Search results

That was my use case, 4 SATA and 1 miniNVME port.

I will add more as we go, and as I find things. Motherboard: - Asrock J3710 incl Corsair DDR3-1333 2x4GB RAM - 40 Euros Could be used for a low-fi router, or something similar has 1 x1 port to populate... https://www.asrock.com/mb/Intel/J3710-ITX/index.us.asp...

Thanks great input. I start to think that AM4 + dGPU might be a vaiable path. I somehow was brain dead while writing the title - I am basically looking for a light home lab server, not an HTPC per se, I stream to the devices but need the transcoding power...

Agree. But with the current market, the 1699 USD, for the smallest unlocked SKU, is in the great picture of the build not even that much money...

I’m currently planning a lightweight build to serve as both an HTPC and a low-power home lab machine. I run a couple of lightweight Docker Containers and Jellyfin for personal serving. I don't need massive compute performance, but I have a few non-negotiable requirements: Needs H.265 hardware...

Sorry my bad - 8x96GB Ram. Reading was never my strength.

I partially agree that you need an RTX 6000 Blackwell or an RTX 5090. It's a delicate balance; self-hosting for a solo operation or an SME is challenging because pooling hardware resources through a larger entity is generally much more efficient. While data privacy is a valid reason to...

Some great ideas here, but some ideas that make economically no sense. For homelabs ok, but I couldn't justify setting up Habana Gaudi2 and pytorch no matter the price - the knowledge is just too scarce...

I have a solid background in Computer Vision, but I’ve also done a fair share of LLM work; not just using them, but also fine-tuning, post-training, DPO, RL, etc. I wanted to share my perspective on current hardware choices… My general take: if you're primarily a user looking for a unified...

Looking for a low-power Intel Core i5/i7 "T" version CPU for a low-power build / homelab node. Ideally looking for something from the 8th gen, or 9th gen, depending on what you have gathering dust. Requirements: Preferably i5, thats enough for me Must be a "T" suffix model (35W TDP). 100%...

Looking for a VROC Premium Key. Shippable in EU

The main issue is the following one: Currently, I have a multi-GPU workstation with dual 4080 Super and a PM9a3 3.8Tb, which is enough for many use cases and quite large datasets. My datasets are often too big to squeeze into my machine's RAM (192GB)—I almost exclusively work on imaging datasets...

Honestly, I don't care about redundancy. We have 100TB+ of storage on RAID in the office for datasets, but for this training, I just need RAID 0, and let's go. If the RAID fails, the dataset will be co-existant on the spinner drives anyway. I am just not willing to fork out a s***load of cash...

usable would be good. How much is 4x 7.3TB in the end in terms of usable? 25-26? Could be an option - but the M.2 have cooling issues in 24/7 use-cases, no?

I am looking for an out-of-the-box solution to get 30TB NVME U.2 Stroage, with a maximum of 4 discs. I'm working on a Computer Vision project and would love to scale the dataset by another factor of 10, which is easily feasible. Unfortunately, it is now at a point where storage becomes a huge...

My reason for 3647 over Epyc SP3 back when I bought it: - Asus C621 WS Workstation board, with Sleep, Audio, and usable bootup times - AVX-512 -> and the Intel-optimized software stack - stability and peace of mind - Intel is still, in eco-system terms, the more complete platform solution...

I use an LGA3647 Xeon Gold 6212u with dual 4080 super as my ML engineering and research rig daily. I run 192 GB RAM and 2 GPUs—stable as fu**. The machine ran for 2 weeks while I was on vacation doing heavy ML training, and when I came back, it was still running. I really like the platform. I...

I wonder if a symmetric 7950x (both CCX with 3D-VCache) Epyc 4000 is not a missed opportunity here? Now we just need a board with 10-25Gbit NIC and we might have an amazing package.

Tldr: Thanks to this fantastic community, my workstation is still air-cooled and runs like a charm. Currently, I am running a C621-64L with a Noctua tower cooler and struggled to get my two GPU setups running with enough spacing between the cards to cool them properly. The first PCI slot is...

My bad. 13900k vs 13900kf, I just found the screenshots. 13900kf: 13900k: