Originally Posted by Thorham
Yeah, I know, but it's the principle. The same for things like firmware upgrades. The moment software can change hardware behavior permanently, I get an itch somewhere.
Not a big fan of FPGA either then ey?
Some general reminders for the nay-sayers in this thread:
20 years ago Intel 486SX were actually 486DX with math co-processor disabled
10 years ago I could soft-mod mod my GF6800 to a (much more expensive) Quadro 4000FX and magically install quadro drivers to get better performance in 3d software. it was possible because it was essentially the same chip.
0 years ago, the same nVidia chip GK110 masquerades as GeForce 780, 780Ti, Titan, Titan Black... etc... difference is in diabling number of CUDA cores, adjusting frequencies.. nVidia is selling them at vastly different prices..
From the beginning it may be that yealds for the top model are bad and instead of throwing the whole chip away it is sold as a cheap chip with lots of disabled cores.. but after yealds improve theu might still artifically disable cores for the cheaper chips.
It is done this way because it makes sense for the manufacturer...
At that point nVidia could also do the same.. pay to unlock CUDA cores. I would actually prefer it to having a zillion physical models to choose from. Its just wasteful.
Ok, Time to... Get.Over.It.. Ya all.