logoalt Hacker News

Qualcomm Cloud AI 100 Ultra – 150W AI Accelerator Card with 128GB at 548GB/s

167 points • by peter_d_sherman • yesterday at 7:55 PM • 60 comments • view on HN

Comments

tokepi • yesterday at 8:49 PM

If the specs seem outdated, it’s because this chip launched in 2024. It’s available at cloud providers including Cirrascale

➕ show 2 replies
WarOnPrivacy • today at 4:29 AM

Buy for the low, low price of - nope. Not for you.

    Cirrascale (8-card Ultra server, 128 GB each, 512 GB system RAM): 
    $4,699/mo (monthly term) or $3,759/mo (annual term)

    Cloud GPU rental (per-GPU): from $5.15/GPU/hr
ref: https://cloud-gpus.com/qualcomm-ai-100-ultra/
➕ show 1 reply
thway15269037 • yesterday at 9:00 PM

What's the price?

➕ show 4 replies
pearl_panel5 • yesterday at 8:29 PM

Curious how the memory bandwidth actually holds up under sustained inference loads compared to the spec sheet numbers.

swiftcoder • yesterday at 9:16 PM

So it's an M4 Max Mac Studio, RAM-wise, with a GPU's worth of compute?

➕ show 1 reply
zitterbewegung • yesterday at 8:47 PM

Sort of interesting design It has half a terabyte of SRAM on the accelerator with a 128 GB of RAM replacing HBM with SRAM but keeping regular RAM.

➕ show 1 reply
g58892881 • today at 10:37 AM

that's a slim card

skavi • today at 3:32 AM

is this hexagon?

➕ show 1 reply
0917236141 • yesterday at 11:11 PM

Seems like a reasonable general-purpose card.

aetherspawn • yesterday at 8:41 PM

Ok so it’s significantly worse than Apple M6 and we don’t know if it’s cheaper or not

Edit: Ah sorry you’re right, the bandwidth is lower but the TOPS is a lot higher. I’m not sure how much effect you get from higher TOPS with such low bandwidth though.

➕ show 4 replies
varispeed • yesterday at 8:31 PM

Aren't these specs already outdated? Seems meh. I would think players would like to bring something new to the table instead of bringing what was okay two years ago or so.

Give me 1TB RAM and 5TB/s or something that exciting.

➕ show 5 replies