geekradar
Int
Explore GeekRadar

More to explore

Good gear All topics About us Say hello
GeekRadar / The comparison desk

COMPARE.

GPUs
Filters

Filter GPUs

Brand
Benchmark coverage
Clear all filters

Best result per entry; configurations may differ.

Type
17 results Benchmark score
NVIDIA

NVIDIA H100 SXM5 96 GB

GPU*
Video memory
96 GB*
Rated memory bandwidth
3,360 GB/s*
Thermal design power
700 W*
GPU cores
16,896*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#1* · 280.74* tokens/sPlatform unreported
No Blender rendering score
NVIDIA

NVIDIA GeForce RTX 5090

GPU
Video memory
32 GB
Rated memory bandwidth
1,792 GB/s
Thermal design power
575 W
GPU cores
21,760*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#2* · 273.68* tokens/sPlatform unreported
Blender 5.1.1 / Windows / OPTIX16,188.89 samples/minRendering · median · 198 complete runs
AMD

Radeon Instinct MI300X

GPU*
Video memory
192 GB*
Rated memory bandwidth
5,320 GB/s*
Thermal design power
750 W*
GPU cores
19,456*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#3* · 232.92* tokens/sPlatform unreported
No Blender rendering score
NVIDIA

NVIDIA GeForce RTX 4090 D

GPU
Video memory
24 GB
Rated memory bandwidth
1,008 GB/s
Thermal design power
425 W
GPU cores
14,592*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#4* · 191.57* tokens/sPlatform unreported
Blender 4.5.0 / Windows / OPTIX10,724.02 samples/minRendering · median · 4 complete runs
NVIDIA

NVIDIA GeForce RTX 4090

GPU
Video memory
24 GB
Rated memory bandwidth
1,008 GB/s
Thermal design power
450 W
GPU cores
16,384*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#5* · 188.96* tokens/sPlatform unreported
Blender 5.1.1 / Windows / OPTIX11,480.17 samples/minRendering · median · 103 complete runs
NVIDIA

NVIDIA RTX PRO 6000 Blackwell

GPU*
Video memory
96 GB*
Rated memory bandwidth
1,790 GB/s*
Thermal design power
600 W*
GPU cores
24,064*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#6* · 104.38* tokens/sPlatform unreported
No Blender rendering score
AMD

AMD Radeon RX 6800

GPU
Video memory
16 GB
Rated memory bandwidth
512 GB/s
Thermal design power
250 W*
GPU cores
3,840*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#7* · 96.65* tokens/sPlatform unreported
Blender 5.1.1 / Linux / HIP1,664.53 samples/minRendering · median · 12 complete runs
NVIDIA

NVIDIA Quadro P620

GPU
Video memory
2 GB
Rated memory bandwidth
64 GB/s
Thermal design power
40 W
GPU cores
512*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#8* · 74.66* tokens/sPlatform unreported
Blender 4.5.0 / Windows / OPTIX81.32 samples/minRendering · median · 4 complete runs
AMD

Radeon Instinct MI250

GPU*
Video memory
128 GB*
Thermal design power
500 W*
Memory options
128*
Architecture
CDNA 2.0*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#9* · 63.94* tokens/sPlatform unreported
No Blender rendering score
AMD

AMD Radeon RX Vega 64

GPU
Video memory
8 GB
Rated memory bandwidth
483.80 GB/s
Thermal design power
295 W*
GPU cores
4,096*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#10* · 47.06* tokens/sPlatform unreported
Blender 3.5.0 / macOS / METAL761.26 samples/minRendering · median · 6 complete runs
NVIDIA

NVIDIA Tesla T4

GPU
Video memory
16 GB*
Rated memory bandwidth
320 GB/s*
Thermal design power
70 W*
GPU cores
2,560*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#11* · 46.38* tokens/sPlatform unreported
Blender 4.3.0 / Windows / OPTIX1,301.55 samples/minRendering · median · 4 complete runs
NVIDIA

NVIDIA Tesla M40

GPU
Video memory
12 GB*
Rated memory bandwidth
288.40 GB/s*
Thermal design power
250 W*
GPU cores
3,072*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#12* · 41.68* tokens/sPlatform unreported
Blender 4.0.0 / Linux / OPTIX515.25 samples/minRendering · median · 5 complete runs
AMD

Radeon Instinct MI325X

GPU*
Video memory
Reported value 1: 288 GB*
Reported value 2: 256 GB*
Reported value 3: 256 GB*
Rated memory bandwidth
6,140 GB/s*
Thermal design power
1,000 W*
GPU cores
19,456*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#13* · 22.41* tokens/sPlatform unreported
No Blender rendering score
NVIDIA

Tesla X2070

GPU*
Video memory
6 GB*
Rated memory bandwidth
177.40 GB/s*
Thermal design power
225 W*
GPU cores
448*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#14* · 14.27* tokens/sPlatform unreported
No Blender rendering score
PNY

PNY Quadro P1000 V2

GPU*
Video memory
4 GB*
Rated memory bandwidth
80.19 GB/s*
Thermal design power
47 W*
GPU cores
640*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#15* · 13.99* tokens/sPlatform unreported
No Blender rendering score
NVIDIA

NVIDIA Tesla K80

GPU
Thermal design power
300 W*
Architecture
Kepler
Manufacturing process
28 nm
Process node
28 nm
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#16* · 9.49* tokens/sPlatform unreported
Blender 4.4.0 / Linux / CUDA225.30 samples/minRendering · median · 3 complete runs
INTEL

Intel Data Center GPU Max NEXT

GPU*
Video memory
128 GB*
Rated memory bandwidth
3,210 GB/s*
Thermal design power
800 W*
GPU cores
20,480*
LLM Inference: llama.cpp Bench: llama-2-7b-Q4_0 (Q4_0) · version unreported · score#17* · 8.21* tokens/sPlatform unreported
No Blender rendering score

* Provisional data