System Builder Marathon, August 2012: $2000 Performance PC

Getting Our Core i7-3930K To 4.6 GHz

We began our CPU overclocking by setting what we felt was a maximum safe voltage, and then trying to find the highest possible stable frequency at that voltage. Jumping up to the 47x CPU multiplier created instability, as did choosing a 101 MHz base clock with a 46x multiplier. We had to settle for a flat 4.60 GHz.

While our memory’s XMP-1600 CAS 8 profile worked as intended, ASRock’s X79 Extreme4 automatically loosened timings when we increased the memory multiplier beyond its XMP rating. This allowed us to choose DDR3-2133 with minimum fuss.

After a few sessions of stability testing, we were able to manually tighten some of the memory timings that the motherboard had automatically relaxed in response to our higher data rate. The board selected 11-11-11-32 timings at DDR3-2133, but we found latencies of 10-11-10-24 to be stable.

Getting to CAS 10 was important to us, since CAS 11 would have resulted in a slightly slower-than-rated response time. Since DDR3-1600 CAS 8 and DDR3-2000 CAS 10 have the same response time, our memory tuning efforts should increase both bandwidth and quickness.

Though we weren’t able to push our CPU frequency higher than 4.60 GHz, we were able to save some wattage by backing down our voltage selection to 1.32 V.

Create a new thread in the US Reviews comments forum about this subject
This thread is closed for comments
124 comments
    Your comment
    Top Comments
  • Crashman
    sarinaidei5-3570k/i7-3770kGigabyte G1 Assassin Z77120GB SSD500GB HDD2xGTX 6702x4GB DDR3 1866And still probably cheaper with obviously better performance.
    Probably not, unless you're only testing games. But we should probably test that anyway. Does anyone else want to see it?
    21
  • Crashman
    namelesstedI am sorry. This is one of the dumbest builds I have seen in a long time...
    No, it's just you (the noob thing). These gaming resolutions are too low to take advantage of SLI, and this one DESTROYS yours in this benchmark set.

    Your config would only be better if there were significant changes to the benchmark set.
    20
  • Crashman
    zander1983Ditch the BR Writer, get a BR combo drive and save yourself $60
    Sorry, I don't see any combo drives for $30 so the savings would be much less than $60. Plus, you'd lose BD-RE backup capability, which can be handy.
    13
  • Other Comments
  • Darkerson
    Interesting setup. I would have favored a way beefier single GPU or a nice dual GPU setup, but I mainly only game, and dont do a lot of encoding or whatnot.
    12
  • Anonymous
    Quote:
    The contest opens on August 20, 2012 9:00 PM PDT and closes on September 3, 2012 9:00 PM PDT.

    So... i notice now that it opens at August 20, not August 19 when the $500 SBM appeared. I submitted my entry at August 19 10:30 PM. So that means that i haven't entered into the sweepstakes, or did i? I am confused, cause only one entry can be accepted.
    4
  • Anonymous
    Nice quality build! Enough said!
    5
  • trumpeter1994
    That has got to be one of the luckiest GTX 670s I've ever seen.
    6
  • sarinaide
    i5-3570k/i7-3770k
    Gigabyte G1 Assassin Z77
    120GB SSD
    500GB HDD
    2xGTX 670
    2x4GB DDR3 1866

    And still probably cheaper with obviously better performance.
    -10
  • Crashman
    sarinaidei5-3570k/i7-3770kGigabyte G1 Assassin Z77120GB SSD500GB HDD2xGTX 6702x4GB DDR3 1866And still probably cheaper with obviously better performance.
    Probably not, unless you're only testing games. But we should probably test that anyway. Does anyone else want to see it?
    21
  • zander1983
    Ditch the BR Writer, get a BR combo drive and save yourself $60
    -4
  • Crashman
    zander1983Ditch the BR Writer, get a BR combo drive and save yourself $60
    Sorry, I don't see any combo drives for $30 so the savings would be much less than $60. Plus, you'd lose BD-RE backup capability, which can be handy.
    13
  • sarinaide
    CrashmanProbably not, unless you're only testing games. But we should probably test that anyway. Does anyone else want to see it?


    It would be very interesting, the IvyBridge chips in productivity numbers hold quite well with the SB-E chips that is the only area which should be a contest.
    2
  • crisan_tiberiu
    16GB ram pointless imo. 2 TB 5400rpm hdd? ...i rather get a 1 TB 7200 rpm hdd. i7 3970k ... i rather get the i7 3770k. From theese i would squeeze in a gtx 680.
    -1
  • mayankleoboy1
    CrashmanProbably not, unless you're only testing games. But we should probably test that anyway. Does anyone else want to see it?


    2x670 is overkill for 1920x1080.
    But a 2000$ build is already overkill. So this SLI setup is OK
    3
  • bawchicawawa
    Should have cut some corners on some 'overkill' items and gone with sli 660 ti's or a single 680.
    -11
  • namelessted
    I am sorry. This is one of the dumbest builds I have seen in a long time. It is almost as if the person who put this together was just a complete noob to PC building. Its like they just went to CPU and found a really expensive one and decided to get it.

    I just tossed a build together on PC Partpicker that would absolutely DESTROY this build.

    CPU - 3570K
    CPU Cooler - Corsair H100
    Mobo - ASUS P8Z77-V
    RAM - Corsair Dominator Platinum 8GB (2x4GB) 1600
    SSD - Corsair Force GS 360GB
    GPU - ASUS GTX670 x2 in SLI
    Case - Corsair 550D
    PSU - Corsair AX750
    Optical - Asus whatever

    If you purchased everything from Newegg it would be $2070. Yes, a bit over the budget, but I am sure I could trim that off somewhere without too much difficulty. I definitely didn't go with the best valued products in my build. The Platinum ram is double the price of their normal RAM, but it fit with the theme espoused by the author of this article of having a QUALITY build, which I totally agree with.

    On the subject of quality, I have chosen a better cooler, better case, and better PSU. How can anybody seriously justify buying a non-modular PSU for a $2k build? That is insane to me. I wish I had the $2k to actually put this build together with a couple of tweaks and put it up against Soderstrom's build and watch him weep as his system gets destroyed.

    Also, I realize I haven't selected a storage drive. I just went with one big SSD. Yes, the 2TB is nice, but I don't think most people actually need that kind of storage, and if you are somebody that does need it, it is a separate cost that should be part of the main build, IMO. Similarly, we typically don't include monitor, keyboard, mouse, etc. pricing into builds. I think mass storage needs to join this category.
    -15
  • Crashman
    namelesstedI am sorry. This is one of the dumbest builds I have seen in a long time...
    No, it's just you (the noob thing). These gaming resolutions are too low to take advantage of SLI, and this one DESTROYS yours in this benchmark set.

    Your config would only be better if there were significant changes to the benchmark set.
    20
  • Mac_McMan
    CrashmanNo, it's just you (the noob thing). These gaming resolutions are too low to take advantage of SLI, and this one DESTROYS yours in this benchmark set.Your config would only be better if there were significant changes to the benchmark set.
    You were right the first time, you can tell by the language he's trolling
    3
  • Crashman
    Mac_McManYou were right the first time, you can tell by the language he's trolling
    OK then, ignoring the hate and going back to rational response, Sarinaide recommended a PAIR of GTX 670's, a 3570K, a Gigabyte Z77 board and 8GB RAM. But this still needs to be a $2000 PC or else it becomes an "enhancement" of yesterday's $1000 build.

    I'm going to take a long shot and say, blow the leftover money on quad SLI support. That could lead to even more tests down the road, no?

    Of course an Ivy Bridge SLI build would still lose under this benchmark set, so it would need to be tested at higher gaming resolutions. Is everyone OK with 5760x1080?
    12
  • namelessted
    CrashmanOK then, ignoring the hate and going back to rational response, Sarinaide recommended a PAIR of GTX 670's, a 3570K, a Gigabyte Z77 board and 8GB RAM. But this still needs to be a $2000 PC or else it becomes an "enhancement" of yesterday's $1000 build.I'm going to take a long shot and say, blow the leftover money on quad SLI support. That could lead to even more tests down the road, no?Of course an Ivy Bridge SLI build would still lose under this benchmark set, so it would need to be tested at higher gaming resolutions. Is everyone OK with 5760x1080?


    Have you looked at the test results? The old build with a GTX680 beats the current build when you look at maximum settings at 2560x1600. Let me say this again, at the highest settings and full resolution the old build with the GTX680 beats the current build on every single game tested in this article. Every single game.

    A 670 SLI setup would only further the performance gap.
    -3
  • Crashman
    namelesstedHave you looked at the test results? The old build with a GTX680 beats the current build when you look at maximum settings at 2560x1600. Let me say this again, at the highest settings and full resolution the old build with the GTX680 beats the current build on every single game tested in this article. Every single game.A 670 SLI setup would only further the performance gap.
    Games make up 30% of the benchmark set. 2560x1600 makes up 25% of gaming scores. Therefore, 2560x1600 performance makes up 7.5% of the overall value score.

    On the other hand, heavily-threaded programs make up 50% of the encoding and 75% of the productivity benchmarks. That's 37.5% (15%+22.5%) of the benchmark totals. 37.5% is a much larger portion than 7.5%, so 3930K+GTX 670 beats 3570+SLI. It's simple math, and the only way to change that math is to change the benchmarks.

    Really, dropping the 3930K only furthers the performance gap.
    8
  • namelessted
    CrashmanGames make up 30% of the benchmark set. 2560x1600 makes up 25% of gaming scores. Therefore, 2560x1600 performance makes up 7.5% of the overall value score.On the other hand, heavily-threaded programs make up 50% of the encoding and 75% of the productivity benchmarks. That's 37.5% (15%+22.5%) of the benchmark totals. 37.5% is a much larger portion than 7.5%, so 3930K+GTX 670 beats 3570+SLI. It's simple math, and the only way to change that math is to change the benchmarks.Really, dropping the 3930K only furthers the performance gap.


    First, I would argue that when it comes to a $2k build, the only thing that matters in terms of gaming benchmarks are the max settings and resolution. When you are spending that much, it literally doesn't matter how well it can do on mid settings and 1080p. Completely irrelevant.

    Secondly, I might have to argue that the benchmarks should be changed to better reflect real-world scenarios instead of reporting Sandra numbers.

    It is just extremely frustrating to see a build like this. With $2k there is so much potential to put together a truly great and balanced machine. This build is far from that. Soderstrom had all that money, and it just feels like he picked out a crazy CPU and then just went down the line and picked random other hardware. I also realize some of the choices are personal preference. I personally think the Phantom 410 just looks awful. When I see a case that looks like that, it makes me think of a 14 year old kid building a "cool" PC.

    There is also that fact that every single component in this build all come from different companies. For budget builds that absolutely makes sense. You have to find deals where they are and that pretty much always means buying different brands. But, he had $2000. It is just something that makes no sense to me, to open up a PC and see that every single component not matching up in any way.

    The whole build just feels like Soderstrom picked a CPU, and then just added the rest of the parts to the cart and he just didn't save the proper budget for a GPU and decided to downgrade it instead of figuring out where money was just being wasted.
    -1
  • Crashman
    namelesstedThe whole build just feels like Soderstrom picked a CPU...
    Soderstrom didn't pick any of the major components. Back in the Q2 SBM, readers complained that they wanted an SB-E and LGA-2011 in the $2000 machine. Readers also complained about the price of its GTX 680, since the GTX 670 performed almost as well for much less money (the GTX 670 wasn't available when the order was placed). Readers also complained that the SSD should not have been shrunken from the Q1 system's 240GB. So you're complaining about reader picks in the 3930K, single GTX 670 (not enough money left for two), and 240 GB SSD.

    So your real beef is with your fellow readers, not the builder of this machine.
    13