Question about cores.

[zachmullin]

I am just honestly wondering why, for example, the rumoured GTX 1050 is being compared performance wise with the RX 470 when the 1050 has 768 cores and the 2048 on the 470. 

Thanks!

[Gravemind]

Because people are dumb.

[Gonio]

Whats inside is not directly that important in comparing I would say if its the same performance.

Best is to see at an certain pricepoint what you can get, and then see what gives the best result. Or a different approach is setting a certain minimal performance it must be able to get and see whats available. Or my personal 2nd opinion besides a certain pricerange is always efficiency combined of results, heat and sound lvl's.

[Mira Yurizaki]

22 minutes ago, zachmullin said:

I am just honestly wondering why, for example, the rumoured GTX 1050 is being compared performance wise with the RX 470 when the 1050 has 768 cores and the 2048 on the 470. 

Thanks!

It all depends on how the implementation in hardware is done. For starters, the RX 470 runs at 926MHz base clock and up to 1.2GHz boost clock. Since I don't have the 1050's rumored specs on hand, I can tell you though that the 1060 is clocked much higher, starting at 1.6Ghz base and up to 1.7GHz boost. So even though it has less shader units, it can be clocked faster to compensate.

 

Otherwise as far as I know about the architectural differences between GCN and Pascal, there isn't really a clear reason as to what makes one perform in comparison to another. And in fact, they're both really different beasts until you get to the execution units themselves.

[WoodenMarker]

The cores are not the same nor equal and the cards cannot be compared by core count alone. Core count is only relevant if they use the same cores. 

Look at benchmarks when they come out. Core count is primarily just a big number to slap on the box and for building hype before release.