Why do so many recommend GTX graphics cards over Quadro? Why do you feel Autodesk is wrong when they recommend Quadro? My choice is nVidia as a brand, since I hope they will be around in the future for any support issues. I’ve had bad experiences with card manufacturers disappearing or being acquired, such as 3Dlabs (acquired by Creative, then disbanded) and Diamond Multimedia (acquired by ATI, support suspended).
I’m in configuration mode for a new workstation, and must carefully select components so the hardware can handle large assemblies (up to 10,000 parts of varying complexity).
I need the latest thinking on this. What I’ve seen so far (I’ve sorted discussions by date.) are dismissals of the value of using Quadro over other high end graphics cards, but no real specifics on what the problems are when using Quadro cards vs GTX, for example. There have been statements that performance of Quadro is no better, or even worse than alternatives. Please explain how Autodesk’s graphics card recommendations fail to include the alternatives, which would support the claims of users. Is it because it is too expensive to research alternatives, or because that is all they use when certifying it to be used with the software? I probably should find any comments by Autodesk that specifically state why NOT to use anything but Quadro. nVidia does have some attractive claims concerning their CUDA technology that seem to indicate it would enhance performance when running Inventor.
I will be running:
Autodesk Product Design Suite 2012
Graphics Card: TBD
Intel Core i7-990X
SSD: OCZ Technology Revo Drive 3 Series 240GB PCI Express
SATA Drive-1 Tbyte: TBD
OS: MS Windows 7, 64-bit
Pwr Sply: TBD
cut and pasted from:
"it's all back to OpenGL v DirectX... Inventor moved to DX about 6 years ago and since then workstation graphics cards and certified drivers aren't really needed.
Certification was needed for OpenGL as there was no official testing procedure of any drivers so Autodesk needed to do it in-house. Once they had tested and checked each driver with each software they would then "certify" it as usable (or not). DirectX drivers are all checked by Mircosoft (WHQL) and thus moves this work-load from Autodesk, as long as they themselves work to DirectX standards. Because of all this there's no need for workstation graphics cards as they are tweaked for OpenGL and excel in that environment.
If you're interested in history here's a pdf created back at the time of the transition with explanation from an Inventor programmer:
bottom line, Workstation (Quadro) cards are ideal (and recommended) for anything using OpenGL but anything using DirectX (as Inventor does now) then a gaming gpu will be fine. Too many of us using gaming cards without problems to suggest the ROI of a workstation card is justifiable.
Other cad SW still use OpenGL so they really need the more expensive workstation cards, but a workstation vendor shouldn't be ignoring gaming cards for use with Inventor (as long as the customer doesn't use any other OpenGL programs) - it's a great advantage of Inventor over other cad software which is rarely touted."
as to whether Autodesk would ever out and out say to use a gaming card or not... tbh, don't think they would as they worked together with Nvidia/ATI to do all the certification in the days of OpenGL. The workstation cards have a pretty hefty mark-up compared to gaming cards, so arguably more profitable for both the manufacturer and the hardware supplier - so obviously both will push/prefer to use them over the gaming cards So, with that in mind, do you think Autodesk would risk annoying Nvidia/ATI by ever officially saying to use the less profitable cards?
I've heard 2 regulalry used arguments from vendors trying to up-sell the workstation cards:
1) tailored/tweaked for use with OpenGL application which is CAD - Moot for Inventor as it's now DirectX. (argument still valid for other CAD software - so the vendor is either ill-educated about Inventor or just profit-grabbing)
2) workstation cards are designed to work 24/7 and last longer than gaming - Personally think this is bull and would love to see any proof. Just think of the number of servers and gaming-rigs around the world and how many of them left on for hours well above the average working day (not to mention kids/ppl leaving their pc's on 24/7 downloading music/films/etc) - all with gaming cards... Come on, if there were reliability problems with gaming cards there would be complaints all over the place.
So, why do ppl recommend gaming (Nvidia GTX) cards over the Quadro - probably price. Comparing the performance over price ratio of gaming cards there's a MASSIVE difference. Using Passmark's tables as benchmark data for the DirectX environment:
The Quadro 6000 performs similarly to the GeForce 560 (non Ti version). The Quadro is £3000 and the GTX 560 is about £120 - so a 25x difference in cost for the same performance... Personally know the GTX 560 is a bit of a waste of cash and would look at the GTX 560Ti (or higher) and thus you've got a card quicker than the Quadro...
Yeah, yeah, there are other factors at play, more than DirectX performance - gpu memory and CUDA performance when you're in the market for the Quadro 6000, but Inventor (apart from the Moldflow and Ansys addons) doesn't use CUDA and believe you really don't need more than 1gb v-ram. So, again, Quadro not needed.
that's an extreme case - for the more average user where the Quadro 600 or 2000 are more likely investments the performance-advantage of a gaming card is even more valid. The Quadro 600 retails for a similar £ to the GTX 560 but ranks #166 (score of 694), so for the same price you've got a card which is 4 times less powerful... hardly a desirable option...
Sure, workstation cards will work so I don't necessarily believe it's a "gaming cards work better than workstation" but a "you don't get any viable ROI for anything above a gaming card." At the end of the day, especially with the present global markets, you can't fritter away cash on expensive parts that don't return that investment thus why I (and many others) use gaming cards, and are happy with them.
Thank, Sam, so much for the comprehensive reply. I’m currently digesting the contents, which critique the Autodesk Quadro claims as follows:
“If you're interested in history here's a pdf created back at the time of the transition with explanation from an Inventor programmer:
bottom line, Workstation (Quadro) cards are ideal (and recommended) for anything using OpenGL but anything using DirectX (as Inventor does now) then a gaming gpu will be fine. Too many of us using gaming cards without problems to suggest the ROI of a workstation card is justifiable.”
The PDF you supplied makes absolutely NO mention of Quadro, which is an interesting omission, since that is what Autodesk is fixated on. I did understand your explanation of why Autodesk does not get outside of the Quadro “box”, as it was a business decision and not a technical one.
There is one thing, however, that your response did not address. I mentioned assembly size of up to 10,000 parts. Does that, in your opinion, still tilt an ROI decision towards gaming cards?
Since I have not yet researched gaming cards, do you have any recommendations on what you would consider the best for very large assemblies? Does the GTX 560 fill that bill, or is the GTX 560Ti preferable? This will not be an “average” workstation geared towards single parts and small assemblies. I do not want to get backed into hardware confined limitations, which can become insurmountable and require graphics card replacement. That would, indeed, be a monumental “headache” that many engineering department managers are ill equipped to deal with, so most all move the blame onto the “troops”, which is certainly not the case here.
I have not as yet followed the performance link you supplied because time is limited. I will do so shortly. The last time I accessed performance links to try to determine from them whether they fit my situation they didn’t address very large assembly files, but rather took a generic example (understandably necessary). I do hope the link makes some revelations on the situation here.
The number of parts is meaningless .. The complexity is what counts. I can make a single part that will bring a computer to its knees.
Just get the fastest directx gaming card out at purchasing time. The graphics card is probably the least of your worries.
Your CPU will be the bottleneck as many parts of Inventor could care less about more than 1 core.
The parts for this project can be represented relatively simply. But there could become a point where simplicity reaches its limits and part count takes over. This could reveal itself in the size of the Bill Of Materials required in conjunction with drawing file output. In the past I’ve noticed many have had issues with part and assembly drawings with their associated BOMs. Hopefully the 2012 release is better than associated comments in previous releases.
Your point is well taken about obtaining the fastest DirectX gaming card and compute intensive CPU. I won’t be doing any overclocking. I think that risks hardware longevity.
As always, memory will be a factor. There are different levels of memory performance as well, which I haven’t even addressed as yet as far as how that will impact performance. I’ll probably populate a board with a preliminary amount of the largest and fastest (?) available module and work up from there. Do you have any memory recommendations? OCZ and Corsair seem to provide some higher end memory solutions.
Good afternoon to all,
Guys, incredibly, I was looking for something with this subject..about 10 days, and today WORKSTATIONresearch decided to post.
I'm configuring a new workstation here, with the following config:
Intel Sandy Bridge Core I7 2600 3.40GHz Quad Box1
Gigabyte (Sandy-Bridge) | GA-Z68MA-D2H-B34
Corsair 16GB(4x4) DDR3 1333MHz.
Hard Disk: Western Digital HDD 1TB SATA3 7200RPM
And now I'm with a doubt about the graphic card. Here I'm gonna do smalls and mediuns plants with Inventor 2012, with in the maximun 15000 parts.
For the graphics card I have 3 options (that i need to know wich one is the best for this case, because i cant spend too much with these machines)
Nvidea GTX560 1GB DDR5 256bits - R$850,00
NVidea Quadro 600(fermi) 1GB DDR3 128bits - R$650,00
NVidea Quadro 2000(fermi) 1GB DDR5 128bits - R$1850,00
In my case, do I really need to buy this Quadro 2000 ? Or with the Quadro 600 I can work without any problem ? Or even GTX 560 ?
In the Passmark link, this GTX 560 is better then a Quadro 2000, is that true ? For works like these, using this PC only for Inventor assemblies, nothing else.
I really need to decide wich card to use, because I'll buy at least 2 machines, and after I buy I cant change these cards.
Thanks, and sorry about my english, I'm still learning.
I have been around the Autodesk/Inv traps a while..... I used to support Quadro all the way when Inv was OpenGL. Now its GTX ALL the way, period. IMO Directx and GTX are fine.... There are a couple of things (esp in idw's with lots of 2D lines) where Quadro is marginally better, but its not worth the cost.
I have moved to a new company that historically has been using Quadro2000 up till a couple of months ago, Of all the Quadro cards I have used in the past the 2000 is by far the worst. Just seems to have heaps of small "Issues" and not really a great performer.
Recent machines have all been correctly speced workstation hardware (Dell or HP) and we order WITHOUT a gfx card and purchase a GTX to go with it.
Safe to say NOBODY here wants to go back to Quadro2000 machines
The spec's in my signature are for my work machine which now features a GTX-590 card. My home machine runs a Quadro FX3800 card that was in my work machine.
My work machine kicks my home machine, which is a Dell Precission 690 with 16Gb with a SAS hard-drive system and the FX3800 card.
As the others have stated with the move to D3D - DX, this has eliminated the need for OpenGL (expensive graphics cards). If you have other software that requires OpenGL, then spend the money.
Thanks for the info, Blair. I really appreciate the specs all have provided on their hardware. It sure looks like I will be looking at an Nvidia GTX-590. Since nothing has been paid out at this point the options are still open.
I just ran acroos nVidia’s Tesla line. The price tag through nVidia is a real turn-off, but no comparison shopping as yet. The power requirements are also a bit much at over 200 Watts, meaning special case cooling requirements.
I’m currently running an Nvidia Quadro FX 4000 by PNY. It has had some problems from the beginning, as I always get a message before entering the desktop that the card will go to a reduced power level to protect it from damage. It has 2 power connectors, and both are attached to the power supply. I have a power supply tester and it does not indicate that the card is getting insufficient power. I’ve decided to ignore the problem since the card does seem to function OK, but I know the hardware has to be upgraded, meaning a whole new workstation. PNY tech support has been helpful and has agreed to replace the card, but that would mean waiting for the replacement and the bigger problem of disconnecting carefully placed and secured wiring, since the power supply manufacturer, PC Power & Cooling, has also agreed to replace the power supply (610 Watt Silencer).
So I will be paying close attention to the Nvidia GTX-590 power requirements. If it also has multiple direct connections to the power supply I’ll need to research if any are having similar issues, as stated above.
The GTX-590 is a dual GPU card. AFAIK this will not be an advantage with IV, correct me if I am wrong?
There are single GPU GTX-580 cards out there that also have extended VRAM to 3GB which would be a better choice I think. The standard GTX-580 is clocked faster than the 590 as well.