3TB enterprise workload for less than a cup of coffee
Not a demo. A production-grade result.
TPC-DS is a decision-support benchmark with 99 complex SQL queries over 24 tables. It reflects real enterprise analytics workloads used in dashboards, ML, and business decisions.
This is not a partial benchmark. Not a hand-picked subset. Not a simplified workload on a tuned cluster with the meter off.
This is TPC-DS at full scale, correlated subqueries, multi-pass window functions, rollups across three sales channels - every query that breaks underpowered engines
01
99 out of 99
Every query completed. Correlated subqueries, multi-pass window functions, rollups across 3 sales channels.
No rewrites. No retries.
02
Sub-dollar at terabyte scale
The full TPC-DS suite against 1 TB for $0.52.
This is what production looks like when the engine is built right.
03
Pure engine performance
Same data. Same queries. Standard cloud hardware anyone can rent.
The only variable is the engine. Turbo is why these numbers exist.
04
Compute headroom for AI
When ML pipelines, feature stores, and real-time analytics compete for the same budget,
efficiency decides who ships and who stalls.
Benchmark Results
1 TB · Total Cost
$0.52
99 queries - 17 min
less than a pack of chewing gum
3 TB · Total Cost
$2.33
99 queries - 40 min
less than a gas station coffee
10 TB · Total Cost
$12.57
99 queries - 3.16 hrs
less than a casual weekday lunch
Infrastructure Details
Standard cloud instances. No custom hardware. The performance comes from Turbo.
Specification
1 TB Run
3 TB Run
10 TB Run
Machine Type
m8gd.8xlarge
r8gd.12xlarge
i8g.12xlarge
vCPU
32
48
64
Memory
128 GiB
384 GiB
512 GiB
Instance Storage
1× 1,900 GB NVMe
3× 950 GB NVMe
4× 3,750 GB NVMe
Network
15 Gbps
22.5 Gbps
Up to 37.5 Gbps
EBS Bandwidth
Up to 10 Gbps
Up to 15 Gbps
Up to 20 Gbps
1 TB Run
Machinem8gd.8xlarge
vCPU32
Memory128 GiB
Storage1×1900 GB
Network15 Gbps
EBS10 Gbps
3 TB Run
Machiner8gd.12xlarge
vCPU48
Memory384 GiB
Storage3×950 GB
Network22.5 Gbps
EBS15 Gbps
10 TB Run
Machinei8g.12xlarge
vCPU64
Memory512 GiB
Storage4×3750 GB
Network37.5 Gbps
EBS20 Gbps
How is Turbo the Fastest Analytical Engine?
SIMD-Accelerated Execution
Processes multiple data values per CPU cycle - the hardware parallelism
that JVM-based engines leave on the table.
Turbo exploits every cycle.
Columnar Vectorized Runtime
Filters, joins, and aggregations run through a native columnar engine on batches.
Lower overhead. Saturated pipelines. Higher throughput.
Smart Scheduling
Fills CPU idle windows during I/O waits with pending work.
2–4× higher utilization across workloads.
Every dollar saved on data infrastructure is a dollar spent on AI
AI initiatives don't fail because of bad models. They fail because there's no budget left after the data platform bill hits.
Feature stores, model training, embedding pipelines, real-time inference, they all compete for the same compute budget. When analytical workloads consume less, everything else gets room.
At $0.52 per terabyte, Turbo doesn't just run your queries. It creates the budget for what comes next
Bring your workload. We'll show you the gap.
Run your actual queries on Yeedu Turbo. Compare the cost to what you pay today The number speaks for itself
