Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Benchmarking 15 “E-Waste” GPUs with Modern Workloads

esologic.com

Post reply on HN
Print view View original

31–40 of 68 posts

Re: Benchmarking 15 “E-Waste” GPUs with Modern Workloads

#31

Post by SoftTalker »

Octoth0rpe wrote: post #29
SoftTalker wrote: post #27

A few years ago we got rid of a bunch of K80s at work, they were not only obsolete but had gotten glitchy as hell. I suspect this is from the many heat/cool cycles they went through. When they were running flat out the exhaust air felt like a hair dryer.

Do you have any #s on how old they were at decomm time? There's some suspicion that part of the AI bubble is companies playing games with depreciation, eg assuming that H100/H200s will survive for 5 years.

They were probably about 8 years old. They were well past reasonable EOL, but they were used for teaching, so performance was not a primary concern, as long as they worked. They had reached the point of not working often enough that we finally scrapped them.

Re: Benchmarking 15 “E-Waste” GPUs with Modern Workloads

#34

Post by BigTTYGothGF »

0xbadcafebee wrote: post #22
loeg wrote: post #18

Earlier quoted context omitted.

"Don't post generated text" - https://news.ycombinator.com/newsguidelines.html

The rules for comments are now so lengthy that probably 90% of comments would violate one. It's like how police can pull you over for any reason and justify it by picking a law you've unwittingly violated

Most of them are just restatements of each other, with the occasional pet peeve thrown in.

Re: Benchmarking 15 “E-Waste” GPUs with Modern Workloads

#35

Post by eso_logic »

wlesieutre wrote: post #30
eso_logic wrote: post #26

Earlier quoted context omitted.

I'm trying bruh fuck!

For an easier change than HTML or SVG, try running them though pngcrush to make the graph images much smaller. Won't give you the lossless vector quality, but you should be able to keep these image files visually indistinguishable at much lower size.

Lesson learned for real, and TIL matplotlib is happy to export SVG. Good to know for next time. I upgraded my lightsail instance size in the meantime.

Re: Benchmarking 15 “E-Waste” GPUs with Modern Workloads

#36

Post by eso_logic »

SoftTalker wrote: post #31
Octoth0rpe wrote: post #29

Earlier quoted context omitted.

Do you have any #s on how old they were at decomm time? There's some suspicion that part of the AI bubble is companies playing games with depreciation, eg assuming that H100/H200s will survive for 5 years.

They were probably about 8 years old. They were well past reasonable EOL, but they were used for teaching, so performance was not a primary concern, as long as they worked. They had reached the point of not working often enough that we finally scrapped them.

This is a great datapoint. Someone else brought up that I should be memory checking the GPUs to understand if things are breaking down.

Re: Benchmarking 15 “E-Waste” GPUs with Modern Workloads

#37

Post by axus »

0xbadcafebee wrote: post #22
loeg wrote: post #18

Earlier quoted context omitted.

"Don't post generated text" - https://news.ycombinator.com/newsguidelines.html

The rules for comments are now so lengthy that probably 90% of comments would violate one. It's like how police can pull you over for any reason and justify it by picking a law you've unwittingly violated

Trying to follow all the rules is fun for its own sake.

Re: Benchmarking 15 “E-Waste” GPUs with Modern Workloads

#39

Post by joe_mamba »

SillyUsername wrote: post #17

No mention of the venerable Tesla P4. 75W peak, 8GB VRAM, about $80 (£60). I have 6x P4s, a Xeon E5 2696v3 (36 threads, 3.8ghz peak but all core turbo unlocked, so 6 cores at 3.8Ghz - about 8 cores at 3.5ghz, or all cores at 3.1ghz), 48GB DDR4, all fit into a micro atx case running on a 650W MSI psu. This gives me a virtual 48GB GPU (llama.cpp ftw) to backup that 48GB of RAM. I typically see scores of at least 7-12t/…

> about $80 (£60)

Man, I wish I lived where you guys lived.

Re: Benchmarking 15 “E-Waste” GPUs with Modern Workloads

#40

Post by 0x20Fearless »

joe_mamba wrote: post #39
SillyUsername wrote: post #17

No mention of the venerable Tesla P4. 75W peak, 8GB VRAM, about $80 (£60). I have 6x P4s, a Xeon E5 2696v3 (36 threads, 3.8ghz peak but all core turbo unlocked, so 6 cores at 3.8Ghz - about 8 cores at 3.5ghz, or all cores at 3.1ghz), 48GB DDR4, all fit into a micro atx case running on a 650W MSI psu. This gives me a virtual 48GB GPU (llama.cpp ftw) to backup that 48GB of RAM. I typically see scores of at least 7-12t/…

> about $80 (£60) Man, I wish I lived where you guys lived.

Same, and hope I can afford that w/$ $/t, 650w is wild to run here24/7 , would cost me 32% of my salary
Post reply on HN