Reading GPU Specs

This lesson shows how to find the generation and architecture of a GPU. It uses the RTX 3090 and the A100 as examples.

Finding GPU Specs

The easiest way is a Google search. For example:

"A100 GPU TechPowerUp"

TechPowerUp is a website that collects detailed GPU specs from many manufacturers. It is one of the easiest places to check GPU details. You can search other GPUs the same way:

"RTX 3090 TechPowerUp"

Open the page to see all the specs.

A Simple Comparison

Compare two GPUs:

  • RTX 3090
  • A100

First, look at the chip name. For example, A100 → GA100.

[!NOTE] Chip design comes in a later lesson. For now, just read the name.

Next, look at the number of cores:

  • A100 → around 7,000 cores
  • RTX 3090 → more than 10,000 cores

This does not mean the RTX 3090 is always stronger, because the core count does not show every kind of core.

Core Counts

A number like "6,912 cores" (the A100) usually counts only single-precision cores. These cores handle standard floating-point math. The number does not include all cores in the GPU.

Modern GPUs have other types of cores too, for example:

  • cores for integer operations
  • cores for double-precision operations
  • special cores for AI (tensor cores)

So do not judge a GPU by this number alone.

Generation and Architecture

RTX 3090

  • Generation → GeForce
  • Architecture → Ampere

GeForce GPUs are built for everyday users in:

  • desktops
  • laptops
  • workstations

Main use cases:

  • gaming
  • content creation
  • general GPU tasks

A100

  • Generation → (historically Tesla, now Data Center GPUs)
  • Architecture → Ampere

These GPUs are built for:

  • servers
  • data centers
  • supercomputers

Key Point

  • RTX 3090 and A100 use the SAME architecture (Ampere)
  • but they are built for completely different use cases

Same architecture ≠ same purpose.

Reminder: - Architecture → technical design - Generation → usage category

Telling Them Apart by Looks

In many cases, you can tell the difference just by looking at the card.

Data Center GPUs (A100, V100, P100)

  • usually NO built-in fan
  • compact, fanless design

They run in data centers with strong external cooling. The server handles the cooling, not the GPU.

GeForce GPUs (RTX series)

  • have built-in fans
  • designed for standalone systems

They run in:

  • desktop PCs
  • personal workstations

These systems need their own cooling, so the card needs fans.

Summary

  • Data Center GPUs → no fan
  • GeForce GPUs → built-in fan

Different environments have different cooling needs. Knowing this helps you:

  • read GPU specs
  • choose the right hardware
  • avoid common beginner mistakes

This becomes more important as you go deeper into CUDA.

Glossary

  • TechPowerUp: a website that collects detailed GPU specs from many manufacturers.
  • chip name: the name of the chip inside a GPU, such as GA100 for the A100.
  • core count: the number of cores in the specs, which does not show every kind of core.
  • single-precision cores: cores for standard floating-point math, usually the only ones in the core count.
  • tensor cores: special cores in modern GPUs built for AI.
  • architecture: the technical design of a GPU.
  • generation: the usage category of a GPU, such as GeForce or Data Center GPUs.
  • Ampere: the architecture shared by the RTX 3090 and the A100.
← previous02 Architecture vs Generation