Infrastructure · Lab 01

High-Performance Computing Lab

Apple silicon, Threadripper Pro, and workstation GPUs for lab research.

Request AccessView Projects
Fleet monitor11 systems online
Fleet FP32
≈379TFLOPS
CPU
320cores
Accel. memory
2.17TB
System memory
2.59TB

01Capacity

Available Compute

Fig. 01—Peak throughput by precision

Dense peak · sparsity off

  • FP4—
  • FP8852 TFLOPS3× RTX 4070 Super
  • FP16≈1.25 PFLOPS4× M3 Ultra · 1× RTX A6000 · 3× RTX 3080 Ti · 3× RTX 4070 Super
  • BF16991 TFLOPS1× RTX A6000 · 3× RTX 3080 Ti · 3× RTX 4070 Super
  • TF32496 TFLOPS1× RTX A6000 · 3× RTX 3080 Ti · 3× RTX 4070 Super
  • FP32≈379 TFLOPS4× M3 Ultra · 1× RTX A6000 · 3× RTX 3080 Ti · 3× RTX 4070 Super

Fig. 02—FP32 by system family

One segment per machine · ≈379.0 TFLOPS total

  • Apple Mac Studio4 × 32.8

    ≈131.2 TF

  • HP Z2 G9 Tower3 × 35.5

    106.5 TF

  • HP Z6 G5 A3 × 34.2

    102.6 TF

  • Lenovo ThinkStation P81 × 38.7

    38.7 TF

Fig. 03—Accelerator memory by family

One segment per machine · Mac Studio memory is unified

  • Apple Mac Studio4 × 512 GB

    2.05TB

  • HP Z2 G9 Tower3 × 12 GB

    36GB

  • HP Z6 G5 A3 × 12 GB

    36GB

  • Lenovo ThinkStation P81 × 48 GB

    48GB

  • Dense peak
  • Sparsity off
  • Neural Engine excluded
  • ≈ M3 estimate
  • FP4 unsupported

02Hardware

Compute Ranking

Ranked by fleet FP32.

  1. 01

    Apple Mac Studio

    4× available · AI Workstation

    ≈131.2TFLOPS

    Fleet FP32 · 34.6% of lab

    CPU

    Apple M3 Ultra

    32-core CPU (24 performance + 8 efficiency)

    Cores32

    GPU

    80-core Apple GPU

    819GB/s unified memory bandwidth

    Accel. mem512 GB

    Memory

    512GB Unified Memory

    32-core Neural Engine

    RAM512 GB

    Storage

    16TB SSD

    Internal flash storage

    Capacity16 TB
  2. 02

    HP Z2 G9 Tower

    3× available · Development

    106.5TFLOPS

    Fleet FP32 · 28.1% of lab

    CPU

    Intel Core i7-14700

    20 cores / 28 threads

    Cores20

    GPU

    NVIDIA RTX 4070 Super

    12GB GDDR6X

    Accel. mem12 GB

    Memory

    32GB DDR5

    ECC RDIMM

    RAM32 GB

    Storage

    2TB NVMe SSD

    PCIe 4.0/5.0

    Capacity2 TB
  3. 03

    HP Z6 G5 A

    3× available · High-Performance

    102.6TFLOPS

    Fleet FP32 · 27.1% of lab

    CPU

    AMD Threadripper Pro 7945WX

    12 cores / 24 threads

    Cores12

    GPU

    NVIDIA RTX 3080 Ti

    12GB GDDR6X

    Accel. mem12 GB

    Memory

    64GB DDR5

    ECC RDIMM

    RAM64 GB

    Storage

    1TB NVMe SSD

    PCIe 4.0/5.0

    Capacity1 TB
  4. 04

    Lenovo ThinkStation P8

    1× available · Flagship

    38.7TFLOPS

    Fleet FP32 · 10.2% of lab

    CPU

    AMD Threadripper Pro 7995WX

    96 cores / 192 threads

    Cores96

    GPU

    NVIDIA RTX A6000

    48GB GDDR6

    Accel. mem48 GB

    Memory

    256GB DDR5

    ECC RDIMM

    RAM256 GB

    Storage

    1TB NVMe + 4TB HDD

    PCIe 4.0/5.0

    Capacity5 TB

03Applications

Research Capabilities

AI & Machine Learning

Train and serve deep learning models.

Scientific Computing

Large-scale simulation and modelling.

3D Visualization

Rendering and visualisation.

Need compute for your research?

Request AccessContact Us