Thank you, Ranlo, you are correct. This is why I stated in terms of raw hashpower, rather than try to give some profitability statistic.
Here are some reliable numbers, for comparison:
GTX 750Ti - first-gen Maxwell architecture
2GB GDDR5 VRAM 128-bit bus width, 1020MHz base clock - 1085 boost clock, maximum 60 watts power consumption
Even with an older version of CCminer, a 750Ti should be pushing between 2.4 - 2.8 MH/sec on the X11 algorithm (depends on the cooling, brand, etc), and in many cases still only using about 40 watts.
GTX 980 - second-gen Maxwell
4GB GDDR5 VRAM 256-bit bus width, 1126MHz base clock - 1216 boost clock, maximum 165 watts power consumption
Again, using a non-optimized version of CCminer typically yields between 6.6 - 8.4 MH/sec on X11. It's a wide variance, because there is a rather large difference between the reference cards and brands with better cooling. Also depends on your machine, and how you've configured the card (OCing and other things), but in general the 980 is about equal to 3 of the 750Ti's (great for density, more output per rig).
And of course, there is a Maxwell-optimized version of CCminer which will greatly boost the performance of both cards, but will not work with any non-Maxwell cards (all of the new 900-series cards are Maxwell). That specific fork of CCminer can be found in this thread:
https://bitcointalk.org/index.php?topic=826901.0 (keep in mind that this version will
NOT work with any non-Maxwell GPU's!)
I'm just trying to compare the profitability of mining different algos. I am not sure of the use of giving a "raw hashpower" number in one algo without comparing it to not only other algos but the exchange rate value of the coins.
The answer to that question is simple. Hashpower-to-wattage ratio (and total hash output) is the only true way to compare the profitability of various GPUs. The reason being, the more hashpower a card has, the more mining output per sec, and therefore more coins/min and more $$$ per hour. So the most accurate way to compare GPUs is to look at their hash output when mining the same algorithm. For example, you could compare the hash output on X11 of the GTX 980, and the AMD R9 290 (in this example, the comparison is from 6.6 - 8.4 MH/sec on the GTX 980 at around 140 - 165 watts, vs 4.8 - 5.6 MH/sec on the R9 290 at around 250 - 275 watts). However, that's only looking at one side of the coin, because total hash output ignores one other very important factor. GPUs use electricity, and electricity costs $$$ (aside from a few unusual exceptions), so the amount of electricity a card uses to get that hash output can be subtracted from your total profits. So efficiency of a GPU is more important than most people think or realize, which is why NVidia has made such a leap in mining popularity with their Maxwell GPUs.
There is a very useful and rather easy-to-use mining tool called Miner Control (thread here:
https://bitcointalk.org/index.php?topic=769239.0 ), which takes care of all those calculations for you, and automatically selects the most profitable pool and algorithm for mining at any given time (near-real-time). It takes some time to get it configured properly for your particular setup or rig, but once configured, it does all the work for you. (admittedly, somewhat easier to set up for NVidia GPUs, than it is for AMD) No more guessing what coin or pool you should be mining, based on day-old figures
just click a single button, and Miner Control does the rest for you. Even better, is that the software is free, and open-source, and donation to the dev is completely optional (has a built-in "donation mining" feature, which you are free to use or not use, as you see fit).
All I wanted was for CUDA miners to chime in and show some numbers, as Deravelas had his 400MHs pointed at my pool and I was not sure what the most profitable algo was for him. This was met with opinions and no numbers. I've since ran the numbers and it is clear what algo CUDA miners should be hashing.
And yes, I already stated previously that NVidia cards perform best on any of the X-algos (X11, X13, etc), and also very well on the M7 algo.