From proof to platform
- dau began with a 5 GB NYSE TAQ market-data dataset and a concrete time-series analytical - task: calculate OHLCV bars. After staging data on dpv1, repeated queries ran in 0.92 seconds - versus 1.5 seconds on the CPU—39% lower per-query latency, with every result matched against - the software golden. That measured workflow now informs the next platform configurations. + dau began on dpv1 with a 5 GB NYSE TAQ dataset and a concrete task: calculate OHLCV bars. + Staged resident, repeated queries ran in 0.92 seconds against 1.5 seconds on the CPU—39% + lower per-query latency, every result matched to the software golden. dpv2 now carries that + work further: a fused temporal-finance workflow—as-of join, derived features, time bars and + rolling moments in a single pass, with no intermediate ever written back to memory—runs in + 11.5 ms against 31.4 ms for the same workload on an 18-core laptop, bit-exact. The + accelerator costs a fraction of the machine it outruns.
dpv1 · Proven
++ dpv1 · Where it started +
Initial platform
+First platform
134K LUTs · 1 GB DDR3
- Proved resident analytical queries can outperform CPU execution on accessible hardware. + Established the core result: resident analytical queries can outperform CPU execution on + hardware you can put beside a laptop.
- dpv2 · Scaling now + dpv2 · Current platform
Next platform
+Running today
204K LUTs · 8 GB DDR3 · PCIe x8
+204K LUTs · 2 GB DDR3 · PCIe x8
- Designed to expand resident capacity and host bandwidth for larger datasets and more - parallel configurations. + Carries the same designs as dpv1 at wider memory reads and eight parallel lanes—where the + fused temporal workflow now beats the CPU baseline.
- dpv3 · Design program -
+dpv3 · Planned
Concept platform
+In design
663K LUTs · 32–128 GB DDR4 · PCIe Gen3 x8 · Fabric links diff --git a/tests/test.ts b/tests/test.ts index 8b323b0..668f093 100644 --- a/tests/test.ts +++ b/tests/test.ts @@ -21,11 +21,13 @@ test('home page presents both ends of the dau platform', async ({ page }) => { page.getByRole('heading', { name: 'A configuration loop that improves around real work.' }) ).toBeVisible(); await expect(page.getByText('explicit fallback', { exact: false })).toBeVisible(); - await expect(page.getByText('dpv1 · Proven')).toBeVisible(); - await expect(page.getByText('dpv2 · Scaling now')).toBeVisible(); - await expect(page.getByText('dpv3 · Design program')).toBeVisible(); - await expect(page.getByText('Concept platform')).toBeVisible(); + await expect(page.getByText('dpv1 · Where it started')).toBeVisible(); + await expect(page.getByText('dpv2 · Current platform')).toBeVisible(); + await expect(page.getByText('dpv3 · Planned')).toBeVisible(); + await expect(page.getByText('In design')).toBeVisible(); await expect(page.getByText('39% lower per-query latency', { exact: false })).toBeVisible(); + // the current platform's measured result, not just the origin story + await expect(page.getByText('11.5 ms', { exact: false })).toBeVisible(); await expect(page.getByRole('link', { name: 'Become a design partner' })).toBeVisible(); });