Slickdeals is community-supported.  We may get paid by brands for deals, including promoted items.
This Deal Has Expired We can notify you when similar deals are posted. Set up Deal Alert
expired
Suryasis 3.1K Views
Nov 29, 2025 5:31 AM

HP Z2 Mini G1a AI Workstation: Ryzen AI MAX+ Pro 395, 128GB LPDDR5X, 2TB Gen4 SSD, 2xTB4, Wi-Fi7, Win11P, 3Yrs Onsite Warranty @ $2249.10 + F/S

$2,249

HP Small & Medium Business
Visit Retailer
Good Deal?
15Comments
Save
Share
Deal Details
Link: https://www.hp.com/us-en/shop/pdp...n8e9ua-aba

Coupon: 10% OFF via CARROT10

Spec:
  • Windows 11 Pro
  • HP Wolf Pro Security Edition (3 years)
  • AMD Ryzen AI MAX+ Pro 395 16C/32T Zen 5 3 GHz (5.1 GHz Boost, 64MB L3 Cache)
  • AMD Radeon 8060S RDNA 3.5 Integrated Graphics (40 CU @ 2900 MHz)
  • 128GB LPDDR5X 8533 MT/s Onboard Memory on 256 Bit Memory Bus (273 GB/s)
  • 1TB PCIe® Gen4 NVMe M.2 SSD
  • MediaTek Wi-Fi 7 MT7925 (2x2) + Bluetooth 5.4
  • Realtek ALC3205-VA2-CG Audio Codec, 2.0W internal mono speaker
  • HP 725 Multi-Device Rechargeable Wireless Keyboard and Mouse combo
  • 300 W internal power adapter, up to 92% efficiency, active PFC
  • Warranty: 3 year (3/3/3) limited warranty includes 3 years of parts, labor and on-site repair
  • Ports:
    • 3x USB-A 3.2 Gen 2 10 Gbps (1 Charging)
    • 1x USB-C 3.2 Gen 2 10 Gbps (USB PD Out 15W, DisplayPort 2.1)
    • 2x USB-A 2.0
    • 2x USB-C Thunderbolt 4/USB4 40 Gbps (USB Power Delivery, DisplayPort 2.1)
    • 2x Mini DisplayPort 2.1
    • 1x 2.5G RJ-45 (Realtek RTL8125BPH-CG)
  • Expansion Slots:
    1. 2x M.2 2280 PCIe Gen4x4 Slots (1 Occupied)
    2. 1x M.2 2230 for WLAN (Occupied)
Community Notes
About the Poster
Deal Details
Community Notes
About the Poster
Link: https://www.hp.com/us-en/shop/pdp...n8e9ua-aba

Coupon: 10% OFF via CARROT10

Spec:
  • Windows 11 Pro
  • HP Wolf Pro Security Edition (3 years)
  • AMD Ryzen AI MAX+ Pro 395 16C/32T Zen 5 3 GHz (5.1 GHz Boost, 64MB L3 Cache)
  • AMD Radeon 8060S RDNA 3.5 Integrated Graphics (40 CU @ 2900 MHz)
  • 128GB LPDDR5X 8533 MT/s Onboard Memory on 256 Bit Memory Bus (273 GB/s)
  • 1TB PCIe® Gen4 NVMe M.2 SSD
  • MediaTek Wi-Fi 7 MT7925 (2x2) + Bluetooth 5.4
  • Realtek ALC3205-VA2-CG Audio Codec, 2.0W internal mono speaker
  • HP 725 Multi-Device Rechargeable Wireless Keyboard and Mouse combo
  • 300 W internal power adapter, up to 92% efficiency, active PFC
  • Warranty: 3 year (3/3/3) limited warranty includes 3 years of parts, labor and on-site repair
  • Ports:
    • 3x USB-A 3.2 Gen 2 10 Gbps (1 Charging)
    • 1x USB-C 3.2 Gen 2 10 Gbps (USB PD Out 15W, DisplayPort 2.1)
    • 2x USB-A 2.0
    • 2x USB-C Thunderbolt 4/USB4 40 Gbps (USB Power Delivery, DisplayPort 2.1)
    • 2x Mini DisplayPort 2.1
    • 1x 2.5G RJ-45 (Realtek RTL8125BPH-CG)
  • Expansion Slots:
    1. 2x M.2 2280 PCIe Gen4x4 Slots (1 Occupied)
    2. 1x M.2 2230 for WLAN (Occupied)

Community Voting

Deal Score
+8
Good Deal

Join The Conversation

Share your experience with the Slickdeals community

Share information with the community. Please follow our Community Guidelines and be kind!

15 Comments

Sign up for a Slickdeals account to remove this ad.

Nov 29, 2025 7:53 AM
3.4K Comments
Joined Jul 2017
fintlewoodlewixNov 29, 2025 7:53 AM
3.4K Comments
Get a DGX Spark clone instead, if you want to do local LLM or train or whatever. It's worth the extra $.
3
Original Poster
Pro
Expert
This user is an Expert in Computers
Nov 29, 2025 4:57 PM
20.7K Comments
Joined Jul 2015
SuryasisNov 29, 2025 4:57 PM
Original Poster
Pro
Expert
This user is an Expert in Computers
20.7K Comments
Quote from fintlewoodlewix View quoted comment :
Get a DGX Spark clone instead, if you want to do local LLM or train or whatever. It's worth the extra $.
No, you have a wrong idea. Go and read what that is for. That is especially targeted for AI developers to develop and optimize Int4 Quantize models, which Nvidia is promoting over Into or Int 16. That thing only works on CUDA. As an end user product, this one is far superior.
Nov 29, 2025 5:35 PM
3.4K Comments
Joined Jul 2017
fintlewoodlewixNov 29, 2025 5:35 PM
3.4K Comments
Quote from Suryasis View quoted comment :
No, you have a wrong idea. Go and read what that is for. That is especially targeted for AI developers to develop and optimize Int4 Quantize models, which Nvidia is promoting over Into or Int 16. That thing only works on CUDA. As an end user product, this one is far superior.
Uh huh. Somebody spends too much time on YouTube.
1
3
Original Poster
Pro
Expert
This user is an Expert in Computers
Nov 29, 2025 7:20 PM
20.7K Comments
Joined Jul 2015
SuryasisNov 29, 2025 7:20 PM
Original Poster
Pro
Expert
This user is an Expert in Computers
20.7K Comments
Quote from fintlewoodlewix View quoted comment :
Uh huh. Somebody spends too much time on YouTube.
Or actually do work and understand, rather than throwing jargons collected from here and there, without understanding a single bit.
Dec 1, 2025 1:27 PM
1.7K Comments
Joined Jun 2009
blakezillaDec 1, 2025 1:27 PM
1.7K Comments
Hey it should be noted that you can knock Windows off. Just buy a cheap key online for $10ish

And then get a 512gb ssd (+$20) and this is down to less than 1800 for the 395max
1
Dec 1, 2025 1:33 PM
1.6K Comments
Joined Nov 2006
ReturnToSenderDec 1, 2025 1:33 PM
1.6K Comments
Quote from fintlewoodlewix View quoted comment :
Uh huh. Somebody spends too much time on YouTube.
Spark isn't for LLM inference. It's been clearly communicated all over YouTube Reddit etc.
Dec 1, 2025 6:42 PM
3.4K Comments
Joined Jul 2017
fintlewoodlewixDec 1, 2025 6:42 PM
3.4K Comments
Quote from ReturnToSender View quoted comment :
Spark isn't for LLM inference. It's been clearly communicated all over YouTube Reddit etc.
Yes, a lot of YouTubers have said so. I'm sure they also stayed in a Holiday Inn Express.
2

Sign up for a Slickdeals account to remove this ad.

Dec 1, 2025 7:45 PM
17 Comments
Joined May 2018
ChayjoemayDec 1, 2025 7:45 PM
17 Comments
1
Dec 2, 2025 6:17 AM
1.6K Comments
Joined Nov 2006
ReturnToSenderDec 2, 2025 6:17 AM
1.6K Comments
Quote from fintlewoodlewix View quoted comment :
Yes, a lot of YouTubers have said so. I'm sure they also stayed in a Holiday Inn Express.
Yes on YouTube, where the accredited tech bloggers exist, and reddit it where the really nerdy tech enthusiasts exists. typically when a bunch of intelligent people reach the same consensus, they are right.

NVIDIA DGX Spark

Best for: Training, fine-tuning, and inference with models specifically optimized for its unique NVFP4 format.

Strengths: 128GB of unified memory, compact form factor, and designed for a complete AI sandbox experience.

Weaknesses: Slower inference speeds compared to high-end GPUs for standard models, with benchmarks showing it being slower than a 5090 and significantly slower than a Pro 6000. The memory bandwidth is a bottleneck for large batch sizes
Last edited by ReturnToSender December 1, 2025 at 10:28 PM.
Dec 2, 2025 6:25 AM
3.4K Comments
Joined Jul 2017
fintlewoodlewixDec 2, 2025 6:25 AM
3.4K Comments
Quote from ReturnToSender View quoted comment :
Yes on YouTube, where the accredited tech bloggers exist, and reddit it where the really nerdy tech enthusiasts exists. typically when a bunch of intelligent people reach the same consensus, they are right.
LOL! Accredited tech bloggers? Typically, reddit is an echo chamber for idiots.
2
Dec 2, 2025 6:30 AM
1.6K Comments
Joined Nov 2006
ReturnToSenderDec 2, 2025 6:30 AM
1.6K Comments
Quote from fintlewoodlewix View quoted comment :
LOL! Accredited tech bloggers? Typically, reddit is an echo chamber for idiots.
I can't believe you're here arguing with the Internet, lol

yeah fam the whole Internet is wrong and you're right. again that box is not for inference. It's for development two different things.

NVIDIA DGX Spark
Best for: Training, fine-tuning, and inference with models specifically optimized for its unique NVFP4 format.

Strengths: 128GB of unified memory, compact form factor, and designed for a complete AI sandbox experience.

Weaknesses: Slower inference speeds compared to high-end GPUs for standard models, with benchmarks showing it being slower than a 5090 and significantly slower than a Pro 6000. The memory bandwidth is a bottleneck for large batch sizes
Dec 2, 2025 6:37 AM
583 Comments
Joined Jan 2013
TerencePDec 2, 2025 6:37 AM
583 Comments
Quote from fintlewoodlewix View quoted comment :
LOL! Accredited tech bloggers? Typically, reddit is an echo chamber for idiots.
Bro, of course info from influencers that shill shit they know jack squat about are trustworthy.
Dec 2, 2025 6:50 AM
3.4K Comments
Joined Jul 2017
fintlewoodlewixDec 2, 2025 6:50 AM
3.4K Comments
Quote from ReturnToSender View quoted comment :
I can't believe you're here arguing with the Internet, lol

yeah fam the whole Internet is wrong and you're right. again that box is not for inference. It's for development two different things.

NVIDIA DGX Spark
Best for: Training, fine-tuning, and inference with models specifically optimized for its unique NVFP4 format.

Strengths: 128GB of unified memory, compact form factor, and designed for a complete AI sandbox experience.

Weaknesses: Slower inference speeds compared to high-end GPUs for standard models, with benchmarks showing it being slower than a 5090 and significantly slower than a Pro 6000. The memory bandwidth is a bottleneck for large batch sizes
Did chatgpt write that for you?
Dec 2, 2025 6:51 AM
3.4K Comments
Joined Jul 2017
fintlewoodlewixDec 2, 2025 6:51 AM
3.4K Comments
Quote from TerenceP View quoted comment :
Bro, of course info from influencers that shill shit they know jack squat about are trustworthy.
YouTube influencers are the best, just like tiktok influencers. Do you love the Kardashians? I do!
1

Sign up for a Slickdeals account to remove this ad.

Dec 3, 2025 1:29 PM
1.8K Comments
Joined Nov 2007
h0va4lifeDec 3, 2025 1:29 PM
1.8K Comments
This could be a steal, how many employees will this replace?

Join The Conversation

Share your experience with the Slickdeals community

Share information with the community. Please follow our Community Guidelines and be kind!