System Builder Marathon, Q2 2013: System Value Compared

[slomo4sho]

i think directly dividing performance by cost makes it easy to misrepresent if a baseline isn't present. sbm uses cheapest machine as a baseline. kinda like measuring without a scale... i think. even a universal measure will need a baseline and even that measurement will be useless outside the scope of the measurement. i hope that made some sense.

Is Zero not an acceptable baseline?

 

[de5_Roy]

Is Zero not an acceptable baseline?

from what i've seen, perf/dollar comparisons are always relative (to something), not absolute. the moment you factor conditions, preferences, etc. the baseline forms. so, zero might not be a baseline.

 

[slomo4sho]

from what i've seen, perf/dollar comparisons are always relative (to something), not absolute. the moment you factor conditions, preferences, etc. the baseline forms. so, zero might not be a baseline.

Not always.

http://www.pcper.com/reviews/Processors/Intel-Core-i7-3930K-Sandy-Bridge-E-Processor-Review/Performance-Dollar

 

[de5_Roy]

Not always.

in this case,
only the gfx card perf/price is being measured.
the baseline is the rest of the pc.

if you take the gfx card price and performance/price, e.g. the gtx680 perf/price figure, by itself, doesn't mean much. the reason gtx670, 7950, 7870 and 7850 perf/price are highlighted because they are being compared to the others, even thought they've been calculated individually. otherwise, the radeon 6870 looks like the best perf/price card (compared to others i.e. relative).
there's always a baseline.

 

[Onus]

^I'm thinking there was a minimum performance requirement, which the HD6870 didn't meet. A high bang/buck ratio is only meaningful if you're getting enough bang.

 

[slomo4sho]

in this case,
only the gfx card perf/price is being measured.
the baseline is the rest of the pc.

if you take the gfx card price and performance/price, e.g. the gtx680 perf/price figure, by itself, doesn't mean much. the reason gtx670, 7950, 7870 and 7850 perf/price are highlighted because they are being compared to the others, even thought they've been calculated individually. otherwise, the radeon 6870 looks like the best perf/price card (compared to others i.e. relative).
there's always a baseline.

I am aware that only one variable is compared in this chart. There are no existing performance/price charts for complete systems currently. Also, the article from which this chart is linked actually examines current high end cards and favors the GTX 670.

^I'm thinking there was a minimum performance requirement, which the HD6870 didn't meet. A high bang/buck ratio is only meaningful if you're getting enough bang.

The author had a clear bias towards the GTX 670 and highlighted criteria to favor the 670.

 

[Crashman]

Sigh. This is why I really, really dislike the system builder marathons; they do nothing but perpetuate the fallacies that already are far too common.

Someone looking at just this article, which isn't that unlikely, would be lead to believe that an i7 is something that an "ultimate" gaming computer has, that an expensive motherboard helps, and that a $2500 PC is going to be far better than a $1500 one.

They really should include performance per dollar figures in this writeup.

They really should include performance per dollar figures in this writeup.

For the parts, or for the computers themselves? Either would be nice, actually.

One thing that would go a long way is stressing how wonky their testing is - most people reading this as advice for building a computer are going to be building a gaming computer purely, rendering 70% of the test bench pointless.

i think directly dividing performance by cost makes it easy to misrepresent if a baseline isn't present. sbm uses cheapest machine as a baseline. kinda like measuring without a scale... i think. even a universal measure will need a baseline and even that measurement will be useless outside the scope of the measurement. i hope that made some sense.

Is Zero not an acceptable baseline?

One can't divide by zero. $ Per Frame might work but then what would you do with the other benchmarks? $ Per Second Saved? But then you need a baseline. Most of these benchmarks don't break down into units per second so easily.

 

[lt_dan_zsu]

First off, I thought this was about trying to build the best valued system in the price range, which they didn't do.
Second, I don't know why they put a 680 in this build. I would have gone for a vapor-x 7970, which you can currently get for $370. Even if you have a vendetta against AMD for some reason, a 770 is a better choice. It's stronger and cheaper. You could argue no free games with the 770, but then the opposite argument needs to be made for the 7970 which comes with 4 free games.
Third, and this has always been a pet peeve of mine about the System builder, the prices for the builds need to be consistent between every quarter. You can't say an enthusiast PC is $1000 one quarter and then say it's $2500 the next. It makes the two build almost impossible to compare. I would say $600, $1000, and $1500 would be a good price point for them.
I use these build as templates when putting together parts lists sometimes because they usually deliver such great value, but definitely not this quarter.

 

[Crashman]

First off, I thought this was about trying to build the best valued system in the price range, which they didn't do.
Second, I don't know why they put a 680 in this build. I would have gone for a vapor-x 7970, which you can currently get for $370. Even if you have a vendetta against AMD for some reason, a 770 is a better choice. It's stronger and cheaper. You could argue no free games with the 770, but then the opposite argument needs to be made for the 7970 which comes with 4 free games.

First, the builder was afraid that the internally-vented 7970 would put too much heat in his little tiny case. Second, the 770 wasn't even on the market when these systems were ordered nearly THREE MONTHS ago. Third, if you'd read the $1300 PC article, you would have known these things. Thanks for reading the roundup though!

 

[slomo4sho]

One can't divide by zero. $ Per Frame might work but then what would you do with the other benchmarks? $ Per Second Saved? But then you need a baseline. Most of these benchmarks don't break down into units per second so easily.

Do you need to divide by zero to understand the value of any real number in respect to Zero(yes, I know you can't divide by zero).

Also consider the inverse of the $/Frames: Frames/$. Can the frames/$ formula not be carried over to every other benchmark? When a lower quantity is preferred in a benchmark, you can inverse the results.

 

[Crashman]

One can't divide by zero. $ Per Frame might work but then what would you do with the other benchmarks? $ Per Second Saved? But then you need a baseline. Most of these benchmarks don't break down into units per second so easily.

Do you need to divide by zero to understand the value of any real number in respect to Zero(yes, I know you can't divide by zero).

Also consider the inverse of the $/Frames: Frames/$. Can the frames/$ formula not be carried over to every other benchmark? When a lower quantity is preferred in a benchmark, you can inverse the results.

Right, the problem is that in dollars per second, more seconds is a good thing. To invert it, one must have a "worst" value that's higher than the expected value (ie, greater than zero). Unless you can think of something else...

 

[-Fran-]

I keep arguing the case of the Optical drives.

Today you can get pretty much any ISO you want for almost anything, included Windows. If you really need one, you can get portable ones which are close in price, but a thousand times more convenient.

Cheers!

 

[beil]

This will be nice to see in future builder marathons results for minimum FPS - not just for average FPS.

 

[jojodadude]

My only comment on this article is that I recently built an HTPC using the Lian Li case with a similar build as shown. Just want to give people a heads up that the case was pretty awkward and very difficult to work in. For instance, you have to take the panels off every single side to get into it to install the components for even a basic build. Its a lot of very tiny black screws. Go for the CM case if you can make it work.