HP was at the International Super Computing 2013 conference in Leipzig, Germany, last week, showing off its ProLiant rack servers as well as the SL6500 Scalable Systems tray servers that have been used to build some of the most powerful supercomputers in the world. And, with no tongue in cheek at all, HP's server geeks were also …
If you want double precision, it drops down to 691 teraflops per rack.
Admittedly this is probably for specialist applications but that is still fucking impressive. We are starting to see how see these systems are going to really shake up the market. This is probably why IBM wants out of the x86 racket, or even out of the business altogether. Though if IBM hasn't got something ARM-based in the closet then I'd be very surprised. They've already done much of the leg work with cell designs.
IBM have already got something based on the PowerPC arch that actually delivers 208 (peak) TFLOPs per rack that you can buy today. It's called BlueGene/Q and it is capable of delivering over 85% of that peak figure on LINPACK, which seems to be well beyond the reach of the CPU-GPU boxes at the moment. Admittedly it sounds a bit feeble compared to the vaporware - but take a look at the Top500 & Green500 and you'll see a bunch of them at the top end of the table. ;)
We are at an interesting point in the HPC wheel of reincarnation. On the one hand with have BlueGene/Q with small cores and tightly coupled vector units on the same die, and on the other we have big CPUs with big discrete GPUs on the other. Meanwhile in the commodity space we have fairly chunky CPUs with (smallish) integrated GPUs, and I suspect it'll be this model that will dominate in the nearish future. It is stating the bleeding obvious though, NVidia and AMD roadmaps point that way, and Intel's MIC stuff is pretty much already there.
The killer micros are coming back to attack the discrete GPUs. :)
Head HPC Bussiness
IBM Bluegene has not been a commercial success. Primarily due to the fact that they are peiced higher. Also the number of applciations that have been made availble on Blue gene is very very low. Except for some niche and IBM committed clients, there are hardly any takers for that. HP Moonshot is also likelyt o face a similar fate. While it sounds good on the space utlization front, many of the X86 ported applications may not perfoorm well on the Atom of ARM processors. HP shoudl explore the design option for putting multiple pure X86 processor blade cards in a chasis.
Re: Head HPC Bussiness
I guess that by "pure X86" you must mean the 8086, as Xeons & Atoms tend to use a derivative of the AMD 64bit ISA. If that's the case I think you'll find just about any ARM running an emulator will outperform an 8086 in terms of IPC, power consumption and bandwidth.
Not for HPC
While it has some exciting possibilities (such as use Xeon E3 chips on each card) even HP say this isn't for HPC.
It's hard for me to see the purpose of stuff like this - if you want/need performance, you go with HPC stuff, whatever is flavour of the day there.
If you need cheap web and DB servers, buy only a few and virtualize.
In either case, the combination of slow(ish) CPU and limited memory makes little sense.
Am I missing something?
- 20 Freescale staff on vanished Malaysia Airlines flight MH370
- Neil Young touts MP3 player that's no Piece of Crap
- Review Distro diaspora: Four flavours of Ubuntu unpacked
- Fee fie Firefox: Mozilla's lawyers probe Dell over browser install charge
- Sysadmins and devs: Do these job descriptions make any sense?