Message boards : Puzzles and science : SPHINX
Message board moderation
| Author | Message |
|---|---|
|
25 years old Send message Joined: 27 Sep 26 Posts: 349 Credit: 5,209,993 RAC: 137,017 |
🗿 New app: Sphinx is live -- hunting Certicom's unsolved ECC Challenge! 🗿 Back in 1997, Certicom published a set of elliptic-curve discrete-logarithm problems with real cash prizes attached, as a public test of how hard elliptic-curve cryptography actually is to break. The smallest ones were cracked years ago. Several are still standing. 🦘 Sphinx runs Pollard's kangaroo algorithm against these still-unsolved instances -- the same proven method Keraunos already uses on our own puzzle range. Every host running Sphinx reports its distinguished points to one shared table, so the whole community is running ONE combined search, not thousands of separate long shots. 💰 All targets Sphinx will work through, in order: 🎯 ECCp-131 -- $20,000 (LIVE NOW) ⏭️ ECCp-163 -- $30,000 (next) ⏭️ ECCp-191 -- $40,000 (next) 🔭 ECCp-239 -- $50,000 (planned, needs wider math) 🔭 ECCp-359 -- $100,000 (planned, needs wider math) 🙏 Honesty first: this offer is over 25 years old and we have no confirmation it's still honored today. We're doing this because it's a real, historically significant problem worth attacking together -- not because we're counting on the payout. If Sphinx ever finds a match, you'll read it here before anyone cashes anything. Thanks for crunching with us -- questions/bugs welcome below. 🙏
|
|
51 years old Send message Joined: 28 Sep 26 Posts: 67 Credit: 45,834,781 RAC: 927,749 |
I have a question about how Sphinx works. My CPU tasks run evenly from 0-100%, end of the story. My GPU tasks run for 42-54% seemingly exclusively on CPU before they are transferred to the GPU which recalculates them in a matter of seconds. At the moment I have 4 running on my AMD iGPU and another 4 running on my 3060m, all carefully manually spaced by a few minutes so that they reach the GPU one after the other.. I could probably run even more at the same time because the GPUs rarely get busy. Is the reason for running them on the CPU first, that you require double-precision at the beginning which can be provided easiest by the CPU? If that is the reason, could you imagine a version for FP64 GPUs which could process them a lot faster like with Nostos?
|