The GPUs might care quite a bit about bifurcation. The RTX Pro 6000s can communicate directly between themselves over PCIe 5.0 16x and saturate the bus doing so. This is used when doing tensor parallel LLM execution for the All-Reduce steps.
Someone can work around this by including a PCIe 5.0 switch like the Broadcom PEX89048. That switch can be used to, for instance, have the two GPUs talk to each other at full PCIe 5.0 16x speeds, while talking to the motherboard at say PCIe 4.0 16x, or even 8x. As an end user, cards using that switch seem to be an AliExpress only item, but I assume an OEM like System76 could get their hands on it to place directly on the motherboard. I imagine the price of adding that to a motherboard could be less than the price of switching to a Threadripper or Xeon 600 platform.
However, I can't find any evidence that System76 has added a switch, and in fact they refer to x8/x8 multi-GPU support, making me think they didn't add a switch. To my mind, this isn't a very satisfactory setup.
I view the agent integration (bearing in mind I haven't tried it yet) as an easier way to replace writing python scripts to do repetitive tasks, for people doing particularly repetitive work. A macro recorder might be more optimal to the anti-AI crowd, but this is probably easier to implement for the developers, while being more flexible for the end users.
I assume that this can be connected to local models, not just cloud models.
1990s calling: there's a language which is strongly typed, and will compile to bytecode that runs both on your server and in your browser as a sort of universal binary. It's named after a type of coffee...
MS(IE) had ActiveScript, of which VBS was just one implementation (although it was the bundled one that always existed). People have built integrations for stuff like Tcl, Python, or Perl, making them available to the browser.
I believe it can be broken down to 200GbE and 100GbE though. It is too bad that it can't route to the management ports for a 10GbE outgoing connection (or maybe it can and I wasn't paying enough attention).
I can see a few uses for this:
1) cluster 4x GB10 machines
2) Connect a small group of video editors via 100GbE to a NVMe file server.
3) Overkill home office network, just because.
I assume you meant that would NOT work super-well with mobile, to which I would agree. I think there are some apps for org-mode files but I've never tried one, but there are plenty of apps for managing notes. I do as little as possible with mobile, which probably makes me an outlier. I rarely use my phone for more than texting or maps/navigation.
I also don't keep dossiers on all my friends, so for me the entire premise is just academic.
Yes, I certainly wouldn't have just skipped past ADS not working. Since samba and zfs support it, this is weird that Ubiquiti wouldn't. If this isn't just a config problem, I would call it cause to return the device.
This support Windows and macOS in addition to Linux. I'm not in a hurry to disrupt things on my Linux machine, but I might just have to try this on the other platforms.
While it does look extremely likely that AI is used for the copy writing, and probably for some of the code maintenance, I'm not convinced that this is being vibe coded.
By my calculation using what I can find of 2019 pricing, a more reasonable solution would have been:
Synology RackStation RS1219+ - $1,200
8x Ironwolf Pro 12TB drives - 8x $450 = $3600
so for a usable RAID-6 capacity of 65TB about $4800 + OSS, or double that if they have two locations they could and replicate between them.
I know there would be cheaper hardware options available (I'm partial to older surplus enterprise hardware myself) but it isn't particularly reasonable to expect a small business to choose a more DIY (at least in terms of software configuration) option. You could step down to 10TB drives, but it seems unlikely they would have wanted to risk that not being quite enough.
On top of that, since 2019 is 7 years ago, it is reasonable to expect they would have had a hardware refresh at some point, I would think 5 years would be more reasonable than 3 year for something like this. At that point, the nas would be more ($1300) but the drives would be less (8 * $300), so add either another $3700 (double if both are replaced, but maybe they figured the offsite backup could be stretched even longer than 5 years)
So, rather than $2500, it would probably have been more like $8500 + OSS or $13,300 without OSS. Still that would have meant a proper are archive and a proper backup and they wouldn't be in this mess, so I completely agree with you about what they should have done, but think you were underselling the price a good bit.
I think most people would find that two or three definition is all they really need. 1080p h264 at high bit rate for compatibility on the local network and low bit rate 1080p for streaming remotely or to save space in local downloads for a trip.
Someone can work around this by including a PCIe 5.0 switch like the Broadcom PEX89048. That switch can be used to, for instance, have the two GPUs talk to each other at full PCIe 5.0 16x speeds, while talking to the motherboard at say PCIe 4.0 16x, or even 8x. As an end user, cards using that switch seem to be an AliExpress only item, but I assume an OEM like System76 could get their hands on it to place directly on the motherboard. I imagine the price of adding that to a motherboard could be less than the price of switching to a Threadripper or Xeon 600 platform.
However, I can't find any evidence that System76 has added a switch, and in fact they refer to x8/x8 multi-GPU support, making me think they didn't add a switch. To my mind, this isn't a very satisfactory setup.
reply