System Builder Marathon: The $5,000 Extreme PC

Conclusion

Tom’s Hardware editors love a great performance value, but did we expect a $5,000 PC to have twice the performance of a $2,500 system? Of course we didn’t. Buyers who break out of the mainstream market should always expect limited return on investment, and the more they spend, the less they get for their money.

But there are always a few builders willing to spend nearly any amount of money to stay on top. For them, practicality means far less than superiority, and our $5,000 PC might not be enough to meet their demands. But even if our $5,000 budget limit can’t create the “ultimate system,” its performance should be similar. Let’s have a look at just how much performance we gained and how much it cost us in value.

3D performance is where the bulk of most high-end system budgets go, and gains of up to 35% over the overclocked December PC are nothing short of amazing.

Encoding gains are fairly small and reflect the 5% increase between our December and February overclocks.

Productivity gains are better than encoding differences and the larger difference is likely a reflection of the new system’s extra RAM.

A combined performance chart shows the February extreme build with a noticeable 13% lead in average performance. The lead was even larger in non-overclocked systems, at 19%.

Even with its 13% performance lead, our $5,000 overclocked PC lost 53% in average value compared to our $2,500 overclocked system. This doesn’t bode well for tomorrow's price/performance shootout, but extreme builders can ignore the price and continue focusing on the system’s performance superiority.

Recent price drops would have allowed us to reconsider our sound card selection and a second 1.5 TB storage drive for redundancy, which are two changes that anyone who likes today’s system should likewise consider. However, neither of these would significantly improve benchmarks, so the added features are just one more step away from the notion of “bang for the buck.”

Create a new thread in the US Reviews comments forum about this subject
This thread is closed for comments
86 comments
Comment from the forums
    Your comment
  • xx12amanxx
    Nice article!

    I wish i could own a rig like that,but it would be that or a 383 stroker for the F-body..lol I can dream cant I?
    1
  • Crashman
    xx12amanxxNice article!I wish i could own a rig like that,but it would be that or a 383 stroker for the F-body..lol I can dream cant I?


    383 strokers are for copycats. Basically, too many bad 400 blocks and people found a cheap way to re-use the cranks to make their 350's bigger. If you have THIS kind of money, you'd might as well go BIG BORE too. Maybe a bowtie block? At any rate, you'll win more races with a real 400 (or larger custom size) so long as the block is good.
    1
  • one-shot
    It is interesting to note on Page 14. On the Sandra XII Multimedia test the Core i7 965 @ 4.2GHz scored 486,971, while the i7 920 @ 4.0GHz scored 386,867. The difference was 200MHz and made such a large difference. Did the Intel SSDs influence such a large gain in performance or the DDR3 @ 1800MHz or perhaps a combination of both?
    2
  • Crashman
    Anonymous said:
    It is interesting to note on Page 14. On the Sandra XII Multimedia test the Core i7 965 @ 4.2GHz scored 486,971, while the i7 920 @ 4.0GHz scored 386,867. The difference was 200MHz and made such a large difference. Did the Intel SSDs influence such a large gain in performance or the DDR3 @ 1800MHz or perhaps a combination of both?


    The 965 has a higher-bandwidth QPI link, so it should be good for boosting at least a few synthetic scores.
    3
  • gim159
    BAH!! Why is it always Intel? Only Servers and people that wish they MAC should use Intel stuffs.. Who else is going to use all of it's features? Not gamers, that is for sure!! Like a gamer is going to spend the extra $$ for ECC memory and not spend it on more important stuff like a GC or a great board. That and an economy minded PC buyer will go AMD anytime, more bang for the buck...

    Also, Ati is better a better card, for linking I mean. The support may be crap, but it is usually worth it to get them! 4850 X2, Water cool one and watch the clock go through the roof!
    -24
  • xx12amanxx
    Intel is the fastest thats why...This is supposed to be an uber rig.

    Ya crashman thats the problem i dont have that kind of money..lol A fresh stock rebuild bolt on's and spray will have to hold me off until better times!
    4
  • gim159
    ¿ "Intel is the fastest" ? Amd 2.0 is a 4.0 in the intel world...
    -20
  • JeanLuc
    I would love to seen those Windows boot times with those RAID 0 Intel SDD's!
    6
  • gim159
    Yeah, Optimize the heck out of it and take all of the unnecessary stuff out.. Probably 4 sec boot, Awesome!
    -8
  • Crashman
    xx12amanxxIntel is the fastest thats why...This is supposed to be an uber rig.Ya crashman thats the problem i dont have that kind of money..lol A fresh stock rebuild bolt on's and spray will have to hold me off until better times!


    I spec'd out a friend's 406 C.I. mouse a few years ago and he got through under $4500 with aluminum heads and a roller cam!
    0
  • DjEaZy
    ... maybe build some AMD based systems too, to see, how they stack up against in price/performance... just for tha fun of it...
    2
  • _horse
    Great article!

    HOWEVER, Should have used an Antec1200 for that much coin on the case. I have one at home using water cooling and its so much easier than any other case I've used to date. Cheaper too, in this instance.
    3
  • Crashman
    _horseGreat article!HOWEVER, Should have used an Antec1200 for that much coin on the case. I have one at home using water cooling and its so much easier than any other case I've used to date. Cheaper too, in this instance.


    I've owned both, the Cooler Master is a nicer case. I mean, we're just throwing opinions around now, right?

    But the Cosmos S is far more portable, and it fits the big radiator perfectly. In fact, it's the only stock case to fit that radiator properly.
    4
  • LATTEH
    With a build like that you guys should have tryed to set Crysis with 16 AA!


    well it probably wont be playable but it would just be neat to see.
    0
  • _horse
    CrashmanI've owned both, the Cooler Master is a nicer case. I mean, we're just throwing opinions around now, right?But the Cosmos S is far more portable, and it fits the big radiator perfectly. In fact, it's the only stock case to fit that radiator properly.


    Thats true, but I didnt know we were going for portability here, especially with a liquid cooled system.
    1
  • rodney_ws
    There's no way that sound "card" is up to the level of a $5k rig.
    2
  • jcknouse
    Nice article.

    Question for the writing staff:

    Have you thought about taking all your review statistics, and assembling an "uber system" based on the best parts based on what you have found in your review tests?

    Just curious. I don't remember that ever having been done before here. And for $5,000, you surely could afford to throw together all of the top notch parts into a system for kind of a "what happens when you put all the best parts together" article.

    Now you guys have me itching to build a new system already...and I just built one back in September!! lol
    3
  • jcknouse
    That is a really good question too:

    With the width of the video cards, where would you plug in a soundcard? Does it use the 3rd PCI-E x16 slot?

    Just curious. I've always noticed how the ATX motherboard size standard hasn't shifted to grow with the growth of the size of components, such as video cards and component heatsinks.

    Limited room bites. lol
    1
  • cah027
    Why not 3 liquid cooled 285's. Maybe drop the Blueray and or go down to a 920 in the next one. Keep the SSD's.
    2
  • Crashman
    cah027Why not 3 liquid cooled 285's. Maybe drop the Blueray and or go down to a 920 in the next one. Keep the SSD's.SSD's are fast but the only benchmarks they matter in are the ones that don't count (synthetics)


    SSD's are fast but the only benchmarks they matter in are the ones that don't count (synthetics)
    1