Posted by admin

Modern DRAM is based on a brilliant design from IBM. But, we're still paying for a latency penalty that's existed since the 60s! In this video, I'm introducing my personal research project (Tailslayer) that immensely reduces p99.99 latency on traditional RAM! By implementing a hedged read strategy taking advantage of (undocumented!) channel scrambling offsets, I've gotten as much as 15x reductions in tail latency. The technique works across Intel, AMD, Graviton, DDR4, DDR5, x86, ARM, you name it. Check out the C++ lib I wrote, watch the video, and try it yourself! --- Timestamps: 00:00 Your RAM goes Blind 01:58 A 400ns tRFC Lockout 04:42 Can we predict it? 08:02 Hedged Reads 15:41 The ROB Trap 19:20 Multicore Threading FTW! 23:19 Memory Controllers Hate You 27:09 A Dark Reason (Rowhammer) 30:07 Reverse Engineering Channel Scrambling 33:44 Where is my Data? (Physically) 38:46 Does it actually work? 45:20 ARM Graviton...the Black Box 48:06 High Frequency Trading --- My project / code (Tailslayer) on GitHub: https://github.com/LaurieWired/tailslayer --- Check out my X account for cool Computer Science stuff! https://x.com/lauriewired