Live data from Hacker News

Mini-Mainframe at Home: A 6-CPU Server from 1997

cpushack.com

11–20 of 27 posts

Re: Mini-Mainframe at Home: A 6-CPU Server from 1997

#12
post #8

When one of my employers closed our site and merged operations with our Chicago office at the end of 2001, we were left with surplus equipment. Nobody wanted about 4 giant quad-PPro servers, so I took them. I never did anything with them and honestly don't remember who made them..we were mainly an HP shop, so maybe? They sat on rollers on the floor, and came up to about the bottom of a desktop, and were square boxes…

Well you have quite the dilemma.

On one hand, they are pieces of vintage hardware.

On the other hand, you may have about 5 grams of gold holding up your father's table (up to .33g/cpu*16).

Re: Mini-Mainframe at Home: A 6-CPU Server from 1997

#13
post #12
post #8

When one of my employers closed our site and merged operations with our Chicago office at the end of 2001, we were left with surplus equipment. Nobody wanted about 4 giant quad-PPro servers, so I took them. I never did anything with them and honestly don't remember who made them..we were mainly an HP shop, so maybe? They sat on rollers on the floor, and came up to about the bottom of a desktop, and were square boxes…

Well you have quite the dilemma. On one hand, they are pieces of vintage hardware. On the other hand, you may have about 5 grams of gold holding up your father's table (up to .33g/cpu*16).

I suppose they could take out just the CPUs for the gold and leave the rest to hold up the tables.

Re: Mini-Mainframe at Home: A 6-CPU Server from 1997

#18

Crazy how these things depreciate to basically zero monetary value over time. Compare that to say, the convenience store on my corner, which is largely unchanged since the 1970s.

Pretty sure the value is negative. You couldn't just throw them in the trash, you'd have to pay someone to take them. 350 nm process, which has gone down almost a factor of 100 by now (apparently 3nm is in prototype), and that's a linear measurement so it's really more like 10,000. Stunning. A silicon atom is about .1 nm in diameter, so a factor of 10 is about it for physical shrinking, definitely not a factor of 100. I hope that we'll be sitting here in 20 years amazed that petaflop computers drew megawatts of power!

Re: Mini-Mainframe at Home: A 6-CPU Server from 1997

#19
post #10

I was wondering why it had 6 CPUs (as opposed e.g. to 8). Seems that ALR did some clever engineering to work around a limitation in the chipset which was only designed for 4. > 6x6 is based on the same 450GX chip set (previously known as Orion) as competitors' four-CPU offerings. The reason everyone else is shipping four-CPU systems is the chip set's 2-bit CPU addressing scheme -- allowing for four-CPU IDs. What ALR…

Of historical note is that more than 8 socket Opteron servers were implemented in similar way. There was an chip (IIRC called AMD Horus) that contained two HyperTransport/CoherentLink controllers connected back to back.

It's a standard approach to making bigger systems, as most built-in cache coherency protocols are optimized for a too-small case. AMD Horus allowed easy build of 8x8 servers, SGI continued their work from Altix in UltraViolet (now owned by HPE?) where they used custom cache coherency chips in between, implementing custom cache directory protocol over NUMAlink connections.

Also, the approach of having "router" and "local area switch" is used in Infiniband, where local subnets use 16bit ID for endpoints, and 128bit addresses for inter-subnet routing (the addresses are IPv6 compatible).

Post reply on HN