Announcement

Collapse
No announcement yet.

Free NVIDIA Fermi Cards To Open-Source Developers

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • IsawSparks
    replied
    Originally posted by crazycheese View Post
    3D acceleration works with Opensource driver for AMD for R300-R500 chips and emerges for R600-700 chips(downloadable and usable).
    VDPAU is port and remap to windows driver part. They didnt invent anything, they just remap it.
    As opposed to ATI who just don't do anything at all with regards to HD video decoding. AVIVO FTL.

    Leave a comment:


  • codestr0m
    replied
    Originally posted by gbeauche View Post
    IMHO, another gpgpu language is a non-sense. Come on, Pathscale is a compiler company, they are good at backends and middle ends, why bother with another language? They should check HMPP out, which can be a better solution (pragmas from plain C code). Though a dedicated code generator could be better than plain CUDA, CAL, OpenCL in certain cases as each implementation has their own set of problems (performance-wise, even on the same platform), curiously.
    Curious how I actually agree with you.. We have extensively looked at HMPP from Caps and yet another open standard does not mean yet another language. I wish I could clarify further, but suffice to say there will be some other major players backing the new standard.

    Leave a comment:


  • gbeauche
    replied
    Originally posted by tuke81 View Post
    On the topic: So PathScale is making new gpgpu standart which is open source and giving some fermi cards to nouveau driver developers. Any toughs of that, do we need a yet another gpgpu language and if yes. Why opencl is not enough? Or is this gpgpu language only suited for HPC industry.
    IMHO, another gpgpu language is a non-sense. Come on, Pathscale is a compiler company, they are good at backends and middle ends, why bother with another language? They should check HMPP out, which can be a better solution (pragmas from plain C code). Though a dedicated code generator could be better than plain CUDA, CAL, OpenCL in certain cases as each implementation has their own set of problems (performance-wise, even on the same platform), curiously.

    Leave a comment:


  • tuke81
    replied
    Hmm thread full of usual amount of bs...

    On the topic: So PathScale is making new gpgpu standart which is open source and giving some fermi cards to nouveau driver developers. Any toughs of that, do we need a yet another gpgpu language and if yes. Why opencl is not enough? Or is this gpgpu language only suited for HPC industry.

    source:

    Leave a comment:


  • smitty3268
    replied
    Originally posted by codestr0m View Post
    While I appreciate your enthusiasm for team red I prefer to support and buy good hardware. I encourage you to look at the lowest level details of the engineering between the two camps and once you're really informed come back and tell others what to do. Nvidia is pushing innovation, but what has team red done in the past couple years?
    Honestly, if you look closely i think you'll see just as much innovation coming from ATI as NVidia. Fermi does finally come with some new architectural features that are pretty cool, but the previous 2xx series had basically nothing new. And ask yourself who is always the first to have a card running on a new manufacturing process - ATI has a ton of more experience with getting hardware to actually run on smaller processes, while NVidia waits a year for it to mature and then takes advantage of some of the work ATI did.

    Leave a comment:


  • crazycheese
    replied
    Originally posted by D?niel Fraga View Post
    Besides that, nvidia offers VDPAU and very good interaction with Composite extension (it's been a long time I don't use any ATI driver, so I can't tell today how it is in this aspect).
    3D acceleration works with Opensource driver for AMD for R300-R500 chips and emerges for R600-700 chips(downloadable and usable).
    VDPAU is port and remap to windows driver part. They didnt invent anything, they just remap it.

    Leave a comment:


  • crazycheese
    replied
    I tell you my opinion, but I may need these questions answered by yourself!

    1) How much Watt does Fermi consume on Idle and on Perfomance?
    2) Does Fermi work in Perfomance mode if not switched my blob driver?
    3) Does Nvidia close source and even lastest opensource thing?

    How much time would it take to create barebone opensource driver that can handle power well? Let it be A
    How much time would it take to create moderately featured 3D driver? Let it be B

    This "offer" is merely a "welcome" to pay HUUUUUGE electricity bill by whatever freelancer takes on!

    Bill size= (HrzA*300 + HrsB*100)*price_per_KWh

    This is sick!

    Leave a comment:


  • Ex-Cyber
    replied
    Originally posted by MartjeB View Post
    Has Nvidia done *anything* apart from rebranding, the pas two years?
    Well, they did Fermi, which seems to genuinely be a novel and worthwhile architecture. On the other hand, its power/yield failings are for reasons relating precisely to "the lowest level details of the engineering" that codestr0m urges us to look at. ATI did low-level workarounds for TSMC process problems, while it looks like Nvidia was mostly leaning on TSMC to just fix the process.

    Leave a comment:


  • Daniel Fraga
    replied
    Originally posted by blackiwid View Post
    What I don?t get is why people are working so hard to make a opensource driver against Nvidias will. Let?s unite and lets buy all AMD cards instead and consentrate development of that driver with no support from Nvidia. If there is a very good free driver for amd hardware, Nvidia gets more under pressure with the last few flopping series of there cards they are under heavy presure at the moment, somedays they can not ignore 1-5% of the market like they did till now.
    Maybe those who bought a nvidia card, including me, just wanted it to work regardless if it requires a closed source drive or not.

    Besides that, nvidia offers VDPAU and very good interaction with Composite extension (it's been a long time I don't use any ATI driver, so I can't tell today how it is in this aspect). AMD/ATI drivers were historically know to be buggy, so maybe this makes people decides not to buy AMD/ATI cards...

    Anyway, you're right, since AMD is more friendly to open source than nvidia. But in practice, I think the situation will not change soon.

    Leave a comment:


  • blackiwid
    replied
    ok this time I mad some mistakes, which are not that small, I am cooking besides writing so sorry about that, the edit button is gone so I cannot correct this.
    But I think you all understand me, and thats the main purpuse of communication

    Leave a comment:

Working...
X