Woot! has via the
Woot! App (
Android or
iOS):
XFX Swift AMD Radeon RX 9060 XT OC 16GB GDDR6 PCIe 5.0 Graphics Card (RX-96TS316BA) on sale for $419.99 - 20% off (max $40) in-app Prime discount =
$379.99.
Shipping is free for Amazon Prime Members (must login with your Amazon account and select a shipping address in order for Woot to apply free shipping) or is otherwise $6 per order.
- Note: Must purchase via the Woot! App. Discount is automatically applied on the final checkout page; up to $40 maximum total discount. Limit one use per account. New Woot! customers are eligible for a 25% Off discount (up to $50 max. discount).
Thanks to Deal Hunter
phoinix for sharing this deal.
About this product:
- Chipset: AMD RX 9060 XT
- Memory: 16 GB GDDR6
- XFX SWFT Triple Fan Cooling Solution
- Boost Clock Up to 3320 MHz
Leave a Comment
27 Comments
Sign up for a Slickdeals account to remove this ad.
generally you get the top of line card for LLM to minimize your watt consumption,
which is probably why top cards are so much more expensive due to LLM demand.
The goal is to get the entire model into vRAM so it doesn't have to spill over to system memory (which is also why DDR is through the roof).
Also, not every model can work/work as well across multiple cards.
Thus, two 16GB cards rarely replace the capabilities of one 32GB card.
However, it does allow you to split the workload in a workflow and run two models simultaneously. You could even have one card running a model that directs other cards running different/purpose built models.
They just don't have the tensor cores to be nearly as efficient.
Sign up for a Slickdeals account to remove this ad.
Why aren't you using MoE or a smaller model?
With that said, he'd have to play games that would benefit from it.
Also, what does the rest of his system look like?
If the CPU is also 6 years old, it may not be able to keep up with this GPU on more demanding games.
They just don't have the tensor cores to be nearly as efficient.
What LLM "doesn't run" on AMD vs Nvidia? Maybe you're talking about CUDA's significantly better support for video and image generation vs ROCm, but that's not an LLM.
The Unsloth models fall flat on their face with AMD.
And no, CUDA isn't just used for image generation. That's leaving out the dozens of binaries and libraries that are geared toward CUDA.
But again, the tensor cores are also a massive advantage.
The proof that there's a clear winner right now should be obvious in the fact that it's still very easy to get AMD cards at close to retail, while even the second hand market for two generation old Nvidia cards is inundated with markups.
Sign up for a Slickdeals account to remove this ad.
Leave a Comment