System Builder Marathon, Q4 2012: System Value Compared

System Builder Marathon, December 2012: The Articles

Here are links to each of the four articles in this quarter’s System Builder Marathon (we’ll update them as each story is published). And remember, these systems are all being given away at the end of the marathon.

To enter the giveaway, please fill out this SurveyGizmo form, and be sure to read the complete rules before entering!

Day 1: The $500 Gaming PC
Day 2: The $1,000 Enthusiast PC
Day 3: The $2,000 Performance PC
Day 4: Performance And Value, Dissected

Introduction

Most people learn from experience, and a consistent benchmark suite makes it easy for the three builders who participate in our SBM every quarter to figure out where they're going to get the best bang for their buck. Call it gaming the system, if you will. Since most of our benchmarks favor the same types of hardware, we typically end up with three builds that reflect a trio of budgets, but employ the same technologies. Our builds became mostly predictable over the past year, with occasional deviations coming up short in both performance and value.

Windows 8 and its impact on the performance of AMD's Bulldozer proved to be underwhelming, but AMD timed the introduction of more capable hardware to coincide with Microsoft latest, giving us Piledriver-based FX CPUs that do help the company's performance story.

We're also adding a number of new threaded benchmarks this time around. So, the CPUs best able to handle taxing workloads are going to rise to the top, naturally. Could Don's FX-8350 upset the balance of this often-familiar retrospective with eight integer cores in a sub-$1000 system, will my Core i7-3770K-based build brute-force its way through the benchmarks, or will the addition of a Radeon HD 7850 help catapult Paul's $500 build to the top of our value charts?

Q4 2012 System Builder Marathon PC Components
 $500 Gaming PC$1,000 Enthusiast PC$2,000 Performance PC
ProcessorIntel Pentium G850: 2.9 GHz, 3 MB Shared L3 CacheAMD FX-8350: 4.0 GHz - 4.2 GHz, 8 MB Shared L3 CacheIntel Core i7-3770K: 3.5 GHz - 3.9 GHz, 8 MB Shared L3 Cache
GraphicsPowerColor AX7850 1GBD5-DH: Radeon HD 7850 1 GBGigabyte GV-N670OC-2GD: GeForce GTX 670 2 GB2 x MSI R7970-2PMD3GD5/OC: Radeon HD 7970 3 GB, CrossFire
MotherboardASRock H77 Pro4/MVP: LGA 1155, Intel H77 ExpressGigabyte GA-970A-D3: Socket AM3+, AMD 970/SB950Gigabyte GA-Z77X-D3H: LGA 1155, Intel Z77 Express
MemoryG.Skill F3-10600CL9D-8GBNT: DDR3-1333 C9, 2 x 4 GB (8 GB)Mushkin Blackline 997043: DDR3-1600 C8, 2 x 4 GB (8 GB)G.Skill F3-1600C8D-8GAB: DDR3-1600 C8, 2 x 4 GB (8 GB)
System DriveWestern Digital WD3200AAKX: 320 GB, 7,200 RPM HDDOCZ AGT3-25SAT3-60G: 60 GB, SATA 6Gb/s SSDMushkin MKNSSDCR240GB-DX: 240 GB, SATA 6Gb/s SSD
Storage DriveUses System DriveHitachi Deskstar 7K1000.C: 1 TB 7,200 RPM Hard Drive
Seagate Barracuda 7200.11 ST31500341AS: 1.5 TB, 7,200 RPM Hard Drive
OpticalLG GH24NS90: 24x DVD±R, 48x CD-RLG GH24NS90: 24x DVD±R, 48x CD-RAsus BW-12B1ST: 12x BD-R, 16x DVD±R, 2x BD-RE
CaseRosewill BlackboneHEC BlitzCooler Master Storm Enforcer SGC-1000-KWN1
PowerAntec VP-450: 450 W, ATX 12V v2.3Corsair CX600 V2: 600 W,  ATX12V v2.3,  80 PLUS
Corsair HX750: ATX12V V2.3 80 PLUS Gold
CPU CoolerIntel Boxed Heat Sink and Fan
Xigamtek Loki SD963Cooler Master Hyper 212 Evo RR-212E-20PK-R2 
 Total Cost $501 $1,009 $1,900
This thread is closed for comments
76 comments
    Your comment
  • lengcaifai
    actually the piledriver based build is more all-rounded, it can be a decent workstation and a decent gaming desktop for those who have tight budget
  • mayankleoboy1
    Which of these builds is the most future-proof ?
  • stickmansam
    The piledriver build should have had some parts swapped out for cheaper ones to reflect the changes in pricing changes from last Quarter. That would have been a more fair comparison to the $1000 build from last quarter.
  • mayankleoboy1
    For the $500 build, why would it have 60% of its value calculated by apps, when it was build for gaming purpose ?
    Just a thought, but shouldnt the percentwise distribution of value for each built based on the purpose for which it was built ?
    Something like : games, apps, storage.

    $500 build : 80%, 15%, 5% (cheapest best gaming with lots of cheap storage. )
    $1000 build : 50%, 40%, 10% (slightly better games over apps. Great apps. fast storage for boot)
    $2000 build. : 42.5%, 42.5%, 15% (equally good games and apps. fast storage should be plenty and fast)
  • mohit9206
    wow its unbelievable to see a $500 gaming pc achieve 50+ fps in Battlefield 3 at 1080p on ultra settings.
    goes to show how even a $500 pc can thrash and destroy xbox 360 and ps3.
  • the1kingbob
    lengcaifaiactually the piledriver based build is more all-rounded, it can be a decent workstation and a decent gaming desktop for those who have tight budget


    I was pleasantly surprised how well it turned out. I believe I would have gone with one that had less cores and spent the money elsewhere. Overall though, it turned out to be a pretty good machine. Now only if they could get the power usage under control.
  • slicedtoad
    You need to use a slightly more complicated performance comparison algorithm. Something that takes into account the fact that over 120fps is useless and doesn't run into problems with things like fps caps at low res.
  • Marcus52
    mohit9206wow its unbelievable to see a $500 gaming pc achieve 50+ fps in Battlefield 3 at 1080p on ultra settings.goes to show how even a $500 pc can thrash and destroy xbox 360 and ps3.


    Yeah the AMD 7850 really pulled its gaming performance up. Very nice too that the Intel G850 didn't choke it off. A sweet build!
  • ojas
    Um i must ask this, BF3, 2560x1600, ultra: is cross-fire not working? Why should 2 7970s serve up the same performance as a single 670? Or am i missing something? Thomas mentions the drop being "extreme" in his build's article but...something's wrong, i think...
  • So we're left with the AMD platform winning out the value comparison for no other reason than the mid-range Intel solution was not permitted to compete - so without filling all the places at the table this test is meaningless. The AMD solution only gets a default win due to no other permitted competition in that price range. This is garbage.

    Firstly, the bulk of sane consumers with even half a clue and with $1000 in their pocket would not have given the AMD platform a 2nd look if given the choice. Are we really suggesting that they would have thrown $1000 at a solution that would not give them a 3770K upgrade option later on if they felt like it?

    Also, this comparison deliberately factored out power consumption, which was rather convenient for AMD. I'm afraid you can't factor this out in this day and age, just because it's hard to quantify the cost across the entire globe. What you could do is produce some sythetics that represent average consumption over a given task and mutliply it up to get the total power over a year - then folk can work out what that would cost them in their own location. What I would like to know is how much that AMD solution would cost me to run for a couple of years when compared to a comparible Intel solution, and then work out what I could have bought with the money saved - it might not be much but I think it's valid - it could be the difference between a decent cooler or a piece of trash.

    Please make these value comparisons tell the whole story by including both platforms within that price bracket - I know that makes life hard for the reviewing team but boo hoo hoo, you're the ones that set out to prove a point, so do a full job please. Tell us the full story, not half of it.
  • ojas
    Hmmm. I wonder what will happen if you put the FX8350 into the $2000 machine?

    pchisholmAlso, this comparison deliberately factored out power consumption, which was rather convenient for AMD.

    second last page?
  • ojassecond last page?


    I know thats there, but you obviously didn't read all of the last page where it clearly states that power was deliberately factored out for the overall value comparison tables.
  • And no, the data I'm referring to isn't there because there is no $1000 intel solution to compare to. Nice troll but no cigar.
  • MxMatrix
    Although I can understand the choices made for all 3 systems I think a X79 & i73970EX should be added for comparison.
    - it would give an interesting value over performance overview
    - there are still people interested in X79 builds today (like me)
    - everyone likes to see them for xmas
  • iam2thecrowe
    ojasHmmm. I wonder what will happen if you put the FX8350 into the $2000 machine?

    It would make it game a fair bit worse, and app performance would be on par or worse than the i7. And there is nothing more you could really spend the extra money on. Maybe a nicer monitor........ but for a gaming build if looking to cut price from the i7, you would just get the i5 for the same price as an 8350.
  • Crashman
    pchisholmI know thats there, but you obviously didn't read all of the last page where it clearly states that power was deliberately factored out for the overall value comparison tables.
    No it doesn't. Show me the word "deliberately" or anything else of similar meaning. Then go back to the last SBM. And the one before it. And the one before that. Efficiency has never been used in the SBM performance-per-price comparison.
    It would be far more accurate to say that the methodology of this SPECIFIC SBM was not deliberately ALTERED to PENALIZE AMD. The site has too much integrity to pull such a stunt.
  • jtd871
    I understand that the $500/$1000/$2000 ratios make a nice exponential progression (factor of two), but the performance and value increases definitely diminish. Unless you are speccing a boutique gaming laptop, I think that I would prefer to see how creative your top-price builder can get for $1500. I am willing to bet that the value could be increased substantially - both in terms of power and price.

    By switching to i53570K, 2x7870, CM Hyper TX3, ~$100 + $260 + 10 = $370 could have been saved. Also, do you really need the Barracuda? Not for SBM (I doubt it makes a difference to the performance one way or the other with a 240GB SSD primary) so another $80 saved. There's $450 saved for a great all-around performer and probably decent overclocker (not that I would ever dream of overclocking - system fans get too loud) too. For the price of the CM case, you could probably get a Silverstone PS07 and a couple of case fan upgrades and have a rig that also looks more appropriate in the office.
  • Crashman
    ojasUm i must ask this, BF3, 2560x1600, ultra: is cross-fire not working? Why should 2 7970s serve up the same performance as a single 670? Or am i missing something? Thomas mentions the drop being "extreme" in his build's article but...something's wrong, i think...
    I think it has something to do with the graphics memory getting used up. It's not a system-wide problem, it's a BF3+CrossFire problem. We used to see a similar thing in Crysis, again only at 2560x1600

    Look at the scaling for BF3 high settings, from 1280 to 1920. Everything looks good up to that point, in single-monitor testing the problem only occurs at 2560x1600 (though it may also affect high Eyefinity resolutions).
  • army_ant7
    Woah! Spotted an inconsistency here... It says in this article that the $2000 build has 16GB of RAM, but in its own article it says it has 8GB. I'm thinking that the former is correct, but this warrants a correction in whichever article needs it. :)
  • salgado18
    /fanboymodeon

    That's so good to read two days after I bought an FX-8120 for the price of an i3! (hoping performance is similar to 8320)

    Go AMD!

    /fanboymodeoff
  • CrashmanNo it doesn't. Show me the word "deliberately" or anything else of similar meaning. Then go back to the last SBM. And the one before it. And the one before that. Efficiency has never been used in the SBM performance-per-price comparison. It would be far more accurate to say that the methodology of this SPECIFIC SBM was not deliberately ALTERED to PENALIZE AMD. The site has too much integrity to pull such a stunt.


    "Today, Paul's machine sets our baseline, and is exceeded by Don's. Yes, as we saw on the previous page, the FX-8350 is less efficient. However, when you factor power out and look only at performance per dollar, this quarter's $1,000 configuration rises to the top in its stock form and even more so after overclocking."

    So you missed this paragraph and the chart underneath it then?

    For these comparisons to be objective and fair they should contain ALL the platform options at the appropriate price point or it is completely subjective. And just because there was a test you didn't do in a previous SBM does not mean you should exclude appropriate data later on in another SBM, especially then the missing data is more pertinent with every month that passes - and if it so happens that this data puts one particular platform at a disadvantage then so be it. Don't bitch about Tom's loosing cred by including data that shows one platform as sucking - thats life. Either man up and do the job properly or don't bother, because half a job does your cred a lot more damage than not doing it at all.
  • salgado18
    pchisholmFor these comparisons to be objective and fair they should contain ALL the platform options at the appropriate price point or it is completely subjective.

    No they shouldn't. This is not a platform comparison, or a processor comparison. It's a System Builder Marathon: I'll give you $1000 dollars to build a machine, can you beat mine if I use $2000? There cannot be two $1000 machines, otherwise it's not what the article means to be.

    The i5 is definitely better at efficiency than the FX-8350, and the later is faster at threaded work, but if you would enter such a competition (not a comparison), knowing that they won't factor power consumption in performance-per-dollar charts, which one would you choose? He chose the FX, more performance-per-dollar, less efficiency, job done.
  • Onus
    mayankleoboy1For the $500 build, why would it have 60% of its value calculated by apps, when it was build for gaming purpose ?Just a thought, but shouldnt the percentwise distribution of value for each built based on the purpose for which it was built?Something like : games, apps, storage. $500 build : 80%, 15%, 5% (cheapest best gaming with lots of cheap storage. ) $1000 build : 50%, 40%, 10% (slightly better games over apps. Great apps. fast storage for boot) $2000 build. : 42.5%, 42.5%, 15% (equally good games and apps. fast storage should be plenty and fast)

    This. You cannot judge a PC's value without considering the purpose(s) for which it was built. You don't score a gamer based on how quickly it can churn out PDF files, and you don't score the professional's PC based on how many FPS it gets in Crysis. For games, applications, and storage, I'd probably weight it more like $500: 85% / 5% / 10%; $1000: 60% / 25% / 15%; $2000: 5% / 80% / 15%.
    The value analysis also has some flaws. For games, the perceived value of a progression from 30-60 FPS might be linear. The progression from 15-30FPS clearly is not, nor is 100-200FPS. The former is a boolean (unplayable vs. playable), and the latter hardly matters. Similarly, if the increased speed of the pro's machine allows an additional $1000 worth of work to get done per week, it's paid for itself soon after construction.
    Looked at with no context, Paul's $500 PC is a good effort that struggles a bit, Don's PC is a buzzkilling letdown, and Thomas' $2K PC is the stuff of wet dreams. Throw in the context, however, and Paul's rises to the level of brilliance, even Don's becomes a true enthusiast build because of all the lessons learned and the tweaking, but Thomas' is just another high-end PC that a lot of us can't afford to build, at least not routinely, so we'd all do it differently. IMHO, Paul is the clear winner this cycle.
    Edit: ...but Don's was pretty cool too.
  • army_ant7
    771306 said:
    "Today, Paul's machine sets our baseline, and is exceeded by Don's. Yes, as we saw on the previous page, the FX-8350 is less efficient. So you missed this paragraph and the chart underneath it then? For these comparisons to be objective and fair they should contain ALL the platform options at the appropriate price point or it is completely subjective. And just because there was a test you didn't do in a previous SBM does not mean you should exclude appropriate data later on in another SBM, especially then the missing data is more pertinent with every month that passes - and if it so happens that this data puts one particular platform at a disadvantage then so be it. Don't bitch about Tom's loosing cred by including data that shows one platform as sucking - thats life. Either man up and do the job properly or don't bother, because half a job does your cred a lot more damage than not doing it at all.
    Come on?! Seriously? They're just presenting comparison charts and when they say one build triumphed over the other two, they give the context, just like "However, when you factor power out and look only at performance per dollar, this quarter's $1,000 configuration rises to the top in its stock form and even more so after overclocking." For crying out loud, they're just saying "when" one does so... What's your problem? Maybe you're the one who should stop "b****ing"?

    You talk like your some kind of hot shot around here. Remember, Tom's Hardware is free (ads aside). We can give feedback, but don't you think we should be polite about it? If someone is rendering a service for you for free, do you act like an A-hole towards that person? If you do, then you really are an A-hole then.

    I apologize for coming off in such an angry way, but come one...