General Discussions

[MikeIAComutation]

Hello,

I am using the AMD Threadripper 1950X 64GB RAM to train a LSTM architecture using Tensorflow ( > 2.0). The CPU is working fine. All benchmarks show optimal performance. However, when I am trying to speed up the training process, I get no better results than with any other CPU. 

I am only using about 5 % of the CPU power. All Cores seem to to something but not much.

I have already been in contact with AMD but they could not help me.

I tried:

import tensorflow as tf
tf.config.threading.set_inter_op_parallelism_threads()
tf.config.threading.set_intra_op_parallelism_threads()
tf.config.set_soft_device_placement(enabled)

But no succsess.

Can anybody help me?

Thank you so much in advance

Mike

[WhyteM]

---

@MikeIAComutation wrote:

Hello,

I am using the AMD Threadripper 1950X 64GB RAM to train a LSTM architecture using Tensorflow ( > 2.0). The CPU is working fine. All benchmarks show optimal performance. However, when I am trying to speed up the training process, I get no better results than with any other CPU. 

I am only using about 5 % of the CPU power. All Cores seem to to something but not much.

I have already been in contact with AMD but they could not help me.

I tried:

import tensorflow as tf
tf.config.threading.set_inter_op_parallelism_threads()
tf.config.threading.set_intra_op_parallelism_threads()
tf.config.set_soft_device_placement(enabled)

But no succsess.

Can anybody help me?

Thank you so much in advance

Mike

---

See here: https://www.tensorflow.org/guide/gpu_performance_analysis

[MikeIAComutation]

Thank you very much. However, the AMD Threadripper 1950X 64GB RAM is a CPU not a GPU. I am running LSTM architectures, which are considered to perform better with a CPU.