GameTales: Cray 6400 (2015) | Hacker News Reader

GameTales: Cray 6400 (2015)

rome.ro

53 pointsmariuz14 comments

The "Cray" 6400 was really the "Sun Sparcenter 4000".

I dont know if Sun paid Cray to get their name on it, but I'm sure Cray did not pick it up because of the fantastic CPU performance.

The entire sun4d series had the same fundamental issue: When a CPU missed local cache and fetched across the "X-Bus", instead of returning a single word as fast as possible in a 2-cycle packet, the memory board did four sequential memory accesses and sent a 9-cycle packet.

The net effect of this was to turn it into an 8MHz machine, if your program did not fit well in the 1Mbyte CPU-cache.

On the plus side, you could stick amazing numbers of S-bus cards into those boxes, for instance if you needed a lot of disks for statistical seek-time reduction.

We had five SS1000's as file servers, they had 9 differential SCSI controllers each, with each controller servicing three Seagate 2GB "Elite" drives - a whopping total of 270 GB, mirrored and striped with Veritas.

CS6400 is not exactly the same thing as SC4000/6000, but the same design slightly improved by people at Cray's business systems division (which originally was FPS Inc.). On the other hand it is still fundamentally sun4d (ie. non-NUMA) machine with partitioning support bolted onto the XDbus architecture.

It is probably of note that 6400 was not supercomputer per-se but Cray's attempt to produce bussines oriented mainframe. It was SPARC based more or less sun4d machine. As it did not exactly fit into SGI's lineup, after the SGI aquisition it was quickly sold to Sun who then made larger UltraSPARC based variant known as Enterprise 10000.

From the technical perspective it is a large single system image multiprocessor system running solaris with hardware support for logical partitioning (which famously involves configuring the backplane logic through JTAG from attached sun4 machine with somewhat non-trivial software).

If it's like the Cray S-MP, it was a very powerful SunOS/Solaris box. Plus a lot of Cray mystique.

I was fortunate as a teen to visit Cray Research Superservers (a cab ride away), to port our workstation software to it. The port was trivial, and took less than a day.

(One of our marketing people was exploring a partnership with Cray, and I guess no real software engineers were available, so they sent the intern.)

> Shortly after I released Quake on June 22, 1996, John decided that developing on Windows NT 3.1 was the way to go. His first project was porting QuakeEd over to Win32. I left id on August 6, 1996.

I guess there's different ways to parse the subtext. He didn't want to work on Windows? I can emphatise.

Also interesting is the Carmack reply in the comments section which says the point of this model was its potential NextSTEP/SPARC compatibility.

> He didn't want to work on Windows? I can emphatise.

You didn't really have a choice when it comes to making games. Other than the consoles.

NeXT was their existing workflow, they built the content on tools running on NeXT and then targeted DOS.

Sure, but NeXT was basically dead by that time and there wasn't any way they could develop a modern (at the time) game on it (say Quake III). And even though they ran the level editor, I doubt whatever the top of the line NeXT machine was at the time could even run Quake 1 at a decent framerate.

Yep, NeXT was long in the tooth, but they could have moved the content tools and cross build env to another other *ix platform. (And even for running the game .. Q3 was OpenGL, and PC gaming GPUs had yet to overtake the workstation 3D hardware).

If they weren't terrible they might have. You've probably read this before:

https://rmitz.org/carmack.on.operating.systems.html

The rest of the owners bought him out and pushed him out of the company. Or something like that.

Huddled around a table at the local Mongolian grill restaurant, 3 young punk nerds cooked up a wild idea:

One of us had found a Cray T3E was for sale on eBay. We mused: What would happen if a truck rolled up on Quakecon, and we asked Carmack "will it play Doom?"

15 years later, I got my answer: https://twitter.com/belthesar/status/1341185861392740358

"Each person would have a hardware interface board that had keyboard and mouse inputs with video output on it."

Interesting. I wonder if Solaris (I assume it would be running Solaris) would support that kind of multi-head thing.

via https://twitter.com/pikuma/status/1634222216463409152

Reply on news.ycombinator.com