NVIDIA CMP 90HX vs NVIDIA Tesla P100 PCIe 12 GB Comparison

NVIDIA
GEFORCE

NVIDIA CMP 90HX

CORE STATE GA102
VRAM 10 GB
CLOCK SPEED 1710 MHz
TDP 320 W
BUS WIDTH 320 bit
ARCHITECTURE Ampere
nm
PROCESS 8 nm
LAUNCH DATE 2021
VS
NVIDIA
GEFORCE

Tesla P100 PCIe 12 GB

CORE STATE GP100
VRAM 12 GB
CLOCK SPEED 1329 MHz
TDP 250 W
BUS WIDTH 3072 bit
ARCHITECTURE Pascal
nm
PROCESS 16 nm
LAUNCH DATE 2016

PERFORMANCE BENCHMARKS

geekbench_opencl
69,000
79,396

Analysis: NVIDIA CMP 90HX vs NVIDIA Tesla P100 PCIe 12 GB

The NVIDIA Tesla P100 PCIe 12 GB and the NVIDIA CMP 90HX are two very different products from NVIDIA’s lineup, separated by five years of architectural evolution and designed for entirely different purposes. The Tesla P100 is a compute-focused accelerator built on the Pascal architecture, while the CMP 90HX is a mining-specific card based on Ampere. The data shows a clear winner in raw compute performance, but the specifications tell a more nuanced story about what each card is actually good for. In the single available benchmark — Geekbench OpenCL — the Tesla P100 scores 79,396 points, which is 15.1% higher than the CMP 90HX’s 69,000 points. This is a significant margin, placing the older card in a different performance class despite its age.

Head-to-Head Benchmarks

The only direct benchmark comparison available is Geekbench OpenCL, and the results are decisive. The Tesla P100 PCIe 12 GB scores 79,396 points, while the CMP 90HX scores 69,000 points. That represents a 15.1% lead for the Tesla P100. In practical terms, this means the Pascal-based card delivers substantially higher compute throughput in OpenCL workloads, which are common in scientific computing, machine learning inference, and rendering tasks that rely on GPU acceleration.

Looking at the nearest rival data reinforces this picture. The Tesla P100 sits at the 92nd percentile of all GPUs, with its closest competitors being the Tesla P100 16 GB (79,605 points, only 0.3% ahead), the GeForce RTX 5090 (79,842 points, 0.6% ahead), and the AMD Radeon RX 6850M XT (78,940 points, 0.6% behind). The CMP 90HX, by contrast, sits at the 90th percentile, with its nearest rivals being the Intel Arc A770 (68,809 points, 0.3% ahead), AMD Radeon Instinct MI25 (68,562 points, 0.6% behind), and AMD Radeon Pro WX 8200 (69,870 points, 1.2% behind). The percentile gap is narrow, but the absolute score difference is meaningful — the Tesla P100 is roughly 15% faster in this test.

The deltaPct values tell an interesting story about performance clustering. The Tesla P100’s nearest rivals are all within 1.9% of its score, indicating it sits in a tightly packed performance band. The CMP 90HX’s rivals are similarly close, with deltas ranging from -1.4% to 0.6%. However, the two cards themselves are separated by a much larger gap. This suggests that the CMP 90HX is competing in a lower performance tier, while the Tesla P100 punches above its weight relative to much newer hardware.

The Verdict

From the data, the NVIDIA Tesla P100 PCIe 12 GB is the clear winner for compute performance. It beats the CMP 90HX by 15.1% in Geekbench OpenCL, and its 92nd percentile ranking places it above the CMP 90HX’s 90th percentile. The Tesla P100 also has a higher launch MSRP of 4,599 USD, reflecting its professional-grade positioning. The CMP 90HX, which has no launch MSRP listed, was designed for a single purpose — cryptocurrency mining — and its specifications reflect that narrow focus.

Who should pick which? If your workload is OpenCL-based compute, scientific simulation, or any task that benefits from raw FP32 throughput, the Tesla P100 is the obvious choice. It offers more shading units, a wider memory bus, and higher bandwidth per byte of memory. The CMP 90HX, despite its newer architecture and higher clock speeds, simply does not translate those advantages into OpenCL performance. The data shows the Tesla P100 wins the only benchmark available, and that is what matters.

However, if you are looking at this from a pure specification standpoint, the CMP 90HX has some advantages that are not reflected in the benchmark. Its 21.89 TFLOPS FP32 throughput is more than double the Tesla P100’s 9.526 TFLOPS, and its memory bandwidth is 760.3 GB/s versus 549.1 GB/s. But the benchmark does not support these specifications translating into real-world OpenCL wins. The verdict is straightforward: for measured performance, the Tesla P100 wins.

Where Each One Wins

The Tesla P100 PCIe 12 GB wins in the only benchmark where both are measured. Its 79,396 OpenCL score is 15.1% higher than the CMP 90HX’s 69,000. It also has a higher memory bus width (3072-bit versus 320-bit), which can be advantageous for certain memory-bound workloads, and more ROPs (96 versus 80) for pixel processing. Its 12 GB of HBM2 memory, while smaller in capacity than some alternatives, offers a different access pattern than the CMP 90HX’s 10 GB of GDDR6X.

The CMP 90HX, on the other hand, wins on raw specification sheets. Its 10 GB of GDDR6X memory runs at 760.3 GB/s, which is 38.4% higher bandwidth than the Tesla P100’s 549.1 GB/s. It has more shading units (6400 versus 3584), more TMUs (200 versus 224 is actually a loss, but the CMP has more shading units), and it supports DirectX 12 Ultimate (12_2) versus the Tesla P100’s DirectX 12 (12_1). The CMP 90HX also has a higher base clock (1500 MHz versus 1190 MHz) and boost clock (1710 MHz versus 1329 MHz). These specifications suggest the CMP 90HX could excel in tasks that are not well-represented by Geekbench OpenCL, such as certain gaming-influenced workloads or tasks that heavily utilize its tensor cores (200 of them) and RT cores (50 of them).

FAQ

Q: Which card is faster in Geekbench OpenCL?

A: The NVIDIA Tesla P100 PCIe 12 GB scores 79,396 points, which is 15.1% higher than the CMP 90HX’s 69,000 points.

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

A: The CMP 90HX has 760.3 GB/s of memory bandwidth, while the Tesla P100 has 549.1 GB/s. The CMP 90HX offers 38.4% more bandwidth.

Q: Which card has a higher FP32 compute throughput?

A: The CMP 90HX has 21.89 TFLOPS of FP32 throughput, while the Tesla P100 has 9.526 TFLOPS. The CMP 90HX is over twice as high on this specification.

Q: How do the two cards compare in terms of percentile ranking?

A: The Tesla P100 is in the 92nd percentile of all GPUs, while the CMP 90HX is in the 90th percentile.

Q: Do both cards have display outputs?

A: No. Both the Tesla P100 PCIe 12 GB and the CMP 90HX have no display outputs.

Q: What is the transistor count and die size difference?

A: The CMP 90HX has 28,300 million transistors on a 628 mm² die, while the Tesla P100 has 15,300 million transistors on a 610 mm² die.

Architecture Differences

The two cards are built on fundamentally different architectures. The Tesla P100 uses the GP100 chip on the Pascal architecture, fabricated on a 16 nm process at TSMC. The CMP 90HX uses the GA102 chip on the Ampere architecture, fabricated on an 8 nm process at Samsung. This process shrink is significant — the CMP 90HX packs 28,300 million transistors into a 628 mm² die, yielding a transistor density of 45.1M per mm². The Tesla P100, by contrast, has 15,300 million transistors on a 610 mm² die, for a density of 25.1M per mm². The newer process node nearly doubles the transistor density.

Architecturally, the CMP 90HX brings features that the Pascal-based Tesla P100 simply does not have. The CMP 90HX includes 50 RT cores and 200 tensor cores, which are absent from the Tesla P100’s specification sheet. The Tesla P100 has no RT or tensor core support at all. The CMP 90HX also supports DirectX 12 Ultimate (12_2) and Vulkan 1.4, while the Tesla P100 supports DirectX 12 (12_1) and Vulkan 1.3. This means the CMP 90HX is compatible with more modern graphics APIs, although neither card has display outputs.

The memory architectures also differ fundamentally. The Tesla P100 uses 12 GB of HBM2 with a 3072-bit bus, while the CMP 90HX uses 10 GB of GDDR6X with a 320-bit bus. The HBM2 implementation provides a much wider interface, while GDDR6X runs at higher effective speeds (19 Gbps versus 1430 Mbps effective). This results in the CMP 90HX having higher bandwidth (760.3 GB/s versus 549.1 GB/s) despite the narrower bus.

Specification Differences

The two cards differ across nearly every major specification. The Tesla P100 has 3584 shading units, 224 TMUs, and 96 ROPs. The CMP 90HX has 6400 shading units, 200 TMUs, and 80 ROPs. The CMP 90HX has significantly more shading units (78.6% more), but fewer TMUs and ROPs. The CMP 90HX also has 50 RT cores and 200 tensor cores, which the Tesla P100 lacks entirely.

Clock speeds favor the CMP 90HX. Its base clock is 1500 MHz and boost clock is 1710 MHz, versus the Tesla P100’s 1190 MHz base and 1329 MHz boost. The CMP 90HX’s memory runs at 1188 MHz (19 Gbps effective), while the Tesla P100’s memory runs at 715 MHz (1430 Mbps effective). This contributes to the CMP 90HX’s higher memory bandwidth.

The TDP difference is notable: the CMP 90HX draws 320 W versus the Tesla P100’s 250 W. The CMP 90HX requires two 8-pin power connectors and a 700 W suggested PSU, while the Tesla P100 needs a single 8-pin connector and a 600 W suggested PSU. The CMP 90HX is also slightly longer at 285 mm (11.2 inches) versus 267 mm (10.5 inches) for the Tesla P100, and it has a specified height of 112 mm (4.4 inches) while the Tesla P100’s height is not listed. The bus interface also differs: the Tesla P100 uses PCIe 3.0 x16, while the CMP 90HX uses PCIe 1.0 x4, which is a much slower interface. The Tesla P100 was released on 2016-06-19, while the CMP 90HX came later on 2021-07-27.

DETAILED SPECIFICATIONS

SPECIFICATION
CMP 90HX
Tesla P100 PCIe 12 GB
Core Specs
Shading Units
6,400
3,584 -44.0%
Shaders
6,400
3,584 -44.0%
TMUs
200
224 +12.0%
ROPs
80
96 +20.0%
SM Count
50
56 +12.0%
Clocks
Base Clock
1500 MHz
1190 MHz
Boost Clock
1710 MHz
1329 MHz
Memory Clock
1188 MHz 19 Gbps effective
715 MHz 1430 Mbps effective
Memory
Memory Size
10 GB
12 GB
VRAM (MB)
10,240
12,288 +20.0%
Memory Type
GDDR6X
HBM2
Memory Bus
320 bit
3072 bit
Bandwidth
760.3 GB/s
549.1 GB/s
Cache
L1 Cache
128 KB (per SM)
24 KB (per SM)
L2 Cache
5 MB
3 MB
Performance
Pixel Rate
136.8 GPixel/s
127.6 GPixel/s
Texture Rate
342.0 GTexel/s
297.7 GTexel/s
FP32 (TFLOPS)
21.89 TFLOPS
9.526 TFLOPS
FP64 (TFLOPS)
342.0 GFLOPS (1:64)
4.763 TFLOPS (1:2)
FP16 (TFLOPS)
21.89 TFLOPS (1:1)
19.05 TFLOPS (2:1)
AI/RT
RT Cores
50
—
Tensor Cores
200
—
Power
TDP
320 W
250 W
TDP (W)
320
250 -21.9%
Suggested PSU
700 W
600 W
Power Connectors
2x 8-pin
1x 8-pin
Architecture
Architecture
Ampere
Pascal
GPU Name
GA102
GP100
Generation
Mining GPUs
Tesla Pascal (Pxx)
Process Size
8 nm
16 nm
Transistors
28,300 million
15,300 million
Die Size
628 mm²
610 mm²
Foundry
Samsung
TSMC
Density
45.1M / mm²
25.1M / mm²
API Support
DirectX
12 Ultimate (12_2)
12 (12_1)
OpenGL
4.6
4.6
Vulkan
1.4
1.3
OpenCL
3.0
3.0
CUDA
8.6
6.0
Shader Model
6.8
6.0
Physical
Slot Width
Dual-slot
Dual-slot
Length
285 mm 11.2 inches
267 mm 10.5 inches
Height
112 mm 4.4 inches
—
Outputs
No outputs
No outputs
Bus Interface
PCIe 1.0 x4
PCIe 3.0 x16
Other
Launch Price
—
4,599 USD
Production
End-of-life
End-of-life
Predecessor
—
Tesla Maxwell
Successor
—
Tesla Volta
View CMP 90HX Details View Tesla P100 PCIe 12 GB Details