Slashdot Mirror
Slashback: Grids, Netscape, AMD
And Campbell's puts glass marbles in their soup pictures. Roland Piquepaille writes "We saw several grid computing announcements in the last couple of days.Of course, Gateway stole the show. In 'Gateway makes store PCs work overtime,' you can read that 'Gateway's network of 8,000 PCs can deliver 14 teraflops.' This is plain wrong. You all know that this number of 14 teraflops is meaningless. It's just the addition of the peak speed of all the PCs -- never reached anyway on individual PCs. You need specialized software to work efficiently with a grid. And two companies are releasing new products to power grids. Avaki rolled out what it believes is the first Java-based data grid software for enterprise-class IT environments. Kontiki, for its part, on Monday released a grid server that brings its content delivery system into the server realm, whereas previously it was only available for PCs. Check this column for a summary, or this article for more details."
Why aren't those things called 'stick-up' ads, anyhow? Internet Ninja writes "Netscape today released version 7.01 of Netscape based on Mozilla 1.0.2. Back in is popup blocking which they got a lashing for in 7.0 as well as tabs as home pages just like Mozilla. Release notes here and there's a couple articles on Netscape devedge which may be of interest to developers."
And they will continue to have produced my Athlon, too. schnoz writes "And you thought AMD was quitting the PC chip market? Then check out this article on Business Week. Not only are they releasing new chips and plan to continue to do so, they're also still very active research wise, working on new transistor making techniques such as the double gate design as well as metal-rather-than-silicon design. Keep going at it AMD!!"
It might interest some to know that physicists are thinking a lot about grid computing, especially those who use computation heavily, such as numerical relativists and fluid dynamicists. An interesting article appeared last year in Physics Today. Let's hope that the academic community's tradition of openness takes root in the Grid.
It's called The Globus Toolkit.
This is not to say that I think AMD doesn't have some good technology. Their FPU is second to none, and they have certainly advanced chip cooling to new levels.
But the fact remains that DDR is a stopgap chip technology. Anything built on top of it will be hampered by its lack of scalability. This includes Intel chipsets based on DDR as well.
With RDRAM and DDR equivalent in price, there isn't any reason to stick with DDR.
I have been pwned because my
here: SGE
its a good piece of software at that.
i have had some experience with it.
"Two things are infinite: the universe and human stupidity; and I'm not sure about the the universe." --Albert Einstein
2k3:
2k and 3
2000 and 3
2003
For the year 2300:
2k and 300
2k3c or 2k300
"Nine times out of ten, starting a fire is not the best way to solve the problem." - my wife
i sure hope not. i don't want the average joe to be blocking pop-ups. once pop-up blocking becomes mainstream then the advertisers are going to switch to a format that is harder to deal with. i like using mozilla and blocking pop-ups, but if the advertisers change their format to a harder to block type, then i'll be seeing ads again.
It's in the Mozilla nightly builds, though I have no idea if it's in 1.2.1
And you can't take a joke. The previous poster was poking fun at Gateway's calculations, where all PCs run at maximum flops, all dedicated to the problem no less. Like the previous poster, they made no accounting for moving data around or other overheads in "grid computing".
That's it. I'm no longer part of Team Sanity.
Yes well I'm sure I could write a browser that really kicks ass, if, like NS3 it ignores all stylesheets, screws up tables and frames and only parses a handful of tags.
Don't bother, someone else already has. It's a GTK-based browser called Dillo.
And it does kick ass.
IBM has a cluster VM just for this occasion.
Can I get an eye poke?
Dog House Forum
AMD is also not quiting it's "True Performance Initiative" Read an update at the Tech Report.
http://www.kubuntu.org/
Don't tell me to get another mail client - Netscape has done the job for me so far.
How about, "Report the bug to Netscape, not Slashdot".
> No no no, you see everybody had Pentiums running at 120 mhz
> when Netscape 3.0 was out.
Err, no, Pentiums didn't run that fast until a year or two later --
at least not the ones anyone could afford to actually buy. A
486 DX4/100 was still considered competitive as a new system even
when Netscape 4.0 came out. (Which, incidentally, tells you how
*old* Netscape 4.x is. Considering that Netscape 6 was really
ony of beta quality, we can be quite thankful that the long wait
is over and Netscape has a decent browser out again (since 7.0PR1,
which "Preview" or not made 6.2.anything look like junk).) This
new Netscape release, from what I've seen of it so far (admittedly,
not extensive use) seems to be quite solid, though of course it
lacks the majority of the features added during the 1.1 and 1.2
milestones. Which is fine; 1.1 lacked stability, and 1.2 is new
enough that it's hard to say (though I'm using 1.2.1 and it seems
very solid to me so far); Netscape is right to go with 1.0.2 for
now. I'm thinking they'll stick with that 1.0.x branch through
several minor releases and go back to the trunk for a new stable
branch around 1.4 or 1.6 or so. (This is not inside information,
just a prediction based on the pattern I've observed in their
behavior over the last couple of years.) By then, the branch
they are using will feel really obsolete to people who have been
testing the Mozilla builds, but that means that when users upgrade
to the next branch they'll notice a sudden influx of features.
That branch could be 7.5, but I'm predicting it will be 8.0
Cut that out, or I will ship you to Norilsk in a box.
Yes, you can run a standard x86 RedHat. That's the attractive thing about the Hammer/Athlon64/Opteron/Whatever. They can run 32- or 64-bit code. In fact, they can run BOTH at the same time. One of the demos that AMD showed was a dual-monitor Opteron, with two spinning 3D objects. One was running as a 32-bit app, the other as a 64-bit app - on the same machine.
However, I believe that RedHat IS going to have a release for the Hammer. Considering that some packages (like Apache) are having a good amount of work done to make them really take advantage of the 64-bit environment, I'm not sure how much of a difference the special distro will make, but there's plenty of time for that.
steve
Oh, you're not stuck, you're just unable to let go of the onion rings.
It's also a complete bear to install.
First, Java is a language. [...] doesn't mean that all other implementations are slow.
Java is an environment as well as a language. Unless this Java grid is planning to throw away the JVM, I think it's fair to say that it's probably using that standard Java environment. I'm not ruling out that a "magic" JVM might come along that somehow overcomes all the baggage of how Java is designed, but so far we've not seen this. Given the current state of technology, it seems foolish to me to throw away all that performance.
Based on my own experience, Java is on the average about 1/10th the speed of an equivalent C program, although clearly it depends on what you're doing. Where Java is particularly bad is very data intensive work, such as string manipulation. Where I was particularly appalled at Java's performance was XML parsing.
Java works best when it's a "glue" mechanism to pass communication between systems. Where it is not appropriate IMO is very computationally intensive applications, which presumably would be what you would use a grid for.
Sometimes it's best to just let stupid people be stupid.
It's in the Mozilla nightly builds, though I have no idea if it's in 1.2.1
Yeah, it's in 1.2.1. Just load up tabs for all the pages you want, then go to Edit | Preferences | Navigator and click "Use Current Group".
Graham "Teach" Mitchell, computer science teacher, Leander HS
AMD stopping competition with Intel was utter BS in the first place. If you've followed their roadmap at all, they've got the Barton core comming out in a couple of months closely followed by ClawHammer and SledgeHammer. They have new cores on the horizon and are researching new technologies. They aren't going anywhere. Forbes is going on my "Company with Idiot Writers List." They're there along with CNet and some other ZD Net and Internet.com companies.
you are right; game consoles do use RDRAM. But in the end, RDRAM is not killed because it's bad technology, but because othere stuff.
first on the tech. (REALLY quick brief)
1) RDRAM has a faster interface (duh)
2) and it has a much more narrow bus
3) but to make chips drive at such a high frequency ON THE CIRCUIT BOARD, the bus interface for RDRAM is totally wacky
explanation: RDRAM is serially connected, *kinda* like... SCSI, or COAX ethernet back in the days. and it's heavily terminated. and because the signal goes so damn fast (remember, circuit board made of FR4 here - not cache->CPU interconnects), the routing of the signal traces, while sparse (something they tout - and it's true, DDR has like 2-4 time the wire density as RDRAM on the board), has very small tolerance for length difference. furthermore because the high speed, the chips must have a very strict output impedance (which is why mem-makers got shitty yields at the beginning and the RDRAM price were so high).
performance wise / practically speaking, since it's the signal routing / RIMM detection and delay adjustment (remember no trace length differences etc) that's difficult and causes trouble - in game consoles where you will never add memory, RDRAM is actually better (easier to work with / better performance - better perf because you don't incur additional delays in the trace by adding more modules, everything is fixed). Same time on PCs, when you do it right, RDRAM still offers better bandwidth than DDR; DDR-2 i am not so sure, but that won't be in massive production for a while so don't wait for it yet. depending on architecture (P4 is, have to say, on the side of "optimized for RDRAM"), you would get better performance out of RDRAM for a little while longer.
now the non-tech side:
RAMBUS charges royalty. 2% i think? now - memory business is not high-margin business (or else there won't be only like 4-5 memory makers left!), so when 2% is actually like 40% from the margin - if you can do away with RAMBUS (even at a performance hit), it would enable you to survive, or make more money - depending on the company.
so... the moral of the story? RDRAM is not bad technology (i.e. has its uses - like in consoles), but it's not GREAT technology, and certainly not good enough to warrent the margin cut and the headaches in engineering (output impedence - and these days they are going to 32/64 bit so the sparse signal lines is less and less of a advertisable benefit). But I expect that it will maintain it's little niche and won't just die off suddenly one day. i mean, heck - even if they only supplied for the game consoles, (especially with the large chunck of change intel gave to RAMBUS) they can survive for quite a while. RAMBUS as a company I think will eventually fail if they continue this path of IP-only, though - for other reasons. but this is getting long already.
My life in the land of the rising sun.