a good PC?

And Alienware is now owned by Dell. Which the OP tries to avoid.

Thomas

On a sunny day (Tue, 21 Nov 2006 21:52:23 +0100) it happened Zak wrote in :

But Seagate will keep selling Maxtor under the Maxtor name.

Is it, really? Wasn't aware of that. Pity.

I've been using Patriot memory in the last three systems I've built.

BTW, nforce chipset fake raid (aka fraid) works fine. In theory, there is a speed penalty since the CPU does the overhead, but on a dual core AMD 4400, I sure don't notice a penalty. I'm running mirrored and striped (4 drives). Raid 10. All the drives are Seagate sata.

rsync works great under Cygwin.

This looks pretty good:

Optional front-panel hot-plug RAID drives, lots of big fans, looks like good, clean packaging and acoustics. I think I'll try one.

Looks like Dell is going cheap low-end to compete on price, and using the Alienware brand for the good stuff. We'll see.

I *want* to pay a lot for a really solid PC. What I don't want is to pay a lot and get the same old junk.

John

I've been using Dow Corning DC-340 (or whatever it is) ever since I lifted a tube from the USAF, ca. 1970. I don't know what it costs, or if there's something better out there, but I've never had a problem with it. In fact, the last couple of new CPUs and heatsinks I've bought, I wiped off their goo and put the Dow Corning.

I guess I'm in the wrong thread - when I need a new computer, I just hop up to Fry's and get a $30.00 AMD, a $79.95 MB that it will run on, a memory stick or two for about $50, and a 120GB drive for about $100.00. Oh, and a case and power supply - the last one was $25.00.

I was very impressed with Fry's not too long ago when I had bought a new MB/CPU, and it didn't work, but I didn't know why. So I took it back, and asked, "Is there somebody that can look at this?" and they did, and told me exactly what the problem was - wrong power supply. At that time, a case/PSU was about $35.00.

Cheers! Rich

servers?

reliable," do you mean:

The occasional hour or so down isn't as scairy as all the time it takes to recover from a hard drive failure or a Windows suicide. It can take a week to truly recover from a crash and get all the apps and setups back where they need to be, and that's worth far more than any PC I could ever buy. That's why RAID1 and lots of disk image copies appeals to me.

I'm stuck with Windows... too many apps that I use need the pig.

John

Fry's, huh? I guess I've just had some bad luck there. I got two FIC mainboards from them when I was in school - they both died within a few years.

Oh, and my first computer - an Atari ST - my dad and I got from the Fry's in Sunnyvale. Bad monitor (left alignment), plus the floppy drives were misaligned or something - got unrecoverable bad sectors, etc.

Does Fry's carry Asus nowadays? Can't go wrong with Asus... only Asus board that died, I killed it by accident - forgot to set the power supply voltage to 220V (trip to the Philippines, was packing in a hurry). Plugged it in, turned on the power supply... big BOOM, big blue flash. (Gotta love that blue electron flash. Ooh... electrons are blue!)

Michael

You might want to take a look at Sun rackmounted x86 servers, they seem well made. Never tried them with MS software, though they do claim it works...

Chris

All things PC are crap!

HP boxes are nice though; everything is click-on and no screws required.

Why put all your eggs in one basket on drive 3 & 4 and one the same box which might get nicked or hit by a meteorite - and it will be noisy as hell???

You pay good money for RAID, so you should use it and not subvert it with D.I.Y partitioning schemes ;-). RAID will protect your data from single drive failure, that is sort of the point of using RAID - so it is IMO better to buy one box with four or five drives in one RAID cluster and spend the money for the second box on a networked tape drive for the backups.

A.F.A.I.K: It is because that you need the server/entreprise edition of Windows XP to officially get the RAID support (Win XP "vanilla" can indeed be hacked to support RAID so it is more or less the same code and obviously more a licensing issue - but why bother, if you are blowing money on proliant servers, the OS is the cheap part).

Maybe stop looking for a do-everything-in-the-same-box solution: Store the data on a network drive sitting in a reliable server with tape backup and RAID running whatever server OS the vendor likes; then work with the data on a cheap, crap PC with windows XP?

I've ordered an Alienware box (see links in my post above) that looks nice. It appears to be very clean, quiet, and has front-panel hot-plug drives that are supported under XP. If I do a system build with all applications set up (a non-trivial piece of work) and RAID that, I can also drop off a couple of image copies and stash those drives in baggies, so if the hardware or software ever gets lunched, I just pop in an image copy of the system and I'm back up. Then I can restore mail and project files easily from offline backups.

I'm not so concerned with losing files as with the hassle of reinstalling everything, getting all the mail and applications and license files and such back working after a crash.

If I get a second identical PC for home, I can also run a set of hot-plug drives anywhere, or image copy one system to the other.

John

That's not my experience - the only hard disk I have seen that has had trouble is on a 12 year old computer, and even that is possibly the result of gradual Win95 FAT32 corruption rather than a physical failure. On the other hand, I have had to replace power supplies on several computers, almost always as a result of fan failure. I've also had to replace several cpu and motherboard fans over the years.

If you do a regular, reliable backup of data to another PC, there is no need for mirroring on the same PC. Mirror "raid" is mostly used for speed (of reading, not writing), not security - you have no way of knowing which of your mirrors is correct in the event of a partial failure.

It doesn't help for bad sectors, as there is no way to tell which copy is good, and which is bad.

The best backup is off-site, so that it covers fire and theft as well. I recommend dirvish, running from overnight scripts to copy snapshots over an ADSL line to an off-site backup.

I've replaced plenty of CPU fans, but at least with Linux a bad fan isn't catastrophic.

The type of bad sectors I see are the ones that can't be read off the disk due to ECC errors. So it's pretty easy to tell which is the good one.

I don't quite see how a better OS helps with bad hardware. I suppose the difference is that when a Linux file system (such as ReiserFS or ext3) claims to use logging to prevent corruption on an unexpected crash, it really works, while on windows, NTFS is only a marginal improvement over FAT for safety.

If that's the case, then fair enough. I don't know whether that is what happens normally when disks fail - I have so seldom seen failed disks that I couldn't say.

Linux seems better at idling the processor when it's not busy, which reduces the total amount of heat generated. Thus, it takes more effort to generate the heat needed to stop the processor.

That could be true - in the Linux server I have that stopped because of heat problems, much of the heat came from the hard disks rather than the processor, as far as I could tell. Linux also has less overheads, especially when running as a server - no gui saves a lot of effort.

On a sunny day (Sun, 03 Dec 2006 19:18:53 +0100) it happened David Brown wrote in :

I notice a few misconceptions here. Many servers run X these days, and X only uses about 1% to 20% CPU resources: Type top, and look for X: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 2646 root 15 -1 162m 67m 9.8m S 14.9 17.9 553:08.76 X

It is not so Linux has lower CPU usage, my win98 (last MS stuff I have) runs cooler then the Linux server on my box. Stopping a procesor because of overheating is a rare thing indeed. I have a script that runs lmsensors every second and sounds alarm if any voltage or temperatue or some fan speeds is out of range. This is _required_ for a server I think, with remote signalling if must be.

As to bad hardware, bad is bad. OS does not help. Do not run bad hardware. Run memory test, from BIOS or whatever, compile a kernel, if all that is smooth and error free you are likely good to go. Bad hardware will show up often during big compiles and memory tests.

Alien P.

If you have speedstep, cpuspeed can monitor the cpu temperature and reduce the cpu speed if it starts to overheat.