down to physical architecture rather than drivers is another matter. (It would be interesting to see a comparison of games/applications on high-end gaming & workstation cards with original and hacked drivers (gaming to workstation, etc).