On Tue, 6 Dec 2016 09:13:41 -0500, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:>On 12/06/2016 05:55 AM, upsidedown@downunder.com wrote: >> On Tue, 6 Dec 2016 08:41:40 +0100, Jeroen Belleman >> <jeroen@nospam.please> wrote: >> >>> On 2016-12-06 00:01, Robert Baer wrote: >>> [...] >>>> Still, a CPU board using light instead of wires/traces could be >>>> designed that would "kick ass" for parallel, multi-processing. >>>> >>> >>> Why? What advantage do you imagine that might have? >> >> If you are building something similar to the Connection Machines >> Hypercube computer, using CAT6 cabling would quickly consumes most of >> the cube module volume. >> >> Using fiber optic cables with multiple fibers and using WDM on each >> fiber will have a huge throughput for a specific cable volume. It >> might even make possible direct connection between any two nodes in >> the cube. > >WDM isn't too helpful in general except for long links, but fibre has >been used inside large computers and data centers for yonks.I was thinking about some rack size fiber loops with an add/drop interface at each box or actually a drop/add fiber system. What is the practical performance in some passive drop/add box along the fiber loop ? Is it realistic to assume that the drop box will attenuate the loop signal enough that a new signal could be inserted into the loop ?
Universal Parallel Bus -- why not?
Started by ●March 20, 2010
Reply by ●December 6, 20162016-12-06
Reply by ●December 6, 20162016-12-06
On 12/06/2016 10:26 AM, upsidedown@downunder.com wrote:> On Tue, 6 Dec 2016 09:13:41 -0500, Phil Hobbs > <pcdhSpamMeSenseless@electrooptical.net> wrote: > >> On 12/06/2016 05:55 AM, upsidedown@downunder.com wrote: >>> On Tue, 6 Dec 2016 08:41:40 +0100, Jeroen Belleman >>> <jeroen@nospam.please> wrote: >>> >>>> On 2016-12-06 00:01, Robert Baer wrote: >>>> [...] >>>>> Still, a CPU board using light instead of wires/traces could be >>>>> designed that would "kick ass" for parallel, multi-processing. >>>>> >>>> >>>> Why? What advantage do you imagine that might have? >>> >>> If you are building something similar to the Connection Machines >>> Hypercube computer, using CAT6 cabling would quickly consumes most of >>> the cube module volume. >>> >>> Using fiber optic cables with multiple fibers and using WDM on each >>> fiber will have a huge throughput for a specific cable volume. It >>> might even make possible direct connection between any two nodes in >>> the cube. >> >> WDM isn't too helpful in general except for long links, but fibre has >> been used inside large computers and data centers for yonks. > > I was thinking about some rack size fiber loops with an add/drop > interface at each box or actually a drop/add fiber system.It limits your connectivity, and doesn't save that much space compared with point-to-point links. Also fibre within racks is all 850 nm multimode (orange sleeve), whereas WDM is done at 1550 nm, single mode (yellow sleeve). Cheers Phil Hobbs> > What is the practical performance in some passive drop/add box along > the fiber loop ? Is it realistic to assume that the drop box will > attenuate the loop signal enough that a new signal could be inserted > into the loop ? >-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
Reply by ●December 6, 20162016-12-06
On Tue, 06 Dec 2016 12:55:44 +0200, upsidedown@downunder.com wrote:>On Tue, 6 Dec 2016 08:41:40 +0100, Jeroen Belleman ><jeroen@nospam.please> wrote: > >>On 2016-12-06 00:01, Robert Baer wrote: >>[...] >>> Still, a CPU board using light instead of wires/traces could be >>> designed that would "kick ass" for parallel, multi-processing. >>> >> >>Why? What advantage do you imagine that might have? > >If you are building something similar to the Connection Machines >Hypercube computer, using CAT6 cabling would quickly consumes most of >the cube module volume.Use something like the TI or Max Video SerDes. One copper pair gets>5Gb, for some distance. > >Using fiber optic cables with multiple fibers and using WDM on each >fiber will have a huge throughput for a specific cable volume. It >might even make possible direct connection between any two nodes in >the cube. > >Most massively parallel computers have direct connection to only a few >nearby nodes in each dimension and relying on mesh networking further >away. This greatly increases latency, when latency prone >serializing/deserialisation is applied in each hop. > >Getting to any node with a single pair of desers would speed up things >greatly.
Reply by ●February 24, 20192019-02-24
Reply by ●February 24, 20192019-02-24
On Saturday, March 20, 2010 at 7:33:19 PM UTC+5, GreenXenon wrote:> Hi: > > I keep hearing about Universal Serial Bus [USB]. Why hasn't a > Universal Parallel Bus [UPB] been implemented yet? > > Wouldn't a UPB be faster than a USB at the same clock rate? If so, > this would mean the same speed with less energy comsumption. Right? > > > Thanks, > > Green Xenonbut
Reply by ●February 24, 20192019-02-24
On Sun, 24 Feb 2019 19:08:10 -0800 (PST), muhammadiftikhar467@gmail.com wrote:>Parallel was faster than serial, why isn't there an UPB - "Universal Parallel Bus"?Because jitter, skew, and cost are bigger problems than bandwidth.
Reply by ●February 25, 20192019-02-25
On Sun, 24 Feb 2019 19:08:10 -0800 (PST), muhammadiftikhar467@gmail.com wrote:>Parallel was faster than serial, why isn't there an UPB - "Universal Parallel Bus"?There was: ISA, and later PCI. Given N wires or twisted pairs, each can carry a self-clocking data stream, and that's generally faster than using those same wires for parallel data. That's why a PCI Express link with N pairs uses each lane for serial data. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●February 25, 20192019-02-25
Reply by ●February 25, 20192019-02-25
John Larkin <jjlarkin@highlandtechnology.com> wrote:> On Sun, 24 Feb 2019 19:08:10 -0800 (PST), > muhammadiftikhar467@gmail.com wrote: > >>Parallel was faster than serial, why isn't there an UPB - "Universal >>Parallel Bus"? > > There was: ISA, and later PCI. > > Given N wires or twisted pairs, each can carry a self-clocking data > stream, and that's generally faster than using those same wires for > parallel data. > > That's why a PCI Express link with N pairs uses each lane for serial > data. > >How many cable pair uses USB3-x? -- Uwe Bonnes bon@elektron.ikp.physik.tu-darmstadt.de Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 1623569 ------- Fax. 06151 1623305 ---------
Reply by ●February 25, 20192019-02-25
On Monday, 25 February 2019 05:15:17 UTC, John Larkin wrote:> On Sun, 24 Feb 2019 19:08:10 -0800 (PST), > muhammadiftikhar467@gmail.com wrote: > > >Parallel was faster than serial, why isn't there an UPB - "Universal Parallel Bus"? > > There was: ISA, and later PCI. > > Given N wires or twisted pairs, each can carry a self-clocking data > stream, and that's generally faster than using those same wires for > parallel data. > > That's why a PCI Express link with N pairs uses each lane for serial > data.It also means, if suitably implemented, the link can keep working when not all the wires are connecting. NT