cat

this chassis has 4 nodes. each node has 1 chip. each chip has 64 Atom cores with AVX-512, 4-way hyperthreading, and 16GB of on-chip MCDRAM (a variant of Hybrid Memory Cube) along with support for 384GB of "far" DDR4.

You can boot Windows 10 on it but it makes no sense to do so; the intention was you boot Linux on it and run OpenMP workloads. This 2U server runs 1,024 threads of x86 code. Xeon Phi was supposed to kill GPUs for number crunching. It didn't

#from the recycling

Huldra @huldratigress4/16/2024, 11:30 PM

I got one of these on eBay for the sole purpose of making `htop` beg for mercy

also you can boot them without an RAM in the DIMM slots

@catball shared with:

#rebug

You must log in to comment.

in reply to @atomicthumbs's post:

Benjamin @cordlessmodem4/16/2024, 10:28 PM

The xeons don't screw around

Vikxin @vikxin4/16/2024, 10:29 PM

Is that a socketed Xeon Phi?? They made those???

walking mirage @atomicthumbs4/16/2024, 10:34 PM

yeah. you can boot windows 10 with it

Vikxin @vikxin4/16/2024, 10:51 PM

Love to have more cores than god

nys @nys4/16/2024, 10:57 PM

wait this is… free recycling??

walking mirage @atomicthumbs4/16/2024, 11:04 PM

I work at an e-waste recycler

nys @nys4/16/2024, 11:34 PM

so much makes sense now i apologize for my critical thinking fail :host-nervous:

in reply to @atomicthumbs's post:

Tamber @tamber4/16/2024, 10:40 PM

McDRAM
I'm lovin' it~

~@modulusshift4/16/2024, 10:45 PM

Intel released Knight's Landing years behind schedule, and then used the sales figures for that outdated product in a highly competitive market to justify killing the product line entirely. Intel is too comfortable to continue working on product lines that aren't instantly successful. I'd argue their strength as such a large company is exactly being able to provide projects like these such a long runway. But I suppose perhaps if they consider Itanium they may have a point that they can't keep doing it. I'd just argue they can't afford not to.

Lapis the Nevrean @lapisnev4/17/2024, 12:34 AM

Intel seems to do this often, Optane lived and died on the two year delay that resulted in releasing the Optane DIMMs in servers that already had faster regular RAM in them and couldn't afford to run them slower for any amount of extended capacity...

walking mirage @atomicthumbs4/17/2024, 1:18 AM

the fucked thing is that you could also use PMDIMMs as block storage which seems like it would be the optimal way to get an absurdly, ridiculously fast SSD. we have a couple of hp z6 g4 workstations on the shelf here that'd work for that and i desperately wish i could try it

Lapis the Nevrean @lapisnev4/17/2024, 1:25 AM

You could possibly use the Optane U.2 drives to get the same benefits? My understanding was that the perceived speed is actually coming from extremely low latency and not bandwidth... *Shrug*

walking mirage @atomicthumbs4/17/2024, 1:52 AM

i wonder how the latency of DDR4 vs. PCIe 4 stacks up

Brian Danger Hicks @ckape4/17/2024, 3:36 PM

Larrabee was designed with one purpose in mind: to make the CEO of Nvidia shut up. So its days we're numbered once it became clear that if anything it was just going to make him even louder.

muffin j lord @muffinlord4/16/2024, 10:53 PM

Sixteen gigs of onboard RAM??

walking mirage @atomicthumbs4/17/2024, 1:21 AM

onboard high-bandwidth RAM. it's 5x faster than the DDR4 and can be used as cache, as RAM mapped to a NUMA node, or both

muffin j lord @muffinlord4/17/2024, 3:19 AM

this shit must light up like the fourth of july under load, holy damn

sirocyl @sirocyl4/17/2024, 1:30 PM

I have to wonder how something like this would perform on e.g., highly parallelized software build workloads. like, run up against a modern many-cored Threadripper or something. I imagine the memory locality helps a bunch for compile times.

walking mirage @atomicthumbs4/17/2024, 6:04 PM

it weirds me out that it has SPARC levels of SMT

in reply to @huldratigress's post:

walking mirage @atomicthumbs4/16/2024, 11:59 PM

oh no wonder none of these had ram in them