The IP stack of Linux and Windows98 already are pretty good latency wise. More tweaking or even rewriting most likely wont give any significant advantage.
Tweaking a winmodem driver however might result to some reasonable gain (connect with sub-max but reliable speed, switch off compression, tweak modem protocol). But this doesnt change anything on the ISPs side. And doesnt help anyone with an external modem. Or ISDN/ADSL/Cable.
Valves PowerPlay looks much more promising: router-priorization of game packets (less ping, less packet loss), UDP header compression, bandwidth guarantees, later a special modem real-time protocol. And some more stuff like multicast for voice etc.
A single P3-700 is most likely faster than any Alpha you can get for 4k. The $1000-question is: How much Alpha MHZ can you get for $4k (as a complete system)?
costs for a Dual P3-700:
p3-700: $800, x 2 = $1600 128mb: $300 board: $300 scsi hd+controller: $700 other stuff: $600 ==== $3500
hm... ok, if you tested it yourself, you have more first hand facts than I have. Still, going back to the original point, PC100 memory doesnt seem to be a significant bottleneck. Right now.
CPUMark, FPUMark: Yes, more or less useless.
PMMX: These are not bad, I run an SQL db on one of them... And for the firewall: I use a 486 doorstopper...
Hm... I remember they had a parallel cable. Pretty fast.
Anyway, the 1541 was intelligent, too, but any data transfer still sucked up 100% CPU. On the other hand, these were no multitasking machines.
And, admitted: the PC disk controller is extremely dumb, you cannot connect more than two drives, it doesnt know about disk zones (and therefore wastes capacity), and you cannot calculate mandelbrots with it...
daisy chaining: sure.
disk drives with own OS: interesting idea. But with some decent SCSI raid controller you basically have that. Of course at a decent price. But disk drives with an integrated filesystem and some high speed protocol would be nice...
disk drives with printer drivers: OK, that makes some sense. But far too much hassle with the chaotic PC architecture.
>The inability of Alpha NT to run Intel NT binaries, and the non-existence of Alpha Win98, probably has much more to do with Intel's continuing dominance than anything else.
Alpha is not really faster except for scientific simulations and similar. And its unreasonably expensive.
Alpha isnt faster for integer performance. And this is what counts, even for Q3. Simply accept it: The fastest Alpha is usually slower than the fastest x86 for most applications (and much more expensive too). Also GCC most likely still delivers pretty slow Alpha binaries.
PCI is adequate for anything except video. We have AGP for this. Or does your HD deliver more than 100MB/s, and you desperately need gigabit ethernet, too? At the same time?
Memory: Ever wondered why the transition from PC100 to PC133 doesnt give any significant performance advantage? Any why Celery with PC66 is pretty fast regardless?
Commodore? The C64 CPU controlled its devices much more directly than PCs today. No DMA, no nothing. 100% cpu usage for anything IO. Be it disk, tape, or printer port.
Disk drives loaded with basic OS and printer drivers? Hu?
Gone are the times when only unix hackers used new Linux releases.
This means: A kernel officially released as "Release" should be VERY stable. Stable enough for anyone to put it on his most important servers, without a second thought.
Remember the problems the 2.2 series went through. Even 2.2.12 is not yet completely there. 2.2.13 or 14 might effectively be.
Maybe a new version tag is needed, additionally to the the odd numbers, and the "pre" and "ac" tags. Like "ea" for "early adopter". Whatever. Anything above "beta".
The distributions would offer two kernels then: "ea" and "stable as hell".
And "stable as hell" is what "release" should mean.
PS: Potential deficiencies in the NT release versions are not really of interest here. Linux can do better. The people in control of the kernel dont need to care about public company quarter results.
since all the freely available mp3 encoders and decoders won't go away anytime in the future.
Any protected digital music format needs to be decoded and played back through the soundcard. There are plenty of soundcards with digital output. It's impossible to hinder people to make high quality MP3s from the digital signal.
And there are still those old fashioned audio CD thingies. They are easy to rip and won't go away anytime soon.
One can restrict future portable MP3 players. Yes. But that's it.
It surely is "innovative" to construct a working tricycle with a 600 ps engine. Especially if it finally is slightly faster than a tricycle powered by a 3 year old.
Slashdot Mirror
The IP stack of Linux and Windows98 already are pretty good latency wise. More tweaking or even rewriting most likely wont give any significant advantage.
Tweaking a winmodem driver however might result to some reasonable gain (connect with sub-max but reliable speed, switch off compression, tweak modem protocol). But this doesnt change anything on the ISPs side. And doesnt help anyone with an external modem. Or ISDN/ADSL/Cable.
Valves PowerPlay looks much more promising: router-priorization of game packets (less ping, less packet loss), UDP header compression, bandwidth guarantees, later a special modem real-time protocol. And some more stuff like multicast for voice etc.
A single P3-700 is most likely faster than any Alpha you can get for 4k.
The $1000-question is:
How much Alpha MHZ can you get for $4k (as a complete system)?
costs for a Dual P3-700:
p3-700: $800, x 2 = $1600
128mb: $300
board: $300
scsi hd+controller: $700
other stuff: $600
====
$3500
hm... ok, if you tested it yourself, you have more first hand facts than I have. Still, going back to the original point, PC100 memory doesnt seem to be a significant bottleneck. Right now.
CPUMark, FPUMark: Yes, more or less useless.
PMMX: These are not bad, I run an SQL db on one of them...
And for the firewall: I use a 486 doorstopper...
Hm... I remember they had a parallel cable. Pretty fast.
Anyway, the 1541 was intelligent, too, but any data transfer still sucked up 100% CPU. On the other hand, these were no multitasking machines.
And, admitted: the PC disk controller is extremely dumb, you cannot connect more than two drives, it doesnt know about disk zones (and therefore wastes capacity), and you cannot calculate mandelbrots with it...
daisy chaining: sure.
disk drives with own OS: interesting idea. But with some decent SCSI raid controller you basically have that. Of course at a decent price.
But disk drives with an integrated filesystem and some high speed protocol would be nice...
disk drives with printer drivers: OK, that makes some sense. But far too much hassle with the chaotic PC architecture.
ram size: yes, the more the better. Up to 128-256MB, depending on the application.
faster hard drive: 22MB/sec and 9ms seek time are enough for me. EIDE. DMA. CPU usage is low. You may need to fiddle with hdparm however...
graphics card: To be honest: For 3D you better run Windows 98. For anything else a low end 8MB card is really fast enough.
AMD cannot compete against Intel with budget CPUs only. They tried and lost. Since Intel financed their low end line with their high end monopoly.
The high end is where the profit is.
And a budget Athlon is underway anyways.
spec: Sure. Now check the 95 versions to get exactly the opposite picture. And now figure where your preferred application is in between...
FPU: Yes, great for running weather simulations. Completely useless for running GCC. Or SQL dbs. Or most other apps.
For 4K you get a dual P3-700. And even a single P3-700 is most likely faster than a single $4k Alpha...
faster than PC100? Probably not. Since RDRAM is a lot slower latency-wise.
And: shortly after DDR is out it will probably be a lot less expensive than RDRAM...
All benchmarks I saw for celeron/PC66 showed a performance similar to P3/PC100.
P200MMX: sure, it doesnt have on-chip L2 cache like the celery.
>The inability of Alpha NT to run Intel NT binaries, and the non-existence of Alpha Win98, probably has much more to do with Intel's continuing dominance than anything else.
Alpha is not really faster except for scientific simulations and similar.
And its unreasonably expensive.
Alpha isnt faster for integer performance. And this is what counts, even for Q3.
Simply accept it: The fastest Alpha is usually slower than the fastest x86 for most applications (and much more expensive too). Also GCC most likely still delivers pretty slow Alpha binaries.
PCI is adequate for anything except video. We have AGP for this.
Or does your HD deliver more than 100MB/s, and you desperately need gigabit ethernet, too? At the same time?
Memory: Ever wondered why the transition from PC100 to PC133 doesnt give any significant performance advantage? Any why Celery with PC66 is pretty fast regardless?
Commodore? The C64 CPU controlled its devices much more directly than PCs today. No DMA, no nothing. 100% cpu usage for anything IO. Be it disk, tape, or printer port.
Disk drives loaded with basic OS and printer drivers? Hu?
seems that Intel has successfully sold you to RDRAM. Forget it. It has higher latency than SDRAM.
And DDR SDRAM gives more bandwidth than single channel RDRAM anyways. Without increasing latency (instead decreasing it at 133 MHZ).
And you dont want dual channel RDRAM with an 840 mobo (except if you always want to buy TWO RDRAMs for upgrades...).
Sure, DDR isnt available. But considering the price RDRAM isnt really available, too.
And besides that: PC100 SDRAM usually isnt the performance bottleneck. The point is that you dont NEED anything above a P200 except for 3D games...
Yeah, and you will hack in ten years still.
I suggest that a True Hacker is someone that looks at the tools available to him and gets something up and running as fast as possible.
I had the same experience, but with Linux.
But not months. To get comfortable with Linux took me four years.
But thats true for any Unix, I suppose.
And four years ago it was a little harder to install and use linux, anyway...
Ok, somewhat offtopic, but not completely.
I dont think its too soon.
As soon as the developers and all other ones daring to use an "odd" kernel cannot break it anymore, they cannot do anything else but "release" it.
But a more fine grained release cycle might be advantageous. Bacically because number and diversity of people using linux has increased greatly.
Gone are the times when only unix hackers used new Linux releases.
This means: A kernel officially released as "Release" should be VERY stable. Stable enough for anyone to put it on his most important servers, without a second thought.
Remember the problems the 2.2 series went through. Even 2.2.12 is not yet completely there. 2.2.13 or 14 might effectively be.
Maybe a new version tag is needed, additionally to the the odd numbers, and the "pre" and "ac" tags.
Like "ea" for "early adopter". Whatever. Anything above "beta".
The distributions would offer two kernels then: "ea" and "stable as hell".
And "stable as hell" is what "release" should mean.
PS: Potential deficiencies in the NT release versions are not really of interest here. Linux can do better. The people in control of the kernel dont need to care about public company quarter results.
Oh! Did he question your illusions?
Oh! Bad guy he is!
Really!
Not to be picky, but this is history, so we should be correct:
While ENIAC probably was the first computer built with tubes, Zuse's machine eventually was the real first computer.
Built with mechanical relays during world war two.
If we define computer as "program driven calculation machine".
Don't know if there were earlier machines from others (not only planned, but existing and functional).
definitely right.
but with a digital sound card you get the same result anyway. Anyway, a dummy driver surely is a cleaner solution.
since all the freely available mp3 encoders and decoders won't go away anytime in the future.
Any protected digital music format needs to be decoded and played back through the soundcard.
There are plenty of soundcards with digital output. It's impossible to hinder people to make high quality MP3s from the digital signal.
And there are still those old fashioned audio CD thingies. They are easy to rip and won't go away anytime soon.
One can restrict future portable MP3 players. Yes. But that's it.
well, this might be similar to those hard drives with an MTBF of 400000h.
I wondered too often how short that period can be.
Just admit it. It's penis envy.
Freud knew it. You know it. We all know it.
So go out and buy a K7. You need it!
It surely is "innovative" to construct a working tricycle with a 600 ps engine. Especially if it finally is slightly faster than a tricycle powered by a 3 year old.
Great. Hallelujah on modern technology!
and had to be expected anyways.
Intel cannot do the same kind of dumping for p3 as they do for the celeron. There is no high end left to pay for that.
Therefore AMD shouldn't have a problem to get reasonable margins at the high end.
Gone are the times of completely insane high end cpu prices!
they probably better use NT. As soon as ms doesn't find anything else to buy, they will surely start to buy porn sites...