Posting up my notes from SC13 is another thing I didn’t get to during the end of the semester. Remedying now.
The main takeaway sequence from conversations on the floor is as such:
- The era of single-core performance gains is already over.
- Furthermore, the era of usable single-die performance for MIMD machines is coming to an end.
- Therefore, big machines are going to be getting physically bigger… to the point where connection lengths are a problem (everything is Infiniband, and Infiniband doesn’t tolerate long runs well)
- There is a LOT of cooling effort to make the necessary density happen – central large fan systems, immersion cooling, closed-circuit water gear, etc.
The other really exciting thing that it seems AMD is going to make it, and more. Their lean period finished when the payoff on the XBone/PS4 came in, and they have a VERY good plan for the next >2 years. It works with the premise above about single-core/die MIMD performance ending, and points in the HSA direction – this is the crazy parts with MMUs so a CPU and GPU can share memory without skew penalty and such. ARM and partners are also generally pointed that way, and have been for some time, though apparently AMD isn’t getting out of the x86 game, but it does look like they are getting out of the fat core game.
Continue reading →
I will be at SC’13 November 16-21 with the aggregate.org/University of Kentucky research exhibit again this year in booth 629. Media and impressions should appear somewhere in my ‘net presence during and after the conference, it is always a good show.
Edit:Pushing photos from the show floor into this album.
I’ve got my pictures from the event up in an album in google’s cloud.
Here are the big, cool things I learned on the floor or at the various evening events:
- Xeon Phi. Xeon Phi everywhere. Intel may have backed off on Larrabee, but the MIC descendants are proliferating quickly, and appear to actually be in use. They really are interesting parts – 60 node Linux x86 SMP box attached to a host system over PCI Express via a network-like interface. Somewhere between a tiny desk-side cluster and a GPU with a programming model you can actually use.
- The population was much less male dominated than is typical for computing events. This is always a good thing.
- The average age of attendees also seemed to be down by the better part of a decade.
- The national labs losing their booths to the government-wide travel restrictions (apparently some folks went junketing in Vegas and it was that bad) changed the feel of the floor. Fewer, but longer and deeper conversations. More open layout, because many of the usual big constructed booths belong to the national labs. Users from the national labs hanging out at vendor’s booths. Not altogether a bad thing, but it was quite different.
- ARM64 (aka aarch64, aka ARMv8). It is happening. It is odd (64KB pages, etc.). Large companies are being bet on it. We’re talking many billions of dollars, biggest bet since Itanium kind of big. The priority seems to be avoiding the Itanium mistakes, making sure the designs arrive promptly, and making sure software support is ready. Dell is talking quietly, Calxeda is gunning for it, Nvidia was showing (but only quietly) plans, AMD is being pointed to as a likely leader, and ARM was is sitting in their little 10×10 booth along one wall of the exhibit floor looking very pleased with themselves.
- AMD is dying. The untimely (and vigorously denied) rumor that they hired J.P. Morgan to begin plans to sell part or all of themselves made it look even worse, but they had almost no presence on the floor, and scheduled a tiny booth next year.
- We talked to a number of networking vendors making interesting things (free-air optical switching, multi-port Ethernet NICs, etc.). Infiniband is so good and so cheap (in a relative sense) for cluster applications right now that everyone else is hunting for an edge. This is a good thing for researchers.
There will be at least one more SC12 post later, when my cube of schwag arrives. The T-shirt harvest was great this year…
I will be at SC12 November 10-16, with the Aggregate.org/University of Kentucky exhibit in booth 631.
I will be posting pictures and impressions through at least one of my online presence mechanisms . I fully expect it to be weird this year with a bunch of the national labs pulled out due to travel restrictions, but it should be interesting.
I learned some really interesting things at SC this year, and now that I’ve had a day to process, I want to share. Many of these observations come from first or second hand conversations, or justifiable interpretations of press releases, so I don’t promise they are correct, but they are plausible, explanatory, and interesting. I apologize for the 1,000 word wall of text, but there is a lot of good stuff.
I think that covers most of the really good stuff coming off the floor this year, although I am still processing and may come up with some other insights when I’ve had more sleep and discussion.
Also, Pictures! WOO! (Still sorting and uploading the last batch at time of posting).