Yesterday I published some benchmarks of Hardwood 1.0 on my Threadripper. Someone suggested I run the One Billion Row Challenge too, to see how it does, so here it is!
Gunnar Morling ran the original benchmarks on an EPYC 7502P, Zen 2, 32 cores with 128 GB of RAM. The official challenge was on 8 cores (sequentially chosen) plus a bonus of all 32 cores.
I chose to run the benchmark using 9 contenders from the published 8 and 32 core results. The 9 contenders I ran were thomaswue, artsiomkorzun, jerrinot, serkan-ozal, abeobk, stephenvonworley, royvanrijn, mtopolnik, yavuztas.
Threadripper 9980X vs EPYC 7502P
Before we go through the results, let’s compare these two machines.
Hertzner AX161 (EPYC 7502P, Zen 2):
32 cores / 64 threads
128 GB ECC DDR4
RAM disk configured for 1BRC
Threadripper 9980X (Zen 5):
64 cores / 128 threads
256 GB ECC DDR5 CL52 (with some minor secondary timing tweaks)
Samsung 9100 PRO, 8TB (no RAM disk)
Cooling: Silverstone XE360-TR + RAM cooling fans
SMT disabled
CPU Monkey reported the following Geekbench results
Fig 1. Geekbench 6 scores, with the Threadripper over 2x better performance
So we’re expecting the Threadripper to do a lot better.
The results
Threadripper results based on Ubuntu 26.04 and EPYC on Fedora 39.
Baseline (single-threaded)
EPYC 7502P = 4:49
Threadripper = 1:15
Threadripper is 4x faster on the single-threaded baseline.
8 cores
Best: thomaswue at 502 ms.
Fig 2. 8 cores, sequentially selected
This is roughly inline with the Geekbench results, with the Threadripper being 2.2-3.1x faster.
32 + 64 cores
Best 32 cores: thomaswue, artsiomkorzun both at 203 ms (jerrinot at 204 ms)
Best 64 cores: thomaswue at 140 ms
Fig 3. EPYC 32 cores, Threadripper 32 (selected sequentially) and 64 cores
A much more varied result this time:
artsiomkorzunandjerrinotsaw no improvement from 32 to 64 cores, whereas the rest saw a non-trivial improvement, with 1.5x being the best.serkan-ozal, the only non-GraalVM entry in this list, saw a 1.8x speed up with the 32 core Threadripper test over its EPYC counterpart.yavuztassaw basically no improvement on the Threadripper 32 core over its EPYC counterpart.
The overall winner is…
The thomaswue submission won on 8, 32 and 64 cores, with artsiomkorzun and jerrinot just behind.
| Cores | Rank | Entrant(s) | Time |
|---|---|---|---|
| 8 | 1st | thomasvue | 502 ms |
| 8 | 2nd | artsiomkorzun | 603 ms |
| 8 | 3rd | jerrinot | 616 ms |
| 32 | 1st | thomasvue / artsiomkorzun | 203 ms |
| 32 | 3rd | jerrinot | 204 ms |
| 64 | 1st | thomasvue | 140 ms |
| 64 | 2nd | artsiomkorzun / jerrinot / abeobk / stephenvonworley | 203 ms |
Can you beat this?
Can I nerdsnipe anyone into trying to beat 140 ms?