Recent content by OSdevr

There's already plenty of LLM AI software that can run on Radeon GPUs much of which is open source. Ollama is a common backend, Msty is a particularly good free (albeit closed source) interface that can use it. Really the only thing notable about this is that it can use the Rysen AI's NPU as...

Sad how impressive Optane was and how Intel fumbled it.

Tariffs?

It is insanely difficult to get accelerated graphics running on today's GPUs. Doesn't help that a lot off it (in particular Nvidia GPUs) is undocumented.

As a solo coder, yeah that checks out.

For manufacturer drivers it's expected. For open source drivers to be available at launch is quite the achievement.

Basically training is creating the AI model (based on an architectural blueprint) and inference is running an AI model that already exists.

Training done on Nvidia, inference done on Huawei.

Ironically x86 was originally envisioned for embedded use and ARM in desktop computers.

Last I checked Nvidia's investment in OpenCL is the absolute bare minimum. You know how DirectX 12 has "feature levels?" Well OpenCL is worse. On paper Nvidia's 4000 series supports OpenCL 3.0 which came out in 2020. However OpenCL 3.0 only requires the complete OpenCL 1.2 functionality which...

Have same problems with Infinite's campaign on my PC. I think it's a LOD problem because sometimes it looks great but other times (particularly after fast traveling) both models and textures seem to be on their lowest detail version.

Cloud VR gaming doesn't sound very fun to me.

I have the last gen version of this on my threadripper. Works great, and when I thought I had a defective unit they sent me a replacement. Turns out my RAM was just barely too tall and prevented proper contact. Been doing well since, unlike the Enermax TR4 cooler that suffered corrosion after a...

Whenever I see "Core 2-series" I think of the Core 2 Quad that was in my first desktop.

Are we on the same forum? I'm pretty sure we all hate Nvidia and justifiably so.