This card has no DRAM. Graphcore states it plainly: "All the memory on the card is contained within the IPU", 900 MB of it, which is why vram_gb is empty and memory_type reads On-die SRAM. Unlike the GC200 and the Bow IPU there is no Streaming Memory option at all, so nothing is being left out of a column here. The two rates are already per IPU, because one C600 card is one IPU, and no division was needed. The 125 GB/s interconnect figure is the datasheet's "1 Tbps bi-directional" across four IPU-Links converted to bytes; the separate 64 GB/s in the Main features table is the PCIe Gen4 x16 host link and is not the same quantity. Maximum power 252 W is the datasheet's own figure and is the thermal ceiling the chassis must remove at the highest power cap setting, against a 185 W TDP at typical workloads and a configurable cap that Graphcore says is typically 180 W or 192 W. Graphcore has no structured sparsity feature, so neither rate has a sparse partner.
Graphcore logo

Graphcore C600

Type: ASICArchitecture: Colossus MK2Form factor: PCIeReleased: 2022Spec confidence: Official
FP8 (dense)
560
TFLOPS
TDP
185 W

Overview

The C600 is the only Colossus MK2 part Graphcore sold as a card, launched in November 2022 and aimed squarely at inference. It is a dual-slot full-height PCIe Gen4 card carrying one Mk2 IPU with FP8 support added, clocked at 1.5 GHz for 560 TFLOPS of FP8 and 280 TFLOPS of FP16 inside a 185 W thermal envelope, passively cooled. All 900 MB of its memory is inside the processor and there is no DRAM on the board at all, which the datasheet states outright: the whole card is one chip and its SRAM. Four IPU-Links carrying 1 Tbps between them let eight cards be cabled into a cluster inside a single chassis without a switch, so a model too large for one card can be split across the group and still stay off DRAM. Against a contemporary inference GPU the memory is tiny and the latency is the product.

Performance

Peak theoretical throughput by precision type

PrecisionPeak
FP64
No verified data available
FP32
No verified data available
TF32
No verified data available
BF16
No verified data available
FP16
16-bit floating point
280TFLOPS
FP8
8-bit floating point
560TFLOPS
FP6
No verified data available
FP4
No verified data available
INT8
No verified data available

The C600 in the GPU landscape

Peak FP16 TFLOPS (dense) against TDP, single-GPU parts tracked by Flopper

06001,2001,8002,4003,0000 W250 W500 W750 W1000 W1250 W1500 WInstinct MI355XC600

Higher and further left is better: more half-precision throughput for less power.

Specifications

Architecture

Colossus MK2

Form Factor

PCIe

Launch Year

2022

Process Node

No verified data available

Memory

No verified data available On-die SRAM

Bandwidth

No verified data available

TDP

185 W

Max power

252 W

Interconnect

125 GB/s IPU-Link

direction and scope not stated by vendor

IPU-Cores

1,472

Spec Confidence

Official

Full Specifications

Compute Engine
IPU-Cores 1,472
Base Clock 1.50 GHz
Memory
Memory No verified data available
Memory Type On-die SRAM
Bandwidth No verified data available
Interface Width No verified data available
On-Die SRAM 900 MB
Interconnect & I/O
GPU-to-GPU IPU-Link
Interconnect Bandwidth 125 GB/s direction and scope not stated by vendor
Interconnect Links 4
PCIe 4.0 x16
Power & Thermal
TDP 185 W
Max power 252 W
Cooling Passive
Physical & Media
Card Length 267 mm
Width 2-slot
General
Form Factor PCIe
Architecture Colossus MK2
Process Node No verified data available
Launch Year 2022

Datasheet & Resources

Flopper Datasheet

Graphcore C600 specifications, generated from our database. Printable.

C600 Datasheet, Overview (Graphcore)

Graphcore · Latest version

View

Data Provenance

Every figure traced to a source

Primary Source

Publisher
Graphcore
Published
No verified data available

Data Quality

Spec confidence
Official
Clock basis
Boost
Core precisions with figures
2 of 9
Normalization
All values in TFLOPS

Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.

Browse GPU Cloud Providers

Similar GPUs

Graphcore Bow IPU

· 2022
FP32: 87 TFLOPS
Compare vs Bow IPU

Graphcore GC200

· 2020
FP32: 63 TFLOPS
Compare vs GC200

Frequently Asked Questions

How many TFLOPS does the Graphcore C600 have?

The Graphcore C600 delivers 280 TFLOPS FP16 and 560 TFLOPS FP8 at peak. Flopper does not currently have a verified FP32 throughput figure for it.

What is the power consumption of the Graphcore C600?

The Graphcore C600 has a TDP (Thermal Design Power) rating of 185 watts.

How much memory does the Graphcore C600 have?

Flopper does not currently have verified memory capacity or bandwidth figures for the Graphcore C600.

What architecture is the Graphcore C600 based on?

The Graphcore C600 is based on the Colossus MK2 architecture, launched in 2022.

Stay Updated on GPU Releases

Get notified when new GPUs are added or specifications are updated.

Loading verification...

No spam, unsubscribe anytime.

Back to GPUs

© 2026 Flopper.io - Compare the GPUs Powering AI