System Builder Marathon, Q2 2013: $2500 Performance PC

[vertexx]

Any thoughts on how this build would have worked in a Silverstone SG08? Reviews on this case rate the cooling extremely favorably. The 180mm fan blowing right down onto the Motherboard could potentially keep the GPU heat away from the CPU. Also, this case has good ducting for the GPU to keep intake air isolated from waste heat air. It would have required an air-cooled CPU though, which probably would have impacted overclocking of the CPU.

 

[wffurr]

Is the 3D Mark graph labeled correctly? The article text says the stock Q1 $1000 PC is supposed to be faster than the stock Q2 $2500 PC, but the graph shows the opposite. I think both sets of top bars are supposed to be overclocks, but one is labeled as the stock $2500 PC.

 

[Crashman]

It was harder to notice with the previous build’s lower frame rate, but both configurations hit a snag at 2560x1600 and the Ultra quality preset. That snag is memory, where 2 GB of graphics RAM simply isn’t enough. The new build's so-called 4 GB card only gives 2 GB to each GPU, and that proves to be a problem.

Um... no? There is no justification for this. 2560x1600 almost twice as demanding as 1920x1080, and you are just seeing the expected performance drop from the resolution increase. There is no evidence of VRAM limitation.

Sorry BigMack70, you'd probably need to look at other articles. We've done the 2GB vs 3GB graphics tests in the past and proven that 2560x1600 and Ultra details are too much for a 2GB card in BF3. We're talking about a huge improvement when switching to 3GB, something around 30% as I recall.

It was interesting to see specific mods / add-ons called out in this build; I'm not sure I've seen that done before, but the results were certainly worthwhile.
I'd love to win this one. My own games (especially at 1920x1080) don't need a GTX690 so I might swap in a lesser card in order to use the drive cage, but this would be one sweet system to sit on my desk.
How loud were the fans?

Not too bad, probably around 3db less than you'd expect from this card, and this is only a moderately-noisy card.

Top mounted fan on page 8. Assembly Part 3 looks to be an exhaust fan in the picture, but in the text is mentioned as blowing down. Figure the text is correct, might want to note the discrepancy.

Like most fans, it draws air in through the open side and blows it out through the side with the frame. Yes, the top fan is installed as an intake.

Any thoughts on how this build would have worked in a Silverstone SG08? Reviews on this case rate the cooling extremely favorably. The 180mm fan blowing right down onto the Motherboard could potentially keep the GPU heat away from the CPU. Also, this case has good ducting for the GPU to keep intake air isolated from waste heat air. It would have required an air-cooled CPU though, which probably would have impacted overclocking of the CPU.

It was briefly considered but didn't appear to have the same proficiency at letting the graphics heat out of the front panel. Also, it's CPU cooler support is VERY VERY VERY VERY BAD as-delivered, thought it might support a few easy modifactions

Is the 3D Mark graph labeled correctly? The article text says the stock Q1 $1000 PC is supposed to be faster than the stock Q2 $2500 PC, but the graph shows the opposite. I think both sets of top bars are supposed to be overclocks, but one is labeled as the stock $2500 PC.

The text doesn't say "less powerful", it says "less than twice as powerful". In other words, the $2500 PC is almost twice as powerful, and all the chart labels are correct.

 

[gm0n3y]

I'm surprised that there was no liquid cooling on this high-end build.

 

[vertexx]

It was briefly considered but didn't appear to have the same proficiency at letting the graphics heat out of the front panel. Also, it's CPU cooler support is VERY VERY VERY VERY BAD as-delivered, thought it might support a few easy modifications

I had the same thoughts regarding CPU cooler support. It would be interesting to try something like the Noctua NH-C14 with just the bottom fan mounted.

With just the bottom fan, you have the top of the heat sink just 5mm away from the SG08 180mm fan. Also, this article (http://www.xbitlabs.com/articles/coolers/display/noctua-nh-c14_5.html) seems to show that the bottom fan only option is highly effective, as long as you don't use the ULNA adapter.

Anandtech recently did an article on the SG08 with some stats against the Prodigy for comparison (http://www.anandtech.com/show/6944/silverstone-sugo-sg08-miniitx-case-review). While more effective at cooling the CPU than the Prodigy, the testing does support your thoughts that the SG08 would be less effective at cooling the GPU.

Here is another comparison, but with a OC'd I5 and a Zalman low profile cooler: http://www.bit-tech.net/hardware/cases/2012/09/17/silverstone-sugo-sg08-review/3

Not to knock your build choice - I actually think the Prodigy cases are one of the better ITX cases on the market. But it would be interesting to see if the SG08/Noctua NH-C14 combo would do the job without requiring any case mods in a much smaller package.

Overall, the article and testing are nicely done - you really proved the value in OC'ing and also why someone would spend the extra $$ for the I7.

 

[fat_panda]

I've also ran across the problem of installing the Asetek back plate on my Mini-ITX motherboard (EVGA Stinger), before that I had a problem with Havik 140 back plate. I hope companies that make CPU coolers recognize the issue of ATX form factor back plates on ITX motherboards and offer ITX version of CPU back plates even at a cost. The ITX systems are growing in popularity, so it would be nice to see some type of solution with CPU back plates

 

[slomo4sho]

I am assuming its because its for M-ITX only albeit that the HAF XB could take ITX and is as big as a Prodigy.

I assume you mean ATX

 

[Crashman]

I don't know what you are thinking of, but you recall incorrectly.

I don't know what you're prattling on about, because as soon as I saw a chart that wasn't mine I zoned out. Could someone delete these two worthless posts (mine and his above)?

 

[lowguppy]

I had a similar issue with that same universal bracket and the bottom mSATA port on my ASRock Z77 mini-ITX board. Had to dremel the bracket to get it to fit.

 

[allanitomwesh]

I should've kept my mouth shut.
The bitfenix prodigy is now mATX,at the same dimensions. Just saying...

 

[Crashman]

I don't know what you are thinking of, but you recall incorrectly.

I don't know what you're prattling on about, because as soon as I saw a chart that wasn't mine I zoned out. Could someone delete these two worthless posts (mine and his above)?

So commenting on an article with relevant data is now making a "worthless post" and is a moddable offense if it disagrees with or challenges an unsubstantiated assertion made by a staff member? Good to know.

No, I'm saying that adding invalid data to the conversation is even worse than adding none at all.

I came across this issue while performing some SLI tests in the past, but I can't search the site. Really, it takes like minutes for me to load a page. I should appoint you to do that.

 

[allanitomwesh]

Please let the $400 rig ACTUALLY BE SMALL

 

[Onus]

Crash, I think I'm going to side with BigMack on this one, at least to the point of saying that some very specific testing needs to be done to suggest that the results he posted are not relevant. That is not saying he is right, but he does have data that appears to back up his assertions. If it is a matter of the settings used (entirely possible), it would be useful to clarify the settings that cause VRAM to make a difference.

 

[RedJaron]

Seems I was pretty close in my initial guess:

And looks like I lost that bet.

I'm surprised that there was no liquid cooling on this high-end build.

Ummmm, perhaps you should read it again?

 

[Joshua Hill]

Great to see the liquid cooling fitting in this case unlike the previous builders efforts. A fan outside the case is an automatic fail imo.

BUT @BigMack70 is right. Of the numerous tech sites I visit going from 1080p to 1440p or 1600p (depending on aspect ratio) results in approximately 60% the performance in non VRAM limited cases (which is pretty much all the time). HardOCP has done some interestingly analysis on VRAM limited situations and like BigMack70 said/quoted the few cases you can force a VRAM limitation are obscure cases that deliver unplayable frame rates even when the GPU has adequate RAM.

This is the first analysis error I've come across at Tom's Hardware in 5 years of regular viewing. Tom's analysis is superior to most tech sites but you made a mistake this time. It happens but @Crashman's response is disgraceful. I hope other Tom's hardware editors/contributors are monitoring this because @Crashman's giving your website a bad image imo.

 

[Crashman]

Crash, I think I'm going to side with BigMack on this one, at least to the point of saying that some very specific testing needs to be done to suggest that the results he posted are not relevant. That is not saying he is right, but he does have data that appears to back up his assertions. If it is a matter of the settings used (entirely possible), it would be useful to clarify the settings that cause VRAM to make a difference.

I had an article, I just can't remember what it was. I spent around half an hour searching old articles before I gave up, it just takes way too long. And I intensely distrust the results from certain other sites.

Battlefield 3 is a weird app, I think it's the one where 4800x1080 sometimes has higher FPS than 2560x1600.

 

[Crashman]

you made a mistake this time. It happens but @Crashman's response is disgraceful.

I'm fairly certain I did not make a mistake. If you're going to prove that the comment about VRAM is wrong, you're going to need to do it using this site's data because other sites test things differently. I tried to find the articles, I gave up after wasting over half an hour on it, do you think it's less disgraceful not to try?

Edit: Found one where 4800x900 outperforms 1920x1080. These are actual results.

http://www.tomshardware.com/reviews/crossfire-sli-scaling-bottleneck,3471-4.html

I still can't find the 2GB vs 3GB or 4GB CrossFire or SLI examples but I do know they exist because I produced them on my own machines. This little search alone took ten minutes, so I don't have any other choice but to "crowd source" the search.

 

[Crashman]

Could you help me understand how you are seeing that article as indicating something about vram capacity affecting performance in BF3? I am seeing 2GB cards (680s) performing about the same as 3GB cards (7970s). Since they're different cards, I'm not seeing that VRAM capacity could be offered as a definitive explanation for the performance difference.

Most of what I see there is crossfire doing unexpectedly poorly at 1080p.

I remember specifically encountering an issue with 2GB multi-GPU configurations dropping off at specifically 2560x1600 and Ultra. Some of these games exhibit odd behavior at 2560x1600 compared to ultra-wide resolutions, Don's explained that one to me but the explanation never "stuck". Chris edits this stuff, so if mentioning the 2560x1600 problem didn't get his attention he's probably seen it to.

But there's always a small chance that I'm remembering the wrong GAME. If that turns out to be the case, I'd still like to know which game I'm thinking of. And then I'd still need to apologize to everyone except the guy who used someone else's charts, because that's unforgivable.

Perhaps the annoyance that site "improvements" have slowed my page load times to a crawl is coloring my responses. But then again, linking to some other sites charts probably turns me a different color.

 

[vertexx]

I don't know what you're prattling on about, because as soon as I saw a chart that wasn't mine I zoned out. Could someone delete these two worthless posts (mine and his above)?

No, I'm saying that adding invalid data to the conversation is even worse than adding none at all.
Really, it takes like minutes for me to load a page. I should appoint you to do that.

And I intensely distrust the results from certain other sites.

If you're going to prove that the comment about VRAM is wrong, you're going to need to do it using this site's data because other sites test things differently.

And then I'd still need to apologize to everyone except the guy who used someone else's charts, because that's unforgivable.
But then again, linking to some other sites charts...

Interesting positions taken here.... Is linking to other sites' data some violation of secret forum etiquette? Are all other sites' analysis really invalid or worthless?

I'm relatively new to Tom's, but this sort of feels like something that would come out of EA.

 

[Crashman]

Interesting positions taken here.... Is linking to other sites' data some violation of secret forum etiquette? Are all other sites' analysis really invalid or worthless?

Are you asking me to validate the data from other sites? Do other sites benchmark things in the same way, using the same settings? Do I even know that the people who write for those other sites are trustworthy?

I'll give you a point on one item from your list though, I phrased the "intensely distrust" part wrong. I intensely distrust a few sites, I trust a few other sites for things not reviewed here, and there are several sites that I only half-trust. Since I'm always calling out others for exaggeration, feel free to bury that particular comment in "thumbs downs". Thanks.

 

[Dugimodo]

I built a recent system in this case using a i7 3770 non-K with stock cooler and a 240mm front fan, 120mm rear exhaust, and sapphire 7970 with 3 fans. Nothing gets excessivley hot even during prime 95 and furmark tests at the same time. I'd have liked to see you reverse your intake and outlet fans and compare the results - I suspect it would make less difference than you think. Also the huge 220 and 240mm Fans that you can fit to the front of this case are what really sets it apart in the itx world, take the hdd cages out, use onyl 2.5" drives, and suddenly it seems very roomy and cool.

 

[ananke]

This is a GREAT build!

 

[jbheller]

I have no idea why Tom's Hardware chooses some of the brands they do. Corsiar were the first closed loop water cooling supplier I am aware of and as far as I am concerned still make the best and easiest to install. I have no idea why every company in existence suddenly has to produce closed loop coolers, or why you even bother with making systems with them. If I had to start modifying my cooler with a file or hacksaw I would put the unit back in the box and throw it away. I have never had these issues with the mounting assemblies of Corsair products, they work correctly first time, every time.

 

[clonazepam]

lol @ Crash... your creative writing skills has had me chuckling up a storm.

 

[de5_Roy]

I have no idea why Tom's Hardware chooses some of the brands they do. Corsiar were the first closed loop water cooling supplier I am aware of and as far as I am concerned still make the best and easiest to install. I have no idea why every company in existence suddenly has to produce closed loop coolers, or why you even bother with making systems with them. If I had to start modifying my cooler with a file or hacksaw I would put the unit back in the box and throw it away. I have never had these issues with the mounting assemblies of Corsair products, they work correctly first time, every time.

corsair doesn't make the device. aseted and coolit make the devices. afaik, swiftech has a semi-customiable clc. other than those 3... i dunno.