Quadro vs. FirePro vs. Consumer GPU for AutoCAD Civil 3D 2012

Status
Not open for further replies.
G

Guest

Guest
Greetings @ all.

As it so happens, I find myself on the horns of a trilemma.
What should I opt for? A professional NVIDIA Quadro or AMD FirePro card, or a generic Radeon / Geforce GPU?

I have to build about half-a-dozen (5-7) systems meant for modeling civil engineering projects (fly-overs, roadways, corridors and the like) using Autodesk AutoCAD Civil 3D 2012 + Revit along with the Adobe CS5 (Photoshop, Premiere), and maybe a little 3ds Max or Maya on the side for visualisation purposes. But the mainstay will be AutoCAD, with the rest being side-effects. The users will be manipulating medium to large data sets.

There's one snag however - these systems are for a small civil engineering consultancy which has a turnover of only about $400,000. Naturally, they are keen to maximize the value they get for their money. I haven't yet decided on the rest of the specs for the machines, but I thought the GPU would be a good place to start.

The GPU's under consideration are:

> NVIDIA Workstation Class: Quadro 2000, Quadro 600
> AMD Workstation Class: FirePro v5800, FirePro v4800
> NVIDIA Consumer Grade: ? (Open to suggestions)
> AMD Consumer Grade: ? (Open to suggestions)

I think I'd prefer a workstation card to a consumer one, unless someone can prove conclusively that a high-end gaming card is better than a low-end professional one. And please, NO SOFT-MODS! :non:

CUDA would be nice for CS5, though I'm unsure if AutoCAD benefits from it. Is this applicable to AutoCAD Civil 3D?

There'll be a maximum of two monitors attached to any system @ 1900x1200 tops, so I don't think Eyefinity is required. Please correct me if I'm wrong.

I've heard that NVIDIA's workstation drivers are superior to AMD's. Is that so? And does the performance gain justify the price disparity between the two?

What do you think would be the minimum for a smooth experience - little to no juddering, crashing, hanging, or frame dropping when zooming, panning and rotating a large data set (say 1000+ points)?

Assuming that the FirePro v4800 and the Quadro 600 survive this brutality (will they?), which one would you take? The gripes I have with the Quadro are that (a) it offers GDDR3 memory, as opposed to GDDR5 for the FirePro v4800 (b) it costs $10 more on Newegg than the v4800, while the user rating it carries is one egg less than the v4800 on the same site and (c) for the life of me, I can't find any reviews about it online. On the plus side, it offers the supposedly better NVIDIA drivers.

In case the FirePro v4800 and Quadro 600 are massacred, which one would you suggest as being the best-bang-for-buck: the Quadro 2000 or FirePro v5800? Any reasons why?

What would be the best GFX card for DCC, which doesn't require you to sever an arm and a leg for it?

I would appreciate it if you could suggest at-least two cards as your first and second recommendations. I'd be most grateful for any links to articles and benchmark tests that you could provide to back up your suggestions and help me make my decision.

Thank you for your valuable time. Kudos to you folks.

Ciao. :)
 
Solution
Our company purchased machines with the Nvidia Quadro 600 to be used with almost the exact same software C3D 2012 & CS5 minus Revit. The performance is good but as stated earlier, this card is only good to handle light 3d renderings. I have had plenty of fatal errors while rebuilding corridors that are about 1.5 miles long, specially if there are complex assemblies in them.

System Specs:
Intel core i7 870 @2.93GHz
8Gb DDR3
Nvidia Quadro 600
G

Guest

Guest
Hi
Quadro cards use ecc ram. They're underclocked consumer cards made for precision not speed. The drivers are fantastic because there were considerable time put in to it so you won't get any crashes like a person now and then get on consumer cards. Coz you know a crash cost companies money.
For Autocad programs, you’ll want to upgrade to a faster Nvidia Quadro card to improve performance. While the Quadro 600 can handle some light workloads, it is not recommended if 3D modeling and Autocad is what you’ll mainly use a workstation for. So the better one even if it costs more it will save you money in the end by performing better.
Nvidia Quadro 5000 2.5GB or the nvidia quadro 4000 2Gb ones are advised

Performance wise, the latest Quadro line-up (Quadro 600, 2000, 4000, 5000 and 6000, avoid the older FX series) has the edge overthe FirePro series. Drivers wise, this is also the case, with Nvidia drivers being more stable, Quadro cards supported by more applications and finally, their drivers are more optimized for greater performance in some applications.

Thanks for the prompt reply.
Unfortunately, I don't think the Quadro 4000 will fit into the budget, let alone the 5000. The 2000 is pretty much the ceiling for now. :(

Odd... it doesn't seem that either of those hi-end cards is certified for advanced AutoCAD features here.

The 1000-5000 points I suggested was the very worst case scenario. Most likely, it'll be something like ~1000 points.

Any comment on the AMD cards?
 
G

Guest

Guest
The Quadro 2000 1gb is the minimum you should get. Below that their not worth it. AutoCad requires a workstation card I think as well. Firepro cards are not up to scratch and supported by a few applications only.
Remember they got more quality components than consumer cards although the gpu might be the same and got a longer warranty normal.

Why do you say that they aren't they up to scratch?
If you look here, it seems that they not only are certified for AutoCAD use, but recommended as well.
 
G

Guest

Guest
Yes they are supported
http://usa.autodesk.com/civil-3d/system-requirements/

I know they're supported.
What I meant was, they weren't specified on NVIDIA's site to be certified. Never mind. ;)
 
G

Guest

Guest


Apparently not all Quadro cards use ECC RAM.
 
ECC ram is not particularly important for your application. It's only really essential in computers being run hard in critical applications 24/7, such as in servers or compute/Tesla clusters. The errors ECC corrects you should never really experience in normal workstation use.

Unfortunately I do not know a ton yet about the workstation cards firsthand due to how expensive they are. I have dabbled in the softmodding and can confirm it fails on all counts.

I do plan to get a Quadro 600 in a couple months for my light modeling needs. As I do more modeling for simulation, my 3D models aren't particularly complex (the simulation is though). For medium to low complexity parts, a gaming or low end Quadro is fine. Most of the 3D modeling I have done is in Autodesk Inventor.

Do you have any samples of their models? If you do, you could buy and build one system with a Quadro 600 and see if it is sufficient. And if it is not, you could return the card for a small restocking fee and buy the Quadro 2000 or an AMD card.
 
G

Guest

Guest


Sounds like a good idea, but I'm lazy and would rather that someone conjure up a review from thin cyberspace comparing the 4 cards. BTW, I was just replying to Dadiggle's claim that NVIDIA cards came with ECC RAM, that's all.
 
G

Guest

Guest


I think its the latest one - if you look at the footnotes, it says SPECViewperf 11, which is AFAIK the latest. Now the problem is I cannot find SPEC Viewperf 11 tests for the v5800 and the v4800. Got a few with SPECViewperf 10, but no 11. Anyone?
 
G

Guest

Guest
I do not claim I know they come with it. my wife use one.

For high-precision, data-sensitive applications, Quadro 2000 is the only professional graphics solution with ECC memory and fast double precision capabilities to ensure the accuracy and fidelity in your results.
http://www.jigsaw24.com/product-details/p132aya/pny-nvidia-quadro-2000-1gb-1-x-dvi-i-2-x-dp-192-core-pci-e-x16-graphics-card

Ecc ram is very important. Its one of the factors that add to the cards precision. Workstation , they need to be perfectly accuate at all times, speeds good but not priority, major differences in the cards are ECC ram (expensive, but a lost bit could be catastophic) and very robust and detailed driver support...the drivers are really where the bulk of the expense comes from and , they have to work and have to work well, we gripe and moan about bugs in the drivers. There are many products out that scale th same way, look at cameras, a $200 point and shoot for the trip to disney world or the $5000+++ used to shoot what we see inmagazines and such, a nice solid sub $500 asian import guitar vs a $3K+ made in the USA model, the latter has better quality components and much more time and care in onstruction.

Thanks. Most likely it will be the Quadro 2000.

Any comments on the FirePro v4800 vs. Quadro 600, assuming the Quadro 2000 costs too much?
 
I don't think that page is right. According to NVidia, the 2000 does not have ECC (hmdhruvarora's link up a few posts). And why would the mid range be the only with ECC (and not the high)? I think only the 5000 and 6000 have ECC, but you won't notice a difference with or without it so don't worry. The 'precision' Quadros bring are in the drivers. The fast double precision is for CUDA, not CAD, and is only in the 4000 and up. 2000 and 600 are artificially limited like the gaming cards. http://www.anandtech.com/show/3961/nvidia-launches-quadro-2000-600

I haven't found much on the 600 vs V4800 yet, but I'll try to look a little more later.
 
G

Guest

Guest


Thanks for taking the time out to help me EXT64. I appreciate it.
Yes, I realize that ECC RAM is not required, and that it may even come with a performance penalty.

And I agree - if you check out NVIDIA's site, it seems that only the Quadro 5000 & 6000 are equipped with ECC RAM. Else, NVIDIA would be touting the fact that even their low-end cards use it.

 
No problem. I wish I could actually find what you are looking for, but I haven't had much luck. I've been trying to decide between the mid/low end pro cards as well for a while and haven't found much real information on the new cards.

I've seen quite a few people say the FirePros are a better value, however they give no evidence to back it up, thus the claim is worthless.

I did find this review, however unfortunately it is for the previous generation of cards: http://www.cgarchitect.com/news/Reviews/Review076_1.asp
 
G

Guest

Guest


That review has some value, albeit limited. For instance, the Quadro 600 can replace the FX 580 and the Quadro 2000 can replace the FX 3800 to give a rough idea of relative performance.

Why is it that workstation cards are the ones which are barely reviewed compared to gaming GPU's whereas they're the ones that have, arguably, the more critical applications of the two? :pfff:

I suppose we could compare the high end GFX cards, and see which one comes out on top. http://www.tomshardware.com/reviews/firepro-v9800-eyefinity-quadro-5000,2780-4.html

I think I'll adopt a simple formula - Quadro over FirePro, if the budget isn't an issue.

Here's my $0.02:

> Ultra Hi-end = Quadro (unless single-card 2+ monitor Eyefinity is required) http://www.tomshardware.com/reviews/firepro-v9800-eyefinity-quadro-5000,2780.html

> Hi-end = Pretty much a tie between the Quadro 4000 and FirePro v7800 as is demonstrated by this review. Depends on your application - if you need CUDA for CS5, then NVIDIA; if you use some program that can utilize APP or would like to save a few bucks and get tri-monitor Eyefinity capability, then AMD.

> Mid range = Might be the Quadro 2000.
1. Here is a review which compares the FirePro v5800 with the Quadro FX 3800.
2. And here is one that compares the Quadro FX 3800 with the Quadro 2000.

In review #1, considering only the vanilla SPECViewperf 10 test (not considering the Multisample and CPU scaling ones), I calculate the FX 3800 yielded an average performance of 100.1% (approx.) w.r.t. the v5800. But the FirePro scaled better than the Quadro in the review. In light of the fact that we are comparing a previous gen product (FX 3800) with a newer gen one (v5800), I'd be prepared to use the scalability factor as only a tie breaker.

In review #2, we have a new version of the test suite (SPECViewperf 11), and no 'real-world' score for the FX 3800, only the official NVIDIA one. The new test suite shouldn't be much of a problem as we already know that the the FX 5800 is 0.1% better than the FirePro - I think we can expect similar figures from the new test suite's benchmarks as well.

What I did was this: I took the 'real' Quadro 2000 scores, divided them by the official Quadro 2000 scores and multiplied them with the official FX 3800 scores, to arrive at what the approximate, average FX 3800 'real-world' scores might look like. (Simple unitary method, that's all.)

Then, expressing the Quadro 2000 scores as a percentage of that of the FX 3800, I came up with 114.306% (approx.). Multiplying that by the FX 3800's performance compared to the v5800 (100.1%), we have: (114.306*100.1)% = 114.42% (approx.)

So we can conclude that the Quadro 2000 is on an average approximately 14.42% faster than the FirePro v5800. Mind you, this is more like a guesstimate than anything else. If NVIDIA's official figures are to be believed, the Quadro's way faster than the FirePro.

A quick check on Newegg finds the Quadro 2000 sitting pretty at $450, while the FirePro v5800 is at $385. Expressing the price difference as a percentage of the v5800's cost, we find that the Quadro 2000 costs 16.88% more than the FirePro at the moment. So would you rather save $65 with a tri-monitor Eyefinity card and perhaps better scalability, or shell out the money and sacrifice a display to get the advantage of CUDA and the supposedly better NVIDIA drivers?

I'm not sure if the Quadro 2000 supports SLI. According to review #2, it doesn't, but the NVIDIA site specifies it supports Multi-OS SLI. Weird... :??: The FirePro v5800 definitely supports Crossfire though.

BTW, I think people aren't very happy with PNY (AFAIK, they're the only ones who make Quadro cards). Customer service, packaging and general impression don't seem to be all too good.

> Low-end = (?)
Buy both the Quadro 600 and the FirePro v4800, bench them for us. :D
Keep the one with better results, and return the worse performing card for a small restocking fee (like you said). The FirePro v4800 comes with GDDR5 as opposed to GDDR3 for the Quadro 600, so that might be a point in its favour. IMHO, the low-end is what it is - not worth it if you plan to do anything more than light 3D work. And if that's the case, why pay the premium? Might as well get a good gaming card for that kind of money.

Thanks for your time.
TTFN.

 
Wow, excellent summary. In one of your other threads, I answered the SLI question (which is no for the 2000. Multi OS is just that, using one card for each OS).

I have heard the same things about PNY, which shocks me. Quadros are supposed to be all about professionalism and support, yet NVidia lets PNY act like this?

I'm more of an AMD person, so I certainly wouldn't mind the v4800. What brought on this upgrade for me was that one of my Finite Element modeling software refuses to load on my ATI 4850. It works fine on my GTX470, but I want it to run on the computer with the 4850. So really any NVidia Geforce or Quadro would work, but since the Quadro 600 is so cheap I was considering it. Now I might try the v4800 like you suggest, and return it if it doesn't work.
 
G

Guest

Guest


Yep, you were right. I asked NVIDIA customer support this question for good measure:

nvidiaquestion.jpg


And here was the response:

nvidiaresponse.jpg


So that's confirmed. Thanks. :)
 

hectorbls

Distinguished
Sep 9, 2009
3
0
18,520
Our company purchased machines with the Nvidia Quadro 600 to be used with almost the exact same software C3D 2012 & CS5 minus Revit. The performance is good but as stated earlier, this card is only good to handle light 3d renderings. I have had plenty of fatal errors while rebuilding corridors that are about 1.5 miles long, specially if there are complex assemblies in them.

System Specs:
Intel core i7 870 @2.93GHz
8Gb DDR3
Nvidia Quadro 600
 
Solution
G

Guest

Guest


Some concrete facts at last...! Thanks a lot hectorbls.

BTW, if you are wondering, that contract hasn't yet materialised which is why I haven't been able to post any benchmarks. :(

It's been long enough, so I think I'll close the thread...

Cheers! ;)
 
Status
Not open for further replies.