Building a box
-
Have built every computer I've ever owned apart from the laptops. Pricewise I generally find it's cheaper than finding a PC of similar specs and more satisfactory because I specify what I want and don't need to waste extra money, just get the absolute minimum I need, for example pilfering spare hard-drives from other machines etc. The hardest bits I feel are getting the software installed, while not particularly hard I've had more gotcha's than with the hardware. As John said the AMDs are cheaper and I find them to be much more configurable. A plus side is that they tend to allow forwards-or-backwards compatability more. For example I stuck with AMD for my latest desktop because unlike Intel (at the time anyway, I think) they said that the new quad-core processors would be backwards compatible with the "old" AM2+ sockets which were the only ones available at the time. You do get a bargain, I remember that my 2.0GHz AMD X2 cost £50 but can be overclocked to something like 2.5GHz of which at the time you were spending almost half-again for the processor (not that I run it that high, it is overclocked to about 2.15GHz just by a simple option in the BIOS). Once you've done it a couple of times it'll become second nature, I think my desktop was up and running inside 30 minutes (minus software). Vista took another ~30 minutes to install so all up and running in just over an hour. Then assign a week configuring Windows and installing software just so :rolleyes:.
Ed.Poore wrote:
they tend to allow forwards-or-backwards compatability more.
not exactly true... but a pleasantly repeated myth. AMD 754? AMD 939? No, both companies "attempt" to think ahead, but this is not easy when you don't know what is going to happen. Both companies have prototypes a generation or two ahead so have plans they can build upon, but do not know the plans of the other company. When competition and technological advancements outside the main chip require updating the socket to take advantage of the new capability, share more with a bridge chip, or any other unplanned expansion, the socket changes. This used to happen often for both companies, now it is getting rarer and rarer as more things are moved to bridge chips and your north-bridge/south-bridge are the non-compatible monsters, and the cpu's move up the chain with no more than a bios upgrade (and rarer and rarer for that).
_________________________ Asu no koto o ieba, tenjo de nezumi ga warau. Talk about things of tomorrow and the mice in the ceiling laugh. (Japanese Proverb) John Andrew Holmes "It is well to remember that the entire universe, with one trifling exception, is composed of others."
-
John Simmons / outlaw programmer wrote:
DO NOT use the heatsink/fan that comes with your CPU. Buy a better one, and preferably one that uses a 120mm fan.
This is true if you go for the AMD CPU. Retail Intel CPUs come with pretty beefy heatsinks and you shouldn't need to worry about buying a bigger one unless you're planning to overclock. Edit: This used to be true, but it may not be anymore. I haven't built an AMD box since the K7 came out.
Imagine that you are hired to build a bridge over a river. The width of the river increases slightly every day, except when it shrinks. Your budget does not allow for you to use concrete or steel - you can only afford timber and cut stone. Gravity changes from hour to hour, as does the viscosity of air. Your only tools are a hacksaw, a chainsaw, a rubber mallet, and a length of rope. Welcome to my world. -Me explaining my job to an engineer
modified on Tuesday, April 29, 2008 1:38 PM
The ones that come with the AMD CPU's are fine - it's just better to use something bigger to cut down on noise (big heatsink means big fan). I haven't used a stock heatsink since 1992.
"Why don't you tie a kerosene-soaked rag around your ankles so the ants won't climb up and eat your candy ass..." - Dale Earnhardt, 1997
-----
"...the staggering layers of obscenity in your statement make it a work of art on so many levels." - Jason Jystad, 10/26/2001 -
AMD CPUs may be slightly less expensive than Intel, but I find the Intel Core 2 Duo line-up to be a far superior processor. I've also had more problems with NVidia chipsets for AMD than with Intel ones.
Anyone who thinks he has a better idea of what's good for people than people do is a swine. - P.J. O'Rourke
I'm not going to be pulled into a which is best discussion. The plain and simple truth is that he needs boxes for developers for as little money as possible (not an unreasonable request). AMD is the best bang-for-the-buck solution available. And I've never had a problem with chipsets for AMD...
"Why don't you tie a kerosene-soaked rag around your ankles so the ants won't climb up and eat your candy ass..." - Dale Earnhardt, 1997
-----
"...the staggering layers of obscenity in your statement make it a work of art on so many levels." - Jason Jystad, 10/26/2001 -
For years I had a local hardware guy, but he went out of business recently. I've bought a couple of boxes from Dell, but find expandability (e.g., proprietery hard drive mounting brackets, etc.) to be an issue. Additionally, it's becoming more of a consideration to be sure I can get XP drivers as I've backed away from Vista, and that's touch and go with the big companies. I know some of you build your own boxes and since I've not travelled that path, I thought I'd ask a question or two. First, if you buy components from someone like NewEgg, etc. how does the price of the finished box stack up to the large manufacturers? Beyond the recording studio, I don't need Screaming Game Machines, just average dev and file serving computers, and I don't want to spend any more money on them than I have to. Secondly, how much of a learning curve is there? I understand that conceptually you get a case, drop in a motherboard, power supply, memory and some peripherals. However, since I'm not a professional hardware geek (and have no time or desire to become one), I don't know what gotchas await the uninitiated. It would be nice to be in complete control of my boxes instead of having to put up with large, draconian or apathetic corporations. However, the point is to eliminate a source of pain, not create a whole new one. How cheap and easy is this?
Christopher Duncan Author of The Career Programmer and Unite the Tribes Coming soon: Got a career question? Ask the Attack Chihuahua! www.PracticalUSA.com
If games aren't important, it's really hard to beat Dell Vostro systems by building yourself. The biggest problem I've seen with the Vostro systems are motherboards that use older Intel 965 graphics chipsets which have limited support for widescreen display resolutions. EDIT: Dell has Vostro Systems for $400, including display, which is more than capable as a development system. They've updated the motherboard over several we have at my office so they actually support the accompanying display in native resolution.
Anyone who thinks he has a better idea of what's good for people than people do is a swine. - P.J. O'Rourke
-
Well, RAID 1 (mirroring) is OK. Generally. You can still run into the problem that drives from the same batch will tend to fail in the same way at the same time, leaving you with no benefit. Try to get drives from different batches or even different manufacturers. RAID 0 is highly unsafe - the failure rate of the array is the failure rate of each drive multiplied together (e.g. a 3% chance of failure of each drive becomes a 9% chance for two drives, 27% for three drives, 81% for four drives, and it's a good thing that 3% is a massive overestimate). It can be OK if you do RAID 10 which is a striped mirror - you have a redundant drive for each stripe, so it reduces the chances of losing the whole set. RAID 5 is a bit of a compromise which gives most of the read benefit of RAID 0, but it heavily penalises writes. If you don't write a whole stripe at a time, the system must read the current contents of the stripe and corresponding parity, then write back the stripe and the parity. A whole-stripe-write only incurs the stripe and parity writes. Caching controllers can give some benefit here in accumulating writes in order to build a whole stripe, or cache the data last read or written to avoid the read penalty on write-back. With any form of RAID but especially RAID 0, make sure you're taking and verifying backups, and testing restores. RAID only protects against hardware failures. Software errors on writing to a disk will get mirrored to the other disk.
DoEvents: Generating unexpected recursion since 1991
Mike Dimmick wrote:
RAID 0 is highly unsafe - the failure rate of the array is the failure rate of each drive multiplied together (e.g. a 3% chance of failure of each drive becomes a 9% chance for two drives, 27% for three drives, 81% for four drives, and it's a good thing that 3% is a massive overestimate). It can be OK if you do RAID 10 which is a striped mirror - you have a redundant drive for each stripe, so it reduces the chances of losing the whole set.
Your math is wrong here. The proper calculation is that you have a 97% chance of non failure raised to the power of hte number of drives. 2 drives is .97^2 = 94.09% chance of nonfailure or ~6% of failure. 3 drives is 92.26, 4 88.53%. The easy way to demonstrate that your calculation is wrong is to look at 2 drives with a 10% failure rate. The odds of both failing isn't 10*10=100%.
You know, every time I tried to win a bar-bet about being able to count to 1000 using my fingers I always got punched out when I reached 4.... -- El Corazon
-
Oh, cool thanks. I didn't know that, I'll look very carefully next time if I do that again. The last combo I brought (my current PC) was quite specific, it had a named brand (biostar) motherboard and PSU that I had checked out beforehand. I haven't had any trouble with them so far.
Simon
There are good combos out there, but there're also rotten ones. Tigerdirect and it's machspeed ones are especially bad, their 'better' ones generally use nVidia chipsets which are currently only worth buying for SLI because otherwise Intels are better in terms of reliability and features.
You know, every time I tried to win a bar-bet about being able to count to 1000 using my fingers I always got punched out when I reached 4.... -- El Corazon
-
For years I had a local hardware guy, but he went out of business recently. I've bought a couple of boxes from Dell, but find expandability (e.g., proprietery hard drive mounting brackets, etc.) to be an issue. Additionally, it's becoming more of a consideration to be sure I can get XP drivers as I've backed away from Vista, and that's touch and go with the big companies. I know some of you build your own boxes and since I've not travelled that path, I thought I'd ask a question or two. First, if you buy components from someone like NewEgg, etc. how does the price of the finished box stack up to the large manufacturers? Beyond the recording studio, I don't need Screaming Game Machines, just average dev and file serving computers, and I don't want to spend any more money on them than I have to. Secondly, how much of a learning curve is there? I understand that conceptually you get a case, drop in a motherboard, power supply, memory and some peripherals. However, since I'm not a professional hardware geek (and have no time or desire to become one), I don't know what gotchas await the uninitiated. It would be nice to be in complete control of my boxes instead of having to put up with large, draconian or apathetic corporations. However, the point is to eliminate a source of pain, not create a whole new one. How cheap and easy is this?
Christopher Duncan Author of The Career Programmer and Unite the Tribes Coming soon: Got a career question? Ask the Attack Chihuahua! www.PracticalUSA.com
I've built one or two with no trouble at all. But not from mail-order; I've gone to the local Fry's to pick out the parts.
-
For years I had a local hardware guy, but he went out of business recently. I've bought a couple of boxes from Dell, but find expandability (e.g., proprietery hard drive mounting brackets, etc.) to be an issue. Additionally, it's becoming more of a consideration to be sure I can get XP drivers as I've backed away from Vista, and that's touch and go with the big companies. I know some of you build your own boxes and since I've not travelled that path, I thought I'd ask a question or two. First, if you buy components from someone like NewEgg, etc. how does the price of the finished box stack up to the large manufacturers? Beyond the recording studio, I don't need Screaming Game Machines, just average dev and file serving computers, and I don't want to spend any more money on them than I have to. Secondly, how much of a learning curve is there? I understand that conceptually you get a case, drop in a motherboard, power supply, memory and some peripherals. However, since I'm not a professional hardware geek (and have no time or desire to become one), I don't know what gotchas await the uninitiated. It would be nice to be in complete control of my boxes instead of having to put up with large, draconian or apathetic corporations. However, the point is to eliminate a source of pain, not create a whole new one. How cheap and easy is this?
Christopher Duncan Author of The Career Programmer and Unite the Tribes Coming soon: Got a career question? Ask the Attack Chihuahua! www.PracticalUSA.com
Having been working a hardware guy / system support for many years before I have to say that you don't build yourself to save money, you simply won't. You do it for the personal control and to get exactly what you want. There's not much of a learning curve I have to admit. But keeping up to date with the constant flow of new products is something for the geeky mind. I personally always put some extra time and money into getting the best most appropriate boxes and motherboards. Don't cheap out on memory. Bad memory and/or overclocking are the most common reasons for BSOD and other odd behaviours. I always buy RAM specifically aimed for the system/motherboard. If uptime is an issue for your fileserver, then go for a brand server with proper support, or make sure you have spare parts on stock. You might get surprised how short shelf life some products have. If RAID is something you want I definitely recommend brands products like LSI or Intel and the usage of hot spares not the least. If the controller card or the second hard drive breaks it's not fun at all.
-
Having been working a hardware guy / system support for many years before I have to say that you don't build yourself to save money, you simply won't. You do it for the personal control and to get exactly what you want. There's not much of a learning curve I have to admit. But keeping up to date with the constant flow of new products is something for the geeky mind. I personally always put some extra time and money into getting the best most appropriate boxes and motherboards. Don't cheap out on memory. Bad memory and/or overclocking are the most common reasons for BSOD and other odd behaviours. I always buy RAM specifically aimed for the system/motherboard. If uptime is an issue for your fileserver, then go for a brand server with proper support, or make sure you have spare parts on stock. You might get surprised how short shelf life some products have. If RAID is something you want I definitely recommend brands products like LSI or Intel and the usage of hot spares not the least. If the controller card or the second hard drive breaks it's not fun at all.
Jörgen Andersson wrote:
Don't cheap out on memory. Bad memory and/or overclocking are the most common reasons for BSOD and other odd behaviours.
agreed. To make sure this isn't a problem run memtest86 overnight, and then run one thread of prime95 per core (affinity is a command line option) in each of the 3 stress testing modes (remembering to scale the ram hog one down to avoid swapping to disk) for 24hrs each. If you're overclocking then back off 100mhz from the max you were prime95 stable. (Not technically required but both of my a64 boxes had once a month bsods until I did this.)
You know, every time I tried to win a bar-bet about being able to count to 1000 using my fingers I always got punched out when I reached 4.... -- El Corazon
-
Mike Dimmick wrote:
RAID 0 is highly unsafe - the failure rate of the array is the failure rate of each drive multiplied together (e.g. a 3% chance of failure of each drive becomes a 9% chance for two drives, 27% for three drives, 81% for four drives, and it's a good thing that 3% is a massive overestimate). It can be OK if you do RAID 10 which is a striped mirror - you have a redundant drive for each stripe, so it reduces the chances of losing the whole set.
Your math is wrong here. The proper calculation is that you have a 97% chance of non failure raised to the power of hte number of drives. 2 drives is .97^2 = 94.09% chance of nonfailure or ~6% of failure. 3 drives is 92.26, 4 88.53%. The easy way to demonstrate that your calculation is wrong is to look at 2 drives with a 10% failure rate. The odds of both failing isn't 10*10=100%.
You know, every time I tried to win a bar-bet about being able to count to 1000 using my fingers I always got punched out when I reached 4.... -- El Corazon
Yeah, I'm an idiot on that one. Percentages do tend to throw me sometimes. (The calculation of 'what should the excluding VAT price be if the including VAT price is X' is non-intuitive, it's X / 1.175 if your VAT rate is 17.5%.) Still, probability of failure of a RAID 0 array is greater than the probability of failure of each individual disk, and that's reason enough to take more care with the array than you would with a standalone disk.
DoEvents: Generating unexpected recursion since 1991
