Pricing Discussions around AMD VEGA *over-flow*

Raja breaks silence on pricing and availability:

HKSCbkz.png


http://wccftech.com/amds-raja-kodur...ing-perf-perfwatt-untapped-hardware-features/
 
Thanks, very interesting, yet sometimes strange tweets.

This for example feels a bit odd, given the circumstances and the timeframe for the launch.
Raja Koduri‏ @GFXChipTweeter:
Seems like a lot of questions around power of Vega. The initial round of reviews didn't showcase the perf/watt dynamic range of Vega well.“

Or this:
Raja Koduri‏ @GFXChipTweeter :
To really understand an architecture - you need to understand the architecture in the content of overall roadmap, tradeoff and constraints.“
Then, by all means, tell us the facts.
 
That tweet is followed up by him stating Vega 10 competes with GP100, GP102 and GP104. Those are the facts.
That's an exhaustive explanation for a 12.5 bln GPU with a host of new features. Thanks for the enlightenment, Raja.

Seriously, you cannot be serious.
 
I think the idea was "we have only one chip, nvidia multiple ones, so you can't expect us to be competitive perfs/watt wise with the nvidia "lower end"" ?

But still, even the biggest nvidia gpu is better perf/watt, so yeah it doesn't make sense...
 
Taken at face value, it reads as if AMD has missed the crossroads where more specialized chips not only start to make sense but become a competitive necessity.

IMHO, AMD shoehorned Vega to compete on GP104 on price, which is nice for customers and hopefully will not hurt AMD too much. It tries to compete with GP100 in terms of FP16-DNN training performance and definitely competes on pricing there. But I fail to see where Vega is competing with GP102? I mean, die size and performance target could be, but that's not serious.
 
I think the idea was "we have only one chip, nvidia multiple ones, so you can't expect us to be competitive perfs/watt wise with the nvidia "lower end"" ?

But still, even the biggest nvidia gpu is better perf/watt, so yeah it doesn't make sense...
Raja clarified those comments are in regards to perf/mm2, not power. Idle hardware doesn't draw that much and perf/watt is highly variable. Forcing the power state down a notch and the result is similar performance with far less power.

But I fail to see where Vega is competing with GP102? I mean, die size and performance target could be, but that's not serious.
Pro market would be valid. FE/WX/SSG targeting various markets. Performance will likely come in time.
 
That's an exhaustive explanation for a 12.5 bln GPU with a host of new features. Thanks for the enlightenment, Raja.

Seriously, you cannot be serious.

I was simply explaining the context of the price/mm2 comment, which you didn't include in your question.
 
Taken at face value, it reads as if AMD has missed the crossroads where more specialized chips not only start to make sense but become a competitive necessity.

I think it's not that they missed the crossroad, it's that the turn is onto a difficult road that requires some off-road capability and AMD only have access to a Prius.
 
The noise around pricing doesn't help us at all. Not sure where that is coming from...
Really? He has no idea where it's coming from? He released Vega, went on vacation, comes back then decided to tweet some ramblings without checking in with his teams on everything that has been going on since he left? uh huh.
 
Back
Top