30 day free trial of Pro/E!

Thatza Spicy GPU! New NVIDIA Quadro FX Graphics Cards. Best One for SolidWorks?

by Josh on March 31, 2009 · View Comments

If you’re a fan of meatballs and Nvidia Graphics Cards or “GPU’s” you’re gonna throw noodles in the air and set the table for the new Quadro FX family that has just come into town.

Nvidia has release their 10th generation cards and the “Strongest Professional Graphics Lineup in History” is the byline every GPU manufacturer wishes they could claim. Does it stand up to the new Quadro line-up that has been a staple of graphics professionals and CAD users?

Let’s take a look and see which one would work best for SolidWorks and other 3D CAD tools.

The Quadro FX family
There are 6 new cards boosting everything from Memory Bandwidth and performance to parallel processing power. The most important aspect for working daily in SolidWorks, from a spec sheet glance, is going to be Memory Bandwidth (the rate at which data can be read) and relative performance.

The 6 new cards are the (in order from hgh-end ot low-end) Quadro FX 5800, 4800, 3800, 1800, 580 and the 380. Here’s a breakdown of the new cards against the card (if Nvidia) that you are most likely using now.

Click to Enlarge

Download PDF (248kb)

The mightiest FX beast of all
The head of the FX family is a massively powerful card which is the first GPU pushing 4GB of memory. To show you what the possibilities are, here’s the main specs on the Quadro FX 5800.

Quadro FX 5800 Specs

  • 4GB Memory
  • up to 240 CUDA programmable parallel cores
  • 102 GB per second Memory Bandwidth
  • 52 billion texels (a pixel on a texture) per second
  • 300 million triangles per second
  • Price: $3,499 MSRP ($3103.99 Buy.com)

To put this in perspective, the mid-range Quadro FX 1700 has:

  • 512 MB Memeory
  • 32 CUDA parallel cores
  • 12.8 GB per second Memory Bandwidth
  • 7.4 Billion texels per second
  • 191 triangle per second
  • Price: $429.96

The FX 1700 is a SolidWorks Certified Graphics Card and from my experience, this mid-range line is one of the most commonly used cards. The step up from the 1700 is the FX 1800, and as seen by the specs above, it raises performance and bandwidth considerably.

The Quadro FX Family Comparison
Click to Enlarge

The best graphics card for SolidWorks?
In the new NVIDIA line-up, the high-end cards are attractive. The thing is, for most people, they’ll be overkill. For a card with great performance and good price the FX 1800 outshines the rest. With a performance score of 60.23* it competes nicely against the high-end cards with roughly the same scores and steps up the performance quit a bit against the FX 1700 at 42.40. The price will come in around $599 MSRP.

Why would you need a high-end card?
If you’re producing HD content or involved in some crazy complex visualizations either with rendering or video, the high-end cards are going to serve you better. Some of the high-end cards also allow you to run multiple OS environments on a single computer using NVIDIA SLI Multi OS Technology. The increased CUDA processing cores will allow for better element analysis and rendering used for computation analytics and graphical data recognition.

Overall
GPU power and price are often a source of contention with SolidWorks users and IT professionals. Nvidia has pretty much always been a great video card option for the SolidWorks users. You want a card that works well for your needs and for the regular SolidWorks user the high-end cards won’t provide much more for what you would pay. Consider your needs and go from there. Generally, it would be better to use some of that money on a system upgrade, more memory or another monitor.

Leo Tolstoy once asked “How much land does a man need?” – If he were alive today he might ask, how much memory bandwidth does a man need?… But then he would see these graphics cards and be like, give me that 5800 dude.

*“represents the geometric mean of the viewperf viewsets and is intended to provide a relative performance difference. Application scaling may vary. SPECviewperf® 10 for more information visit www.spec.org“

Tagged as: fx series, gpu, graphics cards, nvidia, solidworks gpu

  • Ox
    There, finally something useful.
  • D Mytty
    The conclusion of this article is that Solidworks users don't need powerful video cards. The real question is: why not?

    Why is there no mention of CUDA or OpenCL GPGPU coding efforts for Solidworks when discussing video cards for an application? Try googling...you will quickly find that there is scant mention of the concept anywhere.

    IMO, it certainly is newsworthy that elements like 3D Computational Fluid Dynamics modeling (Solidworks Flow Simulation) have been demonstrated to be 16x faster in a parallel GPGPU environment. So why has no one put the question of GPGPU support to DS Solidworks? Is this another example of tame journalists and bloggers looking for handouts?

    To offer futher critical commentary, this 'article' is largely a rewrite of the Nvidia press release. There's no discussion of power consumption, fan noise, stereoscopic support or the link between MultiOS SLI and Parallels Workstation Extreme virtualization. On even the most basic of elements, such as FX1700 being a 'common card', how about a "% of market" number to back that up?

    This 'article' is like most in the IT news: it's fluffy 'low cal' commentary that passes off as 'journalism'. Unfortunately, wasting readers time is the rule, not the exception.
  • Hey there D. Thanks for the criticism... really. let me explain some stuff.

    First, the conclusion is that 'most' solidworks users don't need the high-end cards. most that I know are not doing flow simulations, rendering or FEA.

    I don't have any of these new card. I haven't tested them out. I'm basing everything I wrote on my experience with Nvidia FX series cards I've used. So I can't quite comment on the noise, energy saving, etc, although I wish I could do a thorough review, analysis, comparison of each.

    I posted bits that have helped me choose/review GPU's in the past. It's not going to stand up to the info that everyone needs to know, but I'd think it would help most. at least I hope it does.

    I'm also not a journalist. I write about what I'm interested in and what I think may help other SolidWorks users and people interested in tech and design.

    From what I've found it's a little more specific than the other 'IT news' and I aim to add more info that what you would get a site spouting generalities or copying press releases. If nothing else, you'll usually get my opinion.

    Hope this clears some things up for ya.

    Oh, an on the 'common card' thing, I agree, I looked for stats on the FX line, but could not find the data. I wanted to tell my experience about these cards, and mine is that the FX mid-range series is the most common I see. granted, my view probably a bit limited. anyway, it's my opinion that this mid-range card is still gonna be the best for the $$ for the average SolidWorks user.
  • D Mytty
    My criticism is a compliment...really.

    I don't critique unless it seems that there's actually potential for improvement. While there's a fair bit of 'shooting the shoddy messengers' in my post, I did choose you as the target because there's a sense that you would actually take the criticism to heart and could take it with a little constructive salt. You already have progressed beyond cut and paste, now you gotta 'up the game'.

    You have an excellent writing style and good experience, and that just sets expectations higher. Hopefully you're able to translate your Solidworks experience into more tech postings. More so, I would hope that you could leverage your blog into pressing Solidworks for some answers on issues such as OpenCL and CUDA.

    I maintain that the issue of GPGPU and Simulation could be made relevant for today's GPU purchases given the right disclosure. Simply put, most things which are designed in 3D could actually benefit from Simulation. Once you've used Simulation and watched it buckle down for 16 hours on a fairly simple problem, even an average user wants some speedup. Given the recent advancements in parallel processing involving the GPU, that speedup is available with the right software enhancements to Solidworks. Now we just need disclosure when that's going to happen and whether it will be vendor specific (CUDA, Brook, Intel Larrabee) or standards based (OpenCL).

    As to the Nvidia GPU conclusions of your post, there are several usage scenarios which would make the Quadro FX3800 seem like money better spent. Firstly, if Solidworks announced (or even hinted) that they were making Simulation 'GPU enhanced' (OpenCL, etc) within 1-2 years, then it would definitely make sense to buy a faster GPU to yield a 16x Simulation speedup. Secondly, if users are contemplating stereoscopic design (ie 3D screen), then again the FX3800 is better. Thirdly, if the user has multiple PCs sitting at their desk (with a mix of Windows/Linux apps), then the multi-OS SLI capabilities of the FX3800 could actually mean a quicker ROI by eliminating machines and their upgrades. If a user even has an inkling that they might become a 'power user' along those lines, then the FX3800 is definitely the right buy.

    On the subject of IT coverage, I see an enormous opportunity for an ambitious writer. Given the power of Solidworks (realized and potential), and Solidwork's broad appeal to users that are continually advancing their skills with the software, I am quite amazed by the generally low level of 'IT journalism' coverage. This isn't about reviewing gaming hardware for teenagers, but it can seem like it. The reality is that coverage doesn't come close to meeting the needs of the professional user base that pays big bucks and makes big bucks with Solidworks.

    Time to hit some home runs. Start hammering that keyboard and phonepad like a juiced up A Rod!
  • jheinz2
    D, I think you hit this article on the head. I'm a professional that has been scouring the net looking for a good reason to justify a better GPU for our company. Currently running an FX1700. Although all the benchmarks for this card (results of others and that of my own) look very good, I believe myself and one other individual in my company can be considered a "Power-user" as our models house between 5 and 10 thousand components. I can rebuild a small <1000 piece part in no time at all, but when we open the entire assembly, I usually have to sit around for 10 min saving, or it fails to load at all, and any movement or view change is deadful!

    With that being said, what I'm seeing so far is that the FX3800 is looking like the stand-up winner on my recommendation, partly due to your comments as they reaffirmed my research thus far! I want to be sure what we have now will work for the next couple years as our model gets bigger and more analysis is done on individual components.

    As for you Josh, keep up the good work! There definitely needs to be more "USEFUL" information out there for the professional. All I've been finding is teenagers bragging how good this and that does for Crysis etc...

    Thanks for the help!
  • Thanks J, I'm certainly trying to put out a little more info than what's typically available. While I don't have the resources to do a full-up article, I'm at least trying to get people headed in the right direction with some opinion and links. Thanks for reading and thanks VERY much for the the comment!
  • Ha! awesome. thanks, I'll take take it as such and keep tearing away it with all your input very strongly noted.

    I very much wish I could do a lot more n the tech side, and though my knowledge is somewhat limited, I'm learning quickly what readers are looking for and interested in. I wish I had more comments like yours actually. thanks again for the comments and also thanks for reading.

    Btw, if you're interested in doing a guest post on this stuff, let me know. Sounds like you have the knowledge and insight that could really expose a lot on the GPU side of things. see ya D.
  • What about using 2 cheap cards to get better performance?
  • diverso
    Josh, excellent summary! now where and when can I buy mine?

    Are you working on anything from the Dell's release of the New Generation Precision (T5500) Workstations with Nehalem Architecture CPUs? it would a perfect combo for us.
  • Thanks!! I thought the 1800 would be available now, but there's still nothing up on the NVIDIA site. The best prices I've found online previously are at Amazon or Buy.com.

    I'm trying to wrap my head around the Dell and HP launches as well, but I'm up against a deadline for some drawings and another article. Hopefully I'll get out something as things cool down. So much cool tech coming out!!!
  • diverso
    No worries, part of my job is researching anything SolidWorks (AKA reading SolidSmack) so I'll stayed tuned!
    If there's anything I can help out with, let me know. I usually prepare new tech or SolidWorks News reports myself, but our IT Dept. doesn't get on the bandwagon until I send them a link just like this post. Keep up the good Smacking!!!
  • Why dont they just stick a Quad Xeon processor on the video card and make the thing a computer too.

    Do away with the whole desktop. You just plug your keyboard, mouse and monitor right into the card !
  • I am so down with that. I'm sure the GPU will continually be offloading computations from the CPU, till it's hardly necessary. I'm sure they're playing around with this. I know the ,a href="http://openvidia.sourceforge.net/">OpenVIDIA project is pretty close to this.
blog comments powered by Disqus