System Builder Marathon, August 2012: The Articles
Here are links to each of the five 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 $1000 Enthusiast PC
Day 3: The $2000 Performance PC
Day 4: Performance And Value, Dissected
Day 5: The Surprise $2000 Alternative Build
Introduction
Is it fair to compare differently-priced PCs based on their performance alone? Cheap computers typically lack convenient features and durable parts, which are hallmarks of higher-end machines. Meanwhile, mid-range builders try combining the two worlds, sacrificing some of the more extravagant additions that sometimes go unused in a performance-oriented desktop. Even if Paul, Don, and I are all completely successful at our $500, $1000, and $2000 price points, Don's middle-of-the-road configuration is going to have a huge advantage right out of the gate for its potential to cram balanced performance into a well-built enclosure.
Of course, Paul and I are always challenged to pull your attention away from that middle machine. Paul’s $500 gaming box generates its buzz by generating playable frame rates at 1920x1080 at an extremely modest price, while my $2000 build seeks success by identifying areas where Don might have gone a little too light, and fixing them with an extra thousand dollars worth of funding.

What happens, then, when the $1000 PC has no obvious failings? Traditionally, $700 (give or take $100) is the point where measuring performance and value start tapering off into diminishing returns. Can Don’s $1000 build push the point where money starts flying out the door faster than performance increases?
| Q3 2012 System Builder Marathon PC Components | |||
|---|---|---|---|
| $500 Gaming PC | $1000 Enthusiast PC | $2000 Performance PC | |
| Processor | Intel Pentium G860: 3.0 GHz, 3 MB Shared L3 Cache | Intel Core i5-3570K: 3.4 Base, 6 MB Shared L3 Cache | Intel Core i7-3930K: 3.2 GHz Base, 12 MB Shared L3 Cache |
| Graphics | MSI N560GTX-M2D1GD5: GeForce GTX 560 1 GB | Gigabyte GV-N670OC-2GD: GeForce GTX 670 2 GB | EVGA 02G-P4-2670-KR: GeForce GTX 670 2 GB |
| Motherboard | Gigabyte GA-B75M-D3V: LGA 1155, Intel B75 Express | ASRock Fatal1ty P67: LGA 1155, Intel P67 Express | ASRock X79 Extreme4: LGA 2011, Intel X79 Express |
| Memory | G.Skill F3-10600CL9D-4GBNS: DDR3-1333 C9, 2 GB x 2 (4 GB) | Mushkin Blackline 997043: DDR3-1600 C8, 4 GB x 2 (8 GB) | G.Skill F3-1600C8Q-16GAB: DDR3-1600 C8, 4 GB x 4 (16 GB) |
| System Drive | Western Digital WD5000AAKX: 500 GB, 7200 RPM Hard Drive | OCZ AGT3-25SAT3-60G: 60 GB, SATA 6Gb/s SSD | Mushkin MKNSSDCR240GB-DX: 240 GB, SATA 6Gb/s SSD |
| Storage Drive | Uses System Drive | Seagate Barracuda ST3750525AS: 750 GB, 7200 RPM Hard Drive | Western Digital AV-GP Green WD20EURS: 2 TB, 5400 RPM Hard Drive |
| Optical | Samsung SH-222BB: 22x DVD±R, 48x CD-R | Samsung SH-222BB: 22x DVD±R, 48x CD-R | Asus BW-12B1ST: 12x BD-R, 16x DVD±R, 2x BD-RE |
| Case | Rosewill R218-P-BK | Rosewill Redbone Black | NZXT Phantom 410 Gunmetal |
| Power | Antec VP-450: 450 W, ATX 12V v2.3 | Corsair CX600 V2: 600 W, ATX12V v2.3, 80 PLUS | Seasonic SS-850HT: 850 W, ATX12V V2.3, 80 PLUS Silver |
| CPU Cooler | Pentium G860 Boxed Cooler | Xigamtek Loki SD963 | Scythe Mugen 3 Rev. B SCMG-3100 |
| Total Cost | $501 | $1065 | $2002 |
Just because its shortcomings aren't obvious this time around doesn't make the $1000 machine’s flaws any less serious. It still uses a cheap case better suited to $600 machines, its SSD is too small to hold our test suite, and it does go $65 over budget. Understandably, though, all of those compromises were needed to get a GeForce GTX 670 and Core i5-3570K under its hood. Don bent the rules a little bit, just like any real-world builder would, to get very real performance benefits. Because he did this in response to reader requests, Paul and I are letting him get away with it.
With such robust specifications, we're left with two questions about the $1000 configuration: first, how badly will it destroy the $2000 machine's value, and second, how well will the $500 machine keep up in the benchmarks?
- Chasing Down Diminishing Returns
- Test System Configurations, With Overclocks
- Benchmark Results: 3DMark And PCMark
- Benchmark Results: SiSoftware Sandra
- Benchmark Results: Battlefield 3
- Benchmark Results: DiRT 3
- Benchmark Results: The Elder Scrolls V: Skyrim
- Benchmark Results: StarCraft II
- Benchmark Results: Audio And Video Encoding
- Benchmark Results: Productivity
- Power And Efficiency
- Breaking Down The Value Chart
I mean swapping the G860 for a FX 4100 and a Radeon 7770 *might'* have provided an interesting contrast to the above $500 system.
Productivity up by 20% and games down by 20% I guess. Can only speculate.
Btw, thanks crashman for the tip.
This is just me wondering aloud. So...dunno why the thumbs down
Edit: Ninja'd by chmr
I mean swapping the G860 for a FX 4100 and a Radeon 7770 *might'* have provided an interesting contrast to the above $500 system.
Productivity up by 20% and games down by 20% I guess. Can only speculate.
Btw, thanks crashman for the tip.
This is just me wondering aloud. So...dunno why the thumbs down
And those gaming benchmarks are ridiculous. Most are getting FPS in the 100+ range. So really, there is no comparison between the systems. all values above 60 are the same. How can 150 FPS be better than 120FPS on a 60HZ monitor?
Edit: Ninja'd by chmr
current price-tags feel awkward i have to agree.
to be more specific:
a. Entry level gaming pc ($500): you try to pick the cheapest parts so that you save for the best GPU the rest of your money can buy
b. Enthusiast gaming pc ($1000): how most people try to build, save here and there (either by finding good deals or by dropping quality in RAM and Chassis mostly) so that you can get an awesome CPU & GPU (prolly a SSD too)
c. Hardcore gaming pc ($2000): the tag is too high so you just blindly buy the most expensive parts (like a sheikh on vacation)
what would show more accurate results might be one of the following two:
1. two builds; one of $700-$800 and one of around $1500 (+/- $100)
2. three builds again but with some $150-$200 offset; entry-lvl 650-700, enthusiast 1200-1400, hardcore 1700-1900
Toms did a bunch of game reviews showing how bad AMD is so they don't have to use them for the SBM articles. 11 of the past 12 SBM have all been Intel, and the one AMD was bugged with a cheap cpu.
Even though SBM was I thought to test hardware with different components, apparently as long as its only with Intel.
BF3 as a test needs to be done online, wether its controlled or not, you can at least get a feel of how its going to work. Especially with a dual core cpu.
StarCraft II is the most CPU limited game in our suite. We'll even drop all the low settings and just take a look at the 1920x1080 Extreme chart linked here: http://www.tomshardware.com/reviews/gaming-fx-pentium-apu-benchmark,3120-6.html
(edit: A8-3870K - average 30.4 fps)
Athlon II X4 645 - average 30 fps
Phenon II X4 955BE - average 38.9 fps
Pentium G860 - average 45.6 fps
Core i3-2100 - average 47.6 fps.
openly admitted conspiracy basing all cpu decisions on one article and in particular one game.
Similar to your view, i have one more thought. Since most users are already set/fixed/limited to their monitors or preferred resolutions, i reckon resolution is (or should be) the deciding factor while planning for system components.
For instance, When I built my system
-- I was on a budget of INR 35000 (This converts to around $750 - though if ppl in the USA bought these parts at that time, they would have spent just around $600 - due to various reasons beyond me)
-- I started with my *desired* resolution - 1920x1080.
-- For acceptable gaming AT THIS RESOLUTION within my budget, Radeon 5770 was the choice (NV also had options like the GTX 260 but were too power hungry for me).
-- This in turn decided the CPU for me, the Athlon II X3 435, which balanced the 5770 (meaning the 5770 and the X3 435 provided equal bottleneck to my system)
-- the CPU decided the appropriate motherboard for me
-- all this decided the power supply
-- and so on
Coming to how the reviews are laid here and most other sites, it is by game - i.e. each game on a new page. My humble opinion is to split them by resolution. Have 1 (or 2) page dedicated to games on 1440x900 then come to 16** resolutions and cover all games at this, followed by games at 1920x resolutions and then finally 2560x if needed.
i believe with such a layout, we need to see only one page in any review - the page representing the monitor resolution applicable to us
Sorry for the long post.
Try to think in another aspect. In heavy parts of a game, performance suffers. Let's say, performance cut by half. The 100FPS falls to 50FPS, and the 75FPS falls to 38FPS. In this case the 100FPS actually is more valuable.
This could be a good metric in the next SBM (wherever this is applicable, with scripted benchmarks) minimum FPS achieved.
openly admitted conspiracy basing all cpu decisions on one article and in particular one game.
Pure nonsense. Way to take a quote out of context.
That comment was a direct response to someone who called the use of a Pentium G860 CRAP for our pure gaming system, while instead suggesting a Llano-based Athlon II as a better option, with zero proof offered to back up his claim.
I was simply showing how the Pentium is a far better processor for the intended purpose and provided one example review proving my claim. The RTS I referenced being an extreme case which also happens to make up 25% of our SBM gaming suite. His solution would have been far worse, and probably failed misreably, at the machine's intended purpose, which was to increase our processing fortitude for gaming alone (vs. the G530), and not winning an overall SBM. Winning will take more CPU cores. Specifically of available options in range, 4 overclocked AMD cores (preferably Phenom II) would indeed be better in today's comparison, if that were the machine's goal. I covered that in the text. Shoot, likely the best way to win this thing is with OC'ed i5-2500K and make graphics an afterthought, yet then we'd have to lose the gaming name and accept failure at the higher game settings.
Like it or not, Tom's consistently has found Intel's Sandy Bridge offerings to offer stellar gaming value, and chips like the Pentium G860 have ruled their price bracket and repeatedly earned the site's monthly recommendation for "Best Gaming CPU for the Money". Not based on one game, based on all our time spent testing various CPU's. It has nothing to do with brand preference, we long for either company to trump current top performers, and do NOT like using an enthusiast-unfriendly "locked" dual-core CPU. We are begging for better options, yet appreciate the amazing per-clock performance and affordable gaming abilities Sandy Bridge offers.
So many are bashing a CPU found to offer the best value at this machine's intended purpose, because it is not their preferred brand. Others may not like the Pentium because they wanted a different goal for the $500 system; That is fair enough. But let me ask, I also chose five AMD processors in a row when they best fit the intended purpose for the money. In fact, make that 7 out of 9 systems, spanning 2+ years, were all AMD processors, all my choice. Did you bash the whole site just as much then because I ignored Intel's budget offerings? Or did you understand we do this 4 times a year, having one shot each quarter, covered Intel's best already, and moved on to addressing the weakness we found with what could best serve the purpose (Athlon II and Phenom II X3 & X4).
Well I sure can't ague with that realization.
In my world a sub-$2K rig is a nice terminal. However, I clearly don't represent the 'typical' end user. Most folks (here) are using their PC's to play games, surf the web, check their email and rip media. There is no doubt the occasional person actually uses their PC for something more productive, but the 'productive' uses are so far and wide that you'll be stretched to find a compilation of suitable benchmarks that are going to be useful.
Any decent 2-core+ CPU can game and the differences stem from the games 'physics/computations/etc' (CPU) and last Jan you guys did such an article comparing CPU's; it would be nice to see a 'plus-$200' version - http://www.tomshardware.com/reviews/gaming-fx-pentium-apu-benchmark,3120.html otherwise in the $1K/$2K you'll have fluctuations in GPU's which forces me to read the 'fine' print of the GPU's clocks & settings not to mention you can run the same test 5 times and have 5 different results. So spitting hairs on FPS ±2%~3% might not mean a thing or worst misleading.
Also, a very important note between Marathon's AMD and/or nVidia may update their drivers so retrospective (old) comparisons become a mute point and are indeed very misleading. You can have a 5%~20% jump in performance just as a matter of a driver patch/update.
Statistical Errors & Standard Deviation - Run 3DMark11 10 times and if you achieve the same results you're either a lair or it's a fluke. Further, the duration's of the test are way too short which adds to more potential errors; 3 sec/60 sec vs 4 sec/10 min (600 sec); 5% vs 0.67%.
There are several bench suites that actually will show the differences between then number of cores and hyper-threading so I suggest you explore those benchmarks in future Marathon's.
I could argue for right now. Today NewEgg has a shell shocker deal for an OCZ Agility 3 240GB SSD for only $139.99!!!!!!!!!!
(I know its OCZ and SandForce but I have a 120GB version of same drive and with latest firmware it seems to be doing just fine)
Scrimping here and there would allow for higher graphics options, and better differentiating from the 1k build in gaming benchmarks.
Anyways, great articles as always! Thanks Tom's Team