Why NVidia gpus are slower in computing

[Ubscene]

Hello there, I'm a long time lurker who figured out to register just to write what I found about NVidia GTX 6xx computing performance.

I was browsing various forums for choosing my next GPU and I found a link to a famous site, reporting that:

GeForce GTX 680 trails AMD’s Radeon HD 7900 cards in 32-bit math. And it gets absolutely decimated in 64-bit floating-point operations, as Nvidia purposely protects its profitable professional graphics business by artificially capping perfrmance

So nvidia puts a limiter on his desktop cards (only on not-32bit operations) just because they don't want the professionals to buy them at a much cheaper cost than quadro cards! I'm just speechless.. I don't think I'm ever buying a nvidia card again.

[Rolf]

True, but only for floats.
GeForce GPUs are capped at 12.5% of the DP FP performance, while Tesla GPUs have 50% DP FP perf.
Hashcracking does not use floats.

[forumhero]

correct me if i'm wrong but even their professional cards "quadro, tesla, etc" is slower than AMD cards for hash cracking. regardless if consumer gpus are capped or not, AMD is faster still

[epixoip]

correct. in fact the tesla and quadro cards are even slower than the gtx line.

[atom]

The main reasons for AMD cards are faster than NV cards in general are:

- AMD GPU has more raw integer power (more SP), floating point is not important in crypto.
- AMD GPU/XOP CPU have a much better instruction set (BIT_SELECT, BFI_INT) for calculating crypto stuff.