eCoupon at checkout:
XXQMJ954
Note: Promo codes are typically for one-time use and don't always apply seamlessly. If you run into this issue, refresh the page, re-enter the code, or revisit the cart page.
SPECS:- CPU: AMD Ryzen AI 9 HX 470(12C/24T, Base Clock:2GHz, Up to 5.2GHz)
- GPU: AMD Radeon 890M (Max Graphics Frequency 3100 MHz)
- Memory: 32GB DDR5-5600MHz (SODIMM x2 , up to 256GB)
- Storage: 1TB SSD (Include) M.2 2280 NVMe PCIe4.0 x4
- Connectivity: WiFi 7 (802.11ac), Bluetooth 5.4, Dual 2500Mbps LAN
- Interface: 3x USB3.2, 2x USB2.0, 1x HDMI2.0, 1x DP,1x Type-C (DP/DATA/PD), 1x USB4, 1x OCulink, 2x 2.5G LAN
- Internal Slot: 2x SO-DIMM DDR5 Slot , 2x M.2 PCIe 2280 SSD Slot
- Video Output: HDMI+DP+Type-C+USB4 Quad Display
- Power Adapter: DC 19V 6.3A(AC100~240V, 50/60Hz)
- Other Functions: Auto Power ON, Wake On Lan, PXE
https://www.amazon.com/dp/B0H3ZF1Q85?th=1
Join The Conversation
Share information with the community. Please follow our Community Guidelines and be kind!
18 Comments
Sign up for a Slickdeals account to remove this ad.
Our community has rated this post as helpful. If you agree, why not thank BanzaiBilliam
They've also got the older model of this PC with an HX370 for $969. That might be the best deal given there's maybe a 5% performance difference between the HX370 and HX470.
https://www.bosgamepc.c
https://www.bosgamepc.c
Sign up for a Slickdeals account to remove this ad.
Compared to the Steam Machine though, it's much, much weaker. I believe the Steam Machine is somewhere around laptop RTX 5050 or 4060 level.
Don't understand why someone would say no. The only thing I can think of is you may want more RAM for local LLMS.
And again, the best deal was this past weekend for identical specs for $918.99. I say identical... but I am wrong, the Acemagician was soldered 8000MT/s and this deal is for no soldered 5600 MT/s
Sign up for a Slickdeals account to remove this ad.
The MTP Advantage: Multi-Token Prediction introduces speculative decoding, meaning the system can process and output multiple tokens per memory read cycle. Actual Expected Speed: With MTP active, you will realistically hit 17 to 24 tokens per second. This easily clears the human reading speed threshold and provides a snappy, smooth chat flow.2.
The 27B MoE (Mixture of Experts)The Math: While the entire model file is large, a Mixture of Experts architecture only loads a portion of its parameters per token. Assuming a quantized variant with roughly 8B active parameters, your system only needs to sweep about 5 GB of active weights through the memory bus for each token generated. Actual Expected Speed: Dividing 89.6 GB/s by those 5 GB of active weights yields roughly 15 to 18 tokens per second. It will feel completely conversational—not blazingly fast, but steady and comfortably readable without long pauses. How much total RAM do you have installed in this machine? If it is 32GB, you will have plenty of room for the 12B model, but the 27B MoE might tightly crowd your system resources unless heavily quantized.
I will also add that soldiered memory should run your model faster (like 10 - 15% faster). Especially if you get 8533MT/s.
Join The Conversation
Share information with the community. Please follow our Community Guidelines and be kind!