Show HN: GPU Prices on eBay (gpupricecompare.com)
Howdy!
Keeping with the trend of being influenced by https://diskprices.com, I wanted to make a resource for GPUs on eBay that also take into account a performance metric I tend to look at when checking GPUs.
It's still a work in progress, but it's at a state where I think some of HN might find it useful!
There are a few things that I have planned to add in the coming day(s): 1. Filters for compatible slots and connectors 2. Expand for different eBay regions, rather than just US focus
Let me know if you have any questions or want more / different filters on the page or if I missed something important.
76 comments
[ 3.0 ms ] story [ 134 ms ] threadYour connection is working correctly.
Vercel is working correctly.
1. How can a post with 9 upvotes and 1 comment be number 1 on HN
2. How can a post with so few metrics of engagement be hugged to death.
The first is likely due to the late hour - time is a factor in the HN algorithm and with fewer people interacting with content the bar for getting the number 1 spot is much lower.
The second is probably attributable to the myriad clients using the HN API to scrape content from links shared here or the likely large population of HN users who engage with content but neither comment nor upvote.
What's your scraping process?
Edit: Realised its D3D benched...
https://developer.ebay.com/api-docs/buy/browse/static/overvi...
Currently waiting to hear on upgrade to 1.5m. But 5k works for now!
If you're interested in large language models there's a table of vram requirements for fine-tuning at [1] which says you could do the most basic type of fine-tuning on a 7B parameter model with 8GB VRAM.
You'll find that training takes quite a long time, and as a lot of the GPU power is going on training, your computer's responsiveness will suffer - even basic things like scrolling in your web browser or changing tabs uses the GPU, after all.
Spend a bit more and you'll probably have a better time.
[1] https://github.com/hiyouga/LLaMA-Factory?tab=readme-ov-file#...
Let me know what you think.
Not if it's a brand-new 2024 account and the first sale is an ultra-high-demand NVIDIA graphics card with high VRAM.
Apologies about the item being sold, the list is only updated every 4 hours at the moment. With additional API access (waiting to hear back!), I would be able to get a tighter loop on updating the list so everything is currently available.
This is technically possible, but for ~3,000 products, checking on an hourly basis would make ~72,000 requests per day -- which definitely would get flagged, I think.
So for example, they filter out 5 products, and from the javascript in the browser you perform the request for those 5 products, and then show the user which ones are available. It could also send the data back to your server. Of course you still have to be careful with the number of requests, but it would allow you to get more data without getting flagged.
I think I’m going to keep it relatively simple for now, just to see how it goes. If it becomes a bigger issue, I’ll try out your method!
I suspect it would be meaningful to parametrically filter these inop edge cases out by default.
Finally upgraded in 2024 for $600.
I get maybe 2x the FPS, in the same games at the same settings.
What gives? I thought GPU compute had massively fallen in price.
GPUs have advanced a lot more in tensor computation and GPGPU than in pushing pixels onto a screen
I suppose I'm living up to the username! I do know how to write decent programs and in other languages, but I'm just so much faster when bodging.
I'm thinking of making similar website for more things, in which case I'd clean up the architecture a bit, but my exams are in a few months and this hasn't been particularly successful so we'll see. I probably shouldn't be spending my weekends programming right now :)
I submitted it a few hours ago to two upvotes - I think it might be because I didn't include the link directly, and apparently that downranks it in the algorithm. I've emailed dang to see if it's OK to repost it.
And, up to you, but doing money math with floats is a well known road-to-ruin: "price_with_shipping": 18.990000000000002,
Thanks for pointing out the floats issue. It'll probably be OK because I round to the nearest integer when displaying and only use the extra precision for sorting, but I might round to 2dp on the server side to avoid wasting bandwidth on all those zeroes - I presume floating point imprecision compresses poorly.
decimal.Decimal in Python
java.math.BigDecimal
I don't know if there's an idiomatic way to use this datatype in JavaScript / Node.
Let me know if this helps!
I'm also going to improve the backend a bit to filter out broken GPUs