System Builder Marathon, December 2010: $2000 PC

[hmp_goose]

Haven't we been "warned" about RAIDing SSDs?

 

[Crashman]

[citation][nom]hmp_goose[/nom]Haven't we been "warned" about RAIDing SSDs?[/citation]Warned how? The best explanation I've heard was that "since RAID doesn't support TRIM (true), the cells wear out faster (false?)". Nobody has explained to me how preemptively erasing the cells (TRIM) adds longevity compared to over erasing on-the-fly (No TRIM). The same erase takes place either way, and I've never seen any evidence that wear-leveling algorithms, which are internal on the drive, are negated by RAID.

The only "warning" that makes sense to me is that the drives will slow down after all the cells have been used, making writes somewhat slower...but still faster than we'd see on a single drive (with TRIM).

 

[spellbinder2050]

This was an entertaining read. I just wish that proper memory could have been used.

And where is the Crysis 8x AA that was referenced during the hardware installation? No Metro either? :/

Would a Core I7 x3 OC'd to 4.3 ghz really create a bottle neck for better gpus than two GTX 470's in SLI?

 

[sudeshc]

One of the articles here few days back suggested that 8gb ram is good for 64 bit OS, than why not have that.

 

[Crashman]

[citation][nom]TheCapulet[/nom]Agreed. MW2 is an awful game built on a ancient engine hobbled along to keep making activision $$$. Replacing it with ANYTHING current gen would be better.And honestly... I never realized memory could make that much of a difference. And here I was thinking I knew everything in the known world.[/citation]It was rather stunning, but had the expected memory shown up with the order...the system would have performed better from the outset, no questions would have been asked about memory bottlenecks, and we wouldn't have discovered the issue.

 

[JamesSneed]

The memory issue is an odd one. Could you slap some higher performing memory in there and rerun one of the tests that seem like they were most impacted? Just curious if it really is the memeory that has that profound of an impact or if its multiple issues that are additive like GPU scaling on this platform + memory performance.

 

[Luay]

And to what display will you hook up your 2 GTX 580s to? Add another grand for a 1440 or 1600 display or three 1080 displays. Otherwise GTX 570 will output whatever the GTX 580 is outputting on a 1080 display, even though the 580 is processing more FPS.

Really, isn't it time Tom's should target a certain price point and a certain monitor resolution for your building marathons instead of price alone?

Crosair AX series PSUs and GTX 570 just came out and they are better options, but an Asus Sabertooth X58 review is overdue. These parts have no equal in their categories, for now.

 

[Crashman]

[citation][nom]JamesSneed[/nom]The memory issue is an odd one. Could you slap some higher performing memory in there and rerun one of the tests that seem like they were most impacted? Just curious if it really is the memeory that has that profound of an impact or if its multiple issues that are additive like GPU scaling on this platform + memory performance.[/citation]The tests were used but the numbers weren't reported: It's about a 4% loss in 3 of the games with the 3780 Uncore clock, and 8% loss in one of the games with Uncore at 2x DRAM.

 

[youssef 2010]

Great Article & a Great System, although you can buy it now for $1824 from Newegg. Unfortunately, The fan has been deactivated and the optical drive is out of stock so.

 

[dark_lord69]

If I spent that much on a computer my wife might devorce me.

 

[sonnyisthinkn]

Sound card?... facepalm

 

[kureme]

First time I've seen hard evidence of a memory bottleneck.

 

[tom thumb]

Who spends more on the case than they do on SSDs??

 

[Guest]

It is funny. I would choose the Sabertooth _exactly_ because the UD3R has lane sharing. That way I could use the third slot for e.g. a revodrive and not limit my crossfire/sli setup.

 

[tommysch]

Cheap RAM is cheap...

 

[pcman911]

I too would like to see the difference in the AMD 6850/6870 crossfired. I am no means an AMD fanboy but several sites have been raging about how good a pair of 6850's in CF have been a tremendous value lately. The 6870s also would put some power savings back giving a higher efficiency rating. The 6870s would give the Eyefinity mentioned near the end. Maybe the $1500 build will be based on the i5-760 to see the CPU difference.

 

[NuclearShadow]

If by any chance I win this rig I am going to fix that memory bottleneck.
I'll even do a few benchmarks and see how it fairs after-wards and post the results.

 

[coldmast]

I'm so confused, the June 2010 computer is performing better with lower processor clocks? What is going on (on stock speeds), did someone leave the frame sync on or something?
Same Motherboard
Same Graphics (spec-wise)
Different RAM

 

[cadder]

You mention changing the CPU fan for a lower speed model, and you mention memory bottlenecks in games. I would suggest that you spend the extra money to buy a replacement fan and replacement ram and rerun a few of your tests with the replacement components. I realize that a reader on a fixed budget might not be able to return components and buy new ones, but some enthusiasts would do this, even if they had to sell the almost new components at a slight loss. Even though it doesn't fit completely into the spirit of your build within a certain budget, it would be of value to your readers to see how the optimized components might work better.

Secondly, you are starting to use SSD's in the storage system. You should provide some tests to indicate to the readers how an SSD system performs. I've read your SSD test articles and while interesting, it is hard for me to understand how the various test results would impact real world performance. Maybe you need a few new tests- how long it takes a machine to boot, how long it takes to install a program, how long it takes a program to load, etc., things that the average user could relate to.

 

[James296]

hmmm, I wonder how the $1,000 and $500/$400builds are going to come out, hopefully better, considering prices have drop since last time

 

[Onus]

If you bought that RAM from Newegg based on their photograph of it, I'd hope you'd be able to get them to send you the "real" stuff...

This system is nice, even though I know it would be wasted on me. My apps and games are sufficiently undemanding that I prefer to focus on low power and low noise. The case is fantastic, but I suspect if I win it I'd prefer to see if Tom's and the winner of one of the other two are willing to switch. If not, I'd only use one GPU, a much quieter CPU fan, and probably use the X-560 I already got for my next build for the PSU.

 

[theoutbound]

Great system, but it's a shame about the RAM. I'm looking forward to the rest of the systems from this marathon, but I can't wait until the next round when we should see some new hardware making an appearance. Not that I would turn this system down if I were to win the giveaway or anything.

 

[Silmarunya]

[citation][nom]TommySch[/nom]Cheap RAM is cheap...[/citation]

Expensive Ram is stupid...

There's absolutely no noticeable performance benefit to high end RAM, not even to overclockers.

 

[Crashman]

[citation][nom]cadder[/nom]You mention changing the CPU fan for a lower speed model, and you mention memory bottlenecks in games. I would suggest that you spend the extra money to buy a replacement fan and replacement ram and rerun a few of your tests with the replacement components. I realize that a reader on a fixed budget might not be able to return components and buy new ones, but some enthusiasts would do this, even if they had to sell the almost new components at a slight loss. Even though it doesn't fit completely into the spirit of your build within a certain budget, it would be of value to your readers to see how the optimized components might work better.Secondly, you are starting to use SSD's in the storage system. You should provide some tests to indicate to the readers how an SSD system performs. I've read your SSD test articles and while interesting, it is hard for me to understand how the various test results would impact real world performance. Maybe you need a few new tests- how long it takes a machine to boot, how long it takes to install a program, how long it takes a program to load, etc., things that the average user could relate to.[/citation]Extended file system performance was not recorded for the previous system, so nothing was there to compare this to. The new data will be available in the Day 4 system comparison.

 

[Crashman]

[citation][nom]coldmast[/nom]I'm so confused, the June 2010 computer is performing better with lower processor clocks? What is going on (on stock speeds), did someone leave the frame sync on or something?Same MotherboardSame Graphics (spec-wise)Different RAM[/citation]Read the part at the end of the COD:MW2 test page, and keep these numbers in mind: 4% and 8%.