qupada

joined 1 year ago
[–] [email protected] 3 points 6 months ago (1 children)

Probably best to look at it as a competitor to a Xeon D system, rather than any full-size server.

We use a few of the Dell XR4000 at work (https://www.dell.com/en-us/shop/ipovw/poweredge-xr4510c), as they're small, low power, and able to be mounted in a 2-post comms rack.

Our CPU of choice there is the Xeon D-2776NT (https://www.intel.com/content/www/us/en/products/sku/226239/intel-xeon-d2776nt-processor-25m-cache-up-to-3-20-ghz/specifications.html), which features 16 cores @ 2.1GHz, 32 PCIe 4.0 lanes, and is rated 117W.

The ostensibly top of this range 4584PX, also with 16 cores but at double the clock speed, 28 PCIe 5.0 lanes, and 120W seems like it would be a perfectly fine drop-in replacement for that.

(I will note there is one significant difference that the Xeon does come with a built-in NIC; in this case the 4-port 25Gb "E823-C", saving you space and PCIe lanes in your system)

As more PCIe 5.0 expansion options land, I'd expect the need for large quantities of PCIe to diminish somewhat. A 100Gb NIC would only require a x4 port, and even a x8 HBA could push more than 15GB/s. Indeed, if you compare the total possible PCIe throughput of those CPUs, 32x 4.0 is ~63GB/s, while 28x 5.0 gets you ~110GB/s.

Unfortunately, we're now at the mercy of what server designs these wind up in. I have to say though, I fully expect it is going to be smaller designs marketed as "edge" compute, like that Dell system.

[–] [email protected] 2 points 6 months ago

Unfortunately what's shipping today seems it would offer maybe half that.

For the batteries that were announced this past week, a larger-than-refrigerator-sized cabinet held a capacity of around 15kWh.

Around half the energy density by mass of Lithium batteries, and in the order of a sixth of the density by volume.

Now if only we could come up with a system where your car could be charged while stopped at traffic lights, we might be onto a winner (:

Considering however that the price of sodium is around 1-2% that of lithium, I expect we will see significant R&D and those numbers quickly start to improve.

[–] [email protected] 25 points 6 months ago (7 children)

I've been seeing a lot about Sodium-ion just in the past week.

While they seem to have a huge advantage in being able to charge and discharge at some fairly eye-watering rates, the miserable energy density would seem to limit them to stationary applications, at least for now.

Perfect for backup power, load shifting, and other power-grid-tied applications though.

[–] [email protected] 4 points 7 months ago

Came to post the same. Seems like the most awkward possible way to phrase that.

Your "Disks not included" suggestion, or heck, just "empty" would surely be better.

[–] [email protected] 11 points 7 months ago* (last edited 7 months ago)

Interestingly, looking at Gentoo's package, they have both the github and tukaani.org URLs listed:

https://github.com/gentoo/gentoo/blob/master/app-arch/xz-utils/xz-utils-5.6.1.ebuild#L28

From what I understand, those wouldn't be the same tarball, and might have thrown an error.

[–] [email protected] 11 points 7 months ago

I too have an oddly specific one of these, which is tartare sauce.

I actively dislike all three of mayonnaise, gherkins, and capers. Mix 'em together though? Brilliant.

[–] [email protected] 4 points 7 months ago

Sharp also make great commerical-grade printers that are 100% Linux compatible, we're using these at work: http://global.sharp/products/copier/products/bp_70c65/index.html

They don't really make anything small enough to be a "home" model, this looks like their smallest printer: https://global.sharp/products/copier/products/mx_c358f/index.html (and that's around $1000, if you could even find someone to sell you one).

[–] [email protected] 9 points 8 months ago

eSATAp! What a wild combination.

Not actually a terrible idea, even if it frequently was limited to powering 2.5" drives due to a lack of 12V. Some had extra contacts for that, but most that I saw didn't.

[–] [email protected] 41 points 8 months ago (5 children)

It does however affect getting updates from government agencies, and others who insist on only disseminating real-time information to the public via Twitter.

For instance: https://twitter.com/WakaKotahiWgtn

This is the account for traffic events (road closures, traffic accidents, etc) in my city. Not signed in, the latest visible post is from February 2023.

Since I don't have a twitter account, this is now functionally useless.

[–] [email protected] 33 points 8 months ago

Since the realistic competitor here is probably magnetic tape, current-generation (LTO9) media can transfer at around 400MB/s, taking 12 hours and change to fill an 18TB tape.

Earlier archival optical disk formats (https://news.panasonic.com/global/stories/798) claimed 360MB/s, but I believe that is six, double-sided discs writing both sides simultaneously, so 30MB/s per stream. Filling the same six (300GB) discs would take about an hour and a half.

Building the library to handle and read/write in bulk is always the issue though. The above optical system fit 1.9PB in the space of a server rack (and I didn't see any options to expand further when that was current technology), and by the looks is 7 units that each can be writing a set of discs (call that 2.5GB/s total).

In the same single rack you'd fit 560 LTO tapes (10.1PB for LTO9) and 21 drives (8.4GB/s).

So they have a bit of catching up to do, especially with LTO10 (due in the next year or so) doubling the capacity and further increasing the throughput.

There's also the small matter that every one of these massive increases in optical disc capacity in recent years has turned out to be vapourware. I mean I don't doubt that they will achieve it someday, but they always seem to go nowhere.

[–] [email protected] 30 points 9 months ago (3 children)

From the video description:

I have been a Samsung product user for many years, and I don't plan to stop anytime soon

And all sympathy I had for this person just vanished. If you don't demand better, they will keep doing - and getting away with - shit like this.

Voting with your wallet might be the one voice you have left in this world, what a way to squander it by continuing to buy products from companies whose representatives behave in this manner.

[–] [email protected] 17 points 10 months ago

The estimated training time for GPT-4 is 90 days though.

Assuming you could scale that linearly with the amount of hardware, you'd get it down to about 3.5 days. From four times a year to twice a week.

If you're scrambling to get ahead of the competition, being able to iterate that quickly could very much be worth the money.

view more: next ›