It's good to see a controlled test between several Quadro cards.
It would have been nice to compare with a few GTX. I know the article stated the issues about these cards... but still, 9 out of 10 are choosing GTX because of price AND speed. In a properly cooled system, it's a very cost effective and speedy solution. And I don't hear anyone complaining about a properly cooled GTX cards shutting down because of overheating.
Also, looking at PPBM5, Quadro's aren't the top performers that marketing would lead us to believe (for MPE). If there is a slight gain to be had from them, the price differential turns off most people. In my case, my budget is $12000 for a complete system (receiving it next week... finally). I initially choose the Quadro 5000, but opted for a GTX580 (in a well cooled chassis). It made more business sense. It's also less discouraging to upgrade a GTX then a Quadro on a "yearly" basis (to keep up with the technology for added MPE benefits). Besides, what is MPE? P stands for Playback, and most editors don't playback 24/7. We insert clips, cut them, place effects, playback, tweak the edit points, playback, change filter settings, playback, go get a coffee, playback, edit some more, go on AE or PS, go back to PrPro, playback... any way, you get the idea? MPE is not working 100% of the time. Unlike gaming or 3D modeling and rendering. Granted, multiple, non-stop playback of an effects heavy 2-hour unrendered full feature will probably strain a GTX's MPE capability... the fact of the matter is, most don't use it that way.
Frederic, I hear you and know that most people use GeForce. For us, we spend our time certifying Quadros in large part for the reasons I outlined.
As far as GeForce in a well-cooled machine, you might be trading (in part) where you're spending the money. If you spend a ton on fans or a liquid cooling system, you might end up closer to a Quadro.
In the end, I think it is a matter of choice for the user and the type of work they want to do. If I get a GeForce card I'll try to put it through the tests.
That would be great to see a GeForce in your testing (such as the GTX580). It would certainly answer a lot of questions people have. Not just about price, but performance (speed and quality). Still today, I see way to many questions on forums (not just Adobe). I strongly believe this issue should be put to rest once and for all. The problem with PPBM5, although it does give a good indication, is we can't compare oranges with oranges. The computers are too different, and there are too many variables that alters the final results. Using the same computer, and simply swapping out the cards, is the way to go. The user community is in dire need of having a thorough comparison done.
As for cooling, yes a person could install custom liquid cooling system. But I was simply referring to GPU cards that have their own efficient internal solutions that are more adequate then most; such as the PNY GTX580 liquid cooled which is 30% cooler then the reference board, or the Asus GTX580 Matrix which is 20% cooler then the reference board. As well, a large, well ventilated casing, where airflow is not restricted.
In my case (there was a few changes to my initial configuration), I got a Xygmatek Elysium tower, with extra fans wherever they could fit. And an Asus Matrix GTX580 GPU. I'll be getting it early next week. To date, the stress tests made by my systems integrator are very positive. No Liquid cooling, just a well planned/designed airflow system. I do agree with you that the added costs of extra fans, larger casing, and high end GTX does makes the price to be somewhat close to a Quadro 4000. On the other hand, it's nowhere near the 5000. But considering the cost of the casing and extra fans where mostly due for the overclocked CPU and 12-disk internal RAID.... in my case, it then becomes a card-for-card price battle. Like I said, it made more business sense. And down the short, Moore's Law driven road, I'll get a next gen GTX, and trickle down the 580 to my son's gaming machine.... he'll certainly be very happy about it. And I'll be happy for the extra boost in MPE. Whereas, I'd be locked for at least 3 years on a $2K Quadro 5000.
But to be fair, if and when you do test the GTX, and it turns out that one of the Quadro's does blow out the GTX out of the sky... (although I have my doubts)... then my next card will probably be a Quadro. Just don't count me in for an overpriced Quadro 6000 though... hehehe
Thanks for your helpful article.
Regarding the issue you mentioned with the Quadro 2000, did that look like a software or hardware error? I ask because I am considering this card and would be wary if there were, for example, a bad batch of cards.
edit: Also, I've noticed that only the Quadro 5000 (along with the 6000) has ECC memory. Do you think this made a difference to the test results?
I have not been able to track it down, but it was ONLY with RED 4K footage. So, as long as you're not using that, it shouldn't be a problem. My hunch is that RED 4K was just too much for the Quadro 2000's limited memory.
OK, thanks Dennis.