Performance Rating
AMD Radeon Instinct MI250 outperforms NVIDIA Tesla T4 by 690.22% in the overall GPU ARK performance rating
NVIDIA Tesla T4
Contents:
Memory ML Performance Compute Power Architecture & Compatibility ML Software Support Clocks & Performance Power Consumption Rendering Benchmarks AdditionalMemory
Memory Size
+700%
128
ГБ
Memory Type
Memory Bandwidth
3.28 TB/s
Memory Bus Width
ML Performance
FP16 (Half Precision)
+456%
362.1
TFLOPS
BF16 (Brain Float)
TF32 (TensorFloat)
Compute Power
FP32 (Single Precision)
+456%
45.26
TFLOPS
FP64 (Double Precision)
+17,691%
45.26
TFLOPS
CUDA Cores
+420%
13,312
RT Cores
Architecture & Compatibility
GPU Architecture
SM (Streaming Multiprocessor)
PCIe Version
ML Software Support
CUDA Version
Clocks & Performance
Base Clock
+71%
1,000
Boost Clock
+7%
1,700
Memory Clock
+28%
1,600
Power Consumption
TDP/TGP
-86%
70 W
Recommended PSU
-72%
250 W
Power Connector
Rendering
Texture Units (TMU)
+420%
832
ROP
L2 Cache
+300%
16 MB
Benchmarks
LLM Perf, Llama-2-7b-hf (bnb)
LLM Perf, Llama-2-7b-hf (gptq)
LLM Perf, Llama-2-7b-hf (unquantized)
LLM Perf, Llama-3.1-8B-Instruct (awq)
LLM Perf, Llama-3.1-8B-Instruct (bnb)
LLM Perf, Llama-3.1-8B-Instruct (gptq)
LLM Perf, Meta-Llama-3-8B (awq)
LLM Perf, Meta-Llama-3-8B (bnb)
LLM Perf, Meta-Llama-3-8B (gptq)
LLM Perf, Meta-Llama-3-8B-Instruct (awq)
LLM Perf, Meta-Llama-3-8B-Instruct (bnb)
LLM Perf, Meta-Llama-3-8B-Instruct (gptq)
llama.cpp, llama-2-7b-Q4_0
+38%
63.9 tokens/s
LLM Perf, Mistral-7B-Instruct-v0.2 (gptq)
LLM Perf, Mistral-7B-v0.1 (bnb)
LLM Perf, Mistral-7B-v0.1 (gptq)
LLM Perf, Mistral-7B-v0.1 (unquantized)
Additional
Slots
Release Date
Display Outputs
Renting is cheaper than buying
H100 80Gb SXM (80GB)
NVIDIA
Флагманская GPU на архитектуре Hopper с 80 ГБ памяти HBM3 в форм-факторе SXM. Создана для обучения и инференса больших языковых моделей, HPC задач. Поддерживает FP64, FP32, TF32, FP16, BF16, FP8 и INT8 вычисления с Tensor Cores 4-го поколения.
from 299.11 RUB/hour
