discussion Crappy Used Computer Parts Ads

[lowfat]

Another pay me for my e-waste . Card isn't worth $10 if it even worked.

 

[moocow]

I didn't recognize that CPU at first so I had to look it up...it's an i3, a 7th gen i3 that is 2C/4T. He wanted $250 at first and drop to $195.

 

[sswilson]

Going just by 'feels' it's definitely not the worst I've seen. At $225 list, if you offered $150 that seems... okay? Not exactly a deal, but not offensively overpriced

You called it.... that's what it finally sold for after being listed for a month.

 

[FreeKnight]

You called it.... that's what it finally sold for after being listed for a month.

 

[Izerous]

Another pay me for my e-waste . Card isn't worth $10 if it even worked.
View attachment 45083

P600/P620 cards are like 20-40 and would be a better choice even if the 560 was working so 100% agree unless your after the heatsink or something

 

[lowfat]

 

[FreeKnight]

Holy shit. For a nearly 10 year old GPU lol

 

[Izerous]

300 is high but not to the point of obscurity when they are getting low 200s. Not so far out of touch compared to other listings being shared. Not a lot of cheap GPUs with that much vram so they are fetching some LLM usage/buyers.

 

[moocow]

300 is high but not to the point of obscurity when they are getting low 200s. Not so far out of touch compared to other listings being shared. Not a lot of cheap GPUs with that much vram so they are fetching some LLM usage/buyers.

I'm using one under an EK block for LLM and AI image generation for fun, it's too slow and not enough VRAM. You can still load larger-ish models but anything more complex like using Skyvern to automate web navigation or image generation is too slow. Sage Attention doesn't work on something that old, you need at least RTX 3xxx for it. For $300, go for a 3060 12GB. If Intel B60 24GB is under $600, that may be worth saving money for and stick with VULCAN as the backend.

 

[Izerous]

I'm using one under an EK block for LLM and AI image generation for fun, it's too slow and not enough VRAM. You can still load larger-ish models but anything more complex like using Skyvern to automate web navigation or image generation is too slow. Sage Attention doesn't work on something that old, you need at least RTX 3xxx for it. For $300, go for a 3060 12GB. If Intel B60 24GB is under $600, that may be worth saving money for and stick with VULCAN as the backend.

b50 pro is the goal card for the homelab atm. With the low profile and slot power I can fit 1 in the server without any kind of modifications. With single slot coolers (mod kits exist) I could fit two and split the responsibilities. Anything physically larger or requiring more power will require heavy customization.