Calculator GPU Hardware A compact GPU selection from 12 GB: entry tier at 12–16 GB, main tier at 24 GB and above. Availability, pricing and compatibility need review.
27 records · memory is not a speed ranking
Intel
Arc B580 12 GB GDDR6
Xe2 · 190W · 456 GB/s
Desktop PCIe board; verify exact vendor variant
Source checked: 2026-09-06
Primary source All captured specifications
Memory type GDDR6
Memory Bandwidth 456 GB/s
Power 190W
Form Factor Desktop PCIe board; verify exact vendor variant Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
GeForce RTX 5070 12 GB GDDR7
Blackwell · 250 W · 672 GB/s
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
NVIDIA CUDA® Cores 6144
Shader Cores Blackwell
Tensor Cores (AI) 5th Generation 988 AI TOPS
Ray Tracing Cores 4th Generation 94 TFLOPS
Boost Clock (GHz) 2.51
Base Clock (GHz) 2.33
Standard Memory Config 12 GB GDDR7
Memory Interface Width 192-bit
Memory Bandwidth 672 GB/sec
PCI Express Gen 5
NVIDIA Encoder (NVENC) 1x Ninth Generation
NVIDIA Decoder (NVDEC) 1x Sixth Generation
CUDA Capability 12.0
Length 242 mm
Width 112 mm
Slot 2-Slot
Maximum GPU Temperature (in C) 85
Total Graphics Power (W) 250
Required System Power (W) (5) 650
Required Power Connectors 2x PCIe 8-pin cables (adapter in box) OR 300 W or greater PCIe Gen 5 cable Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system Intel
Arc Pro B50 16 GB GDDR6
Xe2 · 70W · 224 GB/s
Desktop PCIe board; verify exact vendor variant
Source checked: 2026-09-06
Primary source All captured specifications
Memory type GDDR6
Memory Bandwidth 224 GB/s
Power 70W
Form Factor Desktop PCIe board; verify exact vendor variant Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
GeForce RTX 5060 Ti 16GB 16 GB GDDR7
Blackwell · 180 W · 448 GB/s
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
NVIDIA CUDA® Cores 4608
Shader Cores Blackwell
Tensor Cores (AI) 5th Generation 759 AI TOPS
Ray Tracing Cores 4th Generation 72 TFLOPS
Boost Clock (GHz) 2.57
Base Clock (GHz) 2.41
Standard Memory Config 16 GB / 8 GB GDDR7
Memory Interface Width 128-bit
Memory Bandwidth 448 GB/sec
PCI Express Gen 5
NVIDIA Encoder (NVENC) 1x Ninth Generation
NVIDIA Decoder (NVDEC) 1x Sixth Generation
CUDA Capability 12.0
Length Varies by manufacturer
Width Varies by manufacturer
Slot Varies by manufacturer
Maximum GPU Temperature (in C) 87
Total Graphics Power (W) 180
Required System Power (W) (5) 600
Required Power Connectors 1x PCIe 8-pin cables OR 300 W or greater PCIe Gen 5 cable (adapter in box) Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
GeForce RTX 5070 Ti 16 GB GDDR7
Blackwell · 300 W · 896 GB/s
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
NVIDIA CUDA® Cores 8960
Shader Cores Blackwell
Tensor Cores (AI) 5th Generation 1406 AI TOPS
Ray Tracing Cores 4th Generation 133 TFLOPS
Boost Clock (GHz) 2.45
Base Clock (GHz) 2.30
Standard Memory Config 16 GB GDDR7
Memory Interface Width 256-bit
Memory Bandwidth 896 GB/sec
PCI Express Gen 5
NVIDIA Encoder (NVENC) 2x Ninth Generation
NVIDIA Decoder (NVDEC) 1x Sixth Generation
CUDA Capability 12.0
Length Varies by manufacturer
Width Varies by manufacturer
Slot Varies by manufacturer
Maximum GPU Temperature (in C) 88
Total Graphics Power (W) 300
Required System Power (W) (5) 750
Required Power Connectors 2x PCIe 8-pin cables (adapter in box) OR 300 W or greater PCIe Gen 5 cable Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
GeForce RTX 5080 16 GB GDDR7
Blackwell · 360 W · 960 GB/s
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
NVIDIA CUDA® Cores 10752
Shader Cores Blackwell
Tensor Cores (AI) 5th Generation 1801 AI TOPS
Ray Tracing Cores 4th Generation 171 TFLOPS
Boost Clock (GHz) 2.62
Base Clock (GHz) 2.30
Standard Memory Config 16 GB GDDR7
Memory Interface Width 256-bit
Memory Bandwidth 960 GB/sec
PCI Express Gen 5
NVIDIA Encoder (NVENC) 2x Ninth Generation
NVIDIA Decoder (NVDEC) 2x Sixth Generation
CUDA Capability 12.0
Length 304 mm
Width 137 mm
Slot 2-Slot
Maximum GPU Temperature (in C) 88
Total Graphics Power (W) 360
Required System Power (W) (5) 850
Required Power Connectors 3x PCIe 8-pin cables (adapter in box) OR 1x 450 W or greater PCIe Gen 5 cable Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system AMD
Radeon RX 9070 XT 16 GiB
RDNA4 · Power unknown · Bandwidth unknown
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
Name Radeon RX 9070 XT
Architecture RDNA4
LLVM target name gfx1201
VRAM (GiB) 16
Compute Units 64
Wavefront Size 32 or 64
LDS (KiB) 128
Infinity Cache (MiB) 64
L2 Cache (MiB) 8
Graphics L1 Cache (KiB) N/A
L0 Vector Cache (KiB) 32
L0 Scalar Cache (KiB) 16
L0 Instruction Cache (KiB) 32
VGPR File (KiB) 768
SGPR File (KiB) 32
GFXIP Major version 12
GFXIP Minor version 0 Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system Intel
Arc Pro B60 24 GB GDDR6
Xe2 · 120–200W · 456 GB/s
Desktop PCIe board; verify exact vendor variant
Source checked: 2026-09-06
Primary source All captured specifications
Memory type GDDR6
Memory Bandwidth 456 GB/s
Power 120–200W
Form Factor Desktop PCIe board; verify exact vendor variant Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
GeForce RTX 4090 24 GB GDDR6X
Ada Lovelace · 450 W · Bandwidth unknown
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
NVIDIA CUDA® Cores 16384
Shader Cores Ada Lovelace 83 TFLOPS
Ray Tracing Cores 3rd Generation 191 TFLOPS
Tensor Cores (AI) 4th Generation 1321 AI TOPS
Boost Clock (GHz) 2.52
Base Clock (GHz) 2.23
Standard Memory Config 24 GB GDDR6X
Memory Interface Width 384-bit
NVIDIA Architecture Ada Lovelace
PCI Express Gen 4 Yes
NVIDIA NVLink™ (SLI-Ready) No
NVIDIA Encoder (NVENC) 2x 8th Generation
NVIDIA Decoder (NVDEC) 5th Generation
CUDA Capability 8.9
Length 304 mm
Width 137 mm
Slot 3-Slot
Maximum GPU Temperature (in C) 90
Total Graphics Power (W) 450
Required System Power (W) (5) 850
Required Power Connectors 3x PCIe 8-pin cables (adapter in box) OR 1x 450 W or greater PCIe Gen 5 cable
PCI Express Gen 4 Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
L4 24 GB
Ada Lovelace · 72W · 300 GB/s
L4
Source checked: 2026-09-06
Primary source All captured specifications
Form Factor L4
FP32 30.3 teraFLOPs
TF32 Tensor Core 120 teraFLOPS*
FP16 Tensor Core 242 teraFLOPS*
BFLOAT16 Tensor Core 242 teraFLOPS*
FP8 Tensor Core 485 teraFLOPs*
INT8 Tensor Core 485 TOPs*
GPU memory 24GB
GPU memory bandwidth 300GB/s
NVENC | NVDEC | JPEG decoders 2 | 4 | 4
Max thermal design power (TDP) 72W
Form factor 1-slot low-profile, PCIe
Interconnect PCIe Gen4 x16 64GB/s
Server options Partner and NVIDIA-Certified Systems with 1–8 GPUs
GPU Memory 24GB Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system AMD
Radeon RX 7900 XTX 24 GiB
RDNA3 · Power unknown · Bandwidth unknown
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
Name Radeon RX 7900 XTX
Architecture RDNA3
LLVM target name gfx1100
VRAM (GiB) 24
Compute Units 96
Wavefront Size 32 or 64
LDS (KiB) 128
Infinity Cache (MiB) 96
L2 Cache (MiB) 6
Graphics L1 Cache (KiB) 256
L0 Vector Cache (KiB) 32
L0 Scalar Cache (KiB) 16
L0 Instruction Cache (KiB) 32
VGPR File (KiB) 768
SGPR File (KiB) 32
GFXIP Major version 11
GFXIP Minor version 0 Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system Intel
Arc Pro B70 32 GB GDDR6
Xe2 · 160–290W; Intel card 230W · 608 GB/s
Desktop PCIe board; verify exact vendor variant
Source checked: 2026-09-06
Primary source All captured specifications
Memory type GDDR6
Memory Bandwidth 608 GB/s
Power 160–290W; Intel card 230W
Form Factor Desktop PCIe board; verify exact vendor variant Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
GeForce RTX 5090 32 GB GDDR7
Blackwell · 575 W · 1792 GB/s
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
NVIDIA CUDA® Cores 21760
Shader Cores Blackwell
Tensor Cores (AI) 5th Generation 3352 AI TOPS
Ray Tracing Cores 4th Generation 318 TFLOPS
Boost Clock (GHz) 2.41
Base Clock (GHz) 2.01
Standard Memory Config 32 GB GDDR7
Memory Interface Width 512-bit
Memory Bandwidth 1792 GB/sec
PCI Express Gen 5
NVIDIA Encoder (NVENC) 3x Ninth Generation
NVIDIA Decoder (NVDEC) 2x Sixth Generation
CUDA Capability 12.0
Length 304 mm
Width 137 mm
Slot 2-Slot
Maximum GPU Temperature (in C) 90
Total Graphics Power (W) 575
Required System Power (W) (5) 1000
Required Power Connectors 4x PCIe 8-pin cables (adapter in box) OR 1x 600 W PCIe Gen 5 cable Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system AMD
Radeon AI PRO R9700 32 GiB
RDNA4 · Power unknown · Bandwidth unknown
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
Name Radeon AI PRO R9700
Architecture RDNA4
LLVM target name gfx1201
VRAM (GiB) 32
Compute Units 64
Wavefront Size 32 or 64
LDS (KiB) 128
Infinity Cache (MiB) 64
L2 Cache (MiB) 8
Graphics L1 Cache (KiB) N/A
L0 Vector Cache (KiB) 32
L0 Scalar Cache (KiB) 16
L0 Instruction Cache (KiB) 32
VGPR File (KiB) 768
SGPR File (KiB) 32
GFXIP Major version 12
GFXIP Minor version 0 Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
L40S 48 GB GDDR6
Ada Lovelace · 350W · 864 GB/s
4.4" (H) x 10.5" (L), dual slot
Source checked: 2026-09-06
Primary source All captured specifications
GPU Architecture NVIDIA Ada Lovelace architecture
GPU Memory 48GB GDDR6 with ECC
Memory Bandwidth 864GB/s
Interconnect Interface PCIe Gen4 x16: 64GB/s bidirectional
NVIDIA Ada Lovelace Architecture-Based CUDA® Cores 18,176
NVIDIA Third-Generation RT Cores 142
NVIDIA Fourth-Generation Tensor Cores 568
RT Core Performance TFLOPS 212
FP32 TFLOPS 91.6
TF32 Tensor Core TFLOPS 183 I 366*
BFLOAT16 Tensor Core TFLOPS 362.05 I 733*
FP16 Tensor Core 362.05 I 733*
FP8 Tensor Core 733 I 1,466*
Peak INT8 Tensor TOPS Peak INT4 Tensor TOPS 733 I 1,466* 733 I 1,466*
Form Factor 4.4" (H) x 10.5" (L), dual slot
Display Ports 4x DisplayPort 1.4a
Max Power Consumption 350W
Power Connector 16-pin
Thermal Passive
Virtual GPU (vGPU) Software Support Yes
vGPU Profiles Supported See virtual GPU licensing guide
NVENC I NVDEC 3x l 3x (includes AV1 encode and decode)
Secure Boot With Root of Trust Yes
NEBS Ready Level 3
Multi-Instance GPU (MIG) Support No
NVIDIA® NVLink® Support No
*With Sparsity *With Sparsity Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system AMD
Radeon PRO W7900 48 GiB
RDNA3 · Power unknown · Bandwidth unknown
Form factor needs review
Source checked: 2026-09-06
Primary source All captured specifications
Name Radeon PRO W7900
Architecture RDNA3
LLVM target name gfx1100
VRAM (GiB) 48
Compute Units 96
Wavefront Size 32 or 64
LDS (KiB) 128
Infinity Cache (MiB) 96
L2 Cache (MiB) 6
Graphics L1 Cache (KiB) 256
L0 Vector Cache (KiB) 32
L0 Scalar Cache (KiB) 16
L0 Instruction Cache (KiB) 32
VGPR File (KiB) 768
SGPR File (KiB) 32
GFXIP Major version 11
GFXIP Minor version 0 Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
RTX 6000 Ada Generation 48 GB GDDR6
Ada Lovelace · 300W · Bandwidth unknown
4.4” (H) x 10.5” (L) dual slot
Source checked: 2026-09-06
Primary source All captured specifications
GPU Memory 48GB GDDR6 with ECC
Max Power Consumption 300W
Form Factor 4.4” (H) x 10.5” (L) dual slot
Thermal Active Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
RTX PRO 5000 Blackwell 72GB 72 GB GDDR7
Blackwell · 300W · Bandwidth unknown
4.4” (H) x 10.5” (L) dual slot
Source checked: 2026-09-06
Primary source All captured specifications
GPU Memory 48GB or 72GB GDDR7 with ECC
Max Power Consumption 300W
Form Factor 4.4” (H) x 10.5” (L) dual slot
Thermal Active Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
A100 80GB PCIe 80 GB HBM2e
Ampere · 300W · 1935 GB/s
PCIe Dual-slot air-cooled or single-slot liquid-cooled
Source checked: 2026-09-06
Primary source All captured specifications
GPU Memory 80GB HBM2e
GPU Memory Bandwidth 1,935 GB/s
Max Thermal Design Power (TDP) 300W
Multi-Instance GPU Up to 7 MIGs @ 10GB
Form Factor PCIe Dual-slot air-cooled or single-slot liquid-cooled
Interconnect NVIDIA® NVLink® Bridge for 2 GPUs: 600 GB/s ** PCIe Gen4: 64 GB/s
Server Options Partner and NVIDIA-Certified Systems™ with 1-8 GPUs Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
H100 SXM 80 GB
Hopper · Up to 700W (configurable) · 3350 GB/s
SXM
Source checked: 2026-09-06
Primary source All captured specifications
FP64 34 teraFLOPS
FP64 Tensor Core 67 teraFLOPS
FP32 67 teraFLOPS
TF32 Tensor Core* 989 teraFLOPS
BFLOAT16 Tensor Core* 1,979 teraFLOPS
FP16 Tensor Core* 1,979 teraFLOPS
FP8 Tensor Core* 3,958 teraFLOPS
INT8 Tensor Core* 3,958 TOPS
GPU Memory 80GB
GPU Memory Bandwidth 3.35TB/s
Decoders 7 NVDEC 7 JPEG
Max Thermal Design Power (TDP) Up to 700W (configurable)
Multi-Instance GPUs Up to 7 MIGS @ 10GB each
Form Factor SXM
Interconnect NVIDIA NVLink™: 900GB/s PCIe Gen5: 128GB/s
Server Options NVIDIA HGX H100 Partner and NVIDIA- Certified Systems™ with 4 or 8 GPUs NVIDIA DGX H100 with 8 GPUs
NVIDIA AI Enterprise Add-on Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
RTX PRO 6000 Blackwell Workstation Edition 96 GB GDDR7
Blackwell · 600W · Bandwidth unknown
5.4” (H) x 12.0” (L) dual slot
Source checked: 2026-09-06
Primary source All captured specifications
GPU Memory 96GB GDDR7 with ECC
Max Power Consumption 600W
Form Factor 5.4” (H) x 12.0” (L) dual slot
Thermal Double Flow Through Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
H200 SXM 141 GB
Hopper · Up to 700W (configurable) · 4800 GB/s
SXM
Source checked: 2026-09-06
Primary source All captured specifications
FP64 34 TFLOPS
FP64 Tensor Core 67 TFLOPS
FP32 67 TFLOPS
TF32 Tensor Core² 989 TFLOPS
BFLOAT16 Tensor Core² 1,979 TFLOPS
FP16 Tensor Core² 1,979 TFLOPS
FP8 Tensor Core² 3,958 TFLOPS
INT8 Tensor Core² 3,958 TFLOPS
GPU Memory 141GB
GPU Memory Bandwidth 4.8TB/s
Decoders 7 NVDEC 7 JPEG
Confidential Computing Supported
Max Thermal Design Power (TDP) Up to 700W (configurable)
Multi-Instance GPUs Up to 7 MIGs @18GB each
Form Factor SXM
Interconnect NVIDIA NVLink™: 900GB/s PCIe Gen5: 128GB/s
Server Options NVIDIA HGX™ H200 partner and NVIDIA-Certified Systems™ with 4 or 8 GPUs
NVIDIA AI Enterprise Add-on Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
B200 (HGX) 180 GB HBM3e
Blackwell · Power unknown · Bandwidth unknown
HGX module; not a PCIe add-in card
Source checked: 2026-09-06
Primary source All captured specifications
GPU Memory 180 GB HBM3e
Form Factor HGX module; not a PCIe add-in card
Interconnect NVLink; OEM platform required Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system AMD
MI300X 192 GiB
CDNA3 · Power unknown · Bandwidth unknown
Platform-specific accelerator; verify OEM integration
Source checked: 2026-09-06
Primary source All captured specifications
Name MI300X
Architecture CDNA3
LLVM target name gfx942
VRAM (GiB) 192
Compute Units 304 (38 per XCD)
Wavefront Size 64
LDS (KiB) 64
L3 Cache (MiB) 256
L2 Cache (MiB) 32 (4 per XCD)
L1 Vector Cache (KiB) 32
L1 Scalar Cache (KiB) 16 per 2 CUs
L1 Instruction Cache (KiB) 64 per 2 CUs
VGPR File (KiB) 512
SGPR File (KiB) 12.5
GFXIP Major version 9
GFXIP Minor version 4 Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system AMD
MI325X 256 GiB
CDNA3 · Power unknown · Bandwidth unknown
Platform-specific accelerator; verify OEM integration
Source checked: 2026-09-06
Primary source All captured specifications
Name MI325X
Architecture CDNA3
LLVM target name gfx942
VRAM (GiB) 256
Compute Units 304 (38 per XCD)
Wavefront Size 64
LDS (KiB) 64
L3 Cache (MiB) 256
L2 Cache (MiB) 32 (4 per XCD)
L1 Vector Cache (KiB) 32
L1 Scalar Cache (KiB) 16 per 2 CUs
L1 Instruction Cache (KiB) 64 per 2 CUs
VGPR File (KiB) 512
SGPR File (KiB) 12.5
GFXIP Major version 9
GFXIP Minor version 4 Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system NVIDIA
B300 (DGX) 288 GB HBM3e
Blackwell Ultra · Power unknown · Bandwidth unknown
DGX module; not a PCIe add-in card
Source checked: 2026-09-06
Primary source All captured specifications
GPU Memory 288 GB HBM3e
Form Factor DGX module; not a PCIe add-in card
Interconnect NVLink; OEM platform required Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system AMD
MI355X 288 GiB
CDNA4 · Power unknown · Bandwidth unknown
Platform-specific accelerator; verify OEM integration
Source checked: 2026-09-06
Primary source All captured specifications
Name MI355X
Architecture CDNA4
LLVM target name gfx950
VRAM (GiB) 288
Compute Units 256 (32 per XCD)
Wavefront Size 64
LDS (KiB) 160
L3 Cache (MiB) 256
L2 Cache (MiB) 32 (4 per XCD)
L1 Vector Cache (KiB) 32
L1 Scalar Cache (KiB) 16 per 2 CUs
L1 Instruction Cache (KiB) 64 per 2 CUs
VGPR File (KiB) 512
SGPR File (KiB) 12.5
GFXIP Major version 9
GFXIP Minor version 5 Published peak figures are not model throughput. Dimensions and connectors depend on the exact board variant.
Describe my system GB labels follow manufacturer terminology. Calculations use bytes and GiB; shared and dedicated GPU memory are not interchangeable.
How to plan AI hardware How much VRAM does an LLM need? ↗ Estimate LLM memory from weights, quantization, KV cache, context and concurrency. An 8B example and the limits of a preliminary Monsder calculation.
How to choose an AI server ↗ Choose an AI server from workload requirements: model, load, GPU, memory, power and cooling. What a calculator estimates and what needs engineering validation.
Which models fit your GPU? ↗ Match an LLM to your GPU using available VRAM, quantization, context and concurrency. Why fitting in memory does not establish a validated deployment.