ethomaz said:
|
Netyaroze said:
1 gflop old architecture =/= 1gflop new architecture
you cant compare that so easily. 240 Xbox gflop are not as capable as 200 GCN gflop on either ps4 or xbone.Because of efficent design you can put more of the raw compute power to better use than in 8 year old architecture. Achieving the same with less or more with the same.
|
Wrogn.FLOPS are FLOPS.
The difference in architecture is how much it is efficient to use these FLOPS... I mean any in architecture in the best case scenario can use 100% of the FLOPS but there is no best case scenario. In real games a old architecture uses less of these FLOPS at the same due limitations (specifics processs can't be made in parallel, etc) and a new architecture can use more these FLOPS at the same (the process that can't be made in parallel now is possible).
That's the difference in architecture... efficient in the use of FLOPS.
But FLOPS are FLOPS... FLOPS = FLoating-point Operations Per Second... it is like Km/h is Km/h and there is no way to be diffrence between cars but a new can can reach the top speed more easly than a old one (or not... there are bad new architecture like P3 to P4) ... and that will never end.
|
Thats what I am trying to say FLOPS are FLOPS just like HP is HP but if you dont get the power on the street because your tires and transmission suck its worthless.
What good is it to compare 240 gflops from 2005 and say 1.8tflop/240gflop = 7.5 or 1.2tflop/240gflops = 5 if the flops themselfs are not equivalent to each other.
There are problems which can easily be solved 10+ times as fast on the GCN architecture. Its just not comparable at all. The RSX had 400 gflops theoretically. Its totally pointless to compare the floating point operateration performance between architectures that are several generations apart. The theoretical peaks of those old architectures are just that, very theoretical... Nvidia vs AMD gflop comparissons are not as bad as this, the difference between the real world performance of 2005 gpus and 2013 gpus is much higher flop for flop. If km/h are km/h how come a 1.3TFlop Nvidia GPU (GTX480) creams a 2.7 tflop AMD GPU (5870) in everything ? How come a 2013 HD 7870 GHZ with 2.5 tflop just a mere 2 generations younger utterly destroys a HD 5870 in everything ?
You just dont know how they achieved that gflop number its totally pointless to compare a 7 year old AMD GPU to a AMD GPU from today.
I dont see what informational value you get by doing those calculations. Other than saying PS4 is more XBox360s taped together than a XBone its pretty much pointless.
Its fun but everyone should be aware that all of this is a pointless exercise. You could just as easily compare the transistor length between the 360 and the PS4/XB1 and still not get any information of value out of it.
If you can tell me what it means that XB1 is 3-5 360s gflop I am curious to hear what exactly anyone can do with this knowledge. Does it mean better graphics. Does it mean that it runs code 3-5 times faster or that it can display 3-5 times more polygons or that it can actually perform just 3-5 times the floating point operations of a 360 GPU in a game ? Its just nonsensical information...
I should add FLOPS are not a SI unit and not a standardized unit, everyone can come up with a unique way to achieve that flop number and get varying results depending on the test.