Hot Chips 2017: A Closer Look At Nvidia's Volta

[spikey in tn]

Clicking on your videos does nothing. Right-clicking produces "Powered by JW Player 7.11.1". What do I need to do to view your videos?

 

[kyotokid]

...oh what this would mean for GPU rendering. No need for multiple cards.

Apologies AMD, you've just been trumped again

 

[bit_user]

...oh what this would mean for GPU rendering. No need for multiple cards.

Apologies AMD, you've just been trumped again

ORLY? Not if you care about price/performance. Plus, AMD's cache-oriented architecture lets them render bigger scenes than will fit in GPU memory, which I recall being a particular sore spot of yours.

Did you see this?

https://pro.radeon.com/en-us/software/prorender/

 

[njwhite]

The article assumes a great deal of previous knowledge on the part of the reader. It would be nice if each acronym was accompanied by its full name on its first occurrence in the article. Then the previously unaware could benefit from reading as well.

 

[kyotokid]

...oh what this would mean for GPU rendering. No need for multiple cards.

Apologies AMD, you've just been trumped again

ORLY? Not if you care about price/performance. Plus, AMD's cache-oriented architecture lets them render bigger scenes than will fit in GPU memory, which I recall being a particular sore spot of yours.

Did you see this?

https://pro.radeon.com/en-us/software/prorender/

...moot point as the render engines I work with are CUDA based, and I do not work with pro grade software because I cannot afford it (Blender is another story entirely that I won't go into here).

ProRender does not have a plugin for the software I use and personally, I am not about to code one. Also after my experiences with Lux and it's buggy plugins, kind of soured on open source render engines.

Meanwhile, the new AMD 16 GB Vega Frontier runs extremely hot at peak output to the point some sort of exotic liquid cooling solution is not just necessary but almost required to keep from cooking the rest of the system (this was pointed out in a review here the other week).

 

[bit_user]

The article assumes a great deal of previous knowledge on the part of the reader. It would be nice if each acronym was accompanied by its full name on its first occurrence in the article. Then the previously unaware could benefit from reading as well.

The thing to know about this is it's reporting on a talk given at a quasi-academic conference. The audience of that talk is other industry insiders, academics, etc.

So, in order to make that information accessible to someone with no background in the subject, the article could literally turn into an introductory text book. The line has to be drawn somewhere, and I think it's reasonable to assume a slightly less knowledgeable readership than the audience at which original talk was targeted.

I think you do have a point, though. Certain things, like MIO, are not industry standard acronyms or terms. They must've been defined in the talk, but the slides only reference them. It would be helpful if the article did explain those.

That said, most of these terms are pretty easy to find via google and wikipedia. If you're interested in the subject, it would be a good exercise to go through the article and look up each thing you don't know. Do that, a few times, and you'll learn a lot.

 

[bit_user]

...oh what this would mean for GPU rendering. No need for multiple cards.

Apologies AMD, you've just been trumped again

ORLY? Not if you care about price/performance. Plus, AMD's cache-oriented architecture lets them render bigger scenes than will fit in GPU memory, which I recall being a particular sore spot of yours.

Did you see this?

https://pro.radeon.com/en-us/software/prorender/

...moot point as the render engines I work with are CUDA based, and I do not work with pro grade software because I cannot afford it (Blender is another story entirely that I won't go into here).

You can't afford pro-grade software, but you can afford a $10k graphics card? How much does this software cost?

after my experiences with Lux and it's buggy plugins, kind of soured on open source render engines.

ProRender was developed by AMD and then released as open source, as a way of letting users customize to their needs & gathering fixes & enhancements from the community. In contrast, Lux started out as a community effort.

I'm sure various user groups and forums can provide insight into how solid it is, should you wish to look before you leap (assuming they eventually add support for one of the programs you use).