GTX 1080 video cards

appeared on the market in 2016, but have not lost their relevance in seven years. On the contrary, they secured the status of a reliable railway, like all tenth generation GPUs.

Back then, even no one could believe that a GPU with a TDP of 180 W would appear. Low power consumption has hardened the video card, so even mining scars are not visible on it. But how well does the GTX 1080 meet today’s demands in professional tasks? Has it managed to stay relevant in machine learning? Briefly understand the article.

Use the navigation if you don’t want to read the entire text:

→ Why did they like 1080
→ We will conduct testing
→ We analyze benchmarks
→ Conclusions

Why did they like 1080

NVIDIA GeForce GTX 1080. Source.

The GTX 1080 was ahead of its time in terms of specs. Only the RTX 2080 (2944) will be able to surpass it in terms of the number of CUDA cores, but the video card will cost 2.5 times more.

The characteristics of the video card still sound relevant. After all, users are purchasing iron that was once far ahead of its time.

Video card on board

• Architecture: Pascal (GP104).
• CUDA cores: 2560.
• Core frequency: 1607 (1733) MHz.
• Memory: 8 GB GDDR5X.
• Bus: 256 bits.
• Bandwidth: 320 GB/s.

For most games, the GTX 1080 is still enough. But how well does the graphics card cope with work calculations?

We will conduct testing

Let’s take three ready-made configurations

cloud servers

with different video cards and run through GeekBench and AI-Benchmark.

Setting up cloud servers for testing.

1. Go to the Cloud Platform section inside the control panel.

2. We create a server. First, let’s select the configuration in the ru-7 pool: GTX 1080 8 GB, RAM 24 GB, 8 vCPU. For example: such a configuration will cost only about 24 ₽/hour. You can find out the exact price at

configurator

cloud servers.

3. Install

GeekBench

and

AI-Benchmark

and start testing. You can learn about how to do this in

Primate Labs documentation

and

descriptions in PyPi

in accordance.

We analyze benchmarks

AI-Benchmark

You can get acquainted with the test results at

exiled

. If you put everything together, you get the following picture:

Configuration comparison results, AI-Benchmark.

In terms of performance in machine learning, the GTX 1080 performed better than the Tesla T4, but worse than the A2000. Let’s see what results the video cards of other tasks will show.

GeekBench

The results of configuration testing can be viewed using online downloads for

Testa T4

,

GTX 1080

and

A2000

. We select the values ​​and collect them in a summary table:

GeekBench configuration comparison results.

A description of all parameters is available at the link. We highlighted the main points and concluded: in some benchmarks, the GTX 1080 performed at the T4 level and better than the A2000. This can be seen in the Face Detection and Gaussian Blur tasks.

Conclusions

For initial tasks, it is not necessary to take top new video cards. Try to implement your project for the year 1080.

What do you think about this graphics card? Share your experience and opinion in the comments!