Can a GTX 1650 run AI? Measured: about 20 tokens a second
Yes. On a desktop with an NVIDIA GTX 1650 (4 GB) and 16 GB of RAM, Gemma 4 E4B wrote about 20 tokens a second. You read about 5, so answers come out faster than you can read them.
What that means in practice
| Task | Time on this PC |
|---|---|
| Write a normal answer (400 tokens, about 300 words) | 20 s |
| Read a long document you give it (5,000 tokens, about 7 pages) | 0:13–3:28 min |
Worked out from the speeds measured on this PC: time is length divided by speed, and a real answer also spends a moment getting started. For scale, most people read about 4 words a second (Brysbaert, 2019), so a 400-token answer takes you over a minute just to read.
What we measured
Writing is how fast the AI types its answer, in tokens a second; a token is about three quarters of an English word. Reading is how fast it takes in what you give it: your question, a pasted text or a PDF. Reading is always much faster than writing, because the AI can read many tokens at once but has to write them one by one.
| What | Ran on | Result | Measured |
|---|---|---|---|
| Gemma 4 E4B | graphics card, overflowing into RAM | about 20 tokens a second writing · 24–395 reading | 23 September 2026 |
| Gemma 4 E4B (second PC)* | graphics card, overflowing into RAM | 8–25 tokens a second writing · 32–301 reading | 1 October 2026 |
How each number was measured:
- Real chats in GreenCube 0.1.40 and 0.1.42: the speed of every answer of 100 tokens or more, read from the app's log. The range leaves out the slowest and the fastest tenth.
Where we have several runs we show the range: speeds move a little between runs on the same PC. * This tester mostly asked short questions, so answers under 100 tokens are counted too. Short answers measure a little slower.
What it means
Not all of it fits on the card
Gemma 4 E4B is a 4.2 GB file and the card has 4 GB, so it does not fit whole and part of it sits in the normal memory (RAM). Every word it writes has to reach that part over a much slower path, so speed drops and moves around. It still writes about 20 tokens a second. Reading what you paste swung between 24 and 395 tokens a second from one day to the next, most likely because the card was full.
A second PC with the same card
Another tester's GTX 1650 ran Gemma 4 E4B in GreenCube 0.1.42: it wrote 8–25 tokens a second, and read at 32–301. Most of its answers were short, and short answers measure slower than long ones: its longest came near the top of that range.
The bigger Gemma 4 12B
Not comfortably. GreenCube suggests the bigger Gemma 4 12B only on cards with 10 GB or more of VRAM, and this one has 4. On an 8 GB AMD RX 5700 XT it answered short questions, then crashed in a long chat.
Pictures
Very likely, but we haven't timed a picture on this card yet. When we do, the number goes here. A picture uses its own AI of about 6 GB and far more maths than a chat answer, so the graphics matter even more here.
How it compares
Of the 8 PCs we measured with Gemma 4 E4B, this one is number 7 for writing speed. The GTX 1660 Super writes 37–38, about 90% faster. The RX 580 writes 10–11, so this one is about 90% faster.
The test PC
A real tester's own PC, used the way they normally use it, not a lab machine:
- Desktop
- Graphics: NVIDIA GTX 1650
- Graphics memory (VRAM): 4 GB
- RAM: 16 GB
- Processor: 16 threads
Check your own PC
Not sure what your PC has? It takes 30 seconds:
- Press Ctrl + Shift + Esc to open Task Manager.
- Go to Performance.
- Click Memory: the number at the top right is your RAM.
- Click GPU: the name at the top right is your card, and Dedicated GPU memory is your VRAM. Only see Intel or AMD graphics with no dedicated memory? Then you have no separate card.
Then find the closest card in the full table. The same card with more VRAM can fit a bigger AI; a newer card of the same size is usually faster.
Other PCs we measured
- RTX 2080 Ti: 65–71 tokens a second (Gemma 4 E4B)
- RX 5700 XT: 63–72 tokens a second (Gemma 4 E4B)
- RTX 5060: 56 tokens a second (Gemma 4 E4B)
- RTX 3060: 53–55 tokens a second (Gemma 4 E4B)
- RX 6600: 42–44 tokens a second (Gemma 4 E4B)
- GTX 1660 Super: 37–38 tokens a second (Gemma 4 E4B)
- Intel Core Ultra 5 laptop: 14–18 tokens a second (Gemma 4 E2B)
- RX 580: 10–11 tokens a second (Gemma 4 E4B)
- Intel Iris Xe laptop: 5–7 tokens a second (Gemma 4 E2B)
Can my PC run AI? All 10 PCs in one table, and what your PC will feel like.
Questions
Can a GTX 1650 run AI locally?
Yes. On a desktop with an NVIDIA GTX 1650 (4 GB) and 16 GB of RAM, Gemma 4 E4B wrote about 20 tokens a second. You read about 5, so answers come out faster than you can read them.
Can a GTX 1650 make AI pictures?
Very likely, but we haven't timed a picture on this card yet. When we do, the number goes here.
Can a GTX 1650 run Gemma 4 12B?
Not comfortably. GreenCube suggests the bigger Gemma 4 12B only on cards with 10 GB or more of VRAM, and this one has 4. On an 8 GB AMD RX 5700 XT it answered short questions, then crashed in a long chat.
Your own AI, on your own PC
GreenCube runs on your computer: no usage limits, no subscription, works without internet. €9.99, paid once.
Get GreenCube