System Builder Marathon Q1 2015: System Value Compared

[Crashman]

Bigger budgets brought an almost-universal per-builder value decrease, though a few alternative metrics allowed the better machines to top the value curve. Will the trend prevail as we compare all three machines?

System Builder Marathon Q1 2015: System Value Compared : Read more

 

[cknobman]

So my conclusion from this is that SLI is not worth the investment when gaming @1080p even across 3 monitors.

Unless you are going to do 4k gaming the current crop of CPUs combined with the current graphics engines out there are more than capable of 1080p multi monitor gaming at high or even the highest settings.

 

[sumarongi]

I think the performance of your GTX 970 SLI is completely out of line in comparison with the single GTX 970. A quick look around shows most reviewers reported much better scaling than you are showing??

 

[Crashman]

I think the performance of your GTX 970 SLI is completely out of line in comparison with the single GTX 970. A quick look around shows most reviewers reported much better scaling than you are showing??

Pandering perhaps? A reviewer can use a limited range of high settings to find up to 80 or 85% performance gains from the second card. You'll see SOME of those results in this article.

On the other hand, maybe we're pandering to the low-end system by not removing the low-end settings. Look at BF4's 200 FPS cap, and Grid2's DRAM performance bottleneck, both at lower-than-optimal settings for the GTX 970.

Or maybe we just balance out the pandering-to-the-high and the pandering-to-the low. But if it's balanced, we're no longer pandering. Anyway, these details explain the difference between what you've seen there and what you're seeing here.

Edit: Or maybe pandering to the top and bottom cancel each other out to boost the middle. Don's $1300 PC wins!

 

[sumarongi]

I think the performance of your GTX 970 SLI is completely out of line in comparison with the single GTX 970. A quick look around shows most reviewers reported much better scaling than you are showing??

Pandering perhaps? A reviewer can use a limited range of high settings to find up to 80 or 85% performance gains from the second card. You'll see SOME of those results in this article.

On the other hand, maybe we're pandering to the low-end system by not removing the low-end settings. Look at BF4's 200 FPS cap, and Grid2's DRAM performance bottleneck, both at lower-than-optimal settings for the GTX 970.

Or maybe we just balance out the pandering-to-the-high and the pandering-to-the low. But if it's balanced, we're no longer pandering. Anyway, these details explain the difference between what you've seen there and what you're seeing here.

Edit: Or maybe pandering to the top and bottom cancel each other out to boost the middle. Don's $1300 PC wins!

Fair enough, I understand what you are saying. The rating system you are using creates it's own viewpoint on the world that you need to remain consistent within. Maybe value needs to be addressed from the point of view of each individual system and what a real end-user that would buy that system would expect and experience. No one runs SLI GTX 970 on 1600 x 900 resolution so the fact that performance is not great at that resolution has no value at all - negative or positive - and those data points just distort things when it comes down to comparisons.....

 

[Crashman]

Fair enough, I understand what you are saying. The rating system you are using creates it's own viewpoint on the world that you need to remain consistent within. Maybe value needs to be addressed from the point of view of each individual system and what a real end-user that would buy that system would expect and experience. No one runs SLI GTX 970 on 1600 x 900 resolution so the fact that performance is not great at that resolution has no value at all - negative or positive - and those data points just distort things when it comes down to comparisons.....

We have some of that in the final chart, 5760x1080 gaming. I could further refine it to include only 5760x1080 at the highest quality test settings, but it already makes that point.

Perhaps another value chart at 1920x1080 for the low-end system would help, and another for non-gaming if I'd used an X99 system with a cheaper graphics card. I'll toss ideas like that back to the other guys and see what they think.

 

[Onus]

A system has value in the context for which it was built. Even if I were to put a tuner card into it, I'm not likely to judge the value of my PC on its ability to receive broadcast TV, or my TV on its ability to add up numbers. "Is it fit for purpose," though limited, is a fair question, especially on a tight budget.

 

[SinxarKnights]

Whelp that explains the flying pigs I seen today, Paul lost.

 

[Crashman]

Whelp that explains the flying pigs I seen today, Paul lost.

Yes, we can't have any of that. TIME TO REPLACE DON!!!

 

[Kulamata]

You're tying yourself up in knots. Instead of setting an arbitrary limit for each category, I'd suggest designing a build for each category, and then listing the costs, and then, from that, suggesting changes to meet various cost thresholds, with a discussion of why those changes deviate from the "ideal".

This would prevent some of the significant limitations that can occur for want of a nail, or a few dollars. I'd regard a machine with only an SSD, and a smallish one at that, to be crippled, and to not kick the system cost from ~$1750 to ~$1800 to provide a hard drive makes the system a victim of misplaced priorities.

That's bean-countin' son!

 

[OhSnapWord]

Personally, I'm getting tired of seeing all Intel in the SBMs. I would like to see all AMD for an SBM as it would help give a real nice comparison between the two platforms.

 

[chlamchowder]

I'd like to see more AMD too, but right now AMD has nothing against Intel's high end CPUs especially when it comes to single threaded performance. And single threaded performance matters for a lot of the benchmarks here.

Their graphics cards can be good value (there's a R9 280 in the $700 PC), but at the high end Nvidia wins on efficiency, performance, or both.

I really hope AMD can be more competitive in the coming years.

 

[de5_Roy]

I think the performance of your GTX 970 SLI is completely out of line in comparison with the single GTX 970. A quick look around shows most reviewers reported much better scaling than you are showing??

my two cents: reviewers tend to use open case test benches with high end z97 to x99 motherboards and high-o.c.'ed core i7 cpus with high end cooling systems and high end memory kits to remove any kind of possible platform bottlenecks. in a real pc with limited budget, (often time) and options, the results may be different.

 

[Gurg]

I think the dollar threshold is missing the boat and should be replaced by the systems needed to play the majority of games at specified resolutions/screen size at ultra levels. ex 1080 on 24" monitor, 1440 or 1600 on 28" and high end 4k. Let people know the minimum system entree and cost needed to build a system that performs at the most common visuals. As for the lower price level, personally, I would probably just buy a $699 gaming laptop, a $30 laptop cooler and a mouse.

http://www.newegg.com/Product/Product.aspx?Item=N82E16834231974

 

[hasten]

I always love these system builder articles, fun stuff.

For the temperatures, on the O/C $1750 build you are showing 79c over ambient. Is that correct? That seems extremely high, if not reaching the tjmax... is this all done with the stock intel cooler?

 

[Crashman]

I always love these system builder articles, fun stuff.

For the temperatures, on the O/C $1750 build you are showing 79c over ambient. Is that correct? That seems extremely high, if not reaching the tjmax... is this all done with the stock intel cooler?

It is extremely high. I test with both the CPU and GPU maxed, using an AVX-enabled multi-threaded Prime95 and 3DMark 2011 test 1 simultaneously. I also use RealTemp, which shows higher readings than most motherboard software. This CPU just runs hot. TJMax just means thermal throttling, not crashing or destruction, and thermal throttling was the only thing the kept me from using a higher O/C.

Of course I could have set it 100 MHz higher with a little more voltage and just let it thermal throttle. Most apps would have run faster, some would have performed the same due to thermal throttling. But the thing is, I prefer to tell people my clocks with some degree of accuracy.

The lab was 18 °C. Whoever wins this is STILL going to have an issue with the thing clocking down by 100 to 400 MHz, depending on how much warmer their room is. This really can't be helped.

 

[hasten]

I always love these system builder articles, fun stuff.

For the temperatures, on the O/C $1750 build you are showing 79c over ambient. Is that correct? That seems extremely high, if not reaching the tjmax... is this all done with the stock intel cooler?

It is extremely high. I test with both the CPU and GPU maxed, using an AVX-enabled multi-threaded Prime95 and 3DMark 2011 test 1 simultaneously. I also use RealTemp, which shows higher readings than most motherboard software. This CPU just runs hot. TJMax just means thermal throttling, not crashing or destruction, and thermal throttling was the only thing the kept me from using a higher O/C.

Of course I could have set it 100 MHz higher with a little more voltage and just let it thermal throttle. Most apps would have run faster, some would have performed the same due to thermal throttling. But the thing is, I prefer to tell people my clocks with some degree of accuracy.

The lab was 18 °C. Whoever wins this is STILL going to have an issue with the thing clocking down by 100 to 400 MHz, depending on how much warmer their room is. This really can't be helped.

I apologize, if I had used my head last night and referred back to the original article I would have known you used the h100i. In a similar setup with a 4770k the temps were pretty outrageous after 4.2-4.3 and 1.25v or so. I never had the guff to push it as high as you pushed the 4790k. Been trying my luck with a 5820k the last week and after the results I experienced with the 4770k was amazed at high much better thermals Haswell-e has given us (or at least me). 1.35v and 4.65ghz wasn't barely breaking a sweat, but I seemed to have hit a wall @ 4.5ghz (124 bsod) and have reverted back to 4.5ghz with lower vcore for the time being. I'll need to do some more research on that in the coming weeks.

Anyways your build looks great. One thing I have always appreciated is that you go against the "you need 1000w psu forum experts" and use a power supply with sufficient specs for your builds. Would love to give this guy a new home!

 

[Onus]

I think the dollar threshold is missing the boat and should be replaced by the systems needed to play the majority of games at specified resolutions/screen size at ultra levels...

I strongly disagree, primarily because the high-priced system is not, imho, intended to be a gaming PC; it is a professional's PC that is also able to play games [well]. One might suggest the same could be true of the enthusiast's PC, which might have some hobby (media collection/manipulation, folding, maintaining a blog or some other publishing activity) as its primary purpose.
Doing that would only make sense if the entire focus of the SBM series was placed on gaming, which I think would eventually become tiresome. I believe it is other measures than FPS benchmarks which should distinguish these machines from similar builds.

 

[Guest]

Who won the systems?

 

[Crashman]

Who won the systems?

Budget Gaming PC
Christopher Hawkins
Irving, TX

Mainstream Enthusiast PC
Calvin Bowens
Parma, OH

$1750 Performance PC
Nathan Somers
Huntertown, IN

Alternative $1750 PC
Christopher Landers
Syracuse, NY