Slashdot Mirror
Microprocessor Forum
Manufacturers are strutting their stuff at the Microprocessor Forum. Some of the rollouts: Turmoil writes "AMD has demonstrated working SMP. http://www.amd.com/news/prodpr/20165.html" hol writes: "German news site Heise.de reports that a German startup named PACT surprise-announced their processor design at the Microprocessor Forum in San Jose. Apparently this thing is a 128 cpu parallel computing deal which has its roots in the programmable gate array world." infodragon writes "All Linux Devices.com is running a pretty cool article about an X86 chip running on 1 AA battery. They demonstrated it by playing a VCD movie. They also say that mp3s can be decoded/played on it."
This article says that 3D Studio Max was used for the demonstration which means NT/2000 to me.
If dual-proc Thunderbird motherboards were to arrive on the market a month from now, is the code for linux to take advantage already in place?
Did I just get information when I heard that AMD's SMP is different?
--Cycon
Your Brain + EEG + LEGO Robots = Brainstorms
The world of CPU design has been quite stagnant in recent years. Yes, there have been truly massive improvements in an engineering sense, but architecturally speaking, the latest Pentium and the various 64-bit candidates are really no different to a little ol' Motorola 68000 at heart. Harvard, RISC and superscaler designs haven't departed significantly from the same basic and extremely limited architecture which dates back to three decades ago or more.
But PACT's XPP is a different thing altogether, a dataflow computing engine on a chip. This thing is so far outside the current norm that it holds exhalted company with only a very few select others: my list of such exceptional architectures would probably comprise the Intel iA432, the Inmos Transputer, the Crusoe, and now the XPP. (I'm only including real candidates for implementation as micros, not research or demonstrator platforms of which there have been many hundreds of great ones.)
It's beautiful!
My research work on parallel architectures and concurrent languages really needed hardware like this to blossom. I wish the XPP had appeared then!
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra
http://ewh.ieee.org/soc/ias/pub-dept/abbreviation. pdf
Not there.
http://physics.nist.gov/cuu/Units/prefixes.html
Hrm not there either.
http://computer.org/author/style/mno.htm
Fer crying out loud, not there EITHER!
Put up or shut up now, troll.
Come on, that's so last year. In order to gain karma these days you need to say whatever is against the perceived party-line, and complain about how all the moderators will moderate you down for being so avant-garde.
Thankyou for applying to sell your living brain to us. After extensive tests on your cognitive ability, we are happy to offer you $1.80 for your brain. Our trained staff will be with you to perform the extraction in the next hour.
Thankyou, once again, for donating your unused brain.
I see a LOT of comments asking how the AA battery powered CPU stacks up against a Transmeta. While I have no stats on the dinky x86, I have an equivalent that you'll find interesting.
I bought a Micro-ATX Cyrix MediaGX mobo a while back, $59/+shipping, to use as a part of a custom router setup. Well, lo and behold, I'm reading the manual, and it states that it will do full-frame rate MPEG1 (VCD) and DVD playback. Now, don't get me wrong, but this is a 166 chip. What do I do? I pop in a DVD drive, ghost Windows 98SE to it and install the software player (OEM version of PowerDVD, with support for the funny accelerated video chipset.) And it plays 'The Road Warrior' just fine!
It doesn't take much to do VCD/DVD playback. The 166 Cyrix is about equivalent to a Pentium 120. The Transmeta Crusoe is equivalent to a Pentium III-500, for a max of seven times faster.
.sig: Now legally binding!
You wish.. MPEG1 video decompression is dead cheap for CPU power... A lot of people have spent the last 10 or so years writing assembly decoders for MPEG1 video on the x86.
.88 magnum -- it goes through schools.
It probably isn't all that powerful. I'd rather see someone do power-reduction like this on more power-friendly platforms. By the time you get some of the other archs out there down to one-battery power levels, they'll probably put this thing to shame.
--
It's a
--
It's a
-- Danny Vermin
"Traditionally the battery-powered handheld appliances, like PDAs and mobile phones, are based on RISC microprocessors, which normally consume only a few hundred mini-watts while the traditional X86 microprocessors consume more than ten times of that figure."
Ask Slashdot - What is a mini-watt?
Al Gore - I invented mini-watts.
G. Bush - My daddy gave me mini-watts.
Remember, You are unique...just like everyone else.
Motorola also annouced 1Ghz G4's, as planned. Coming to a Mac near you.
Burn Hollywood Burn
An entire VCD with an AA Battery? Now that is slick... and even though the Voltage slipped from 1.5V to 1.1V, the processor remained functional? Man, this is very cool... Anyone know how the Transmetta chips take a slipping Voltage? can it handel this?
I am become Troll, destroyer of threads
Well, here's another answer to the recent question about why companies still tinker with x86 devices. Apparently the recent technologies still can improve upon the original behavior.
When it comes to single processors, the Athlon and the PIII are pretty much equal on a clock for clock basis (Athlons reach higher speeds).
But when it comes to SMP, the Athlons should have a BIG advantage; the EV6-protocol.
The EV6 which is the bus used by the Athlon-arcitechture, and licensed from Compaq (Alpha)
provides DEDICATED bandwith for each of the CPUs,
(to the chipset). The PIII's however must all share the bandwith, which is not really sufficient for optimal operation. Don't know what Intel has done for the P4 though..
Go take a look at the XPP page. It sits on a PCI card, it runs linux (it really does! there's both the weird XPP bit and a 32bit RISC core programmed with gcc), and it delivers amazing processing power. It sounds awfully familiar to the SETI accelerator PCI card hoax! Except this seems to be for real. So, can I get it to do SETI? I think I recall from the last set of comments that they don't treat ports as legit; that's no fun, cuz the thing is intended for DSP/image analysis, which is all SETI is anyway. I think I can get it to distributed.net though; after all, it's also suggested for crypto stuff. I mean heck, it can sit there, talk to the net every so often to get a new key, and with 256MB of RAM and 51.2 GFLOPs available I think it will do just fine at moving me up the ranks in distributed.net, or the great internet mersenne prime search. Too bad its expensive... and no software yet (plz? I don't think this thing should die too soon). Does anyone have pricing info?
Just out of curiousity, how much power does that chip have? I have an old Zeos Palmtop(that runs dos pretty good) that ran off of 2 1.5 volt batteries. Mind you, it only lasted a half hour and couldn't play a movie. You could play Gauntlet 2 on it with a bit of tweaking.
Just saw that M$ isn't voicing any upcoming support for AMD's Hammer. Heaven forbid anyone accuse M$ of using that old Monopoly power to kill a product...
--
Chief Frog Inspector
A feeling of having made the same mistake before: Deja Foobar
"Always"?
Translation: the last six months since you started reading slashdot and grew pubic hair.
History lesson: Go back before AMD bought nexgen in mid-90's. If it wasn't for Nexgen, AMD would have folded long ago. On their own, all they had was a K5, which sucked. Once they bought all of NexGen's brainpower with the chump change they made from ramping flash (Intel fumbled the flash market in 93-94, which let AMD win some capital $$$), Nexgen all but handed over the k6 and k7.
So does AMD deserve credit for the K6 and K7?
Reworded: does a company deserve credit for intellectual property that it purchases?
Tough call. I would say, yes, they do. It all comes from talented engineers, and whatever faceless company that they're a slave to owns the rights to their intellect. AMD had very little talent in the late 80's-early 90's, then they bought some. The exact same way that Intel is trying to buy dominance in the networking world. Isn't that capitalism?
Personally, I love the competition, and I want to see AMD get spanked because Sanders is an idiot and the underdog deserves a boot to the head once and while... but then again I'm a big fan of schadenfreuda (sp?)
---
Unto the land of the dead shalt thou be sent at last.
Surely thou shalt repent of thy cunning.
https://www.accountkiller.com/removal-requested
The whole field of Configurable Computing has been trying out architectures like this for some time (don't know why /. hasn't covered this technology more).
/. hasn't covered the technology more: because FPGAs on their own are mainly just an amorphous sea of gates. You need a lot more than that before you have a viable (and therefore interesting) computing engine. StarBridge came up with such a working system and so it was no surprise to see it featured on Slashdot.
I'll tell you why
In contrast, Xilinx's strides in FPGA and RC technology tend not to feature because there's a gulf between a beautiful RC chip like the 6200 and actually being able to compute with it. Even Xilinx know that now -- their newer devices are more advanced FPGAs but they don't even attempt to carry the generic RC mantle like the 6200 tried to do, unsuccessfully. It came close, but you need a lot more than just an FPGA to make a useful RC: you needs a preconfigured computing architecture to start with, otherwise the programmer needs to think in terms of gates, and that's one paradigm shift too far. The 6200 suffered from not being specific enough. That's a peculiar observation to make in the FPGA field, but it reflects reality in the computing field, and even RCs need to take that into account.
And that's what PACT seem to have done with their XPP. Sure, its reconfigurable parts are based on FPGA technology (the only sensible way of doing it), but they've created a whole new dataflow computing engine with that RC resource, and it's the latter that's interesting for computing people, not the FPGA itself nor the internal RC mechanism.
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra
Another Slashdot feature was posted on the same topic within the same day, probably by mistake. Here's a pointer to that continuation of the subject thread, for those that like to browse the archives.
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra
I believe a single 760MP Northbridge can only handle 2 processors, but I think AMD has hinted that you can have 2 (or more?) 760MP's on a single board, so in theory 4 or 8 way should be possible, but I expect we'll see 2-way first.