1080ti

[Jajhaj]

Is the 1080ti the best card bang for buck? (Aside from the p102-100)
Has the same power consumption and hash rate as a 3070 and only cost $100 per card.

[Snoopy]

where did you get your infos about same hashrate on 1080ti and 3070?

[slyexe]

Is the 1080ti the best card bang for buck? (Aside from the p102-100)
Has the same power consumption and hash rate as a 3070 and only cost $100 per card.

4090 is best bang for your buck, but if you're on a budget than anything else can be considered. Depends on what HASH and what your purpose of use is. Judging by your other post you are interested in WPA Hashes so probably just a hobby and not looking at building a dedicated cracking rig. So have to think about your workspace and how many cards you are looking to invest into. More cards means more RAM to equal VRAM and more PCIE Slots + Power. So its hard to say what your specific best bang for the buck is.

[Jajhaj]

where did you get your infos about same hashrate on 1080ti and 3070?

Just compare the hashrate between them, is almost the same.

[Jajhaj]

Is the 1080ti the best card bang for buck? (Aside from the p102-100)
Has the same power consumption and hash rate as a 3070 and only cost $100 per card.

4090 is best bang for your buck, but if you're on a budget than anything else can be considered. Depends on what HASH and what your purpose of use is. Judging by your other post you are interested in WPA Hashes so probably just a hobby and not looking at building a dedicated cracking rig. So have to think about your workspace and how many cards you are looking to invest into. More cards means more RAM to equal VRAM and more PCIE Slots + Power. So its hard to say what your specific best bang for the buck is.

4090 isnt the best bang for buck, excluding electricity costs, it is 2x more efficient.
You could buy 14 1080ti’s for the price of one 4090 and they would all smoke 4090 by over 3 times. And ram, well you can get a 16x4 ddr4 ram kit for like $50.

[Chick3nman]

It might be important to note that while you CAN buy many 1080Ti's for the price of a 4090, you can't actually run that many 1080Ti's without spending extra money on supporting hardware that isn't needed for a single 4090 or even a few 4090's. Just comparing the cost of the GPUs without comparing everything else necessary to run them is an incomplete cost comparison. With the 4090 being anywhere from 3 to 12x as fast depending on algorithm, it's also not easy to just blanket compare cards across generations like that. Also, perhaps important to understand is that the 1080Ti is a very old card at this point and expecting long term support for them is probably not an ideal thing to do. Driver EOL is not the only thing to consider, as the CUDA Toolkit also has it's own versioning problems and when it comes to getting older cards to work properly you are likely to be stuck on older versions that essentially phase those cards out first in terms of being "up to date" enough to be useful.

If you are cost optimizing a build, you should be buying from newer generations that the 10XX series in general, for more reasons than just raw "hashes per second per dollar" unless you have a specific algorithm you intend to run and you are looking only for that metric.

For some quick examples of why this isn't so easy:
4090 can do 240kH/s bcrypt, 1080Ti can do 20kH/s. 12x faster.
vs
4090 can do 285GH/s NTLM, 1080ti can do 51GH/s. 5.6x faster.

Comparing these cards requires a little bit more work to get accurate numbers.