Going digital: the cost of photo storage


imageStoring photos costs next to nothing. 

You can take over twelve (12) high-resolution, properly backed up digital photos for a single penny.  

For the cost of a Starbucks latte, with tip, you can take around five thousand (5,000) high-res shots.  

Snap away

So not only should you feel free to go wild taking pics – you can also go digital for next to nothing, and free yourself from reams of paper documents every year.  

What’s the catch?   You have to make sure they’re backed up, and you have to learn to use photo management software (e.g. Google’s Picasa). 

Back story

I’ve enjoyed digital photography since 2001, when I got my first digital camera, a Nikon CoolPix 950.  I loved this little camera:  it had a cool swivel-body design and felt great in my hands. 

I ended up taking 10,000 shots with it in the first year.    Nikon Coolpix 950

Since then, I have upgraded to a Canon EOS-20D (a digital SLR and a fantastic camera) and, for around-the-house and other daily subjects, a conveniently small Canon SD1100IS, also highly recommended. 

My current DSLR is a Canon EOS-20DSomewhere around 2006, with the debut of Google’s free – and extremely easy to use – Picasa photo management software, I realized that snapping a quick picture of a paper document was an easy way to preserve it.  

Instead of makingimageand filing paper copies of my company expense reports, I snapped digital pictures of them and tagged them in Picasa as “expense reports.”   

Instead of filing auto repair bills, I snapped and tagged.   And so on.

Today, the only paper documents I keep are official legal documents, such as car titles and signed non-disclosure agreements (NDAs).   The rest are on my computer’s hard drives as digital photos – backed up and archived as well. Picasa’s incremental search enables me to find any tagged photo in seconds. 

I love being paperless now.     

From time to time, I wondered about the “real” cost of all these digital photos.  It was cheap, I knew — but exactly how cheap was it?   Could I take photos at high resolutions like eight megapixels, or should I downsize to save space? 

I decided to find out.  

Figuring the storage cost of a photo

If you’re not interested in technical number-crunching related to figuring the cost of storage, you can skip to the next section.

I’m not even going to consider the cost of the electricity needed to power the camera or the computer:  it’s surely almost nothing.  (Maybe a helpful reader will do the math?)  Nor am I going to consider the cost of the camera, since that’s a personal choice.   

Hard drive detail

Instead, the cost of a digital photo is simply the cost of the hard drive space needed to store it.   Here are some conservative assumptions of the cost of disk space in a personal computer (all currencies in U.S. dollars):

Cost of 1 terabyte (1 TB = 1,000 GB) of hard drive space = $100.  
Easy to find terabyte drives at this price point or below.

Cost of 1 gigabyte (1 GB = 1,000 MB) of storage = $0.10 = one dime.   
Just ten cents per raw gigabyte!  That is amazing to me.  

Now that we know the cost of raw hard drive storage, we need to make some assumptions about the size of each photo. Then we can simply divide one into the other to get the cost per photo: 

Size of a high-resolution (8 MP) photo is five megabytes (5 MB).   
I got this figure by looking at some actual JPEG photos I took with my Canon EOS-20D.  Since JPEG is a compressed format, picture sizes will vary for different subjects, but this seems like a reasonable, conservative number.

Now we can calculate the raw storage cost of a digital photo:

Cost of a high-resolution digital photo is $0.0005 = 1/20th of a penny.
To get this, I multiplied $0.10 per GB storage cost by  0.005 GB (that’s 5 MB for a photo).  

Put another way, you can take 20 high-resolution pictures for a single penny.

So there’s our first answer: twenty photos for a penny, or an amazing two thousand photos for a dollar.   That’s insanely inexpensive!  

But hold on – as with many things, reality is just slightly more complex.  

Figuring total cost of ownership (TCO)

For a more accurate storage cost, we should really account for storing multiple copies of our pictures.    In reality, we don’t just take the photos and store them on a single hard drive;  they need to be backed up. too.   A good rule of thumb is that anything worth preserving should be saved in at least three places.  

With only a single backup (the “originals” and a backup copy), if your primary hard drive fails or is otherwise compromised (drenched, burned, stolen, etc), you have to cross your fingers and hope that your single backup copy is good.   Not pleasant.

To protect against a single hard drive failure – namely, the hard drive on which you store your precious photos, as well as a lot of other stuff – you can set up a mirrored drive configuration (RAID-1).   Everything the system writes to a mirrored hard drive, it also writes at the same time to a second hard drive (the mirror), maintaining the two as exact copies of each other.   I use RAID mirroring on all my storage drives and highly recommend it.   Unfortunately, RAID is a slightly advanced technology – you may want to ask your favorite tech buddy for help getting started with it.  

So let’s assume that for every hard drive you buy, you’ll buy at least another two for backup, mirroring, and/or archival.   

But this still isn’t completely accurate:  it models cost as if the photos were “paying for” the entire hard drive.  In reality, the hard drive is shared: with the operating system (e.g. Windows OS, Mac OS, etc.), with installed applications, and generally with a number of  other, non-photographic data like documents and databases.   (Hopefully you are backing all this up, too.)  Accordingly, if we discount the cost of the hard drive space by something reasonable like 50%, then photo storage becomes even less expensive (twice as cheap).  

With this updated and more realistic model of keeping three copies of our photos around and sharing the “cost model” of storage within the computer, we now get:

Cost per PHOTO gigabyte (1 GB = 1,000 MB) of storage = $0.05 = one nickel.   
Divide the raw storage cost of $0.10 per GB by two to discount it. 

Storage needed for a 8 MP photo is fifteen megabytes (15 MB).   
Since we always want three copies around, we multiply the photo size (5 MB) by three.

And now we can divide one into the other to find out the “real cost” per digital photo: 

“Real cost” of a persistent high-resolution digital photo is $0.00075 = ~1/12th of a penny.
To get this, I multiplied $0.05 per GB storage cost by  0.015 GB (that’s 15 MB) per photo stored in triplicate.    “Persistent” is used here in the software engineering sense of “can or will not go away easily.”

So if there’s any bad news, it’s that the real cost of a digital photo is really a little more than what we calculated initially as the “raw storage cost.”    The good news is that a little more than “incredibly small” is still incredibly small:  we can take about twelve “persistent” photos for a penny.

Life is good.


Any way you cut it, photo storage is super-cheap!    What does this mean in our everyday life?

We can take twelve eight-megapixel photos for a penny.    That’s twelve hundred high-resolution, doubly backed up digital photos for, oh, about a buck. 

For the price of one Starbucks grande cappuccino, dry, with tip ($4), we can take and store almost five thousand high-resolution photos.   If it’s pretty, take a picture of the cappuccino, too.   

Latte art

Five thousand photos per year is about fourteen per day, every day.   Or about one hundred photos every weekend of the year.  

Five thousand high-resolution pictures will still only consume about 25 GB of primary hard drive space.  This is only a few percent of today’s desktop drives, which are 1,000 GB  (1 TB) and rising in size. 

The incremental cost of digital pictures is almost zero.  You have already sunk big dollars into the camera, the computer, and its hard drives.   Using digital photos to capture anything now costs almost nothing, whether it’s your kid drooling on the rug or every bleedin’ piece of paper that you would have filed somewhere, only to never, ever be able to find it when you need it…. 

So snap away!   If you do decide to go digital, just make sure that you back it all up.   The only question about hard drive failure is:  when?

Six failed hard drives    Think it won’t happen?  Here’s six
of my hard drives that failed.

On digitizing documents…

Don't be like thisI’ve had flatbed scanners over the years, and although I will admit it’s been a while, I don’t like them nearly as well as digital cameras for capturing documents.  

For one, a scanner is another piece of hardware to buy and maintain.  Number two, it’s another piece of software to figure out and run when you want to scan.   Thirdly, a scanner takes up valuable desk space.   And finally, a digital camera is extremely portable and can be used anywhere to “capture” a document.

To capture a paper document with a digital camera, I simply take a picture of it, making sure to include the entire document — you can always crop it later.   I try to use a flash whenever possible to minimize movement and shadow; it also makes text sharper.  Then every so often, I transfer the images to my computer and “tag” them in Picasa (e.g. “expense report” or “financial” or “bill”) so that I can find them almost instantly in the future. 

More reading…
  • Backup Tips IOMega.com
    A good primer: “A comprehensive data backup strategy is vital to your data security. Too many people wait until disaster strikes before they think about a backup. A good backup is an excellent way to protect against viruses, deteriorating hard drives, disasters and human errors. If it would not be cost effective or even possible to re-enter data, then you need a backup strategy.
  • Introducing Picasa 3  A video tutorial by Google on YouTube
    If you don’t currently use photo management software, Picasa is an excellent, free starting point.   The install is painless – try it today at Picasa.com!
  • What’s new in Picasa 3.5  Another video tutorial from Google on YouTube
  • How to store digital photos  PhotoShelter.com
    A good look at issues, cost, capacity planning, and total cost of ownership for the beginner all the way up to professional photographers.
  • Mozy.com  
    If you’re not backing up yet, this is a cheap, free, and painless way to get started.  Mozy is very easy to use, just download the software and follow the basic instructions.
  • Digital photography Wikipedia
    "Almost all of the cost of digital photography is capital cost, meaning that the cost is for the equipment needed to store and copy the images, and once purchased requires virtually no further expense outlay.”

New Core i7 PC: Selecting the Components

[tweetmeme source="KeithBluestone"]This article is part of a series in 2010 on custom-building a high-performance computer with the latest Intel Core i7 processors:

  1. Build or buy a new Core i7 supercomputer?
  2. Choosing a New CPU: Intel Core i7-920/930 vs. i7-860/870
  3. New Core i7 PC: Selecting the Components
  4. New Core i7 PC: The Build


After deciding to build my new computer instead of buy it (see “Build or buy a new Core i7 supercomputer?”), the next task was to figure out which components to buy.  Armed with a thorough review of the choices and some spare PC parts, I ordered a balanced mix of new components that, for only $600, should make me the owner of one of the faster PCs on the planet.

Before you race off to build your custom PC, peek into the future to see some of the lessons I learned on my build; hopefully it will save you some time.

The Winners

I already had a spare tower case, a lightly used and excellently rated Antec 430W True Power supply, and some enterprise-class SATA hard drives.   That left me in need of a motherboard, a processor, some memory, and a graphics card.

Here were the components I picked (and bought) as a fantastic performer that should last for at least five years (prices include tax and shipping):

  Component Cost
image Core i7-860 Processor
Powerful quad-core CPU
image ASUS P7P55D-E ATX Motherboard
Solid performer future-proofed with USB 3.0 and SATA 6 Gbps
image G.SKILL Ripjaws Series 4GB DDR3 1600 RAM
Fast, reliable memory
image XFX Radeon HD 4650 Video Card
Enough to get the job done well for me
image LITE-ON Black 24X DVD RW Player
With the old one a little finicky – why not?
  Total: $595


I’ll admit it: I spent a whole lot more time researching the system components than I ever thought I would.  I read hundreds of peoples’ comments on NewEgg and Amazon;  browsed a score of product reviews;  and sifted through countless conversations on top tech forums.

I included rating, cost, power, noise, reliability, efficiency, and good ol’ ease of use when looking at components.

=> It’s definitely a whole lot easier to go buy a Dell – particularly a top desktop like the XPS, and especially their PowerEdge servers.   You’ll pay for the convenience, naturally; but you’ll get a high-performance system instantly.  For me, taking the time to learn about the latest in computer technologies and hand-select a top-performing system was a lot of fun.

=> As a software architect by profession, it’s important to understand the details of computer architecture: how the components operate, communicate, and coordinate their jobs within the computer.   As a tech guy, I’m frequently asked for advice by friends and family on what systems or components to buy, so it’s nice to be informed and able to share well-reasoned advice.


If you’re trying to make sense out of this post (and a million other out there), it will surely help you to understand what I am looking for in this system.

  • I don’t need an “extreme” gaming-style rig, which would spike the price, suck more power, and almost certainly make more noise.   No super-overclocking motherboards, gotta-be-faster-than-you RAM, or frag-you-more dual Crossfire video card configurations.
  • I wanted a quiet, powerful, flexible system, just short of extreme: it would do everything well and last for about five years at least.

I wanted to be mindful of a computer’s primary performance bottlenecks, attacking them in a balanced fashion.  I would no more want to pair a world-class processor with slow memory than I would want a fancy graphics card when I really need a faster hard drive.

Without further ado, the components and their runners-up…

imageThe processor: Intel Core i7-860

The Core i7-860 is an all-around top performer that bests its close competitor the i7-920 in most benchmarks and uses less energy.  For a summary of the differences and why I chose the i7-860, read this article.

    Pros Cons
This is the one I chose Core i7-860, $250 Well-loved everywhere, bests i7-920 in most benchmarks, uses less energy Costs a little more initially than i7-920
  Core i7-920, $220
Core i7-930, $250
Excellent processor, almost statistically identical to the i7-860 Uses more energy, fewer & more expensive motherboards, requires RAM in banks of three
  Core i5-750, $180 Very close to i7-920 and i7-860, while less expensive. A great option if you’re cutting costs. No hyperthreading, so appears to OS as four cores instead of eight

image The motherboard: ASUS P7P55D-E

The ASUS P7P55D-E motherboard is a midrange, classic choice from a top manufacturer.  While the “midrange” designation stung for a while – who wants to be midrange? — it’s midrange only in that it lacks extra bells and whistles (dual Crossfire/SLI, dual network cards, 10-channel sound) that most of us don’t need; and it doesn’t cost as much.

This “E” series motherboard from ASUS has everything I need, plus adding a bit of future-proofing by supporting the next round of USB and SATA standards: USB 3.0 and SATA 3.0.

=> I didn’t have any problem paying for features I needed;  but I’d rather buy features that would really contribute to the overall performance of the computer.  Think fast GPU or solid-state drive (SSD).

Selecting the motherboard was the most difficult task after selecting the processor itself.  There are a myriad of options to sort through, including support for RAM, number and type of PCI card slots, max number of hard drives supported, and various permutations of audio, network, and USB/FireWire support.

My primary sources were the reviews on NewEgg.com, supplemented by expert reviews on top technical sites like AnandTech.com , TomsHardware.com , and a few others.   I looked through the NewEgg reviews and discounted boards that had too many DOA comments (board died), compatibility issues, or just too many negative ratings.

=> When it comes to user product reviews, there’s always a sprinkling of haters who are apparently never satisfied with anything.   I keep this in mind when reading product reviews.   Some people are determined not to be happy…

In the LGA 1156 arena (the socket type mandated by my choice of the Core i7-860 processor), there are a lot of choices.   Reviewing the many options, I decided that what I didn’t want was:

  • I won’t need dual GPU’s (graphics cards) running at full speed.  A major feature divide in the motherboard set is whether it supports dual Crossfire or SLI, meaning you can have two GPUs installed and running at full tilt.   Most GPUs today support dual monitors, which is fine for me.
  • I won’t need maximum overclockability.  I do definitely want to be able to play around with overclocking.  It seems like it would be fun to tweak system settings.   Some motherboards (ASUS Maximus III) have featuresets created with the extreme overclocker in mind.  Not me.
  • I won’t need anything too fancy.  Dual network interface cards (NICs), huge number of PCI Express x16 slots (the best & fastest), flashy LED lights on the motherboard, 10-channel sound – all cool, but not required.

What I did want out of the board:

  • I did want to future-proof it with USB 3.0 and SATA 6 Gb/sec interfaces.  USB 3.0 has a massive bandwidth and power increase over USB 2.0.  In three to five years, USB 3.0 devices will be cheap and plentiful.  The same can be said for SATA 6 Gb/sec interfaces.  I could always buy a USB 3.0 card later for probably $30, but why not get it integrated now?
  • I did want RAID support.  RAID makes it easy to defend against a single hard drive failure (RAID-1, “mirroring”) , as well as increase performance (RAID-0, “striping”).  I plan to have four 500 GB drives striped and mirrored in a RAID-10 array, giving me the best of both worlds.  (Note: if you just want basic RAID variants, Windows 7 and WS 2008 have support for RAID-0, 1, 5, etc.  More advanced RAID style  like RAID+10 – mirroring and striping – require motherboard support or 3rd-party RAID adapters.)
  • I did want support for fast memory.  Today’s major computing bottleneck is not in the processor, it’s in the communication between the CPU and the memory.  While giant strides have been made in processor architectures, memory latency has seen far less improvement.  I wanted fast memory and the ability to use it;  this translated to selecting boards that supported at least DDR3 1600.  Most of the boards do support this, btw.

Here are the finalists in the motherboard category, narrowed down from many more.  They’re all top-rated boards:

    Pros Cons
This is the one I chose ASUS P7P55D-E, $140. Highly rated version of the P7P55D standard, but adds support for USB 3.0 and SATA 3.0. None for me!
  Asus P7P55 SuperComputer, $240. Excellently rated;  offers huge extensibility through its five PCIe x16 slots.  Reviewers raving about the excellent build quality. Relatively expensive and generally overkill. I just don’t need that much extensibility.
  Asus Maximus III, $250. Well liked, with tons of overclocking options Aimed at the enthusiast overclocker;  overkill for me.
  GIGABYTE GA-P55A-UD4P, $195. Excellently rated board with USB 3.0, SATA 6 Gb/sec, and supporting high-speed dual GPUs. None;  this was a runner-up, and in the end, I wanted to buy an Asus board because of their reputation.

I also considered other ASUS motherboards in the P7P55-E family. These included the “Pro” and “Premium” designations.   In general, they all were excellently reviewed, but simply had more features than I needed: most of the options were fast dual GPU support (e.g. dual x8 Crossfire: two PCIe x16 slots that degrade gracefully to x8 in dual GPU config); enhanced audio; more PCIe slots (and typically fewer legacy PCI slots).  In my case, I felt the extra money could better be put to use for extra RAM, another hard drive (or a faster one), or a better graphics card.

  • ASUS P7P55D-E Premium, $290.  Dual GPU-capable @ x8, USB 3.0 + SATA 6 Gbps, 4 x PCIe, 2 x PCI, 10-channel audio, dual NICs.
  • ASUS P7P55D-E Pro, $200.  Dual GPU-capable @ x8, USB 3.0 + SATA 6 Gbps,  5 x PCIe, 2 x PCI, 8-channel audio, single NIC.
  • ASUS P7P55D-E, $140.  Dual GPU-capable @ x4, USB 3.0 + SATA 6 Gbps,  5 x PCIe, 2 x PCI, 8-channel audio, single NIC.  (This was the board I chose.)

Lastly, I ruled out all the non-“E” ASUS motherboards (e.g. P7P55D/Pro/Premium) because they did not have USB 3.0 and SATA 6 Gbps.

image Memory (RAM): G.SKILL RipJaws

With the motherboard selection out of the way, the RAM was pretty easy.   With memory, I was looking for:

  • Compatibility.  I wanted to throw the memory in my new motherboard and have no issues or hassles.
  • Speed.  I wanted the fastest possible RAM without being “extreme” and hockey-sticking the price.  DDR3 1600 seemed to be the standard here.
  • Robustness.  While overclocking was not my prime objective, I did want the flexibility to play with it.  So I wanted RAM that would tolerate OC’ing well.

There are a ton of RAM choices out there, so feel free to browse away.  I spent the least amount of time looking at RAM options, since I just wanted it to best fast and stable.  My selection and the runners-up:

    Pros Cons
This is the one I chose G.SKILL RipJaws Series 4GB (2 x 2GB), $115. Excellently reviewed, fast RAM with heat management. Plus, they look cool. None for me!
  G.SKILL 4GB (2 x 2GB), $105. Excellently reviewed, fast RAM. None
  Corsair Dominator 4GB (2 x 2GB), $150. Excellently reviewed, high-end RAM with great heat management for OC’ing. A little expensive

XFX Radeon HD 4650 Graphics card (GPU): XFX RADEON 4650

My goal for the graphics card was to be capable and well matched to the rest of the system.  Since I’m not a gamer or a professional videographer, I wouldn’t need a top-of-the-food-chain GPU.   But with the rise of video and the convergence of TV and the internet, I wanted to be able to at least play full-screen HD content flawlessly.  On my old PC, a Dell PowerEdge 400SC server with an aged ATI Radeon 9600 card, I couldn’t play HD content on YouTube without an occasional stutter.

Desktop GPUs have become big business:  modern GPUs are basically little computers-on-a-card.  They have dedicated processors on them, up to 1 GB RAM, dedicated cooling systems, and in some cases, even require dedicated connections from the power supply.

In fact, there are a ton of cards out there which are power hogs and can significantly increase the energy consumption of the entire PC.  The high-end graphics market seems to be dominated by gamers (more power to you), who are typically playing mano-a-mano first-person combat games with each other over the internet.  They need high frame rates and blazing graphics speed.   But some of these top-end graphics cards explicitly require 500-watt or 600-watt power supplies.

What I personally wanted of out a graphics card:

  • Dual monitor outputs. If you haven’t experienced the joy of a dual-monitor setup, you don’t know what you’re missing.  The good news is that most cards out there today support dual outputs.
  • DVI interface. There are two basic types of connections from a PC’s graphics card: VGA (older; analog) and DVI (newer; digital).  Moving forward, I won’t be needing the older analog VGA connections.  If you have digital displays, e.g. an LCD or other non-tube display, there’s little sense in sending anything but a digital signal to it.  With a VGA signal, the GPU card has to first convert the digital info (from the computer) to an analog signal (the VGA output), then the monitor has to take the analog VGA signal and convert it back into digital form again.
  • Ability to play full-screen HD content. There’s a tsunami of HD content on the way, but even watching YouTube videos at HD is fun now.
  • Reasonable power consumption. No power hogs.
  • Quiet. I didn’t want a GPU with a noisy fan.  Silence is golden.

There are a ton of highly rated choices out there between $100 and $200.  The NewEgg crowd seemed to especially like cards by XFX and EVO.   I chose the XFX because it seemed to have everything I wanted.  At the very worst, I could buy a more capable card if needed, and I’d have a spare graphics card.

My choice, with the runners-up:

    Pros Cons
This is the one I chose XFX Radeon HD 4650, $55. Very capable, very quiet, and very affordable. Reviewers seemed to love it; they confirmed that  it could run full 1080P HD content (1920 x 1280), and many mentioned it did fine with gaming. This is a budget card as GPUs go
  XFX Radeon HD 4850, $140. Highly rated, a ton of memory (1 GB), quiet, reasonably priced, and moderate energy requirements (450W power supply).  Looks beastly cool. More expensive
  EVGA GeForce 9800 GTX, $135. Similar to the XFX HD 4850 (above). Fast and powerful.  NewEggers bought this in droves, judging by the number of reviews (900+). More expensive; half the memory (512 MB) of the XFX HD 4850; more noisy than the XFX card?

Power supply

I had a relatively new Antec True Power 430W lying around; after doing a little research on the web, it turns out the Antec 430W supply is an excellent, high-quality, low noise power supply (if you’re interested, you can see AnandTech’s review from 2003).   It seemed like a waste to just leave it sitting around, so I decided to try it.  If the Antec reviews had been anything but stellar, I would have invested in a new supply.

Getting a good power supply is important for the life of a computer: it has to provide clean, stable power to the sensitive system components.  If you want your home-built system to last, don’t scrimp on the power supply.  Apparently most RAM failures are due to electrical overvoltage issues.  It has a reasonably tough task to do: converting the oscillating 120V signal (in the US) from the wall into varying DC voltages of 3.3V, 5V, and 12V.

=> It’s highly recommended you use some form of UPS, battery backup or power conditioning to protect your system.  For my critical servers, I have an APC Smart-UPS, which provides backup battery and power conditioning;  but at minimum, use something like a APC Back-UPS, which provides outage, surge and spike protection.

A power supply should be rated to support the maximum demands of your system at full load.  If the power supply cannot keep up, it will just shut down the system.  The main power-hungry components by far are the CPU and the graphics card.  Reasonable guidelines today would seem to be 450W (minimum), 500W to 650W (mid-range), and 700W to 1000W for the high end.

=> When in doubt, buy more power supply than less.  A high-rated power supply does not use more energy than a lower-rated supply for the same load, in general.  E.g. a 650W supply doesn’t use more electricity than a 450W supply.

Besides providing adequate power to run the computer, here are some desirable features in power supplies today:

  • Active PFC. I don’t understand all the details of active PFC, but it seems to enable more efficient power supplies. Widely available today.
  • Quiet operation. Achieved mainly through the use of fans designed for silent operation.
  • Modular cables. Prevents case clutter:  unneeded modular cables can be detached and stored, whereas non-modular cables must be tied and otherwise managed within the case.
  • Under/overvoltage protection.  Protects the system against sudden spikes or drops in AC voltage.  Like when you turn the vacuum cleaner on.
  • Energy efficiency. A power supply is converting from AC voltage to DC voltage, and the conversion is not perfect. The higher the efficiency, the less power wasted (and heat generated).  Greener power supplies will have a Bronze, Silver, or Gold energy certification, but will cost more than non-certified PS’s.
  • Connectors.  How many SATA hard drive power plugs are there, or fan plugs, etc. coming off the PS?  Not really a big issue, since the real issue is power, and cheap adapters can easily easily be bought to provide more plugs of any type.  Be aware that some of the mid- to high-end graphics cards require a direct four- or six-pin connection from the power supply.

Antec BP550 is a great value There are a lot of choices out there.  A basic i7-860 system (or other Core i3/i5/i7 system, such as a i7-920 or i5-750) without an over-torqued graphics card will probably never use more than about 200-250W;  but to be safe and to enable extensibility, it seems prudent to pair it with a 450W or so power supply, minimum.

So I began my search for power supplies with active PFC and 500-600 watts of power.  Some of the favorites the NewEggers love include Corsairs (voted “Best Power Supply Manufacturer in 2009” by PC Magazine) and Antecs (“Most Reliable Power Supply Brand,” PC World France), Zalman, and Rosewill, among others.

In the mid-range arena, these were the most highly rated, reliable, and quiet power supplies that I found.

    Pros Cons
  Antec BP550 Plus 550W, $70. Highly rated, modular cables.  Not advertised as having a quiet fan, but reviewers’ consensus was “very quiet.”   (I bought this one for my next build) Not energy certified – but seems close enough.
  Antec earthwatts EA500 500W, $70. Highly rated, 80 PLUS-certified energy efficiency, low noise cooling fan. Non-modular cables.
  Corsair CMPSU-650TX 650W, $100. Very highly rated, thermally controlled fan, 80 PLUS-certified, lots of cables. Slightly more expensive; non-modular cables.


I sifted through hundreds of user feedback comments, scores of products, and a lot of detailed product reviews on the web to build a stable, powerful, quiet, affordable supercomputer based around an Intel Core i7-860 processor.

There are a ton of computer components out there with every conceivable option you could want.   When I chose this system, I wanted to make sure I had the dollars invested in all the right places to end up with a balanced set of matched components without any major bottlenecks.  Since I had some spare parts available, I was able to put together for around only $600.

Hopefully this article will help you save some time!

=> In the next article:  I’ll provide an update on the actual build experience.



Power Supplies

Graphics Cards (GPUs)

Additional Notes

In browsing and selecting from the many components available, I made heavy use of NewEgg’s wish lists.  I created a wishlist for each of the system component types: motherboard, RAM, GPU, and power supplies.  As I searched through products and read reviews, I would add leading contenders to the wish list for later consideration. This way, I could efficiently narrow down my search.

What NewEgg could really use is a product feature comparator to compare similar products side-by-side.  For example, Intel’s CPU comparison wizard is great.  Are you listening, NewEgg?

How to Buy a New PC, Part 1



This article is for non-technical folks who want to buy an affordable new computer that will perform well for around three to five years.  How do you figure out which is the right new computer for you with all the confusing options available today?   While I can’t tell you exactly what to buy, I can show you how to greatly simplify your search.

=> In the next article (Part 2), I’ll go over guidelines for the minimum component configurations, including how much RAM to get, how big a hard drive, and so on.

So you need a new computer…


Very exciting, isn’t it?  But today’s computer technology is mind-boggling for the average person.

Which multi-core, hyper-threaded, obscurely-named processor to get with the new system?   The Core 2 Duo T6600 or the dual-core Pentium T4400?   The i5-650, the i7-860, or the i7-920?    Do you need a graphics card with a 512 MB GDDR3 and a 128-bit data path, or the one with 1024 MB GDDR5 and a 256-bit data path?   You might need an advanced degree in physics just to figure it all out.

But by applying a few basic concepts – and perhaps doing a little research — you can simplify the choices to two or three potential systems.

=> I’m assuming you have a reasonable budget to play with. If you feel that you must buy a true budget PC at the current time, then just go to your local trusted retailer or online computer vendor and pick a system!  Any modern computer will likely run circles around your old one.

Divide and conquer

Yes, it’s easy to be overwhelmed when you look at today’s computer technology, even for me – and I have thirty years’ of experience in technology!  But you can cut through a lot of the confusion by taking a divide-and-conquer approach.

Simply look at the purpose, budget, and desired lifespan for your new computer.   You’ll need a general idea of how you want to use it, how much you want to spend on it, and how long you want to have it.

Then follow a few simple rules. Chances are decent that you’ll end up with a system that will perform well for years to come.

Rule #1: Treat it as an investment

image34Your computer is what a business would call capital equipment: a physical item that delivers certain value to the business, has an associated cost, and has an expected lifespan.   If you just look at the computer as a big, expensive purchase today, then you may scrimp & cut corners – and the system you get may not live up to your expectations in three years.

To get the proper perspective, amortize the cost of the computer over its intended lifespan.  This may be a rather ugly, Latinate word with roots in death (a + mortire, meaning “to die”), but that’s exactly the point.  If you want the computer to last five years, for example, then divide the cost of new system (or any extra options or upgrades) by the lifespan to get the cost per time.

This big-picture view will make it easier to make the right choices and justify your investment.  You’ll see that an extra $100 on a faster processor, more RAM, or a bigger monitor amounts to barely pocket change every month.

Don’t handicap yourself from the start by being penny-wise and pound-foolish.  Take a strategic viewpoint if you want your new system to last.

Example 1:  You’re considering spending $1,000 on a new system.  For most people, that’s a lot of money.  But over a five-year lifespan, that’s $200 per year, $17 per month, or about fifty cents a day.   There are a lot of things you could do to save $17 a month, if you had to…  but how cool will you feel every day with an awesome new laptop?

Example 2: You’ve narrowed it down to an affordable desktop computer for $600. But you’d really like to future-proof it; Dell offers an upgrade to the “i7-860” processor for $149.  You’ve done your research and consulted your local tech guru, but this still seems like a large premium for a $600 system.    If you’re hoping for a five-year lifespan, then this extra $149 is only $30 per year, a $2.50 per month, or around a dime a day.  Choosing meditation over your Starbucks no-foam caramel soy latte for just one day will buy you about two months of the the premium processor.


Choose wisely in life

Rule #2: Set out your goals

What do you want to do with the system?   This is the first question you should ask, and it’s the one that can save you the most time and effort.   Knowing what you want to get out of a system immediately narrows down the field.

When you’re thinking about what you want the computer to do, remember to fast forward a few years and imagine what you might be doing in three to five years.  Streaming high-definition video content comes to mind, for instance.

Here are a few general examples, along with the category or class of computer you’ll need, in ascending cost order.   When you go to select a specific computer, you’ll start in the category that fits you (e.g. Multimedia PC).

  • Surf the web, read e-mail, do light document work: Budget PC. Browsing the web or editing documents are not demanding tasks for modern computers.  Almost any PC will do this with ease, from a budget PC to a netbook on up.
    Price range: $350 to $500    Examples: netbooks / budget desktopslaptops
  • Surf, docs, plus… watch video programming:  Multimedia PC. In case you haven’t heard, video is on the rise.  Within 3-5 years, we’ll surely see much more HD content available from providers like YouTube.  As more and more internet-connected personal computers are capable of displaying high-definition video, TV programming will make rapid moves online.  ;  see Hulu, etc.  “Multimedia PCs”  are equipped with better graphics cards and better audio, and generally be hooked up to a home theater setup.  Which is very cool.
    Price range: $500 to $1000   Example: Dell Inspiron 560
  • Surf, docs, plus…  high performance: Workstation PC. If you don’t really care about home theater and you do want a more powerful computer,  look at the workstation category.  These computers will be able to do more computationally intensive tasks, and typically have more memory, a faster processor, and often more expandability.  Workstation PCs are a very good future-proofed choice:  they do everything well and are more likely to perform well for quite a while.
    Price range: $650 to $1000
  • I want it all!  Performance PC. If you want a computer that can do it all, then you’re in the performance PC category…  bless your heart.
    Price range: $800 to $1500
    image54The other question you should be asking is: laptop or desktop?
    A laptop is a quintessentially modern choice. It has clear advantages: It’s wonderful to not be chained to a desk:  with a laptop, you can compute anywhere.   Think kitchen, couch, bed,coffee shop, library, hotel room.  But it also has distinct disadvantages:  you’ll pay a premium for a laptop, and lifespan in general is not as good as a desktop’s.
    The netbook is just a smaller, less powerful and less expandable laptop. It’s a great example of why you need to pin down your goals:  if all you need to do is surf the web, check your mail, and do light document work, consider a netbook.  But don’t expect it to be too fast, or to have as long a lifetime.

image60A regular desktop computer on the other hand – also variously known as a studio, mini-tower, or tower computer, depending on the size of its case –  will give you more performance for your dollar, as well as increased expandability and upgradability.   This is why desktops generally last longer than laptops, too.

=> The section at the end titled “Laptops vs. Desktops” compares the major pros and cons of each.   Generally, you will know in your heart whether you want a laptop or a desktop.  Go on…  look in your heart.

=> Laptop advice.  I personally would avoid laptops with HUGE screen sizes like 17” or 18”.   I’ve seen a few folks with these, and they just look too bulky.   Unless you feel compelled by a special need, stick to a standard 14” or 15” screen.

Rule #3: Expand your horizons

Invest in computers and learn to use them well. Technology is awesome and we’re lucky to be living in this age.  An investment in technology is an investment in yourself – and indirectly your family, friends, co-workers, and so on.   Get a good system that won’t be frustratingly slow in two years!


That’s the basic formula. A little forethought will make selecting a new computer much easier, and increase the chances that it will serve you well for a good number of years.

A quick recap:

  • Understand that buying a computer is a strategic investment. This will enable you to see the big picture.  Realize that an extra $100 or $200 for the right option is not really a lot over five years.
  • Know what you want to use it for. Among all the choices you’ll face, this will enable you to narrow down the field quickly and focus on a few select systems.   It will also help you get the proper advice from more technical folks.
  • Expand your horizons. An investment in a computer is far different from an investment in almost every other thing you’ll buy; get a good computer and explore!

image If you need a place to start, here’s your homework:

  1. Figure out purpose, budget, and desired lifespan. You should come out of this knowing the category of PC you want, e.g.. budget, multimedia, performance, etc.
  2. Narrow down the choices to a couple that are within your budget. Browse your favorite online vendor or local store; start your search in the PC category you’ve identified.  If you have questions about technology choices (you will), then do at least some light research on them: read any help pages offered by your online vendor of choice, and Google the technologies you have questions about.  You’re looking to triangulate a handful of strong, consistent opinions from folks who seem to know what they’re talking about.  If there’s no consensus, or few results, then either you’re asking the wrong question (rephrase it?) or it doesn’t matter.
  3. Optional but recommended:  bounce the final choices you’re considering off your favorite tech buddy. Do your homework first and pick a few decent finalists before asking, and typically people are happy to help and pleased that you thought of them.   DON’T ask silly, open-ended questions like “I’m looking for a new computer, are there any you can recommend?”  This is like saying “I’m looking for a car, are there any you can recommend?”     And if you get good advice from your tech, remember to follow up with a choice six-pack of handcrafted brew or a pound of premium coffee.

=> As one of those “techs” myself, I love it when people take the time to put their thoughts and candidate systems in an e-mail.  It makes it so much easier to come up with a good answer.  Recap your fundamentals, e.g. “my budget is $800, I want a multimedia PC, and I want it to last five years.”  E-mail your tech as complete as possible a description of your systems;  many online vendors these days will allow you to e-mail a wish list or shopping cart (Dell certainly offers this  option).  At the very least provide links to the web pages with the computers on them.

Have fun selecting your new computer and I hope this guide has been of help to you. If you have your own advice or there’s something you think I’ve missed, I’d love to hear it in the comments.

Laptops/notebooks/netbooks vs. desktops

This mini-guide should give you a decent idea of the major pros & cons of the two types of computer.

Laptop Desktop Comments
Portable Fixed location  
More expensive Less expensive Miniaturization has its costs
Runs on batteries No battery needed Lithium notebook batteries are good for around 2-3 years, tops.  It’ll cost you ~$100 to $150 for a new one.
Uses less power Uses more power A desktop might be around $50 more per year in energy costs.
Generally slower processors Generally faster processors Laptop CPUs are designed to save power but give up performance in return
Generally smaller and slower hard drives Excellent hard drive capacities Unless you take a lot of video, hard drive sizes are large enough today for most apps
Minimally expandable Can add expansion cards, memory, and additional hard drives more easily The expandability of desktop systems is typically related to the physical size of its case:  more room to put stuff.
Monitor is built-in Needs a separate monitor You may have a monitor that will work;  but consider a bright, large new monitor (at least 19”) as part of the investment.

imageReferences and more reading

Choosing a laptop vs. a desktop:

General system help:

  • Dell Computer Selector Dell.com
    A very nice wizard-style approach.  Enables you to select system type (laptop or desktop), budget, and intended use.  Sounds familiar…

Build or Buy a New Core i7 Supercomputer?

[tweetmeme source="KeithBluestone"]This article is part of a series in 2010 on custom-building a high-performance computer with the latest Intel Core i7 processors:

  1. Build or buy a new Core i7 supercomputer?
  2. Choosing a New CPU: Intel Core i7-920/930 vs. i7-860/870
  3. New Core i7 PC: Selecting the Components
  4. New Core i7 PC: The Build


I’m usually a strong proponent of choosing to “buy” standardized computers (e.g. Dell) rather than “build” a custom server.  But in this case, the stars are aligned in favor of a building a handcrafted system that should be a real screamer – and hopefully a lot of fun in the process.

The Landscape

In a prior post, I discussed the merits of computers based around the latest, greatest CPUs, concluding that the benefits of raw computing power and energy efficiency made it pretty much an open-and-shut case for investing in the latest technology.

image I thought I would seal the deal by simply ordering a Dell PowerEdge server, as I’ve done with great success in the past: I own a PowerEdge 1600SC and a PowerEdge 400SC, both really excellent machines for their time.  But I was curious to see what the latest and greatest advances in mainstream chip technology had brought us.

I followed up with a little research on the Intel Core i7 family of processors with the new Nehalem architecture.  In particular, I looked at the two primary choices at around a $200 price point:  the Core i7-860 and the Core i7-920.   (The i7-860 edged out the i7-920…  read the post if you’re interested in the details.)  These quad-core processors are basically supercomputers on a chip.

Being a software architect, it’s one of my “core” beliefs (no pun intended) that a deep understanding of the target platform is a great way to design the best software solutions.

The Decision…

It’s been roughly ten years since I built a custom PC.   Reading about the latest motherboards and components, I realized that the industry has had ten more years to perfect the art of producing standardized pc hardware components.   Reviews of motherboards on NewEgg.com rave about “overclocker’s dream” and “works flawlessly.”

I was intrigued:  how hard could it be, now, to throw a new motherboard into an old computer chassis, plug in a CPU and some memory, and power it up?   The alternative was a pretty tasty Dell system that I could snap my fingers and have delivered.

Which to choose? Let’s back up for a minute and look at what I want out of this system.


I wanted to get the following out of this new computer system:

  • Be future-proofed enough to run well for five years: this includes expandability and performance.
  • Be able to run computationally intensive modeling and pattern detection applications in support of the new stock market analysis platform I’m creating with some folks.  So it has to be fast.
  • Be able to take advantage of new virtualization technologies (VT).  This is nothing short of a revolution for the IT crowd:  the ability to run complete “images” of computers on one physical box.  So for instance, I could have the equivalent of  complete Windows 7 and Windows XP machines running on the same box.    Amazing.


I sketched out a pretty capable system meeting these needs.  The key components looked like this:

  • A Core i7-860 processor. Has four separate physical cores, hyperthreading (which doubles the effective cores to a total of eight), awesome memory bandwidth, and great energy efficiency (95W TDP).  See my post comparing the Core i7-860 vs i7-920.   It’s one of the fastest desktop CPUs available today:  definitely meets the “future-proofed” criterion.
  • 8 GB of RAM, expandable to 12 GB or 16 GB. By simply adding more RAM, I can use virtualization to “host” multiple complete virtual computers on this one “supercomputer” – no need to buy a whole new system.   So RAM expandability is key.
  • Support for SATA RAID on the motherboard. Using redundant arrays of disks  is absolutely a best practice to avoid the headaches that come with a single hard drive failure.   I read a great quote the other day: “The difference between a good hard drive and a bad one is that the good one hasn’t failed yet.

I’m not into video games, so I don’t need anything special in the graphics category like “Crossfire” or “SLI” display technology:  a solid & capable dual-DVI graphics card should do the trick.

Picking a system: the pre-built road

Dell PowerEdge T310My prebuilt system of choice was a Dell PowerEdge T310, configured with a Xeon X3440 @ 2.53 GHz (more or less a slightly slower i7-860), 8 GB RAM, 250 GB hard drive for around $1,050.    Add a graphics card for $150 and the total system cost is $1,200.

I generally choose the PowerEdge servers (vs. consumer desktops like Dell XPS’s) because of their high quality and reliability;  I also install the OS myself so I avoid all the adware/trialware crap that bloats these new machines.

The T310 is a pretty awesome machine for a home server.   The Dell motherboards and bioses are engineered for business server use – which most of the time is fine.

Picking the system: a custom build

I have a spare case, an extra Antec True Power supply lying around, and four enterprise-class 500 GB SATA drives, and I wanted – I as mentioned above Corsair Dominator DDR3– to see what a custom build experience would be like, ten years later.    In addition, it might be fun to play with overclocking the i7-860 processor (in all my spare time!). A lot of folks having fun with this based on the comments in the NewEgg reviews and other forums on the web.

Since I have the spare case, power supply, and the hard drives, I would just need CPU, RAM, motherboard, and a graphics card.  I could put more into the specific components I wanted, specifically a great motherboard and as much high-quality RAM as possible.

A quick breakdown of a custom build looked something like this:

Component Price
Motherboard $200
i7-860 CPU $250
8 GB RAM $250
Graphics card $150
Total: $850

At $850 for a custom build vs. $1,200 for the Dell, this was starting to look like a good deal.  While I would end up with an overclockable, high-performance computer, I would miss the convenience and build quality of a new Dell server.


I love buying standard Dell PowerEdge servers, but in this case it’s hard to resist playing:  I’m going with the custom build this time.

For fun, I’ll hand-pick a quality set of components, including a socket LGA 1156 motherboard, RAM, and a display card.  I’ll post updates here and on Twitter as I look around.

Do you have any advice on i7-860 system components (or any LGA 1156-based), including motherboard, RAM, or video card selections?    Post a comment.

References and more reading…

Virtualization technologies:

Choosing a new CPU: Intel Core i7-920/930 vs. i7-860/870

[tweetmeme source=”KeithBluestone” only_single=false]This article is part of a series in 2010 on custom-building a high-performance computer with the latest Intel Core i7 processors:

  1. Build or buy a new Core i7 supercomputer?
  2. Choosing a New CPU: Intel Core i7-920/930 vs. i7-860/870
  3. New Core i7 PC: Selecting the Components
  4. New Core i7 PC: The Build

[Update 8/25/2010: the Core i7-870 is now available and has matched the price point of the i7-860. It’s absolutely identical to the i7-860 except that it runs at 2.93 GHz instead of 2.8 GHz. ]

[Update 3/21/2010: the Core i7-930is now available. It’s absolutely identical to the i7-920 except that it runs at 2.8 GHz instead of 2.66 GHz. Where I mention the i7-920 by name, it also applies to the i7-930.]


imageThe new Intel Core i7-920/i7-930 and i7-860/i7-870 processors are extremely close in most aspects,  but the i7-920/930 offers features that only enterprise users, professional videographers, or heavy gamers will need.   The i7-860/870 matches or betters the i7-920/930 in almost all benchmarks – and uses less power!

So unless you have an extreme need, you can save a few bucks with the i7-870 and plow it back into more memory, a faster/bigger/more reliable hard drive, or a better graphics card.

Note from the future: Don’t forget the CPU cooler!

Quad-core: the current sweet spot

Currently, my choice for a reasonably snappy system with a life expectancy of around five years is an i7-870 or an i7-930.  The higher-performing, extreme versions of these processors cause the price to shoot radically upwards.  Why pay for that performance when overclocking will accomplish the same thing?

In a related blog post,  I made the case for upgrading to the latest and greatest processors instead of taking advantage of fabulous deals on enterprise-class computing hardware on eBay.   In short, the newest Intel Core i7-family processors have much to offer over older processors, including improved power efficiency, general computing power and flexibility, and total cost of ownership (TCO).     Both processors have support for virtualization (VT), allowing you to host a virtual computer within these systems that looks like a real, separate PC.

These processors both have four physical cores, and with Intel’s HyperThreading technology, these four cores appear as eight separate processors to the operating system.  The new Nehalem processor architecture improves memory speed (latency and bandwidth), which can be a major performance bottleneck as improvements in the speed of processors over the last thirty years have far outpaced those in the speed of memory chips.

Why multi-core computing is such a good idea. Somewhere around mid-2004, Intel introduced the first dual-core processors for the desktop, initially in high-end systems and gradually working their way into most computers.  One main reason these multi-core systems (especially a quad-core) are going to be so much more responsive than your old single-core Dell is because there are simply more applications and background processes running on a modern computer.  And they all chew up available processor power.  In particular, big anti-virus suites (Norton, Symantec) are some of the worst offenders: in troubleshooting systems for friends and family, I’ve seen them consuming up to 100% of CPU cycles.  But why get mad?  Get even.  With a properly spec’d quad-core box you’ll have plenty of capacity to run everything.  Chances are, you won’t even notice it’s there.  (Make sure you have enough memory!  At the date of this writing – June 2010 -  you should have no less than 4GB RAM.)

So which processor to get:  the i7-930 or the i7-870? In a nutshell, you could probably quit obsessing over the details, buy either one and get about the business of enjoying your snappy new PC.   For me, being a detail-oriented software architect who specializes in high-performance and high-throughput architectures, I wanted to understand the finer differences between the two and make an informed decision.

Quad-core Nehalem architecture

Quad-core Nehalem architecture


First of all, the processors’ names are more or less no help to us. You might think that the “i7-9×0” is a later, more advanced version of whatever the lesser “i7-8×0” might be.   But understanding Intel’s processor nomenclature and numbering system is a reasonably sized effort in itself, and in this case, 930 is not necessarily greater than 870.

After several days of trawling through a ton of web pages, from Intel’s site (very good!) to AnandTech to Tom’s Hardware to discussion forums, these are my conclusions.

The prime differences, with the winner in each category, would seem to be:

  • Clock speed (i7-870).  The i7-930 runs at 2.8 GHz vs.  the i7-870 at 2.93 GHz, giving the i7-860 about a 5% advantage in raw clock speed, plus a more aggressive turbo mode which pushes the i7-870 to a max of 3.6 GHz vs. the 930’s max of 3.06 GHz – almost a 20% increase.
  • Power consumption (i7-870).  The i7-930 has a TDP of 130W, which is about 50% higher than the i7-870 at 95W.   In AnandTech’s benchmark reference systems, the i7-860 system at idle uses 85W and the i7-920 system uses 115W;  see the AnandTech i7-860 review, power consumption page.   At about $1 per watt-year over a computer lifetime of 5 years, 30W could cost you up to $150 more.
  • imageMotherboards (i7-870). Motherboards  for the i7-8×0 Lynnfield processor family (socket LGA 1156) are less expensive and more prevalent than for the i7-9×0 Bloomfields (socket LGA 1366).
  • Overall system throughput (i7-930). The i7-9×0 has a faster bus speed, utilizing a QPI bus with a max bandwidth of 4.8 gigatransfers per second (GT/s).   The i7-8×0 uses a DMI bus with a bandwidth of 2.5 GT/s.  However, this is only an advantage if you’re maxing out the bus… rare unless you’re in an enterprise server setting or doing graphics or other data/compute-intensive work.  According to the Intel specs, the i7-9×0 also has 20% higher memory bandwidth (25.6 GB/s) than the i7-8×0 (21 GB/s), also important for high-performance applications.
  • Extensibility (i7-930). The i7-9×0 enables direct dual PCI-e for crossfire & SLI applications: important to gamers mostly for high-performance graphics setups.


Neither of these processors supports ECC memory.  This is not a huge issue for most folks, but for the more critical and most stable systems (servers and scientific computing installs), ECC memory protects against data corruption caused by – believe it or not – cosmic rays.


While again you could probably go out and buy either the Core i7-870 or the i7-930 and be happy, I found little to justify the i7-930 over the i7-870 in general, with the exception of the truly hard-core crowd.

How to Buy

My first choice for PC hardware is NewEgg.com, which has an excellent online store, good prices, and fantastic customer service.

Naturally, you can also find pre-built Core i7-based systems at Dell and other vendors;  for example, the Studio XPS systems.

  • Also: Xeon X3440 on NewEgg.com ($240).  As mentioned above, the X3440 is the rough equivalent of the i7-860 (Lynnfield core), but supports ECC RAM.