low end quadro is 5 times faster than high end gaming card

[deadrats]

for years i have heard various people claim that quadro are the exact same card as their geforce gaming cousins, with the only difference being the drivers. a similar claim is made with regards to intel's consumer cpu's and xeons as well as amd's consumer grade cpu's and opterons.

on the cpu side it's easy to disprove these claims, especially with current generation cpu's; the workstation versions have larger caches (L2 and L3) as well as higher core counts (opteron is currently up to 12 cores, soon to be 16, xeon is up to 10 cores with 20 virtual cores thanks to HT).

the reality is that the way the caches are used is also changed via different microcode and the workstation variants also have different schedulers, and if you read through the developer's documentation you will note that there are other subtle differences that the layman is incapable of appreciating, all designed at extracting maximum performance under heavy workloads, such as with heavily threaded applications or numerous single threaded tasks, way more than any desktop cpu will ever face.

i have often theorized that the same holds true with workstation cards verses their gaming counterparts. over the years i have "soft modded" a few gaming cards into "workstation cards", using custom bios so that they would be recognized as the more expensive workstation cards and thus would allow me to install and use the workstation drivers with said cards. the reality is that despite the newly modded cards being seen as workstation cards both during boot up and by the drivers and testing applications, time after time these modded cards failed to perform as a true workstation card, clearly their is more to it than just optimized drivers, there has to be significant hardware difference as well.

today i believe i have conclusive proof that workstation class cards are in fact much more than re-branded gaming cards:

http://www.xbitlabs.com/news/video/display/20110406235438_Nvidia_Quadro_400_for_169_Ou...ive_Times.html

without quoting the entire article, the quadro 400 is five times faster than a gtx 580 in a industry standard benchmark. when you consider that the quadro 400 is a dx10.1 part, powered by a gt216gl chip, with 64 bit path to 512 mb ddr3 and only 48 cuda cores compared to the gtx 580's dx11 chip with a 384 bit path to 1.5 gb of ddr5 and 512 cuda cores, driver optimizations don't come close to explaining how a chip with less than 1/10 the cores, a third the ram and 1/6 the ram interface width can be 5 times faster, clearly there are some significant hardware difference nvidia doesn't tell it's customers about.

i wish i could see some cuda powered encoding benchmarks and image quality comparisons between a quadro and a gaming card.

[jagabo]

That comes right out of NVIDIA's press release:

http://www.nvidia.com/object/product-quadro-400-us.html

No doubt it's comparing the Quadro 400 with GPU based code to the GTX 580 with non GPU code. Because their Quadro drivers don't work on non-Quadro cards. NVIDIA would never compare the two on equal footing.

[johns0]

Title should be 'low end quadro is 5 times faster than high end gaming card in industry leading CAD/CAM applications'.

[Munster1]

Nvidia wouldn't dare fubar the gaming cards to give poor performance in professional applications to protect their market segmentation, would they?

[jagabo]

The gaming cards don't come with the application specific drivers (and the Qaudro drivers don't work with non-quadro card unless you hack them). That's what you're paying for when you buy a Quadro. Some Quadro cards do have more memory.