I just found out something that is very troubling.

Apparently, nVidia is CRIPPLING the OpenGL of cards > GeForce200.

So that a OpenGL calls on a GTX480 will be ~4 times slower than on a GTX285!!!!

From the reading I'm doing, it appears that nVidia is also cripping CUDA on these cards.

The Quadro has fast OpenGL but crippled CUDA - and the Tesla has full CUDA performance.

I'm including a few links, but a search for '580 opengl crippled' pulls up a lot of them

http://blenderartists.org/forum/showthr ... ost1987116

http://blenderartists.org/forum/showthr ... ost1848257

As my GTX-580 3GB arrives tomorrow, I find this VERY disturbing!!!!

Is there any official info about this? Just curious.

This is only my opinion but I find that plan a bit unlikely. To cripple existing Nvidia cards, they would have to do it through voluntary driver updates, which would only infuriate existing Nvidia customers, both gaming and commercial. Existing customers would have no incentive to update their Nvidia drivers and if they did, they can always regress to a version that didn't cripple OpenGL. There would also be no incentive to purchase another Nvidia product in the future for fear of getting burned again. It doesn't make sense that Nvidia would actively try to alienate the majority of their userbase.

Again this is just my opinion but I wouldn't worry about it much. Enjoy your new GTX 580--I'm envious! :)

G.

_________________
Greenlaw
Artist/Partner - Little Green Dog | My Demo Reels (2013,) (2015,) (2017,) and (2019)

Watch a one minute excerpt on Vimeo now!

http://www.opengl.org/discussion_boards ... 014&page=5
http://forums.nvidia.com/index.php?showtopic=166757

They did this after the 285, so unless you compared opengl and cuda to that card, you wouldn't notice. They may even have done it in hardware, someone mentioned using a laser on the chips to knock something out in one thread.

But, the benchmarks don't lie. Just google for 'gtx opengl crippled'.

There's an active thread on this at newtek:

http://forums.newtek.com/showpost.php?p ... stcount=27

Don't be envious, I may have to pay 15% to return it - and then hassle with AMEX to get that back. But, I'll be out of shipping at any rate.

All I can say is that I upgraded from a GTX 260 to a GTX 460, and the 460 totally smokes the 260. Most definitely with iPi DMC tracking.

G.

_________________
Greenlaw
Artist/Partner - Little Green Dog | My Demo Reels (2013,) (2015,) (2017,) and (2019)

Watch a one minute excerpt on Vimeo now!

I ?think? that iPi uses CUDA acceleration?

Maybe even throttled, the CUDA is faster as there would be a LOT more CUDA cores? The 580 has 512 cores, the 460 has 288 cores, the 285 has 240 cores, the 260 has 192 cores.

For 3d-coat, I can run in DX rather than opengl, and someone posted that 3ds and other AD tools will run DX (I wonder if you can specify DX or opengl?).

But, I'm wondering about Lightwave, Fusion, etc.

I'm very glad to hear the 460 is running faster than the 260, as this gives me hope at least for CUDA. I'd hate to have to return it, but ... I'm a bit unhappy with nVidia at the moment.

** EDIT *** this benchmark seems to indicate CUDA is getting faster in newer GTX cards!

http://kernelnine.com/?p=218

BTW - if OpenGL calls like “glReadPixels()” on a GTX480 are ~4 times slower than on a GTX285, how does this actually affect 3D users day to day?

If we’re using opengl to to display an estimation of the render for tweaking, I would guess 3d apps would be using the glWritePixels - much like a video game would to display the game.

Is this only going to affect users to want to read what opengl has ‘rendered’ - or is this going to affect normal usage of 3D apps? When would you need to read those buffers, rather than display them?

Unless you’re using ?? is it vRay ??, I thought all actual rendering was done in software, at least for final output, but ... I could be confused?

I don't think iPi DMC takes advantage of CUDA but I haven't heard officially.

FWIW, here's what I recall in my three years of using iPi DMC with at least three different graphics cards:

I started out using a Geforce 9600GS and tracking then took about 10 seconds per frame. At the time I didn't know if this was good or bad because I had nothing to reference it against.

When I upgraded to the GEforce GTX 260, my time improved significantly to about 2 to 3 seconds per frame. Part of this can be explained by a major software update in iPi Studio at that time, when it switched from CPU to GPU processing; but even before the update, I recall seeing an improvement in speed. Around that same time I tested the software on my computer at work, which at the time had a very modest GT240. Tracking times were a bit slower than GTX 260 but not too horrible and definitely better than the 9600GS.

A few months later, I upgraded my personal computer at home to a GTX 460 and my tracking speed improved to 0.67 second per frame! Definitely a big improvement, and this time I couldn't blame it all on the iPi devs. :P

At the time I wanted to get a GTX 480 but I'm actually at my limit with this computer. The case is on the smallish side, the power supply is modest (350w I think,) and I don't want to buy a new computer yet, so the lower-wattage version of the GTX 460 pretty much maxed my computer specs. That said, I can't complain about having a GTX 460 in this box--for what I paid for this card (about $170, and I think you can get it a lot cheaper now,) it's performed way beyond my expectations.

That said, I don't know anything about Nvidia's official plans for 'crippling' Geforce. It's just my opinion that doing so in a significant manner is not going to increase sales of Quadro cards, and it doesn't make a lot of sense to do so--the two cards are aimed at different markets, consumer users and industrial CAD users respectively. The majority of Geforce users are video game enthusiasts and the majority of Quadro users are industrial designers and engineers; Visual FX artists fall somewhere in-between and I don't think they're considered a significant market by comparison.

I guess we'll see though.

G.

_________________
Greenlaw
Artist/Partner - Little Green Dog | My Demo Reels (2013,) (2015,) (2017,) and (2019)

Watch a one minute excerpt on Vimeo now!

Greenlaw - Yes, I think you are right and ipi uses GPU rather than nVidia specific CUDA. Turbulence, Premiere, etc. do use CUDA.

I did find some Lightwave benchmarks with the GTX-480 and quadro 5000.

http://forums.newtek.com/showpost.php?p ... stcount=42

It looks like the 480 is holding it's own.

nVidia did slow down the glReadPixels and some other functions

http://www.opengl.org/discussion_boards ... Post291789

from the GTX-285. But, I cannot evaluate how that will effect real world performance. It's a shame that they have done that, as the bad blood that generates will spur people (like the fellow who originally alerted me to this on the Lightwave forum) to totally bash nVidia - which will cost them a whole lot of money they can't afford to misplace.

I've always gone with nVidia, so this whole thing has troubled me and right or wrong, made me a bit suspect of nVidia. Slowing your customers down (to try to make more money from them) is not a good long term strategy for profit.

I'd like to come up with some real-world benchmarks that I could test while I have the 285 in my box for Lightwave, Turbulence, etc. I've been waiting on the head motions (saw your great post where this may not be as big of a deal, but I want to mocap ballet and dance) - so I've gotten out of practice with using ipi. But, I should be able to come up with a 'test' for that.

Onwards thru the fog!

I've known about the openGL issue with nvidia for quite some time. All they are doing is limiting what the card can do in the driver so that they can sell a professional card (aka Quadro) which is the same exact board as a GT or GTX for WAYYYY more money but with support and un-throttled openGL. It has kind of become a non issue as most to all programs such as 3DS Max now runs better on DirectX.

_________________