Comments for Phoenix Game Development

Comment on AI: Memory Bandwidth comparison for selected DDR4 CPU’s by PhoenixGames

PhoenixGames — Tue, 04 Aug 2026 23:08:28 +0000

In reply to Julio. Thank you for your comment, I'm very glad I was able to help!!

Comment on AI: Memory Bandwidth comparison for selected DDR4 CPU’s by Julio

Julio — Tue, 04 Aug 2026 20:08:06 +0000

Thanks a lot for the excellent review!! It helped a lot in my research i will go to with the 3975!!!

Comment on Spiritus Astrum [P0-B10]: p1488: Towards Early Access by Z Image

Z Image — Sat, 03 Jan 2026 11:39:30 +0000

The navmesh detail really caught my attention! It sounds challenging to have NPCs follow commands reliably, especially considering terrain issues

Comment on Spiritus Astrum [P0-B10]: p1488: Towards Early Access by AI ASMR

AI ASMR — Sat, 27 Dec 2025 10:35:13 +0000

I love reading about the complex navmesh generation and NPC commands-a real look into game development! It’s fascinating how these systems evolve over time

Comment on AI: Estimated token generation rates for selected model sizes (and algorithm to calculate same) by PhoenixGames

PhoenixGames — Thu, 30 Oct 2025 16:00:41 +0000

In reply to Name.

Thank you for your reply.

The numbers and algorithms posted above were always rough rules of thumb, and having tested them in real world conditions, they are certainly not entirely accurate, however, they are not entirely inaccurate either.

I was running 120b Q8 models with two 3090s (The rest of the model being on the CPU) and I was getting anything from 1-2 tk/s depending on the context, etc.

The graph above shows 3.43 tk/s for 140 gb/s RAM bandwidth, which is noticeable higher than I was getting in reality, but again, it’s a rough estimate.

The formula I used (memorybandwidth / model size) was based on internet searches, reddit, etc. I admit, not the most reliable sources, but I couldn’t find anything better at the time. Can you tell me where you got yours?

Another problem that I have noticed is that you are dividing by the GPU bandwidth, so, as the GPU bandwidth increases, the token generation rate decreases, which, surely, should be the opposite?

For example, you said:

(2 * 120 * 1000) / (936) = 256.410

So, for a GPU with a bandwidth of 936 gb/s, we would get 256 tk/s.

Ok, lets now assume that we double the bandwidth, to 1800 gb/s. We would expect to get twice the tokens, right? But with your algorithm:

(2*120*1000)/(1800), we get: 133.33

So instead of doubling, the tk/s is halving? Which can’t be right surely, unless I’m missing something?

Comment on AI: Estimated token generation rates for selected model sizes (and algorithm to calculate same) by Name

Name — Thu, 30 Oct 2025 15:41:40 +0000

Hellom the information in this article is absolutely wrong. If your inference was really memory bound and you could fit the model into single GPU, you could get away with:

(2 * n * 1000) / (b_w)

where

n – number of billion parameters
b_w – GPU bandwidth in GB/s

Taking your example:

> So, if the model is running entirely on the GPU (With a bandwidth of 936 GB/s) and the model size is 120GB, then the token generation speed would be:

we would have

(2 * 120 * 1000) / (936) = 256.410

which again, would be close if the model wouldn’t need to be split across several GPUs and the card was 100% memory bound – this depends on the model as much as on the card

Comment on Dependency Hell: The Linux Experience by PhoenixGames

PhoenixGames — Wed, 15 Oct 2025 13:26:18 +0000

In reply to Disponat. What areas are you stuck on? I might be able to offer some advice?

Comment on Taming the Beast: Running DeepSeek V3-0324 Locally by PhoenixGames

PhoenixGames — Wed, 15 Oct 2025 13:25:28 +0000

In reply to Miss Byron. You are very welcome, thank you for your comment!

Comment on Taming the Beast: Running DeepSeek V3-0324 Locally by Miss Byron

Miss Byron — Wed, 15 Oct 2025 13:12:04 +0000

OMG Thank you SO MUCH for such detailed description! I’ve been banging my head against the wall for 3 months, trying to figure out how to run it on my fairly good setup! Your insights helped me A LOT. Wishing you all the best!!

Comment on Dependency Hell: The Linux Experience by Disponat

Disponat — Mon, 13 Oct 2025 01:50:18 +0000

As someone currently stuck in dependency hell (also related to ComfyUI), this was cathartic to read.