NVidia to support Mango project & developers with high-end cards, including Tesla

NVidia mailed as Siggraph follow-up, to support Mango project & developers with high-end cards, including Tesla. With AMD/ATI an appointment has been made at IBC conference Amsterdam (early September). Cooperation with Intel OpenGL team will be set up too.

it’s gonna be insane :smiley: I hope the Tesla card goes to brecht and Cycles.

I think this will accelarate the cycles evelolution x-factor.
No need for boost library anymore :wink:

i just hope they dont make it nvidia or cuda dependent. all 3 game consoles have went amd for next gen. i dont think the future for nvidia is going to be very good. the money amd will make off 200 million consoles over the next 10 years is going to give them a huge r&d advantage so cuda will become a thing of the past. better to go with opencl .

Still wouldn’t give them my money after the absolutely horrendous fermi performance.

OpenCL is needed regardless of the card, since both AMD and Nvidia support it. I hope AMD gives Mango some support aswell.

Nvidia my ass.

I hope the BF gets a lot of expensive quadro and tesla cards, then sell them immediately and fund the development of OpenCL implementation with it.
It´s obvious what Nvidia is doing there. Give a Tesla card out for free, taged with a 4 digit price, worth a regular crippled GF100 GPU for 300 Euro, just uncrippled. That done, ensure Blender jumps on the CUDA train, generating a new market for their cards for pocketmoney.

But I am stating the obvious and I am certain no one at BF falls for such a cheap trick…

P.S.: I am nvidia user for years, fan of the product, hate the copany. But also have amd cards so I don´t sound the drums for either of them, IMO they are both just bad to the core :wink:

With AMD/ATI an appointment has been made at IBC conference Amsterdam (early September). Cooperation with Intel OpenGL team will be set up too.

guess the meeting with them is make sure it’s opencl, infact brecht has already started on getting the opencl parts to work (afaik there’ always been in his code but disabled in the early alpha releases of cycles)

besides opencl also enables CPU+GPU not only one or the other.

You guys have too much of a lone ranger anti-corporate attitude that is at times a little dissociated from the truth on the ground. Anyways Thumb up to Nvidia for supporting Mango. I can’t wait to see what aid AMD and Intel offer.

Kudos to Nvidia for offering some of their equipment.

People can rant and rave all they want about their recent antics, and personally I think they can be justified, Fermi performance doesn’t seem so good with OpenGL/ Blender.

But at the end of the day now that Nvidia will be providing equipment and support, plus I’d imagine talk will go back and forth between Ton/ the developers, and Nvidia so for driver releases we may see some good improvements in speed etcetera.

If AMD, and Intel join in to with equipment (Intel cards would just be testing I’d imagine) then we’ll see Blenders OpenGL speed really become a lot more solid, and faster!

Not to mention all of the improvements that OpenCL will bring to things like Cycles, the compositor, and hopefully in the not too distant future things like OpenCL simulations in the viewport, and things of that nature.

This is great news. Hope Blender can support OpenCL soon. We’ll need lot of particle FX with the help of OpenCL technology for Project Mango.

arexma has a point though. Blender users are probably most likely to have Geforce rather than Quadro or Tesla. As nice as those cards are if you can afford them we want Cycles and other GPU stuff to continue to work effectively on Geforce. I think Octane already had a set back due to Nvidia putting a strategic spoke in their works.
Given the deliberately obstructive behaviour of Nvidia in crippling their cards and then being able to charge a bunch for a properly working one I think Blender should be wary of corporate gifts. We dont want to assist them in any way to make Cuda features in Blender inaccessible or effectively useless to anyone without a Quadro or Tesla. They can feel free to donate some 590 or something although the last Geforce series have been especially anemic for things like Blender use even though they rip along on cuda tasks.
Besides which it is already hard to have a Geforce and Quadro driver together so down the road a little you might end up having to buy 2 Quadro instead - one for modelling etc and another on hand especially for cuda tasks.:mad:
The typical user gets pushed out of Blender then unless they have $$$ or model <100 tris.:yes:

It´s not about lone ranger, dissaociated or rant and rave guys, you don´t seem to see the danger.

With this generation of cards they showed that they simply cripple cards to sell expensive professional cards.
Have you thought about the next generation of GeForce cards?

It´s no nice gesture of nivida, they don´t care, that´s peanuts compared to their PR and marketing budget.
They try to establish a monopoly in the CG sector, which the almost have by now counting the few rare occurances of FireGL cards. Once they established it you either buy their hardware or don´t use the software.
What if the next generation of GeForce is crippled to 1/10 of the OpenGL performance of a Quadro and 1/100 of the CUDA performance of a Tesla and the Quadro is crippled 1/100 in CUDA performance?
They have shown that they have no scruple to do so with Fermi already and all outcrys are waved aside… conspiracy theory, rave, rant…

Is the average Blender user going to buy a Quadro and a Tesla card to have sculpting viewport performance and render with cycles?
Or you just buy the Quadro, then you can at least sculpt but not render with CUDA.
Or just the Tesla? Then you can render fast, but only the lowpoly stuff your GeForce card is able to handle in OpenGL.

Nvidia is smiling in your face, while stabbing your guts.
And they don´t care nor will they. They sell plenty of GeForce to gamers, more than enough Quadro to CG studios and Tesla galore for scientific and CG applications and currently all three have the same GPU.

The only upside I see is that the BF gets free hardware and other manufactures might jump in now with hardware, libraries, papers or know how.
I most certainly don´t want to see Blender polluted with proprietary stuff.

I don’t think this has a lot to do with anti-corporate attitude. But off course there is a motive behind it is as I don’t think they give these things because of ton’s beautiful eyes. Off course the wish is to push CUDA as these things really rocks with it.

There is nothing wrong with that although as an non nvidia user (don’t have the money for a pro card that is needed to run Blender efficiently these days) I hope the OpenCL part doesn’t get hold back because of this.

OpenCL is more a danger for Nvidia as there is competition, while in the CUDA market they do have a monopoly. And software is the most important part, so pushing for CUDA only solutions is in their best interest.

But at the end of the day now that Nvidia will be providing equipment and support, plus I’d imagine talk will go back and forth between Ton/ the developers, and Nvidia so for driver releases we may see some good improvements in speed etcetera.

  1. they are aware of the problem for some months now
  2. the problem is easy to reproduce and has been documented to hell
  3. the “bug” isn’t a bug because Nvidia deliberately cripples the drivers.
  4. #WONTFIX - upgrade to pro versions

The reason they crippled their drivers is to sell their overpriced pro cards as a lot of studios just bought gamer cards instead. And for all that I have heard their only reaction has been to upgrade to the pro versions if you want decent OpenGL performance.

They really don’t care for use going for ATI instead when they can sell 1 card for the price of 5-10 gamercards (have you seen the prices of those things ?) What I have paid for my 2GB GDR5 RAM 6950 I can have a quadro version with a whooping 256MB GDR3 RAM and with performance that would make you cry :slight_smile:

I think another reason was not to pull down the pants of smaller Quadro customers and loose them to FireGL or Radeon. Imagine you just bought 20 high end quadros for your studio, which you newly opened 3 months before the fermi launch because the launch was postponed almost half a year and you couldn´t wait.
Suddenly a gamer card is on the market that outperforms all your cards. :slight_smile:
And I think the GF100 OpenGL performance was crippled too much because they had to go with 100% performance for the Quadro cards. But you need a huge benefit over the last Quadro to sell it, so you have to make sure the new GeForce is slower than the old Quadro in OpenGL and at the same time make sure the new Quadro is dunnow, twice as fast as the old one. We all know the result.

And personally I think professional users are pissed as well. The Quadros are crippled in their CUDA performance as well, and the Tesla cards have no signal out, they are pure computing slaves.
I don´t know though if it has any technical reasons why part of the CUDA cores are disabled on Quadros/GeForce, but I highly doubt it.

Great news! It’s nice to see such a big player in our industry supporting blender. Now if we could just stop the rebel attitude, we might make something of ourselves after all.

I don’t know maybe I am been naive on this but I think they are not going to impose any conditions on how Cycles or anything gets developed. I think this is been done to perhaps gain some marketing material and gain some good will among the blender community on their part. If Mango is rendered on their cards that is a big marketing coupe for them. If they can lend their help making Blender run better on theirs cards than even better.

I also don’t thinkt hey will impose restrictions but Tesla cards tend to work better with CUDA then OpenCL I have read. If I’m not mistaken Nvidia claim was that CUDA could better utilize Geforce GPU’s then OpenCL.

When you are one time constraints like in project mango I think the possibility arise that the focus will lay on CUDA if there is a lot to gain… and who would blame them ?

I could be wrong but maybe that is the situation Nvidia is hoping on.

Maybe if AMD/ATI (the OpenCL route) would also donate some hardware the situation will be a whole other story.

yeah. the last one i saw was anti-paypal. i don’t get it. as if them hating those companies would make any difference.
it’s nice that Blender is getting interest and support from the graphics hardware industry.

That’s because CUDA is developing at a faster rate than OpenCL, not because of the design of the cards. In almost every way Nvidia is winning the battle on the gpu front. The very nature of an open source project like OpenCL slows it down. That’s why we have so much to be thankful for with Blender, Ton’s leadership and the small core team keep it nimble.