add 2024Jan benchmark (#39)

* add 2024Jan benchmark

Signed-off-by: Keming <kemingyang@tensorchord.ai>

* add sidebar

Signed-off-by: Keming <kemingyang@tensorchord.ai>

* Update src/faqs/benchmark.md

Co-authored-by: Jinjing Zhou <VoVAllen@users.noreply.github.com>
Signed-off-by: Keming <kemingyang@tensorchord.ai>

* update low precision indexing link

Signed-off-by: Keming <kemingyang@tensorchord.ai>

---------

Signed-off-by: Keming <kemingyang@tensorchord.ai>
Co-authored-by: Jinjing Zhou <VoVAllen@users.noreply.github.com>

.vitepress/config.mts

@@ -102,6 +102,7 @@ export default defineConfig({
           items: [
             // { text: 'pgvecto.rs vs. pgvector', link: '/faqs/comparison-pgvector' },
             { text: 'pgvecto.rs vs. specialized vectordb', link: '/faqs/comparison-with-specialized-vectordb' },
+            { text: `benchmark`, link: '/faqs/benchmark' },
           ]
         },
         {

src/faqs/benchmark.md

@@ -0,0 +1,21 @@
+# Benchmark
+
+`pgvecto.rs` has been focused on performance from the beginning. We have continuously monitored the performance of `pgvecto.rs` and compared it with other vector search libraries. The following benchmark results are from January 2024.
+
+The test is done on Google Cloud [n2-standard-8](https://cloud.google.com/compute/docs/general-purpose-machines#n2_series) (8 vCPUs, 32 GB RAM) with the `laion-768-5m-ip` dataset.
+
+## `pgvecto.rs` v.s. `pgvector`
+
+With the HNSW index, `pgvecto.rs` can achieve **2.5x** responses per second as `pgvector` can do with a slightly better precision of around 97%. This advantage increases when higher precision is required.
+
+![pgvecto.rs_vs_pgvector](./images/2024Jan_pgvectors_compare.png)
+
+When the `vbase` mode is enabled, `pgvecto.rs` can achieve over **2x** more responses per second compared to `pgvector` on various filter probabilities.
+
+![pgvecto.rs_vs_pgvector_filter](./images/2024Jan_pgvectors_filter_compare.png)
+
+## `pgvecto.rs` with different quantization methods
+
+`pgvecto.rs` supports several [quantization methods](../usage/quantization.md) and different low precision [indexing](../usage/indexing.md). These methods can help to reduce memory usage at different scales. The following figure shows the performance of `pgvecto.rs` with different quantization methods.
+
+![pgvecto.rs_quantization](./images/2024Jan_pgvectors_quantization.png)