Data types and precision support#

2025-10-20

12 min read time

Applies to Linux and Windows

This topic summarizes the data types supported on AMD GPUs andROCm libraries, along with correspondingHIP data types.

Integral types#

The signed and unsigned integral types supported by ROCm are listed inthe following table.

Type name

HIP type

Description

int8

int8_t,uint8_t

A signed or unsigned 8-bit integer

int16

int16_t,uint16_t

A signed or unsigned 16-bit integer

int32

int32_t,uint32_t

A signed or unsigned 32-bit integer

int64

int64_t,uint64_t

A signed or unsigned 64-bit integer

Floating-point types#

The floating-point types supported by ROCm are listed in the following table.

Supported floating-point types

Type name

HIP type

Description

float4 (E2M1)

__hip_fp4_e2m1

A 4-bit floating-point number withE2M1 bit layout, as describedinlow precision floating point types page.

float6 (E3M2)

__hip_fp6_e3m2

A 6-bit floating-point number withE3M2 bit layout, as describedinlow precision floating point types page.

float6 (E2M3)

__hip_fp6_e2m3

A 6-bit floating-point number withE2M3 bit layout, as describedinlow precision floating point types page.

float8 (E4M3)

__hip_fp8_e4m3_fnuz,
__hip_fp8_e4m3

An 8-bit floating-point number withE4M3 bit layout, as described inlow precision floating point types page.The FNUZ variant has expanded range with no infinity or signed zero (NaN represented as negative zero),while the OCP variant follows the Open Compute Project specification.

float8 (E5M2)

__hip_fp8_e5m2_fnuz,
__hip_fp8_e5m2

An 8-bit floating-point number withE5M2 bit layout, as described inlow precision floating point types page.The FNUZ variant has expanded range with no infinity or signed zero (NaN represented as negative zero),while the OCP variant follows the Open Compute Project specification.

float16

half

A 16-bit floating-point number that conforms to the IEEE 754-2008half-precision storage format.

bfloat16

bfloat16

A shortened 16-bit version of the IEEE 754 single-precision storageformat.

tensorfloat32

Not available

A floating-point number that occupies 32 bits or less of storage,providing improved range compared to half (16-bit) format, at(potentially) greater throughput than single-precision (32-bit) formats.

float32

float

A 32-bit floating-point number that conforms to the IEEE 754single-precision storage format.

float64

double

A 64-bit floating-point number that conforms to the IEEE 754double-precision storage format.

Note

  • The float8 and tensorfloat32 types are internal types used in calculationsin Matrix Cores and can be stored in any type of the same size.

  • CDNA3 natively supports FP8 FNUZ (E4M3 and E5M2), which differs from the customizedFP8 format used with NVIDIA H100(FP8 Formats for Deep Learning).

  • In some AMD documents and articles, float8 (E5M2) is referred to as bfloat8.

  • Thelow precision floating point types pagedescribes how to use these types in HIP with examples.

Level of support definitions#

In the following sections, icons represent the level of support. These icons,described in the following table, are also used in the library data type supportpages.

Icon

Definition

NA

Not applicable

❌

Not supported

⚠️

Partial support

✅

Full support

Note

  • Full support means that the type is supported natively or with hardwareemulation.

  • Native support means that the operations for that type are implemented inhardware. Types that are not natively supported are emulated with theavailable hardware. The performance of non-natively supported types candiffer from the full instruction throughput rate. For example, 16-bitinteger operations can be performed on the 32-bit integer ALUs at full rate;however, 64-bit integer operations might need several instructions on the32-bit integer ALUs.

  • Any type can be emulated by software, but this page does not cover suchcases.

Data type support by hardware architecture#

AMD’s GPU lineup spans multiple architecture generations:

  • CDNA1 such as MI100

  • CDNA2 such as MI210, MI250, and MI250X

  • CDNA3 such as MI300A, MI300X, and MI325X

  • CDNA4 such as MI350X and MI355X

  • RDNA2 such as PRO W6800 and PRO V620

  • RDNA3 such as RX 7900XT and RX 7900XTX

  • RDNA4 such as RX 9070 and RX 9070XT

HIP C++ type implementation support#

The HIP C++ types available on different hardware platforms are listed in thefollowing table.

HIP C++ Type

CDNA1

CDNA2

CDNA3

CDNA4

RDNA2

RDNA3

RDNA4

int8_t,uint8_t

✅

✅

✅

✅

✅

✅

✅

int16_t,uint16_t

✅

✅

✅

✅

✅

✅

✅

int32_t,uint32_t

✅

✅

✅

✅

✅

✅

✅

int64_t,uint64_t

✅

✅

✅

✅

✅

✅

✅

__hip_fp4_e2m1

❌

❌

❌

✅

❌

❌

❌

__hip_fp6_e2m3

❌

❌

❌

✅

❌

❌

❌

__hip_fp6_e3m2

❌

❌

❌

✅

❌

❌

❌

__hip_fp8_e4m3_fnuz

❌

❌

✅

❌

❌

❌

❌

__hip_fp8_e5m2_fnuz

❌

❌

✅

❌

❌

❌

❌

__hip_fp8_e4m3

❌

❌

❌

✅

❌

❌

✅

__hip_fp8_e5m2

❌

❌

❌

✅

❌

❌

✅

half

✅

✅

✅

✅

✅

✅

✅

bfloat16

✅

✅

✅

✅

✅

✅

✅

float

✅

✅

✅

✅

✅

✅

✅

double

✅

✅

✅

✅

✅

✅

✅

Note

Library support for specific data types is contingent upon hardware support.Even if a ROCm library indicates support for a particular data type, that typewill only be fully functional if the underlying hardware architecture (as shownin the table above) also supports it. For example, fp8 types are only availableon architectures shown with a checkmark in the relevant rows.

Compute units support#

The following table lists data type support for compute units.

Type name

int8

int16

int32

int64

CDNA1

✅

✅

✅

✅

CDNA2

✅

✅

✅

✅

CDNA3

✅

✅

✅

✅

CDNA4

✅

✅

✅

✅

RDNA2

✅

✅

✅

✅

RDNA3

✅

✅

✅

✅

RDNA4

✅

✅

✅

✅

Type name

float4

float6 (E2M3)

float6 (E3M2)

float8 (E4M3)

float8 (E5M2)

CDNA1

❌

❌

❌

❌

❌

CDNA2

❌

❌

❌

❌

❌

CDNA3

❌

❌

❌

❌

❌

CDNA4

❌

❌

❌

❌

❌

RDNA2

❌

❌

❌

❌

❌

RDNA3

❌

❌

❌

❌

❌

RDNA4

❌

❌

❌

❌

❌

Type name

float16

bfloat16

tensorfloat32

float32

float64

CDNA1

✅

✅

❌

✅

✅

CDNA2

✅

✅

❌

✅

✅

CDNA3

✅

✅

❌

✅

✅

CDNA4

✅

✅

❌

✅

✅

RDNA2

✅

✅

❌

✅

✅

RDNA3

✅

✅

❌

✅

✅

RDNA4

✅

✅

❌

✅

✅

Matrix core support#

The following table lists data type support for AMD GPU matrix cores.

Type name

int8

int16

int32

int64

CDNA1

✅

❌

❌

❌

CDNA2

✅

❌

❌

❌

CDNA3

✅

❌

❌

❌

CDNA4

✅

❌

❌

❌

RDNA2

✅

❌

❌

❌

RDNA3

✅

❌

❌

❌

RDNA4

✅

❌

❌

❌

Type name

float4

float6 (E2M3)

float6 (E3M2)

float8 (E4M3)

float8 (E5M2)

CDNA1

❌

❌

❌

❌

❌

CDNA2

❌

❌

❌

❌

❌

CDNA3

❌

❌

❌

✅

✅

CDNA4

✅

✅

✅

✅

✅

RDNA2

❌

❌

❌

❌

❌

RDNA3

❌

❌

❌

❌

❌

RDNA4

❌

❌

❌

✅

✅

Type name

float16

bfloat16

tensorfloat32

float32

float64

CDNA1

✅

✅

❌

✅

❌

CDNA2

✅

✅

❌

✅

✅

CDNA3

✅

✅

✅

✅

✅

CDNA4

✅

✅

✅

✅

✅

RDNA2

✅

✅

❌

❌

❌

RDNA3

✅

✅

❌

❌

❌

RDNA4

✅

✅

❌

❌

❌

Atomic operations support#

The following table lists which data types are supported for atomicoperations on AMD GPUs. The atomics operation type behavior is affected by thememory locations, memory granularity, or scope of operations. For detailedvarious support of atomic read-modify-write (atomicRMW) operations collected ontheHardware atomics operation supportpage.

Type name

int8

int16

int32

int64

CDNA1

❌

❌

✅

✅

CDNA2

❌

❌

✅

✅

CDNA3

❌

❌

✅

✅

RDNA3

❌

❌

✅

✅

RDNA4

❌

❌

✅

✅

Type name

float4

float6 (E2M3)

float6 (E3M2)

float8 (E4M3)

float8 (E5M2)

CDNA1

❌

❌

❌

❌

❌

CDNA2

❌

❌

❌

❌

❌

CDNA3

❌

❌

❌

❌

❌

CDNA4

❌

❌

❌

❌

❌

RDNA2

❌

❌

❌

❌

❌

RDNA3

❌

❌

❌

❌

❌

RDNA4

❌

❌

❌

❌

❌

Type name

2 x float16

2 x bfloat16

tensorfloat32

float32

float64

CDNA1

✅

✅

❌

✅

❌

CDNA2

✅

✅

❌

✅

✅

CDNA3

✅

✅

❌

✅

✅

CDNA4

✅

✅

❌

✅

✅

RDNA2

❌

❌

❌

✅

❌

RDNA3

❌

❌

❌

✅

❌

RDNA4

✅

✅

❌

✅

❌

Note

You can emulate atomic operations using software for cases that are notnatively supported. Software-emulated atomic operations have a high negativeperformance impact when they frequently access the same memory address.

Data type support in ROCm libraries#

ROCm library support for int8, float8 (E4M3), float8 (E5M2), int16, float16,bfloat16, int32, tensorfloat32, float32, int64, and float64 is listed in thefollowing tables.

Libraries input/output type support#

The following tables list ROCm library support for specific input and outputdata types. Select a library from the below table to view the supported datatypes.

Category
ML & Computer Vision
Communication
Math Libraries
Primitives
Library
Composable Kernel
MIGraphX
MIOpen
RCCL
hipBLAS
hipBLASLt
hipFFT
hipRAND
hipSOLVER
hipSPARSE
hipSPARSELt
rocBLAS
rocFFT
rocRAND
rocSOLVER
rocSPARSE
rocWMMA
Tensile
hipCUB
hipTensor
rocPRIM
rocThrust

For more information, please visitComposable Kernel.

Data Type

Support

int8

✅

int32

✅

float4

✅

float6 (E2M3)

✅

float6 (E3M2)

✅

float8 (E4M3)

✅

float8 (E5M2)

✅

float16

✅

bfloat16

✅

float32

✅

float64

✅

For more information, please visitMIGraphX.

Data Type

Support

int8

⚠️

int16

✅

int32

✅

int64

✅

float8 (E4M3)

✅

float8 (E5M2)

✅

float16

✅

bfloat16

✅

float32

✅

float64

✅

For more information, please visitMIOpen.

Data Type

Support

int8

⚠️

int32

⚠️

float8 (E4M3)

⚠️

float8 (E5M2)

⚠️

float16

✅

bfloat16

⚠️

float32

✅

float64

⚠️

For more information, please visitRCCL.

Data Type

Support

int8

✅

int32

✅

int64

✅

float8 (E4M3)

✅

float8 (E5M2)

✅

float16

✅

bfloat16

✅

float32

✅

float64

✅

For more information, please visithipBLAS.

Data Type

Support

float16

⚠️

bfloat16

⚠️

float32

✅

float64

✅

For more information, please visithipBLASLt.

Data Type

Support

int8

✅

float4

✅

float6 (E2M3)

✅

float6 (E3M2)

✅

float8 (E4M3)

✅

float8 (E5M2)

✅

float16

✅

bfloat16

✅

float32

✅

For more information, please visithipFFT.

Data Type

Support

float32

✅

float64

✅

For more information, please visithipRAND.

Data Type

Support

int8

Output only

int16

Output only

int32

Output only

int64

Output only

float16

Output only

float32

Output only

float64

Output only

For more information, please visithipSOLVER.

Data Type

Support

float32

✅

float64

✅

For more information, please visithipSPARSE.

Data Type

Support

float32

✅

float64

✅

For more information, please visithipSPARSELt.

Data Type

Support

int8

✅

float8 (E4M3)

✅

float8 (E5M2)

✅

float16

✅

bfloat16

✅

float32

✅

For more information, please visitrocBLAS.

Data Type

Support

float16

⚠️

bfloat16

⚠️

float32

✅

float64

✅

For more information, please visitrocFFT.

Data Type

Support

float16

✅

float32

✅

float64

✅

For more information, please visitrocRAND.

Data Type

Support

int8

Output only

int16

Output only

int32

Output only

int64

Output only

float16

Output only

float32

Output only

float64

Output only

For more information, please visitrocSOLVER.

Data Type

Support

float32

✅

float64

✅

For more information, please visitrocSPARSE.

Data Type

Support

float32

✅

float64

✅

For more information, please visitrocWMMA.

Data Type

Support

int8

✅

int32

Output only

float8 (E4M3)

Input only

float8 (E5M2)

Input only

float16

✅

bfloat16

✅

tensorfloat32

✅

float32

✅

float64

✅

For more information, please visitTensile.

Data Type

Support

int8

✅

int32

✅

float8 (E4M3)

✅

float8 (E5M2)

✅

float16

✅

bfloat16

✅

tensorfloat32

✅

float32

✅

float64

✅

For more information, please visithipCUB.

Data Type

Support

int8

✅

int16

✅

int32

✅

int64

✅

float16

✅

bfloat16

✅

float32

✅

float64

✅

For more information, please visithipTensor.

Data Type

Support

float16

✅

bfloat16

✅

float32

✅

float64

✅

For more information, please visitrocPRIM.

Data Type

Support

int8

✅

int16

✅

int32

✅

int64

✅

float16

✅

bfloat16

✅

float32

✅

float64

✅

For more information, please visitrocThrust.

Data Type

Support

int8

✅

int16

✅

int32

✅

int64

✅

float16

⚠️

bfloat16

⚠️

float32

✅

float64

✅

Note

The meaning of partial support depends on the library. Please refer to the individuallibraries’ documentation for more information.

Note

As random number generation libraries, rocRAND and hipRAND only specify outputdata types for the random values they generate, with no need for input datatypes.

Note

hipBLASLt supports additional data types as internal compute types, which maydiffer from the supported input/output types shown in the tables above. WhileTensorFloat32 is not supported as an input or output type in this library, itis available as an internal compute type. For complete details on supportedcompute types, refer to thehipBLASLtdocumentation.

hipDataType enumeration#

ThehipDataType enumeration defines data precision types and is primarilyused when the data reference itself does not include type information, such asinvoid* pointers. This enumeration is mainly utilized in BLAS libraries.The HIP type equivalents of thehipDataType enumeration are listed in thefollowing table with descriptions and values.

hipDataType

HIP type

Value

Description

HIP_R_8I

int8_t

3

8-bit real signed integer.

HIP_R_8U

uint8_t

8

8-bit real unsigned integer.

HIP_R_16I

int16_t

20

16-bit real signed integer.

HIP_R_16U

uint16_t

22

16-bit real unsigned integer.

HIP_R_32I

int32_t

10

32-bit real signed integer.

HIP_R_32U

uint32_t

12

32-bit real unsigned integer.

HIP_R_32F

float

0

32-bit real single precision floating-point.

HIP_R_64F

double

1

64-bit real double precision floating-point.

HIP_R_16F

half

2

16-bit real half precision floating-point.

HIP_R_16BF

bfloat16

14

16-bit real bfloat16 precision floating-point.

HIP_R_8F_E4M3

__hip_fp8_e4m3

28

8-bit real float8 precision floating-point (OCP version).

HIP_R_8F_E5M2

__hip_fp8_e5m2

29

8-bit real bfloat8 precision floating-point (OCP version).

HIP_R_6F_E2M3

__hip_fp6_e2m3

31

6-bit real float6 precision floating-point.

HIP_R_6F_E3M2

__hip_fp6_e3m2

32

6-bit real bfloat6 precision floating-point.

HIP_R_4F_E2M1

__hip_fp4_e2m1

33

4-bit real float4 precision floating-point.

HIP_R_8F_E4M3_FNUZ

__hip_fp8_e4m3_fnuz

1000

8-bit real float8 precision floating-point (FNUZ version).

HIP_R_8F_E5M2_FNUZ

__hip_fp8_e5m2_fnuz

1001

8-bit real bfloat8 precision floating-point (FNUZ version).

The full list of thehipDataType enumeration listed inlibrary_types.h.