71–80 of 90 posts
Earlier quoted context omitted.
Can you link because 8k is the price of two A100s so that would be a steal.
4x A100s with the baseboard for $9k (USD). https://www.ebay.com/itm/336632412718 Buy one and let me use it when you're sleeping.
Earlier quoted context omitted.
Last year you could buy a AI Max 395+ with 128G for 2.5k, now it's almost $4k. Or maybe you're right, I originally remembered 2k as well. I wanted to wait for the AI Max 395+ upgrade of my laptop, and now it makes no sense to upgrade.
> Last year you could buy a AI Max 395+ with 128G for 2.5k, now it's almost $4 Only if you pay the Framework premium. https://www.bosgamepc.com/products/bosgame-m5-ai-mini-deskto... I don't have access to the USD price, but it's 2500€ (tax included), up from 1600€ in November when I ordered mine.
As a sibling comment says, the author has a video of testing GLM 5.2 q2 on the pair of machines at https://www.youtube.com/watch?v=PkKXm_mKCCM . He also has a video about getting Deepseek 4 Flash on 1 or 2 AI Max machines using antirez's ds4 at https://www.youtube.com/watch?v=Cfl3TS7ME5s .
Edit: I actually have a website tracking the price of the AI Max 395+ mini pcs (on amazon, anyway). They can be had for about $3300 in the US and about 3400 euro in Germany. ( https://gpuquicklist.com/apus?models=AI%20Max+%20395%20128GB and germany: https://gpuquicklist.com/apus_de?models=AI%20Max+%20395%2012... )
Earlier quoted context omitted.
4x A100s with the baseboard for $9k (USD). https://www.ebay.com/itm/336632412718 Buy one and let me use it when you're sleeping.
how do you even connect these? i assuming expensively
You'd need at least threadripper cpu/motherboard to handle that many pcie slots/lanes.
Not to mention that a baseboard like that is not going to work with a standard ATX power supply, so you'd have to provide your own power solution.
If that doesn't sound fun to you, you're probably better off buying Claude tokens.
Earlier quoted context omitted.
ignoring the fact one would need a bit of a different setup (chassis, PSU) to run it, I casually looked and there's nothing below $25-50k euros for such a board decked out, depending on the config. TBH even that doesn't sound bad, but I wouldn't even know where to start how to run it.
This is from the USA site: https://www.ebay.com/itm/157742745616 It's always a gamble buying used electric but per the description, fully decked out server with 256gb vram. Though at 2U it is going to sound like a 747 taking off from your office.
I have two 128gb Strix Halos and have been extremely excited about Antirez's (Redis author) work on DS4, especially with 4bit quant using two machines: https://github.com/antirez/ds4 Right now the speed isn't good for GLM 5.2, Deepseek V4 Flash speed is okay for me (actually reading the output) and quite usable. See kyuz0's great recent video here: https://www.youtube.com/watch?v=PkKXm_mKCCM With a bit more speed and…
I'm about to move house and one of the things I want is better local AI. Ideally running a variety of smaller models at the same time, but occasionally one larger one. 128GB unified is pretty appealing, but for the price I can also get a handful of 5060 ti 16GB which give vastly better compute/inference speed.
Seen a few reviews of the strix halo boards from people complaining the tps is too low so definitely keen to hear others' experiences with em.
I have two 128gb Strix Halos and have been extremely excited about Antirez's (Redis author) work on DS4, especially with 4bit quant using two machines: https://github.com/antirez/ds4 Right now the speed isn't good for GLM 5.2, Deepseek V4 Flash speed is okay for me (actually reading the output) and quite usable. See kyuz0's great recent video here: https://www.youtube.com/watch?v=PkKXm_mKCCM With a bit more speed and…
I had a Strix Halo laptop with 128GB which unfortunately died last week. I paid 2800 euro for it. If I buy the same machine today, the sticker price is 7899. The device was not perfect by any means, but the ability to run fairly large models is some kind of magic.
Earlier quoted context omitted.
>sticker price is 7899. It's not even worth it at that point. You can get a used enterprise grade SXM baseboard with 4-8 V100/A100 GPUs off eBay at a similar price. That will even get you actual HMB ram and NVlink. Along with 10x the AI performance, assuming you don't care about your electricity bill of course.
You can get a new M5 Max MacBook pro with 128 GB unified ram (targeted by Antirez for DwarfStar4) even after the Apple price increases, it's less than 7899 by at least $1000. And you probably won't pull more than 100 Watts.
I've played with ds4 on such a machine, and the battery drains when using only a 96W power brick. But you can put it in low power mode and the fans won't even turn on while it delivers something like a third of the performance.
Earlier quoted context omitted.
The Strix Halo mini PCs use the exact same chip, and have a much smaller footprint than any laptop. Have you seen the size of these machines? I can and have easily popped my daily driver computer into my very small backpack to attend a demoparty for example. With the laptop you probably won't get silent operation at the peak 100-140w, i.e. you've now massively overpaid for lower performance.
Can you get these from vendors like Asus and lenovo these days? The ones I've seen on aliexpress are from unknown Chinese vendors.
I've also got a Beelink minipc, which also had some issues, and the unknown Chineese vendor did respond meaningfully to support email, and created a BIOS update in ~2 weeks after a driver update surfaced a firmware bug.
I'd recommend skipping ASUS. Lenovo may be fine.
Earlier quoted context omitted.
how do you even connect these? i assuming expensively
The connection does not look standard so probably need to find some kind of OEM interposer card, which would allow you to connect some slimSAS cables that in turn connect to pcie adapter cards. You'd need at least threadripper cpu/motherboard to handle that many pcie slots/lanes. Not to mention that a baseboard like that is not going to work with a standard ATX power supply, so you'd have to provide your own power so…