Evaluating CPU performance for PRINCE

[MrWonderful]

Let's say I want to evaluate the number of candidates generated by PRINCE on my CPU.

There doesn't seem to be a benchmark argument that can be passed to PRINCE, so my solution was to pipe the candidates to hashcat using the 99999 (plaintext) hash type. I choose this hash type because it seems unlikely to me that the GPU (a RTX 3090) would be the bottleneck in this case.

My first question: is there a better way to evaluate the number of candidates generated by PRINCE on my CPU?

Now, let's say I want to upgrade my CPU for PRINCE to generate more candidates. Unfortunately, there seems to be few benchmarks where people have tested their CPU against PRINCE and reported their throughput. This is unlike GPUs where I can find benchmarks for many GPUs against many hash types.

The best thing I have found is a website where CPUs are benchmarked for various operations. According to this website, my CPU has a score of 12,889 and has the following stats:

Code:

Integer Math: 41,613 MOps/Sec
Floating Point Math: 25,736 MOps/Sec
Find Prime Numbers: 45 Million Primes/Sec
Random String Sorting: 23,625 Thousand Strings/Sec
Data Encryption: 4,364 MBytes/Sec
Data Compression: 185,673 KBytes/Sec
Physics: 814 Frames/Sec
Extended Instructions: 15,721 Million Matrices/Sec
Single Thread: 2,393 MOps/Sec

Now, say I pick this CPU with a score of 32,822 with the following stats:

Code:

Integer Math: 128,567 MOps/Sec
Floating Point Math: 79,878 MOps/Sec
Find Prime Numbers: 118 Million Primes/Sec
Random String Sorting: 70,197 Thousand Strings/Sec
Data Encryption: 12,795 MBytes/Sec
Data Compression: 559,735 KBytes/Sec
Physics: 1,617 Frames/Sec
Extended Instructions: 42,170 Million Matrices/Sec
Single Thread: 2,650 MOps/Sec

My second question is: can I expect PRINCE to generate candidates at least three times faster than my current CPU? Is there a better way to estimate how many candidates would be generated by this new CPU?

Thanks!

[lapsikmees]

I think problem is not your CPU, it's how fast data moves between PCIe lane from CPU to GPU, so best performance should be CPU that supports PCIe 5 and also GPU that supports that. But I might be wrong, that is how I understand this.

[MrWonderful]

I think problem is not your CPU, it's how fast data moves between PCIe lane from CPU to GPU, so best performance should be CPU that supports PCIe 5 and also GPU that supports that. But I might be wrong, that is how I understand this.

I'm unsure about that. My GPU is connected to a PCIe 3.0x16 slot. On my current CPU, I average 7,000 kH/s candidates on hashcat (that is, without any rules) with the 99999 hash type.

Do you know any way to verify how fast data is moved from the CPU to the GPU?

[buka]

If you're using the straight dictionary attack and no rules, see
https://hashcat.net/faq/morework
princeprocessor is fast enough. It's hashcat that is the bottleneck.

[MrWonderful]

If you're using the straight dictionary attack and no rules, see
https://hashcat.net/faq/morework
princeprocessor is fast enough. It's hashcat that is the bottleneck.

That's a bummer. Thanks!

To be clear, I do realize I can use rules as an amplifier to increase the throughput; and that is usually what I do.

However, in this specific case, I don't need so much of an amplifier than more "base" candidates.