Earlier quoted context omitted.
Will we be able to get similar bandwidth with socketed ram with CAMM / LPCAMM modules in the near future?
Maybe, but due to the physics of signal integrity, socketed RAM will always be slower than RAM integrated onto the same PCB as whatever processing element is using it, so by the time CAMM / LPCAMM catches up, some newer integrated RAM solution will be faster yet. This is a matter of physics. It can't be "fixed." Signal integrity is why classic GPU cards have GiBs of integrated RAM chips: GPUs with non-upgradeable RAM…
The Framework Desktop is a beast
331–340 of 464 posts
Re: The Framework Desktop is a beast
#332Earlier quoted context omitted.
Apple base models tend to be fairly competitive but they have some of the most extreme margins on RAM and SSDs in the industry. They charge $600 CAD to go from 16GB -> 32 GB. They charge $900 CAD for 512 GB -> 2 TB SSD.
Unless I'm misunderstanding their store page, Framework charges $687 CAD ($500 USD) to go from 16GB → 32 GB RAM. They do only charge $214 CAD ($156 USD) to 512 GB → 2 TB SSD, thanks to it just being an NVMe stick.
That doesn't seem accurate for any of their computers? There is a pretty big leap from 32GB -> 64GB for the Desktop, but that is also a different processor.
Re: The Framework Desktop is a beast
#333I could do with someone explaining to me (sorry not very advanced knowledge in this area), how did Framework manage this performance at a lower price point? And why can't average Joe at PCPartPicker do something like this?
Re: The Framework Desktop is a beast
#334I’ve been agonizing over getting the Framework Desktop for weeks as a dev machine/local LLM box/home server. It checks a lot of boxes but the only reason to look at the Framework Desktop over something like a Minisforum MS-A2 is for the LLM and that seems super janky right now. So I guess I’ll wait a beat and see where we are later in the year.
How quiet is the Minisforum?
[1] https://store.minisforum.com/products/minisforum-bd790i-x3d
Re: The Framework Desktop is a beast
#335Earlier quoted context omitted.
Framework also charges quite a surplus on the SSD. For example "WD_BLACK SN7100 NVMe 2TB" is €229, but a more typical price is around ~€140, so that's about €90 extra. Don't know how that compares to Apple.
Framework absolutely does fleece you on the SSD, so don’t buy SSDs from them if you’re price-sensitive. But compared to Apple it’s nowhere close: per Apple’s website, upgrading a current Mac Mini or Mac Studio from 512 GB to 2 TB costs $600. You can buy a WD_BLACK SN850X 8TB for that much money (or a WD_BLACK SN8100 4TB and still have a quarter of it left over, if you really want PCIe 5.0, but there’s no point at the…
Re: The Framework Desktop is a beast
#336Earlier quoted context omitted.
Not to defend Apple here, but it's also a bit apples to oranges (heh) because the power consumption is not easily comparable. I would hazard a guess and say: at that spec, if you're looking at 1Y TCO, the Apple could easily be cost competitive performance per dollar. Since they're in spitting distance of each other, just get the one you're most comfortable with. Personally I prefer Linux and I'm really happy that non…
You have to account for resale when it comes to TCO. Resale value for a non-Apple PC is essentially zero - i.e. no one will buy it, or you'll get pennies on the dollar if they do. Whereas there's a strong market for used Apple hardware, and you can easily recover 50% or more.
The used market for high performance PC gear is quite efficient. You're not going to get a high-end CPU or GPU from the past several years for pennies on the dollar.
Likewise, the resale market for Apple products doesn't guarantee 50% or more unless you're only looking at resale value of very new hardware. For example, you can pick up an M1 Max MacBook Pro (not that old) for closer to 1/3 of the original price.
Re: The Framework Desktop is a beast
#337Earlier quoted context omitted.
Yes, the Strix series of AMD uses a similar architecture as M series with massive memory bandwidth and big caches. That results in significantly better performance.
Isn't this the desktop architecture that Torvalds suggested years ago?
Re: The Framework Desktop is a beast
#338> The Framework Desktop with 64GB RAM + 2TB NVMe is $1,876. To get a Mac Studio with similar specs [...] you'll literally spend nearly twice as much [...] The Framework Desktop is simply a great deal. Wow, someone managed to beat Apple on price?? I don't know that it logically follows that anything is a great deal when it undercuts Apple. Half sounds about right -- I thought Apple was a bit more competitive these day…
He didn't pick the equivalent Mac to compare to. The closest Mac would be a an M4 Pro Mac Mini, not a Mac Studio.
Right now I see a 64GB M4 Pro Mac Mini with 2TB SSD is $2600. Still more expensive, but not double the price. The Apple refurbished store has the same model for $2200.
The M4 Pro Mac Mini probably has the edge in performance, though I can't say how much. The Ryzen platform driver and software support is still early.
I think these Strix Halo boxes are great, but I also think they're a little high in the hype cycle right now. Once supply and demand evens out I think they should be cheaper, though.
Re: The Framework Desktop is a beast
#339I’ve been agonizing over getting the Framework Desktop for weeks as a dev machine/local LLM box/home server. It checks a lot of boxes but the only reason to look at the Framework Desktop over something like a Minisforum MS-A2 is for the LLM and that seems super janky right now. So I guess I’ll wait a beat and see where we are later in the year.
My main worry about all the Minisforum, Beelink, etc. PCs is: potential lack of UEFI firmware updates (does anyone have experience with how good they are with updates?) and potential backdoors in the UEFI firmware (either intentionally or unintentionally). A China-aligned/sponsored group has made an UEFI rootkit targetting ASUS/Gigabyte mainboards: https://www.spiceworks.com/it-security/vulnerability-managem... Why n…
Re: The Framework Desktop is a beast
#340How is AMD GPU compatibility with leading generative AI workflows? I'm under the impression everything is CUDA.
CUDA isn't really used for new code. Its used for legacy codebases. In the LLM world, you really only see CUDA being used with Triton and/or PyTorch consumers that haven't moved onto better pastures (mainly because they only know Python and aren't actually programmers). That said, AMD can run most CUDA code through ROCm, and AMD officially supports Triton and PyTorch, so even the academics have a way out of Nvidia he…
I don't think this is particularly correct, or at least worded a bit too strongly.
For Nvidia hardware, CUDA just gives the best performance, and there are many optimized libraries that you'd have to replace as well.
Granted, new ML frameworks tend to be more backend agnostic, but saying that CUDA is no longer being used, seems a bit odd.