The Vega FE is already in stores. (not the cheaper RX)

AMD has been advertising their High Bandwith Cache Controller on their new Vega cards. It is supposed to allow for VRAM objects to be offloaded to slower system memory or storage so that the device doesn't hit a memory limit.

Will this be integrated into Blender? It would be nice to not run out of VRAM. I understand that a "pagefile" style memory dump would be slower than VRAM, but wouldn't it still be faster than using a CPU? If in theory the card can handle memory requests wouldn't blender only need to not stop at the physical VRAM limit?

This is a question diving into how the main release of blender handles memory objects. Also, the FE Professional cards with HBCC have been for sale for weeks. (They are just really expensive) No one has probably looked into this yet. I can see that https://developer.blender.org/diffusion/B/ has the source code. I just did a git clone of the master project. Opened the GPU_framebuffer header. It looks like it uses the OpenGL framebuffer framework. I don't think it would take long to adapt to the HBCC. Anyone look at this yet?

enter image description here


closed as off-topic by Duarte Farrajota Ramos, cegaton, David Aug 2 '17 at 0:07

  • This question does not appear to be about Blender within the scope defined in the help center.
If this question can be reworded to fit the rules in the help center, please edit the question.

  • 4
    $\begingroup$ How can we know? The cards aren't even out yet. $\endgroup$ – Duarte Farrajota Ramos Aug 1 '17 at 19:28
  • 1
    $\begingroup$ I'm voting to close this question as off-topic because because it is about hardware, not Blender $\endgroup$ – Duarte Farrajota Ramos Aug 1 '17 at 19:29
  • 3
    $\begingroup$ Feature requests could go to rightclickselect.com I doubt anyone here has already acquired one and even so I doubt any devs have bothered to support this in such short notice. If this is a question about Cycles Memory management, that may indeed be interesting, so I'd suggest rewording exclusively as so. $\endgroup$ – Duarte Farrajota Ramos Aug 1 '17 at 19:47
  • 1
    $\begingroup$ @johnsonjp34 this site is not related to the blender foundation or to the developers of blender (maybe some read from time to time), but is run largely by users of the program. A feature request made here might never get to the right person. Please use the following link to find out where to direct your question: blender.stackexchange.com/questions/1190/… $\endgroup$ – cegaton Aug 1 '17 at 23:36
  • 1
    $\begingroup$ I'm voting to close this question as off-topic because it is a feature request to support specific hardware. $\endgroup$ – cegaton Aug 1 '17 at 23:40

A lot of implementation is needed for this function to be included in Blender. And taking advantage of the VRAM on your GPU is no simple task, as it's mainly dedicated to storing temporary information from your GPU's CPU.

Also, this would cause an issue with users with HDDs, as they are slower to write and read information than a SSD is. There are even some drives out there that do not send out the information detailing if it's a HDD or SSD, so there is no foolproof way of doing this.

This feature would be nice to be added indeed, but I don't think we'll see this implemented until the HBCC technology has at least become partially industry standard.

But to answer your question; Yes, it will most likely be added at some point in time. But if it's in 1 year, 2 years, 5 years or 10 years, I don't know, and that's hard to answer.

NOTE: I run my own computer company, so I know a lot about these technologies, but I'm not a Blender developer. I can give you my opinions and answers, but I can't speak for the Blender team.

Hope this helps.

  • 1
    $\begingroup$ Thanks. That's kind of what I was thinking. I was hoping that it could be tied to only system RAM to prevent the HDD/SSD issue. Indeed, not a simple task, but I'm happy it is probably on the radar. $\endgroup$ – johnsonjp34 Aug 1 '17 at 18:55

Not the answer you're looking for? Browse other questions tagged or ask your own question.