AMD Radeon Pro 570X vs NVIDIA P104-100 Comparison

AMD
RADEON

AMD Radeon Pro 570X

CORE STATE Ellesmere
VRAM 4 GB
CLOCK SPEED 1105 MHz
TDP 150 W
BUS WIDTH 256 bit
ARCHITECTURE GCN 4.0
nm
PROCESS 14 nm
LAUNCH DATE 2019
VS
NVIDIA
GEFORCE

P104-100

CORE STATE GP104
VRAM 4 GB
CLOCK SPEED 1733 MHz
TDP
BUS WIDTH 256 bit
ARCHITECTURE Pascal
nm
PROCESS 16 nm
LAUNCH DATE 2017

PERFORMANCE BENCHMARKS

geekbench_metal
40,066
N/A
geekbench_opencl
27,604
52,368
geekbench_vulkan
28,859
45,165
3dmark_3dmark_steel_nomad_dx12
N/A
1,413

Analysis: AMD Radeon Pro 570X vs NVIDIA P104-100

The NVIDIA P104-100 and AMD Radeon Pro 570X are both end-of-life GPUs, but they occupy very different positions in the benchmark hierarchy. The data shows a clear performance gap: the NVIDIA part wins both head-to-head tests, with an average benchmark score of 32,982 against 32,176 for the AMD part. While the overall averages are close, the specific workload results tell a story of divergent architectural priorities, with the P104-100 dominating in compute-heavy APIs while the Radeon Pro 570X remains competitive in its niche.

Head-to-Head Benchmarks

The most striking result in the head-to-head comparison comes from Geekbench OpenCL, where the NVIDIA P104-100 scores 52,368 against the Radeon Pro 570X’s 27,604. That is a 89.7% delta, meaning the P104-100 delivers nearly double the raw compute throughput in this cross-vendor API test. The magnitude of this win is substantial, not a marginal edge — it indicates that in OpenCL workloads, the P104-100 is in a completely different performance class.

The Vulkan results narrow the gap but still favor NVIDIA decisively. The P104-100 posts 45,165 while the Radeon Pro 570X manages 28,859, a 56.5% difference. Both tests show the same winner, but the smaller delta in Vulkan suggests the AMD architecture is relatively more competitive in graphics-oriented APIs than in pure compute scenarios. Still, a 56.5% lead is not a close contest.

Looking at the overall averages, the P104-100’s 32,982 sits just 0.4% above the NVIDIA T600 Mobile’s 32,849 and 0.5% below the T550 Mobile’s 33,161. The Radeon Pro 570X’s 32,176 is 0.9% above the Intel Arc Pro A30M’s 31,894 and 0.7% below the AMD FirePro S10000’s 32,388. Both cards slot into the same general performance tier, yet the head-to-head deltas reveal that the P104-100’s strength is concentrated in specific APIs rather than spread evenly.

The P104-100 also has a third benchmark score not present in the head-to-head set: 1,413 in 3DMark Steel Nomad DX12. This result, combined with its Geekbench scores, yields a 77th percentile ranking among all GPUs. The Radeon Pro 570X’s 76th percentile is nearly identical, but its benchmark portfolio is different — it includes a Geekbench Metal score of 40,066, which is not available for the NVIDIA card.

Architecture Differences

The underlying silicon explains much of the performance split. The NVIDIA P104-100 uses the GP104 chip on a 16 nm TSMC process, packing 7,200 million transistors into a 314 mm² die. The AMD Radeon Pro 570X uses the Ellesmere chip on a 14 nm GlobalFoundries process, with 5,700 million transistors on a 232 mm² die. The transistor density tells an interesting story: AMD achieves 24.6M transistors per mm² versus NVIDIA’s 22.9M, meaning the AMD die is more densely packed despite having fewer total transistors.

Clock speeds differ significantly. The P104-100 runs at a 1607 MHz base and 1733 MHz boost, while the Radeon Pro 570X operates at 1000 MHz base and 1105 MHz boost. This 600+ MHz boost advantage for NVIDIA is a major contributor to its compute lead. However, the AMD card’s TDP is listed at 150 W, while the P104-100 has no TDP figure in the data — instead, it shows a suggested PSU of 200 W and requires a single 8-pin power connector, whereas the Radeon Pro 570X uses no external power connectors and is classified as an IGP (integrated graphics processor).

Memory configurations are similar in capacity but different in type and speed. Both have 4 GB on a 256-bit bus, but the P104-100 uses GDDR5X at 10 Gbps effective, yielding 320.3 GB/s of bandwidth. The Radeon Pro 570X uses GDDR5 at 6.8 Gbps effective, producing 217.0 GB/s. That is a 47.6% bandwidth advantage for NVIDIA, which directly impacts memory-bound workloads.

The compute unit counts favor NVIDIA as well. The P104-100 has 1,920 shading units, 120 TMUs, and 64 ROPs, against the Radeon Pro 570X’s 1,792 shading units, 112 TMUs, and 32 ROPs. The ROP difference is particularly stark — 64 versus 32 — which explains the pixel rate gap: 110.9 GPixel/s for NVIDIA versus 35.36 GPixel/s for AMD. Texture rates also favor NVIDIA at 208.0 GTexel/s versus 123.8 GTexel/s.

FP32 throughput reinforces the pattern: the P104-100 delivers 6.655 TFLOPS against the Radeon Pro 570X’s 3.960 TFLOPS. However, FP16 performance flips the script. The AMD card achieves 3.960 TFLOPS in FP16 (a 1:1 ratio with FP32), while the NVIDIA card manages only 104.0 GFLOPS (a 1:64 ratio). This makes the Radeon Pro 570X dramatically better at half-precision workloads.

API support differs in edge cases. Both support DirectX 12 and OpenGL 4.6, but NVIDIA lists Vulkan 1.4 while AMD lists Vulkan 1.3. NVIDIA’s DirectX support is 12_1, while AMD’s is 12_0. The P104-100 has no display outputs, making it a pure compute card, while the Radeon Pro 570X’s outputs are "Portable Device Dependent," reflecting its intended use in Mac systems. The bus interface also differs: PCIe 1.0 x4 for NVIDIA versus PCIe 3.0 x16 for AMD, which could bottleneck the P104-100 in some data-transfer scenarios.

Where Each One Wins

The NVIDIA P104-100 wins decisively in OpenCL and Vulkan compute workloads, with margins of 89.7% and 56.5% respectively. Its higher clock speeds, larger ROP count, and greater memory bandwidth make it the clear choice for any application that stresses FP32 compute, pixel fill, or texture throughput. The 3DMark Steel Nomad DX12 score of 1,413, while not directly comparable to the AMD card, further indicates strong performance in modern graphics APIs.

The AMD Radeon Pro 570X wins where NVIDIA is weak or absent. Its FP16 performance is effectively 38 times higher in raw throughput when normalized (3.960 TFLOPS versus 0.104 TFLOPS), making it suitable for workloads that leverage half-precision arithmetic. The Metal benchmark score of 40,066 is available only for the AMD card, suggesting it has a role in Apple-centric environments. The lower TDP of 150 W and lack of external power connectors also make it more suitable for compact or integrated systems.

In terms of portability, the Radeon Pro 570X’s PCIe 3.0 x16 interface is far more modern than the P104-100’s PCIe 1.0 x4, which could be a significant advantage for moving data to and from the GPU. The P104-100’s dual-slot form factor and 267 mm length also make it physically larger, while the AMD card is an IGP with no specified dimensions.

The Verdict

Based strictly on the benchmark data, the NVIDIA P104-100 is the superior choice for anyone prioritizing raw compute performance in OpenCL or Vulkan. Its 89.7% lead in OpenCL and 56.5% lead in Vulkan are not marginal — they represent a generational gap in throughput for those APIs. The higher FP32 rate, greater pixel and texture rates, and higher memory bandwidth all point to a card that is simply faster in conventional GPU workloads.

The AMD Radeon Pro 570X is the appropriate pick for scenarios where FP16 performance matters, where Metal API support is required, or where power and physical constraints dominate. Its 150 W TDP and IGP form factor make it adaptable to systems where a dual-slot, 8-pin-powered card would not fit. The PCIe 3.0 x16 interface is also a practical advantage over the P104-100’s PCIe 1.0 x4.

The percentile rankings are nearly identical — 77th for NVIDIA versus 76th for AMD — but the average scores (32,982 versus 32,176) slightly favor the P104-100. The nearest rivals for each card reinforce the parity: the P104-100 trades places with the T600 Mobile and T550 Mobile within 0.5%, while the Radeon Pro 570X sits within 1.1% of the FirePro S10000 and RX 7900 GRE. This suggests that in aggregate, these are comparable GPUs, but the P104-100’s wins are concentrated in the tests that matter most for compute, while the AMD card’s strengths are more specialized.

The data does not support a conclusion that one card is universally better. The P104-100 wins the head-to-head tests 2-0, but the Radeon Pro 570X offers capabilities the NVIDIA card simply does not have, particularly in FP16 and Metal. For a general-purpose compute card, the P104-100 is the stronger pick. For a Mac-compatible, low-power, half-precision-capable GPU, the Radeon Pro 570X has clear advantages.

FAQ

Q: How much faster is the NVIDIA P104-100 than the AMD Radeon Pro 570X in OpenCL?

A: The P104-100 scores 52,368 in Geekbench OpenCL versus 27,604 for the Radeon Pro 570X, a delta of 89.7%.

Q: Does the AMD Radeon Pro 570X win any head-to-head benchmark?

A: No. In the two head-to-head tests (Geekbench OpenCL and Geekbench Vulkan), the NVIDIA P104-100 wins both. The AMD card has a Geekbench Metal score of 40,066, but no Metal result is available for the P104-100.

Q: What is the memory bandwidth difference between the two cards?

A: The P104-100 has 320.3 GB/s bandwidth using GDDR5X memory, while the Radeon Pro 570X has 217.0 GB/s using GDDR5 memory. Both have 4 GB capacity on a 256-bit bus.

Q: Which card has better FP16 performance?

A: The AMD Radeon Pro 570X achieves 3.960 TFLOPS in FP16 (1:1 ratio with FP32), while the NVIDIA P104-100 achieves only 104.0 GFLOPS (1:64 ratio). The AMD card is significantly better for half-precision workloads.

Q: What are the physical power requirements for each card?

A: The NVIDIA P104-100 requires a 1x 8-pin power connector and a suggested PSU of 200 W, with a dual-slot form factor. The AMD Radeon Pro 570X has a 150 W TDP, uses no external power connectors, and is classified as an IGP.

Q: How do these cards compare to their nearest rivals in average benchmark score?

A: The P104-100’s average score of 32,982 is 0.4% above the NVIDIA T600 Mobile and 0.5% below the T550 Mobile. The Radeon Pro 570X’s average of 32,176 is 0.9% above the Intel Arc Pro A30M and 0.7% below the AMD FirePro S10000.

DETAILED SPECIFICATIONS

SPECIFICATION
Pro 570X
P104-100
Core Specs
Shading Units
1,792
1,920 +7.1%
Shaders
1,792
1,920 +7.1%
TMUs
112
120 +7.1%
ROPs
32
64 +100.0%
Compute Units
28
SM Count
15
Clocks
Base Clock
1000 MHz
1607 MHz
Boost Clock
1105 MHz
1733 MHz
Memory Clock
1695 MHz 6.8 Gbps effective
1251 MHz 10 Gbps effective
Memory
Memory Size
4 GB
4 GB
VRAM (MB)
4,096
4,096 0.0%
Memory Type
GDDR5
GDDR5X
Memory Bus
256 bit
256 bit
Bandwidth
217.0 GB/s
320.3 GB/s
Cache
L1 Cache
16 KB (per CU)
48 KB (per SM)
L2 Cache
2 MB
2 MB
Performance
Pixel Rate
35.36 GPixel/s
110.9 GPixel/s
Texture Rate
123.8 GTexel/s
208.0 GTexel/s
FP32 (TFLOPS)
3.960 TFLOPS
6.655 TFLOPS
FP64 (TFLOPS)
247.5 GFLOPS (1:16)
208.0 GFLOPS (1:32)
FP16 (TFLOPS)
3.960 TFLOPS (1:1)
104.0 GFLOPS (1:64)
Power
TDP
150 W
TDP (W)
150
Suggested PSU
200 W
Power Connectors
None
1x 8-pin
Architecture
Architecture
GCN 4.0
Pascal
GPU Name
Ellesmere
GP104
Generation
Radeon Pro Mac (500X Series)
Mining GPUs
Process Size
14 nm
16 nm
Transistors
5,700 million
7,200 million
Die Size
232 mm²
314 mm²
Foundry
GlobalFoundries
TSMC
Density
24.6M / mm²
22.9M / mm²
API Support
DirectX
12 (12_0)
12 (12_1)
OpenGL
4.6
4.6
Vulkan
1.3
1.4
OpenCL
2.1
3.0
CUDA
6.1
Shader Model
6.7
6.8
Physical
Slot Width
IGP
Dual-slot
Length
267 mm 10.5 inches
Outputs
Portable Device Dependent
No outputs
Bus Interface
PCIe 3.0 x16
PCIe 1.0 x4
Other
Production
End-of-life
End-of-life
View Radeon Pro 570X Details View P104-100 Details