What is the Linpack's "Highly Parallel Computing" benchmark?
The third benchmark is called the Highly Parallel Computing Benchmark and can be found in Table 3 of the Benchmark Report. (This is the benchmark use for the Top500 report). This benchmark attempts to measure the best performance of a machine in solving a system of equations. The problem size and software can be chosen to produce the best performance.
http://www.netlib.org/benchmark/hpl/
Please note the words "highly-parallel" and "best performance" and the following phrase from the link from the quote:
Nonetheless, with some restrictive assumptions on the interconnection network, the algorithm described here and its attached implementation are scalable in the sense that their parallel efficiency is maintained constant with respect to the per processor memory usage.
In a round-about way those quotes mean "if it does X FLOPs on 1 processor, it'll do 10X FLOPs on 10 processors because it's embarrassingly
parallel. We're talking SETI- or RC5-type parallel.
Regarding the 5% efficiency, please reference the following paper to see some numbers. In STREAMS Triad, the Cray X1 has 24% efficiency, while a P4 had 3.4% of a peak rated at 2x the X1 processor. The paper was done by the Army HPC Research Center in Minneapolis, MN: http://www.ahpcrc.org/publications/X1CaseStudies/C luster_CrayX1_Comparison_Paper.pdf
Run down the list and look at processor counts. We've got 5120 at the top (vector), but number 2 needed 8192 to get the job done. BigMac at #3 drops to 2200 and the processor counts hover in that 2000+ category. Until #19, when Cray's X1 jumps in at 252 processors.
Having a fast computer is cool and all, but if you can do it with 252 CPUs instead of 1024 (#22, P4 2.4), isn't that a win?
Besides, LINPACK doesn't stress interconnect latency and bandwidth, only cache and memory performance. When you run a "real" codes on these Mac/Xeon clusters and get 5% efficiency, suddenly the Earth Simulator (and the small Cray X1's) look good when they blow well past the 50% efficiency mark.
If there's more than one PCB with 1 or more processors on it hooked together by wires (custom ASIC routers, Myrinet, ethernet, bluetooth, 9600 baud serial, whatever) people will call it a supercomputer. These days a 19" rack of Sony Clie PDAs will get the nod.
But even this fails because Apple called a G4 cube a supercomputer because it was as fast as a Cray was 10 years before.
It appears that DevChannel, a part of OSDN, adheres to the/. article format: 500 words or less.
Great comparison? He only tried 2 things! What about random file access? large/small file access? streaming/random? multiple streaming? I don't pretend to know all aspects of Hard drive performance, but it's sure as hell bigger than that article supports. What a waste of bandwidth.
If I was SUED for a ton of money, I'd be soiling my pants over the potential outcome that I COULD LOSE. Everyone talks tough, but if I could get out of a potential jam for $2000, I'd probably do it.
I think at the age of 27 I'm finally starting to outgrow gaming, so the concept of a "lightweight" PC that can be used for digital imaging and video storage interests me. My thirst for FLOPS isn't totally gone, so the Athlon64 has my attention as well. It's like trying to choose between a Civic and a monster truck.:(
but you can't get blood out of a turnip
(summary: S&P downgrades SGI to "negative", rates corporate credit as CCC-, states they have $141M cash on hand and are maxed out on their lines of credit).
Re:VNC merged with screen
on
GTK+ TTY Port
·
· Score: 1
That's what VNC does, but in a graphical sort of way. I do all my work through VNC servers, so when the inspiration hits, I can log in from home and obtain the same environment even through I'm WindowsXP at home and GNU/Linux at work.
VNC merged with screen
on
GTK+ TTY Port
·
· Score: 2, Informative
This looks like VNC that's been merged with screen. screen was a great text-based virtual login back in the day, and is still useful when latency is too high for VNC. However, it's still a pain to use.:)
With GTK++ TTY mode, you could have a virtual text-based desktop capable of controlling (via mouse) any thing you'd want without opening many virtual screens.
I've seen stories of Intel showing off their 10 GHz CMOS transistor (or inverter gate, which would be 2 transistors), but at that signaling rate *nothing* can get done between @(posedge clock). P4's 20-stage pipeline would grow to 400 or more (latch, and-gate, latch, repeat). Imagine the size of the Out-of-Order execution units on that CPU.
So, while the clock rate is impressive, it probably isn't doing anything effective in the first place. They're just red-lining their cars in neutral.
Ordinary off the shelf microprocessors don't have the bandwidth to memory or bandwidth to other processors to simulate complex problems. NEC's machine is a Vector architecture (SX-6), similar to the kind you see from the Cray X1. Vector architectures are a SIMD-style processor.
Most, if not all, profs request to be notified if they will be taped recorded. Some have book aspirations, and they don't want their thoughts to be sold without their consent.
Linus might think about working remotely from Finland until this mess gets straightened out.
Does this desperate attempt seem like SCO's struggling for survival, or now that their original plan has holes they must keep plugging away to reach their goals?
Dual-width can be nicer than dual channel due to the latter's ability to return data seriously out of order. A good processor design can account for it, but not fix it. Dual width gives you nice bandwidth at a decent latency (don't kid yourself, SDRAM has never been synonymous with "seriously low-latency") and shouldn't break the logic cell bank since there's still only 1 logic controller. The only down side I see to this is that you're locked into one type of memory technology, basically guaranteeing the fixed lifespan of the product. What happens if Yellowstone takes off? What about DDR2 at speeds of 533 667 Mbit?
Slashdot Mirror
It's the Inquirer, so grain of salt required:
Rumour of big HP Opteron deal rolls around
http://www.theinquirer.net/?article=9911
Run down the list and look at processor counts. We've got 5120 at the top (vector), but number 2 needed 8192 to get the job done. BigMac at #3 drops to 2200 and the processor counts hover in that 2000+ category. Until #19, when Cray's X1 jumps in at 252 processors.
Having a fast computer is cool and all, but if you can do it with 252 CPUs instead of 1024 (#22, P4 2.4), isn't that a win?
Besides, LINPACK doesn't stress interconnect latency and bandwidth, only cache and memory performance. When you run a "real" codes on these Mac/Xeon clusters and get 5% efficiency, suddenly the Earth Simulator (and the small Cray X1's) look good when they blow well past the 50% efficiency mark.
from the sounds-like-a-james-bond-plot dept.
ASCI Red Storm google search
The aesthetics of the iPod appear to fetch a $100 USD premium (at least for the 20G version-- DDJ is $300 and iPod is $400). To each his own, indeed.
If there's more than one PCB with 1 or more processors on it hooked together by wires (custom ASIC routers, Myrinet, ethernet, bluetooth, 9600 baud serial, whatever) people will call it a supercomputer. These days a 19" rack of Sony Clie PDAs will get the nod.
But even this fails because Apple called a G4 cube a supercomputer because it was as fast as a Cray was 10 years before.
It appears that DevChannel, a part of OSDN, adheres to the /. article format: 500 words or less.
Great comparison? He only tried 2 things! What about random file access? large/small file access? streaming/random? multiple streaming? I don't pretend to know all aspects of Hard drive performance, but it's sure as hell bigger than that article supports. What a waste of bandwidth.
http://www.fileshack.com/file.x?fid=3832
I'm not positive, but the tone of the article's summary led me to believe it was made tongue-in-cheek. Where's that foot icon?
So can this little guy get a Linux install that autoboots for a MAME cabinet?
If it were a real Microsoft patch, it would have executed without you knowing about it. Only these rogue virii actually *ask* you to run it.
If I was SUED for a ton of money, I'd be soiling my pants over the potential outcome that I COULD LOSE. Everyone talks tough, but if I could get out of a potential jam for $2000, I'd probably do it.
I think at the age of 27 I'm finally starting to outgrow gaming, so the concept of a "lightweight" PC that can be used for digital imaging and video storage interests me. My thirst for FLOPS isn't totally gone, so the Athlon64 has my attention as well. It's like trying to choose between a Civic and a monster truck. :(
but you can't get blood out of a turnip
(summary: S&P downgrades SGI to "negative", rates corporate credit as CCC-, states they have $141M cash on hand and are maxed out on their lines of credit).
That's what VNC does, but in a graphical sort of way. I do all my work through VNC servers, so when the inspiration hits, I can log in from home and obtain the same environment even through I'm WindowsXP at home and GNU/Linux at work.
This looks like VNC that's been merged with screen. screen was a great text-based virtual login back in the day, and is still useful when latency is too high for VNC. However, it's still a pain to use. :)
With GTK++ TTY mode, you could have a virtual text-based desktop capable of controlling (via mouse) any thing you'd want without opening many virtual screens.
I've seen stories of Intel showing off their 10 GHz CMOS transistor (or inverter gate, which would be 2 transistors), but at that signaling rate *nothing* can get done between @(posedge clock). P4's 20-stage pipeline would grow to 400 or more (latch, and-gate, latch, repeat). Imagine the size of the Out-of-Order execution units on that CPU.
So, while the clock rate is impressive, it probably isn't doing anything effective in the first place. They're just red-lining their cars in neutral.
Fileshack
Ordinary off the shelf microprocessors don't have the bandwidth to memory or bandwidth to other processors to simulate complex problems. NEC's machine is a Vector architecture (SX-6), similar to the kind you see from the Cray X1. Vector architectures are a SIMD-style processor.
If you're 75 miles north of Chippewa Falls, stop down and we'll try to get you in for a tour.
Most, if not all, profs request to be notified if they will be taped recorded. Some have book aspirations, and they don't want their thoughts to be sold without their consent.
Linus might think about working remotely from Finland until this mess gets straightened out.
Does this desperate attempt seem like SCO's struggling for survival, or now that their original plan has holes they must keep plugging away to reach their goals?
Dual-width can be nicer than dual channel due to the latter's ability to return data seriously out of order. A good processor design can account for it, but not fix it. Dual width gives you nice bandwidth at a decent latency (don't kid yourself, SDRAM has never been synonymous with "seriously low-latency") and shouldn't break the logic cell bank since there's still only 1 logic controller. The only down side I see to this is that you're locked into one type of memory technology, basically guaranteeing the fixed lifespan of the product. What happens if Yellowstone takes off? What about DDR2 at speeds of 533 667 Mbit?
by this year they meant this decade.