System Builder Marathon Q3 2015: Value Comparison

The Results

Synthetics

3DMark & PCMark

3DMark rewards Eric’s expensive graphics card and my own expensive CPU, while also showing Chris’ combo as equally viable to Eric’s in its overall score. As our own benchmark suite also produces overall scores, I evidently don’t stand a chance.

PCMark rewards Chris’ use of an SSD, while also giving my machine high work and home scores for its exceptional CPU. Unfortunately, only the storage score is used in our final evaluation.

SiSoftware Sandra

My Core i5-6600K reigns triumphant over Sandra’s Arithmetic and Cryptography benchmarks, while Eric’s low-cost Athlon X4 860K puts a good show in Arithmetic alone. The competition heats up!

And then there’s Sandra Memory Bandwidth. I used to go after Don on the overclocking configuration for not even attempting to push his RAM, yet this time I actually purchased RAM that wouldn’t. That’s why I left well enough alone concerning Chris’ configuration. As for Eric, his lower-end memory controller was still capable of some O/C optimization.

Gaming

Arma 3 & Battlefield 4

Arma 3 appears CPU-limited at our lowest test setting, giving my entry-level GPU a rare chance to top Chris’ mainstream and Eric’s high-end parts. CPU limitations remain in effect for the AMD Athlon build through 1920x1080. Chris’ medium-GPU/medium-CPU combo take the lead at triple-screen resolutions, but only at standard quality settings. Eric’s big GPU finally pays for itself at high resolutions, but only with “Ultra” quality enabled.

Battlefield 4 gives Eric’s GPU a far greater workout, nearly ignoring the poor IPC of his CPU. Gaming PC is starting to look like a misnomer for Chris’ build.

Far Cry 3 & Grid 2

Eric’s GTX 970 takes all comers in Far Cry 3, though its lead surprisingly disappears at Ultra quality and 4800x900 and returns at 5760x1080. My Prosumer PC falls far behind both of the other machines with its entry-level gaming card, while the Gaming PC splits the difference at triple-screen resolutions.

High quality is a fairly low GPU load in Grid 2, and the game becomes DRAM-bottlenecked, then CPU-bottlenecked, before finally leaning harder on the graphics card in triple-screen resolutions. I’ll gladly add the low-end wins to my overall score, just as Eric compiles all of his his triple-screen, ultra-quality wins.

Applications

Encoding & Creativity

We know that the dual-core i3 in Chris’ Gaming PC has better IPC than the quad-core Athlon in Eric’s machine, yet we’re still surprised that the higher-clocked quad-core wins only one video encoding benchmark. As for audio encoding, it’s single-threaded and thus favors both higher clocks and higher IPC, giving the Prosumer PC a win.

Chris’ dual-core gaming PC beats up Eric’s quad-core even in Photoshop OpenCL, where he might have hoped the GTX 970 would help. Shifting our attention to processors of similar IPC, my true quad-core Prosumer PC barely outpaces the virtual quad-cores of Chris’ Gaming machine in Adobe After Effects.

Productivity & File Compression

Adobe Acrobat is single-threaded, so my faster clock barely outpaces Chris’ slower clock, and Eric’s faster clock can’t compensate his lower IPC processor design. I was hoping Eric’s Mini PC would outpace the Gaming PC in 3ds Max, but it can’t.

Chris was the only builder to use an SSD, and it paid off with a surprisingly large win in 7-Zip.

Power & Heat

I built my Prosumer PC for efficiency whilst seeking the lowest cost high-end power supply, as well as compensating for the effect of graphics heat on CPU temperature within a poorly-ventilated case. Eric had similar thoughts, choosing a lower wattage Nvidia card over a similarly high-end AMD part. Chris went all in, yet still kept power consumption low with a dual-core processor. When the electric bill arrives, everyone wins!

An exceptionally high CPU overclock heated up my core, while a tight chassis heated up Eric’s GPU. Though a better case helped Chris’ Gaming PC remain cool on both fronts, but his middling overall performance and power consumption kept his efficiency in the middle position.

Thomas Soderstrom
Thomas Soderstrom is a Senior Staff Editor at Tom's Hardware US. He tests and reviews cases, cooling, memory and motherboards.
  • TNT27
    Console PC ~ 300-400$ build, Average Pc 800, and 1600$ ultimate
    Reply
  • Crashman
    How about we throw out the budget limits next time and ask everybody to fix all the problems with their current build specs? We'd still need to compete for value, so the "no budget" part should still yield machines in the $900 to $1200 range, because once you go "too far" you start to loose the "price/performance" math.
    Reply
  • TNT27
    16706286 said:
    How about we throw out the budget limits next time and ask everybody to fix all the problems with their current build specs?

    Im kool with that!
    Reply
  • synphul
    I can appreciate the results of these specific systems in terms of performance but I'm seeing some serious flaws here. The 'gaming' build opted for an ssd which has little impact on gaming performance and settled for a lower end gpu while the amd mini build went with a gaming gpu and ditched the ssd? No wonder the amd mini system got best marks for performance/price in gaming. It would have been a much different story the other way around.

    Unfortunately prices are volatile and some of the prices listed aren't accurate. The gtx 970 used in the amd mini build must have been having an incredible sale as it's still offering a rebate card and yet right now on newegg it's $325 ($305 with rebate) which alone blows the $800 budget. I was thinking the newegg links automatically updated to reflect current pricing via newegg's site but it doesn't appear to.

    Similarly I don't recall the wd blue 1tb being close to $70, most times in pc partpicker it's around $48-52 and currently on newegg it's $53. If that gtx 970 were actually $250 it could've been added to the i3 gaming build but at it's actual price blows the budget. Then again it blows the amd mini pc budget based on current price anyway.
    Reply
  • Onus
    I like the idea of seeing what would happen if the current builds were "fixed." This will give useful information on the future-resistance of each build.
    Otherwise, three machines at the same budget has provided some excellent data points. IMHO, this is the best SBM we have seen in a good while now. They're one of my favorite features of the site, and this quarter's series was exceptional.
    Another idea would be to keep the same purposes (and budgets) for which these three PCs were built, but randomly assign each one to a different builder; same purpose, same budget, ok who wins?
    Reply
  • AndrewJacksonZA
    I'd like to propose lowering the budget to $400 and aiming the "gaming" and "HTPC" builds to 1080p. Have a "pro" PC build to target pro and casual gaming workloads with a resoluion of 1440p.
    Reply
  • AndrewJacksonZA
    How about we throw out the budget limits next time and ask everybody to fix all the problems with their current build specs? We'd still need to compete for value, so the "no budget" part should still yield machines in the $900 to $1200 range, because once you go "too far" you start to loose the "price/performance" math.
    In my estimation, (read: guess ;-) slap a Fury or a Fury X on the AMD mini PC and the "gaming" PC and you're pretty much done. That falls within the $900 - $1200 price range.

    I guess the builders could also do a Crossfire or SLI experiment if they really wanted to test out wild ideas (or heck, just look around for a cheap 295X2 :-)
    Reply
  • Crashman
    16706789 said:
    I like the idea of seeing what would happen if the current builds were "fixed." This will give useful information on the future-resistance of each build.
    Otherwise, three machines at the same budget has provided some excellent data points. IMHO, this is the best SBM we have seen in a good while now. They're one of my favorite features of the site, and this quarter's series was exceptional.
    Another idea would be to keep the same purposes (and budgets) for which these three PCs were built, but randomly assign each one to a different builder; same purpose, same budget, ok who wins?
    Randomly...but we each picked our own, it's not as if Eric was assigned the gaming build just because he's filling in for game-system-builder Paul :)

    Reply
  • Aspiring techie
    How about we throw out the budget limits next time and ask everybody to fix all the problems with their current build specs? We'd still need to compete for value, so the "no budget" part should still yield machines in the $900 to $1200 range, because once you go "too far" you start to loose the "price/performance" math.
    I like it! I'd like to see what would happen if the builder can spend whatever he wants. One thing I have never seen here is a build that is completely and totally overkill. That would be cool to see.
    Reply
  • Crashman
    16706970 said:
    How about we throw out the budget limits next time and ask everybody to fix all the problems with their current build specs? We'd still need to compete for value, so the "no budget" part should still yield machines in the $900 to $1200 range, because once you go "too far" you start to loose the "price/performance" math.
    I like it! I'd like to see what would happen if the builder can spend whatever he wants. One thing I have never seen here is a build that is completely and totally overkill. That would be cool to see.
    We started out with a completely overkill $4400 build the first time. No OS in the price either :) Everyone hated it.
    Reply