Just out of curiosity, has anyone out there in Slashdotland had good luck with enterprise IDE solutions? Who knows. Perhaps some success stories might change my pro-SCSI/fibre view.
As always, if you treat the disks properly, they're going to be almost as reliable as SCSI/FC. Just don't expect the same performance (you'll need roughly 2x the number of spindles... maybe more). But you get a lot of capacity for 1/3 to 1/4 the cost of SCSI.
Make sure the drives have (a) quality power feeds and (b) proper cooling and they'll do just fine.
(And I prefer to go after the 3-year / 5-year warranty drives which puts the manufacturer on the hook.)
I wouldn't consider the 750GB drives to have an insane price premium. At least, not compared to historical prices. They're down under the $0.50/GB mark (but not as cheap as the ~$0.28/GB price point of lower-capacity drives). If they were still up around $500, I'd have more of a complaint.
Are they worth the price premium for reduced heat / power requirements? Maybe... especially if it lets you pack double the density into an existing enclosure without spilling over to another enclosure. (Enclosures probably add $200/disk as a base cost, so putting the largest possible drives in makes some sense.)
Note that the above is a full-tower using a standard ATX motherboard. Price for the PSU would be around $600 for a redundant 750W unit (which I *think* will fit inside the case). We're bonding some NICs together to give better bandwidth and fail-over capability. Also, we can fit (3) additional hard drives inside the case but they won't be as easy to swap.
The dual-core CPU is going to become a bottleneck at some point. We'll probably upgrade to a quad-core AM2 CPU if they ever become available down the road. I've seen loads of 25-40% total CPU time while building the RAID10 array.
Total space is 4200GB in RAID1/RAID10 setup (possibly 4900GB if we use 2 of the internal bays). Figure roughy $1.50/GB for a fully populated unit. But we're only starting with a (7) drive unit (700GB RAID1 + 1400GB RAID10 + hot-spare) and will add the other (8) drives down the road.
Rebuild rate for a RAID1, 2-drive, 750GB SATA set is around 75MB/s. (Raw read/write rates for 750GB drives are around 75MB/s as well.) So figure 3 hours to rebuild a RAID1 array.
Not sure what rebuild rates would be on a RAID5, probably about half of that? So 6 hours to rebuild the array?
(That's using 750GB SATA drives with Software RAID on a PCIe motherboard.)
300GB RAID1 - data, databases, subversion
600GB RAID10 - backups (rsync snapshots or Second Copy via Samba)
Offsite backups are a set of 300GB external USB/Firewire drives, rotated periodically. Those drives pretty much only hold a snapshot at a current point in time. They also concentrate on non-replaceable data rather then system configuration.
System config is a combination of storing any edited config files in Subversion along with weekly snapshots via Dirvish / RSnapshot / RSync. The SubVersion allows me finer control over config changes while the snapshots catch more drastic errors where I have to restore the entire O/S. The O/S snapshots can also be offloaded to read-only media or tape for long-term storage.
At those prices, we'll definitely not bother to upgrade our existing systems from WinXP Pro to WinVista Business. If those upgrade prices were about 50% of what is listed, it would be worthwhile. There's just no worthwhile gain in moving from WinXP to WinV.
(And to think I already thought that WinXP Pro prices were too expensive.)
I won't hand-build machines for other people either (support and warranty issues). I'll offer advice, parts-lists and hand-holding, but they have to do the grunt work and the O/S installation. I also make sure we capture an image of the drive using Knoppix+NTFSClone as well as installing a 2nd HD and some sort of backup software (Second Copy 2000 or XCOPY).
And you can do better then Dell's prices while still getting good quality parts. The downside is the 1-2 hours of assembly (depends on familiarity with the case and motherboard) and the time spent installing the O/S and applications. But the O/S and application installs can happen on the side while you do something else.
To give you an idea of what we're building for the office:
$0131 AA15070 WindowsXP Pro OEM
$0299 AA24200 Microsoft Office Pro 2003 OEM
$0148 MB-BA22656 AMD ATHLON 64 X2 3800+ AM2
$0084 (bundle) Asus M2NPV-VM
$0138 (bundle) Mwave 2GB DDR2 533 2x1GB RAM
$0009 (bundle) Assemble and Test
$0103 BA30107 Antec Sonata II w/ SmartPower 2.0 450W PSU
$0084 AA27440 SEAGATE 200GB ST3200827AS SATA 7200RPM 8MB
$0035 AA32590 NEC ND-3550A BLACK
$0008 AA36600 NEC Floppy (BLACK)
=====
$1039 ($740 w/o MS Office 2003)
Add another $75-$100 for a half-decent PCIe video card (in which case you should probably go with the ASUS M2N-E Socket AM2 motherboard for another $10). Shipping costs will be around $75 (order everthing at the same time).
Estimated useful lifespan on these units is 8-12 years. Assuming no hardware defects such as bad capacitors on the motherboard. We're giving up the Dell warranty because it doesn't last past 3 years anyway. (And most parts come with 3yr or 5yr warranties.) Everything is commodity parts, which means a quick trip to the local CompUSA is the worst-case solution to fixing an issue.
Pricing something equivalent in power from Dell results in prices that are $1250-$1550. Plus we're able to re-use older components (floppies, DVD-ROMs, CD-ROMs, hard drives, cases, power-supplies) from older systems to cut some costs.
So it's not overly expensive to DIY. Motherboard bundles make it almost fool-proof compared to a few years ago. Hardest part is figuring out which motherboard to use and what case to put it all in.
Sidenote: I heard the phrase "I couldn't care less" on a BBC radio show from the mid-1950s (The Goon Show).
So the phrase has been around for a while and doesn't seem to be a very recent invention.
The older Antec p160 case was nicer to work with then the p180b. The problem with the older p160 is (a) motherboard tray that limits you to exactly ATX spec boards and no larger (b) only so-so cooling in the front drive bays even with a dedicated intake fan. Still, the older p160 is a dream to get the side panel off and swap out a drive.
The p180/p180b does a nicer job of cooling. But replacing drives in the lower 2 drive bays involving disconnecting multiple drives. I don't care for the blowhole on top though and feel that it invites disaster (spill a drink on top of that case and it goes directly to the motherboard). Still, I can pack 10 drives into one of these units (using a Coolermaster 4:3 bay cooler up top). The other issue with the blowhole is that it allows CPU fan noise to exit directly out of the case and almost a straight shot towards the user. (I have a dual-Opteron setup in mine.)
For normal builds, I prefer the Antec Sonata II or the Antec Overture II. The Minuet 300 is also an interesting size but with a few more tradeoffs. The Sonata's are a god fit for a gamer case (while being quiet), while the Overtures are more about small size.
The ThermalTake Armor case with the solid side panel is also interesting. Figure I can get 15 drives into that case along with a heavy-duty motherboard and PSU. Three internal drives and 12 drives in hot-swap chassis up front.
I have at least one of every case I've mentioned. The Sonata and p160 cases I have multiple of because I like them so much.
The thing is, I've built my own PC roughly every 3-4 years since forever. I'm getting too old (or at least too impatient) to mess around with heatsink glue, dealing with a mobo component vendor who insists it must be the other vendor's fault for supplying a dodgy processor, and all that jazz.
Motherboard bundles fix a lot of that particular issue. The seller puts the CPU/RAM/MB together and tests it (for a very small fee). Prior to finding someone who sold motherboard bundles, it was a pain to figure out which CPU goes with which RAM and which motherboard.
(I build around a system per month, currently 2-per month since we're doing a big upgrade cycle.)
I remember not that long ago when socket 939 came out that AMD said that this was going to be the socket they were going to stay on for a looong time and that the sacrifices of obsoleting the 754 and 940 were totally worth it: when AM2 came out so soon after it really made me wonder, why is there a need for a new socket right now? It's not like X2AM2 chips are that much different from X2939 ones...
I sorta remember that too, so I went and looked. Socket 939 came out in June 2004 (or thereabouts). This could be part of the downside of having a built-in memory controller. Since DDR reached the end of it's reach (PC3200 seeming to be the fastest commonly available) and DDR2 commonly available, they decided to go with a new pin-out so that you couldn't mistakenly mix/match the wrong CPU with the wrong memory. Less confusion for the customer.
At least, I think that's why the pin-out was changed... (according to the AMD FAQ it was).
What AMD has said at this time is that the new AM3 chips (which support DDR3) will be compatible with AM2 and AM3 motherboards. So you can put an AM3 chip into an older AM2 motherboard, but not the other way 'round. We'll see if that holds true...
Intel is already on 65nm. The last I read about AMD's move to 65nm on the tech sites was that it was next year, by the time Intel will already be moving to 45nm. AMD is officially a generation behind in that department.
I'm reading rumor and news leaks that say more along the lines of Q4/2006 for the 65nm production. The new Opteron socket F CPUs are due out soon and I'm not sure whether they're going to be 90nm or 65nm. And that could slip until 1Q/2007 or later of course. (I confess that I mostly scan the news looking for when AMD's next price cut is going to take place. Because we mostly buy CPUs in the lower half of the Athlon64/Opteron lineup.)
AMD is always going to be a generation behind on process size. It's the nature of being a smaller company (have to make do with less). So the downside is that AMD is always going to lag 6-24 months behind the process shrink of Intel. On the other hand, they've done very well with 90nm technology. So while the 65nm shrink won't be a magic bullet, it will hopefully at least narrow the gap (in terms of overall performance, power consumption and thermal output).
It's a good time to be a customer. For either side of the fence. Reasonably priced chips that are multi-core, hardware virtualization (making things easier to run multiple O/S instances), and increased competition between the two sides.
(We're mostly rolling out AM2 X2 3800+ CPUs in our desktops because it's reasonably priced. If they fire-sale the old 939 X2s, we may pickup a few more of them. Odds are low that we'll upgrade CPUs down the line, but it's a reason to favor AM2 over 939. The Core 2 Duos are still more then we want to pay for a desktop CPU and we don't want to buy end-of-life Pentium D 805/9xx dual-cores. If a low-end Core 2 Duo was $120-$150, I'd consider it for these particular machines.)
For me it was decided in favor of the non-Linksys solution because I had a VIA C3 600MHz fanless system sitting idle. So I dropped 2 laptop HDs in, rebuilt the unit and installed Smoothwall. Smoothwall configuration took about an hour since I wasn't doing anything complex at the start.
I'm not exactly sure how much power it draws, but the attached 1000VA Smart-UPS estimates that it has about 3 hours of run-time during a power outage. There's also a KVM switch, a DSL modem, and a 16-port gigabit switch plugged into that UPS. One of these weeks I'll drag out the Kill-A-Watt meter and find out for sure how many watts that C3 unit is drawing.
Yes, and before that they had probably dropping really, really slowly down to $200 for quite some time as processor prices has been stagnant for the last 1-2 years. People who bought these processors just when prices stagnated have had them almost "for free" until a few weeks ago. People who bought them a few weeks ago have already seen the price of their new processor cut to half.
You can also pull data like that off of pricescan.com for individual parts (but not overall). For example, the Athlon64 X2 3800+ was ~$350 in Oct 2005 and only dropped to ~$300 by Jun 2005. The X2 4200+ started around $500 in Aug 2005 and was only down to $400 by Jun 2005.
Which was a bit annoying from the consumer-side, because the prices never dropped like we hoped they would after the initial rollout. (Heck, we still haven't seen price breaks on the Opteron lines. And I think those are now overpriced compared to the Intel Core 2 Duo chips.)
Still, I've heard rumors that there's another round of price breaks coming from AMD within a few months (October). But that only seems to be on the Turion mobile chips and the low-end Sempron chips.
A price of $80 for the X2 3800+ would be nice after all this waiting... the $150 price is pretty good, but more is always nicer.
You find that you spend less time planning your suicide than you used to.
Heh, no kidding, I've found that to be a very useful benchmark of whether or not the meds / therapy are working or not. I much prefer to be able to go to sleep at night rather then tossing and turning for hours as I debate the ways and means of suicide.
I'd recommend the newer Practical Cryptography (by Niels Ferguson and Bruce Schneier) over Applied Cryptography. The forward is an interesting read and highlights some of the problems with the Applied Cryptography text (and focus of the earlier book). Applied Cryptography is probably better if you're going to write the encryption algorithms yourself.
Practical Cryptography is written with a bit more hindsight because of watching all the sub-standard encryption systems that have appeared over the last decade because folks think encryption is "magic" that solves all ills. There's a few pages where the authors basically say "well, we thought it was a good recommendation at the time, but we didn't fully account for the human factor".
10x improvement over the last 10 years is about right (give or take 2 years).
A Celeron 566MHz has a memory bandwidth around 150-175MB/s. An Athlon64 3200+ has a memory bandwidth around 1600MB/s. (Approximate numbers, with all sorts of undocumented assumptions.) So it's a pretty good bet that modern single-core CPUs are at least 10x as fast as the ones from 8-10 years ago.
Of course, now we have things like dual/quad CPU machines with 2-4 cores per machine... so we could up that number to 150x as fast as a machine from 8-10 years ago. So instead of 3000 machines, we might only need 20-30.
I consider easy-to-camp spawn points to be a design issue with the map.
On maps that I design (for CoD:UO), I try to strike a balance between it being possible to spawn-kill (a valid part of FPS tactics in my book) while making it possible for a team to drive off a spawn-killer. That means spawn area designs that:
- Don't spawn everyone in the same location (in maps where there are distinct "sides")
- Spawn players in locations that are mostly sheltered with some cover
- Spread the spawn points in a way that they mutually support each other
- Players should spawn in a location that allows them to size up the environment from cover
- There should be at least 2 ways out of the spawn location
- Provide access to anti-tank weapons if tanks/vehicles are a factor in a mass spawn-kill situation
Generally, we'll see mass spawn-kill situations that only last about 2 minutes before the team takes back their spawn points. By that point, the team will rally back because they spawn next to anti-tank weaponry making the life of an enemy tank very short in the spawn area. A lone rogue spawn-killer has to get inside the spawn area to do damage is generally limited to one kill here and there.
Unfortunately, it's more of an art then a science and there are no hard and fast rules. Plus you have to understand how the spawn system works behind the scenes. And play testing will uncover all sorts of situations that you didn't anticipate.
Is there some strange reason that Slashdotters think that the only use for wireless networks is browsing the Internet? None of you have ever used wireless to print, or copy a file off a server, or play a LAN game, or stream video, had more than one wireless device running?
Sometimes I think that all of the real-world geeks have got up and left. You know, the folks who work on networks at places that are willing to pay them money to do so?
That or newbie Friday has been held over into Saturday this week...
(Goes back to researching FC, iSCSI, SANs, switches, mesh topology...)
Is everyone else downloading 1GB files that they need Right Now Dammit?
Yes, aren't you?
Try telecommuting when a particular job involves media files. Fortunately it's rare, but it's getting more and more common at our company where job folders are starting to climb into the double and triple digit megabyte range.
(I'm starting to wonder if a 6TB SAN is going to be big enough...)
That assumes that he has shell access on the server. Which isn't all that common at the low-end of web hosting.
I probably would've run an FTP client with multiple threads (say 4 or 8) to better maximize the throughput. That way while some threads are waiting for responses, you've still got data transfering. Unfortunately, that requires that your provider allows for multiple connections to the FTP server (somewhat uncommon).
the 3600x2 might get squeezed into a system at that pricepoint.
Since the X2 3800+ sells for ~$148, does anyone have information as to what the X2 3600+ chips are going to sell for? ne article says that they're being released into the channel for $149 but that there would probably be price cuts.
It would have to be down around $100 retail in order to compete with the low-end clearance of the old Intel 805 dual-core Pentium 4s.
Good for 5-8 years of use? Go with a business-class machine from Toshiba (Tecra M5), Lenovo (Thinkpad T60/T60p) or Apple (MacBook Pro and maybe the MacBook). All of those options are going to cost you $2000-$3000 for a fully configured and well-built machine with a 3 year or 5 year warranty.
Good for playing games on? Look for a machine that has lots of bells and whistles but will only last 2-3 years.
Slashdot Mirror
Just out of curiosity, has anyone out there in Slashdotland had good luck with enterprise IDE solutions? Who knows. Perhaps some success stories might change my pro-SCSI/fibre view.
As always, if you treat the disks properly, they're going to be almost as reliable as SCSI/FC. Just don't expect the same performance (you'll need roughly 2x the number of spindles... maybe more). But you get a lot of capacity for 1/3 to 1/4 the cost of SCSI.
Make sure the drives have (a) quality power feeds and (b) proper cooling and they'll do just fine.
(And I prefer to go after the 3-year / 5-year warranty drives which puts the manufacturer on the hook.)
I wouldn't consider the 750GB drives to have an insane price premium. At least, not compared to historical prices. They're down under the $0.50/GB mark (but not as cheap as the ~$0.28/GB price point of lower-capacity drives). If they were still up around $500, I'd have more of a complaint.
Are they worth the price premium for reduced heat / power requirements? Maybe... especially if it lets you pack double the density into an existing enclosure without spilling over to another enclosure. (Enclosures probably add $200/disk as a base cost, so putting the largest possible drives in makes some sense.)
Rough parts list for a sizeable server that we're slowly building:
$4200 (12) 750GB drives
$0159 Thermaltake Armor VA8000BNS Black Chassis: 1.0mm SECC
$0190 Thermaltake toughpower W0117RU ATX12V/ EPS12V 750W
$0035 DVD-RW (BLACK)
$0050 misc parts (fans, cables)
$0182 MB-BA22658 AMD Athlon64 X2 4200+ AM2 (WINDSOR)
$0200 XXXXXXXXXX Asus M2N32-SLI Deluxe motherboard (ATX)
$0210 XXXXXXXXXX 1GBx2 ECC memory modules PC2 4200 DDR2 533
$0009 XXXXXXXXXX Assemble & Test
$0334 (2) INTEL PRO/1000 PT DUAL PORT EXPI9402PT gigabit PCIe x4
$0500 8-port SATA-II PCIe x4 RAID adapter (undecided)
$0367 4:3 SATA hotplug backplanes (iStarUSA BPU-340SATA)
=====
$2236 + $4200 for drives
Note that the above is a full-tower using a standard ATX motherboard. Price for the PSU would be around $600 for a redundant 750W unit (which I *think* will fit inside the case). We're bonding some NICs together to give better bandwidth and fail-over capability. Also, we can fit (3) additional hard drives inside the case but they won't be as easy to swap.
The dual-core CPU is going to become a bottleneck at some point. We'll probably upgrade to a quad-core AM2 CPU if they ever become available down the road. I've seen loads of 25-40% total CPU time while building the RAID10 array.
Total space is 4200GB in RAID1/RAID10 setup (possibly 4900GB if we use 2 of the internal bays). Figure roughy $1.50/GB for a fully populated unit. But we're only starting with a (7) drive unit (700GB RAID1 + 1400GB RAID10 + hot-spare) and will add the other (8) drives down the road.
Rebuild rate for a RAID1, 2-drive, 750GB SATA set is around 75MB/s. (Raw read/write rates for 750GB drives are around 75MB/s as well.) So figure 3 hours to rebuild a RAID1 array.
Not sure what rebuild rates would be on a RAID5, probably about half of that? So 6 hours to rebuild the array?
(That's using 750GB SATA drives with Software RAID on a PCIe motherboard.)
300GB RAID1 - data, databases, subversion
600GB RAID10 - backups (rsync snapshots or Second Copy via Samba)
Offsite backups are a set of 300GB external USB/Firewire drives, rotated periodically. Those drives pretty much only hold a snapshot at a current point in time. They also concentrate on non-replaceable data rather then system configuration.
System config is a combination of storing any edited config files in Subversion along with weekly snapshots via Dirvish / RSnapshot / RSync. The SubVersion allows me finer control over config changes while the snapshots catch more drastic errors where I have to restore the entire O/S. The O/S snapshots can also be offloaded to read-only media or tape for long-term storage.
At those prices, we'll definitely not bother to upgrade our existing systems from WinXP Pro to WinVista Business. If those upgrade prices were about 50% of what is listed, it would be worthwhile. There's just no worthwhile gain in moving from WinXP to WinV.
(And to think I already thought that WinXP Pro prices were too expensive.)
I won't hand-build machines for other people either (support and warranty issues). I'll offer advice, parts-lists and hand-holding, but they have to do the grunt work and the O/S installation. I also make sure we capture an image of the drive using Knoppix+NTFSClone as well as installing a 2nd HD and some sort of backup software (Second Copy 2000 or XCOPY).
And you can do better then Dell's prices while still getting good quality parts. The downside is the 1-2 hours of assembly (depends on familiarity with the case and motherboard) and the time spent installing the O/S and applications. But the O/S and application installs can happen on the side while you do something else.
To give you an idea of what we're building for the office:
$0131 AA15070 WindowsXP Pro OEM
$0299 AA24200 Microsoft Office Pro 2003 OEM
$0148 MB-BA22656 AMD ATHLON 64 X2 3800+ AM2
$0084 (bundle) Asus M2NPV-VM
$0138 (bundle) Mwave 2GB DDR2 533 2x1GB RAM
$0009 (bundle) Assemble and Test
$0103 BA30107 Antec Sonata II w/ SmartPower 2.0 450W PSU
$0084 AA27440 SEAGATE 200GB ST3200827AS SATA 7200RPM 8MB
$0035 AA32590 NEC ND-3550A BLACK
$0008 AA36600 NEC Floppy (BLACK)
=====
$1039 ($740 w/o MS Office 2003)
Add another $75-$100 for a half-decent PCIe video card (in which case you should probably go with the ASUS M2N-E Socket AM2 motherboard for another $10). Shipping costs will be around $75 (order everthing at the same time).
Estimated useful lifespan on these units is 8-12 years. Assuming no hardware defects such as bad capacitors on the motherboard. We're giving up the Dell warranty because it doesn't last past 3 years anyway. (And most parts come with 3yr or 5yr warranties.) Everything is commodity parts, which means a quick trip to the local CompUSA is the worst-case solution to fixing an issue.
Pricing something equivalent in power from Dell results in prices that are $1250-$1550. Plus we're able to re-use older components (floppies, DVD-ROMs, CD-ROMs, hard drives, cases, power-supplies) from older systems to cut some costs.
So it's not overly expensive to DIY. Motherboard bundles make it almost fool-proof compared to a few years ago. Hardest part is figuring out which motherboard to use and what case to put it all in.
Sidenote: I heard the phrase "I couldn't care less" on a BBC radio show from the mid-1950s (The Goon Show). So the phrase has been around for a while and doesn't seem to be a very recent invention.
The older Antec p160 case was nicer to work with then the p180b. The problem with the older p160 is (a) motherboard tray that limits you to exactly ATX spec boards and no larger (b) only so-so cooling in the front drive bays even with a dedicated intake fan. Still, the older p160 is a dream to get the side panel off and swap out a drive.
The p180/p180b does a nicer job of cooling. But replacing drives in the lower 2 drive bays involving disconnecting multiple drives. I don't care for the blowhole on top though and feel that it invites disaster (spill a drink on top of that case and it goes directly to the motherboard). Still, I can pack 10 drives into one of these units (using a Coolermaster 4:3 bay cooler up top). The other issue with the blowhole is that it allows CPU fan noise to exit directly out of the case and almost a straight shot towards the user. (I have a dual-Opteron setup in mine.)
For normal builds, I prefer the Antec Sonata II or the Antec Overture II. The Minuet 300 is also an interesting size but with a few more tradeoffs. The Sonata's are a god fit for a gamer case (while being quiet), while the Overtures are more about small size.
The ThermalTake Armor case with the solid side panel is also interesting. Figure I can get 15 drives into that case along with a heavy-duty motherboard and PSU. Three internal drives and 12 drives in hot-swap chassis up front.
I have at least one of every case I've mentioned. The Sonata and p160 cases I have multiple of because I like them so much.
The thing is, I've built my own PC roughly every 3-4 years since forever. I'm getting too old (or at least too impatient) to mess around with heatsink glue, dealing with a mobo component vendor who insists it must be the other vendor's fault for supplying a dodgy processor, and all that jazz.
Motherboard bundles fix a lot of that particular issue. The seller puts the CPU/RAM/MB together and tests it (for a very small fee). Prior to finding someone who sold motherboard bundles, it was a pain to figure out which CPU goes with which RAM and which motherboard.
(I build around a system per month, currently 2-per month since we're doing a big upgrade cycle.)
I remember not that long ago when socket 939 came out that AMD said that this was going to be the socket they were going to stay on for a looong time and that the sacrifices of obsoleting the 754 and 940 were totally worth it: when AM2 came out so soon after it really made me wonder, why is there a need for a new socket right now? It's not like X2AM2 chips are that much different from X2939 ones...
I sorta remember that too, so I went and looked. Socket 939 came out in June 2004 (or thereabouts). This could be part of the downside of having a built-in memory controller. Since DDR reached the end of it's reach (PC3200 seeming to be the fastest commonly available) and DDR2 commonly available, they decided to go with a new pin-out so that you couldn't mistakenly mix/match the wrong CPU with the wrong memory. Less confusion for the customer.
At least, I think that's why the pin-out was changed... (according to the AMD FAQ it was).
What AMD has said at this time is that the new AM3 chips (which support DDR3) will be compatible with AM2 and AM3 motherboards. So you can put an AM3 chip into an older AM2 motherboard, but not the other way 'round. We'll see if that holds true...
Intel is already on 65nm. The last I read about AMD's move to 65nm on the tech sites was that it was next year, by the time Intel will already be moving to 45nm. AMD is officially a generation behind in that department.
I'm reading rumor and news leaks that say more along the lines of Q4/2006 for the 65nm production. The new Opteron socket F CPUs are due out soon and I'm not sure whether they're going to be 90nm or 65nm. And that could slip until 1Q/2007 or later of course. (I confess that I mostly scan the news looking for when AMD's next price cut is going to take place. Because we mostly buy CPUs in the lower half of the Athlon64/Opteron lineup.)
AMD is always going to be a generation behind on process size. It's the nature of being a smaller company (have to make do with less). So the downside is that AMD is always going to lag 6-24 months behind the process shrink of Intel. On the other hand, they've done very well with 90nm technology. So while the 65nm shrink won't be a magic bullet, it will hopefully at least narrow the gap (in terms of overall performance, power consumption and thermal output).
It's a good time to be a customer. For either side of the fence. Reasonably priced chips that are multi-core, hardware virtualization (making things easier to run multiple O/S instances), and increased competition between the two sides.
(We're mostly rolling out AM2 X2 3800+ CPUs in our desktops because it's reasonably priced. If they fire-sale the old 939 X2s, we may pickup a few more of them. Odds are low that we'll upgrade CPUs down the line, but it's a reason to favor AM2 over 939. The Core 2 Duos are still more then we want to pay for a desktop CPU and we don't want to buy end-of-life Pentium D 805/9xx dual-cores. If a low-end Core 2 Duo was $120-$150, I'd consider it for these particular machines.)
It's definitely a toss-up for the hardcore geek.
For me it was decided in favor of the non-Linksys solution because I had a VIA C3 600MHz fanless system sitting idle. So I dropped 2 laptop HDs in, rebuilt the unit and installed Smoothwall. Smoothwall configuration took about an hour since I wasn't doing anything complex at the start.
I'm not exactly sure how much power it draws, but the attached 1000VA Smart-UPS estimates that it has about 3 hours of run-time during a power outage. There's also a KVM switch, a DSL modem, and a 16-port gigabit switch plugged into that UPS. One of these weeks I'll drag out the Kill-A-Watt meter and find out for sure how many watts that C3 unit is drawing.
Yes, and before that they had probably dropping really, really slowly down to $200 for quite some time as processor prices has been stagnant for the last 1-2 years. People who bought these processors just when prices stagnated have had them almost "for free" until a few weeks ago. People who bought them a few weeks ago have already seen the price of their new processor cut to half.
You can also pull data like that off of pricescan.com for individual parts (but not overall). For example, the Athlon64 X2 3800+ was ~$350 in Oct 2005 and only dropped to ~$300 by Jun 2005. The X2 4200+ started around $500 in Aug 2005 and was only down to $400 by Jun 2005.
Which was a bit annoying from the consumer-side, because the prices never dropped like we hoped they would after the initial rollout. (Heck, we still haven't seen price breaks on the Opteron lines. And I think those are now overpriced compared to the Intel Core 2 Duo chips.)
Still, I've heard rumors that there's another round of price breaks coming from AMD within a few months (October). But that only seems to be on the Turion mobile chips and the low-end Sempron chips.
A price of $80 for the X2 3800+ would be nice after all this waiting... the $150 price is pretty good, but more is always nicer.
You find that you spend less time planning your suicide than you used to.
Heh, no kidding, I've found that to be a very useful benchmark of whether or not the meds / therapy are working or not. I much prefer to be able to go to sleep at night rather then tossing and turning for hours as I debate the ways and means of suicide.
I'd recommend the newer Practical Cryptography (by Niels Ferguson and Bruce Schneier) over Applied Cryptography. The forward is an interesting read and highlights some of the problems with the Applied Cryptography text (and focus of the earlier book). Applied Cryptography is probably better if you're going to write the encryption algorithms yourself.
Practical Cryptography is written with a bit more hindsight because of watching all the sub-standard encryption systems that have appeared over the last decade because folks think encryption is "magic" that solves all ills. There's a few pages where the authors basically say "well, we thought it was a good recommendation at the time, but we didn't fully account for the human factor".
10x improvement over the last 10 years is about right (give or take 2 years).
A Celeron 566MHz has a memory bandwidth around 150-175MB/s. An Athlon64 3200+ has a memory bandwidth around 1600MB/s. (Approximate numbers, with all sorts of undocumented assumptions.) So it's a pretty good bet that modern single-core CPUs are at least 10x as fast as the ones from 8-10 years ago.
Of course, now we have things like dual/quad CPU machines with 2-4 cores per machine... so we could up that number to 150x as fast as a machine from 8-10 years ago. So instead of 3000 machines, we might only need 20-30.
Just don't camp the spawn points
I consider easy-to-camp spawn points to be a design issue with the map.
On maps that I design (for CoD:UO), I try to strike a balance between it being possible to spawn-kill (a valid part of FPS tactics in my book) while making it possible for a team to drive off a spawn-killer. That means spawn area designs that:
- Don't spawn everyone in the same location (in maps where there are distinct "sides")
- Spawn players in locations that are mostly sheltered with some cover
- Spread the spawn points in a way that they mutually support each other
- Players should spawn in a location that allows them to size up the environment from cover
- There should be at least 2 ways out of the spawn location
- Provide access to anti-tank weapons if tanks/vehicles are a factor in a mass spawn-kill situation
Generally, we'll see mass spawn-kill situations that only last about 2 minutes before the team takes back their spawn points. By that point, the team will rally back because they spawn next to anti-tank weaponry making the life of an enemy tank very short in the spawn area. A lone rogue spawn-killer has to get inside the spawn area to do damage is generally limited to one kill here and there.
Unfortunately, it's more of an art then a science and there are no hard and fast rules. Plus you have to understand how the spawn system works behind the scenes. And play testing will uncover all sorts of situations that you didn't anticipate.
Mmm, that reminds me... it's time to update my Knoppix+NTFSClone image...
Ah is that how that works... (been wondering why for the same item, shipping charges can be triple that of the other listings).
Is there some strange reason that Slashdotters think that the only use for wireless networks is browsing the Internet? None of you have ever used wireless to print, or copy a file off a server, or play a LAN game, or stream video, had more than one wireless device running?
Sometimes I think that all of the real-world geeks have got up and left. You know, the folks who work on networks at places that are willing to pay them money to do so?
That or newbie Friday has been held over into Saturday this week...
(Goes back to researching FC, iSCSI, SANs, switches, mesh topology...)
Is everyone else downloading 1GB files that they need Right Now Dammit?
Yes, aren't you?
Try telecommuting when a particular job involves media files. Fortunately it's rare, but it's getting more and more common at our company where job folders are starting to climb into the double and triple digit megabyte range.
(I'm starting to wonder if a 6TB SAN is going to be big enough...)
Why not upload one big tar/zip file?
That assumes that he has shell access on the server. Which isn't all that common at the low-end of web hosting.
I probably would've run an FTP client with multiple threads (say 4 or 8) to better maximize the throughput. That way while some threads are waiting for responses, you've still got data transfering. Unfortunately, that requires that your provider allows for multiple connections to the FTP server (somewhat uncommon).
the 3600x2 might get squeezed into a system at that pricepoint.
Since the X2 3800+ sells for ~$148, does anyone have information as to what the X2 3600+ chips are going to sell for? ne article says that they're being released into the channel for $149 but that there would probably be price cuts.
It would have to be down around $100 retail in order to compete with the low-end clearance of the old Intel 805 dual-core Pentium 4s.
In particular, who make good laptops?
Define "good".
Good for 5-8 years of use? Go with a business-class machine from Toshiba (Tecra M5), Lenovo (Thinkpad T60/T60p) or Apple (MacBook Pro and maybe the MacBook). All of those options are going to cost you $2000-$3000 for a fully configured and well-built machine with a 3 year or 5 year warranty.
Good for playing games on? Look for a machine that has lots of bells and whistles but will only last 2-3 years.