System Builder Marathon, Sept. 2010: Value Compared

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
[citation][nom]Crashman[/nom]The CPU is throttling the rest of the system. Most of the benchmarks show a CPU-capped pattern.[/citation]
But I bet that cpu wasn't using anywhere near 100%... It'd be interesting to find where the real limits are being imposed there. Is Nvidia playing some nasty driver tricks again or can they seriously just not write drivers that can efficiently use a hexacore processor. Seriously, I know clock for clock AMD is behind but it had two extra processors that were apparently un-utilized and somehow it still seemed gimped compared to what a Phenom II vs core i5/i7 should look like.
 
okay... i got caught up on the dirt results. Must have been a driver bug probably a SLI problem rather than anything else. Elsewhere the Phenom II looks to have done quite well.

AMD's Fermi power usage campaign seemed to have been spot-on. That power draw is scary. HAHAHA
 

1. The cheapest build DID have a decent video card. For the price, it does a great job. Can settings in the latest titles be maxed at the highest resolutions? Of course not. For people playing older and/or less demanding games, or willing to lower some settings, this card is a great mainstream choice.
2. I'd like to see GTX460 SLI too, but they weren't available when this build was done. Maybe next time, although with the HD6xxx coming out, who knows...?
3. Using the current benchmark suite, a SSD would make no difference; but I see that as a problem too, especially where balance is a (or the) goal. I'd like to see some benchmarks created or added that would show how much a SSD improves the user experience. Having used one, I don't see ever again building a rig for myself without it. If I've got 10-15 minutes in the morning to check email, Slickdeals, and the daily Woot (and the Tom's lead story), I probably don't have 2-3 minutes to waste just booting up.
 
Learned lots of stuff this week:

1. Considering the timing of the $2000 build, I'd upgrade the processor and go for 2x470s in SLI. Or 2x5870s in Crossfire. Now that the GTX 460s are out and after reading this article, I'd spend my $2000 (if I had 'em) on an i7 platform and 2x460s. I agree with cknobman's plans for this build, though: Strip out the power hungry 480's and sell them for a grand and buy a 5850 or 460 for 300 bucks. For the third time this week: I love the case!

2. The $1000 build was marred only because the builder forgot to tweak the memory properly (forgot? memory? oxymoron?). I'm now thinking what could be skimped on the setup to put in 2x460s. Nevermind. It's great as it is.

3. For the $400 build, I already gave my two cents in the comments section of that article. Personally, I think it was well built. Yes, it's wanting in several departments but then folks who can only afford such entry-level builds have a lot of compromises to deal with: I'd choose a different PSU, wrack my brains on alternative GPU options, get better cooling, etc., etc...

Too bad I'm not a resident of the US of A. I'd like to have a shot at any of these systems.

Hey Tom's, for the Christmas SBM, could let those abroad join, too? You don't have to do it on all succeeding SBMs... just once a year, during the Holidays.

Amen.
 
An interesting comparison indeed. Sad to see the X6 being a bit too weak though. While the X6 does seem to do well for threaded apps, the disappointing game performance certainly seems to make the dual 480s a bit of a waste. It just confirms that for gaming the Phenom II X4s will do just as well as the X6 and for the price of an X6 an Intel Setup is a better gaming option. AMD needs Bulldozer now.
 
I think something is wrong with the Dirt 2 results. I benched my pair of 5850s to over 80 fps average...

As for the builds themselves, they're ok but nothing spectacular. I think the forum builds have better.
 
i've been reading a lot about the great three way sli scaling with the gtx 460, i wonder why they chose two 480's for the 2000 sb
 
I like that the $2000 system simultaneously illustrates the folly of design by committee AND my desire to own said machine. I like it.
 
Like I said in another comment, This will be the first month that people sign up hoping to win the 1k machine instead of the 2k.

Well, I hope to win any of them. I wouldn't use the 2 higher systems as is, though. If I won the low end machine, it would become my TV/Hulu/Netflix machine, with a change in the hard drive for one of my left over 1TB drives. If I won the 1k machine, I'd throw the GTX470 into my main machine and run SLi, then put the rest together for a system for my dad with a Radeon 4870 I have sitting around. If I won the 2k machine, the video cards would go into my machine, that machine would get my current GTX470 and go to my dad. My dad would likely love to get a new machine right now.

Only the low end machine is really balanced. To much GPU and not enough CPU for both the 1k and 2k machines. They just can't get the performance out of the video cards they should because of gimped CPUs. I've neveer build machines like that.

Why don't people just go for the Core i7 930 as a base? It's a good value, overclocks well, and can handle the higher end video cards. With the 2k machine, I'd go with the 930 with a 6GB memory kit and offset the cost with a pair of GTX470s instead of GTX480s. It would wipe the floor with this 2k machine.
 
Always love these articles. It's interesting to see the same debates happening that we see in the forums about choices.

Bummer about the big rig not being quite what you'd hoped. Better luck with the next one.
 
I think you guys are just trying to establish a baseline so that Sandy Bridge chips look good when dropped into a low-end system in lieu of separate low-end CPU/GPU combinations 😉
 
In the beginning of reading of this article I tend to explain inefficiency of the 2000$ machine as the strong dependence of 480s on CPU power. But at the end after going through some very strange benchmark results I tend to think there is much more going on there. For instance, how is that in Crysis on Very High details 1000$ machine with a single 470 card beats 2x480 for 2000$ one? AMD was running 4GHz, thats not such bottle neck that 2 cards should be slower than one SLOWER card. This just does not make sense unless there is something more going on. Could you test this issue more? Otherwise, people may tend to think its impossible to pair AMD with SLI from now on, but it is not clear why. Is it only CPU? If yes, does it means because Nvidia uses for the SLI calculations only single core? Or is it some bottle neck of motherboard, where 2x480 in SLI were not running full speed? Is there an issue with drivers, after all 480s are quite new, etc.

It would really help to have CPU and GPU usage plotted while benchmarking this AMD machine and it would also help to put there only one 480 and 470 from 1000$ machine for comparison. Also, changing of motherboard. We have plenty of reports on 2x480 running on Intel platforms (mobo+cpu) pretty fine, but there is no deep study on SLI on AMD platform and these published results tell it does deserve some more time.

Is it possible that SLI is broken/gimped on motherboards for AMD cpus in general or something like that?
 
I bet there are some guys at AMD pulling their hair out right now.. That Hexa Core is pure fail. Even the i5 in the second build would have done a much better job. I was a big AMD fan boy back in the Athlon days, but I gotta admit, AMD ain't up to snuff now. Hope they come up with something to even out the playing field soon! One thing I have to say is that they could've stuck an HD5970 in the place of those 2 GeF480s and invested that extra cash on the MoBo and CPU, an i7 930 OCed can give even the i7 6 Core a run for it's money in some things, and it doesn't break the bank! Would have made a lot more sense.. But, their intentions were good: give people a choice besides Intel, unfortunately Core i5 & i7 are just to good for AMD right now..
 
For the next SBM I would love to see Tom's attempt to build a DIY water cooled rig. I have been thinking about switching from air cooling to water cooling in order to make my rig less loud, but the number of choices is a bit daunting. What pump is quiet, powerful and reasonably priced? How many low speed fans do I need for a 4x140 radiator? In what order should I hook up the CPU/GPU/MB. A 2500 budget might be appropriate for a good dual GPU water cooled rig.
 
[citation][nom]mattmock[/nom]For the next SBM I would love to see Tom's attempt to build a DIY water cooled rig. I have been thinking about switching from air cooling to water cooling in order to make my rig less loud, but the number of choices is a bit daunting. What pump is quiet, powerful and reasonably priced? How many low speed fans do I need for a 4x140 radiator? In what order should I hook up the CPU/GPU/MB. A 2500 budget might be appropriate for a good dual GPU water cooled rig.[/citation]

That and making SSD obligatory for 2000$+ builds.
 
I'm glad to see Tom's attempt more balanced builds this go around. Personally I only play games 25% of the time and its always frustrating to see builds where 50% of the money is spent on graphics cards, although it seems like the $2k build didn't quite hit the mark. Maybe next go around you can make a better effort at hitting a balanced build on the high end.
 
[citation][nom]mcvf[/nom]That and making SSD obligatory for 2000$+ builds.[/citation]Nope, they're worthless in the benchmarks and they add cost. I don't care how many times people ask, I can continuously refer to past SBM's where they were used, and made the system fall even lower in the value analysis.

You'd might as well be asking for a $700 case...ooops, been there done that too.
 
[citation][nom]mcvf[/nom]That and making SSD obligatory for 2000$+ builds.[/citation]

This has been explained over and over. The point of the Marathon is to build the best performance on their benchmark suite under the price limit. SSDs do not help it. The end.

[citation][nom]mcvf[/nom]There's something fishy with the AMD, needs more testing[/citation]

I agree with this. Despite the fact that the build was meant to appease the AMD fans clamoring for a spot in high-end testing, I think it will only make them feel angry and mis-represented. It was nice of Tom's to respond and own up to their mistake, but at the same time the disparity is so great as to be pretty strange, and imply there's some deeper problem with the system than "LOL MASSIVE AMD FAIL." There have been some pretty enlightening discussions regarding bottlenecks and AMD's interface with SLI.. I'd really like if Tom's held onto the system to do some more testing and fed us some more comparisons.

It almost seems like the poor AMD was set up to appear a failure, paired with a way overpowered graphics card configuration and running applications it wasn't designed to shine in. Maybe the conclusion, as seeemed to be at the end of the $2k machine's article, should just be "for powerful graphics solutions, AMD is insufficient." I'm not sure if it deserves being dismissed like that, though, especially with that really wonky stuff going on in the benches.
 
[citation][nom]Crashman[/nom]Nope, they're worthless in the benchmarks and they add cost...[/citation]
In the current benchmarks, true. Particularly when discussing balanced builds, however, please introduce one or more benchmarks that can show how much a SSD can improve the user experience. It can't always be about FPS.
 
[citation]This has been explained over and over. The point of the Marathon is to build the best performance on their benchmark suite under the price limit. SSDs do not help it. The end.[/citation]

Yes, we understand. This isn't some immutable law though. What people are doing is expressing interest in a SBM that includes other things. Perhaps some loading/booting benchmarks could be added that would show the effect of HDD performance. System sound level could be measured.
 
Status
Not open for further replies.