Nvidia Pascal Titan X will not feature faster FP64 or FP16 performance

No increase in memory capacity means I can't see a lot of people using Octane, Vray RT etc upgrading, they will just try and just grab second hand Maxwell Titan X's. Bit of a missed opportunity all round for me.
 
I don't think this paints an accurate picture of all TITAN X use cases. Plenty of users access TITAN X compute without the double precision. To say it's nothing more than a gaming GPU is disingenuous at worst.
 
30% performance increase over a 1080 for twice the amount a 1080 costs, No FP performance over a 1080 and not even the fully enabled GP102 chip, Yep, That's a pass from me.
 
30% performance increase over a 1080 for twice the amount a 1080 costs, No FP performance over a 1080 and not even the fully enabled GP102 chip, Yep, That's a pass from me.

Don't expect a full GP102 chip. That's reserved for HPC and other higher profitable markets. Hence the newly announced Quadro P6000 which is the full GP102 core.
 
Don't expect a full GP102 chip. That's reserved for HPC and other higher profitable markets. Hence the newly announced Quadro P6000 which is the full GP102 core.

At some point they will release the full GP102 core, Probably call it the Titan X Black Edition and charge an extra 200 for it over the standard TX.
 
At some point they will release the full GP102 core, Probably call it the Titan X Black Edition and charge an extra 200 for it over the standard TX.

Really doubt they will. They would be wasting units. It won't be any noticeably faster and people won't buy it, may as well charge more for it where you know it will be bought for much higher prices.
 
Really doubt they will. They would be wasting units. It won't be any noticeably faster and people won't buy it, may as well charge more for it where you know it will be bought for much higher prices.

Either way I'm not making the mistake of getting a Titan again, It's either the full fat xx80 chips or the Ti variants for me here on out, Unless AMD hit one out of the park with Vega then I'm going to be all over that like white on rice ^_^
 
Either way I'm not making the mistake of getting a Titan again, It's either the full fat xx80 chips or the Ti variants for me here on out, Unless AMD hit one out of the park with Vega then I'm going to be all over that like white on rice ^_^

I'm just waiting for custom 480s.. then i'll ride the wave until Vega 11 comes out if it's worth getting or Vega 10 if it ends up having more perf/$. Never again will I spend Fury X money. Just get burned to fast in the tech world. Things move to quickly.
 
I don't think this paints an accurate picture of all TITAN X use cases. Plenty of users access TITAN X compute without the double precision. To say it's nothing more than a gaming GPU is disingenuous at worst.

Even at it's price point? For compute workloads it's fairly easy to access multiple GPUs in parallel and your algorithms need to be optimized for parallel execution in the first place to even run properly on GPUs. So the only point in using a Titan X over a pair of 1080s or even 1070s is probably the need for its 12GB memory. And even that is only true if one dataset is larger than those 8GB the 10x0 cards have. If data can also be partitioned, than a pair of 10x0 cards do have 4GB more memory.
 
Back
Top