System Builder Marathon Q4 2015: System Value Compared

[Crashman]

Three builders with practically unlimited budgets compete for best performance-per-dollar across our multi-faceted benchmark suite. Which build will win?

System Builder Marathon Q4 2015: System Value Compared : Read more

 

[Onus]

I like all three of these machines. Any changes I'd make would be niggles, and mostly a matter of scale. I'd be happy to win any of them, and would probably do some minor mixing and matching with what I already have.
For all of them, I'd upgrade the SSD to 500GB just because I can.
I'd pull the power-scarfing R9 390 out of the gaming PC, put in my GTX970, and probably add another drive for a RAID1 pair. I'm not anti-AMD, but another 200W+ under load is a horrific waste. I really hope AMD can get back in the low-power competition; the HD4670, HD5670, HD6670, and HD7750 were all amazing cards for how little power they used.
The prosumer machine would get a RAID1 pair of drives added.
The LAN box is actually a tough call. I could just add one more drive to it also, but I'd be sorely tempted to put a H170 mITX board and Skylake i3 in it instead, if only because those are available to me. I would not need to though; it's an excellent machine as it is.

 

[Crashman]

I like all three of these machines.

Maybe that's why the individual builds are still generating feedback and the 3-way comparison isn't

 

[RedJaron]

The Windows I bought for my build isn't download only. It's on a DVD with the rest of my parts. I threw in the USB drive so the winner wouldn't have to install Windows over a USB 2.0 ODD.

That said, I really thought I had a shot at the value win when I saw you two spending $200 more than me. Skylake's better memory bandwidth makes a big difference. We may need to re-visit memory bandwidth in gaming because I'm pretty sure I saw mine get affected with the machines I built this time.

But I shot-myself in the foot by limiting what I could build last quarter. There's only so many ways you can take a LAN box. A general purpose PC would've been wide open for value oriented changes and improvements. If I get a shot next quarter, I may well game the system to make something designed to win value. That actually could be an interesting experiment. Would a machine purpose built for value alone still provide enough pure performance to satisfy an end user?

 

[Crashman]

The Windows I bought for my build isn't download only. It's on a DVD with the rest of my parts. I threw in the USB drive so the winner wouldn't have to install Windows over a USB 2.0 ODD.

That said, I really thought I had a shot at the value win when I saw you two spending $200 more than me. Skylake's better memory bandwidth makes a big difference. We may need to re-visit memory bandwidth in gaming because I'm pretty sure I saw mine get affected with the machines I built this time.

But I shot-myself in the foot by limiting what I could build last quarter. There's only so many ways you can take a LAN box. A general purpose PC would've been wide open for value oriented changes and improvements. If I get a shot next quarter, I may well game the system to make something designed to win value. That actually could be an interesting experiment. Would a machine purpose built for value alone still provide enough pure performance to satisfy an end user?

Dude, isn't it enough to tie me at stock speed? I mean, I spent the OC money to OC right? I'd better get some advantage from it!

Sorry about the DVD thing, Newegg isn't actually telling me that the Win10 package you bought comes with media.

 

[RedJaron]

Well, no, like you said, part of your cost was the ability to overclock. If you didn't use that, it's like paying for a dGPU but not installing it. You were right, that is a fair comparison. I lost.

 

[Onus]

It's great to get a high bang/buck, but you still need some minimum amount of "bang" to be useful. Those with 144Hz monitors will disagree, but for mainstream 1080p gaming, counting FPS over 60 as no better than 60 might hand that value win to a GTX950.

 

[Crashman]

Well, no, like you said, part of your cost was the ability to overclock. If you didn't use that, it's like paying for a dGPU but not installing it. You were right, that is a fair comparison. I lost.

Yes, and your climb rate means next time you'll probably win. But I have a plan:

I CAN'T win if I spend big, so how about next time we give me a big somewhat-specific budget where I'm required to spend most of the money, and likewise an excuse not to win? Then I'd basically be putting all my O/C effort in to reach 80-90 percent of your score :-D

These head-to-heads are brutal, I don't want anyone to lose the value competition unless they have an excuse such as "chasing better quality and more user-friendly features."

 

[Onus]

Make it more brutal; cut the budgets down to $550-$600.

 

[Crashman]

Make it more brutal; cut the budgets down to $550-$600.

Remember Paul? We figured out that you can find perfect performance-value around $600, and that the closer you got to $800 by trying to cram in performance parts, the more the rest of the system held you back. That you had to keep upgrading the rest of the system until you reached balance, before you'd get the value score back up at a higher performance level. And that the second value zenith was around $1100.

People want to see $800 builds, but we knew long ago that everything between $650 and $1050 was a "dead zone".

Did this experience with Paul give me an unfair advantage? It depends on what you consider fair. I would have used this CPU anyway, and a $250 to $320 graphics card, so my path was set without consideration of the final price. Fortunately, reader feedback on the Q3 SBM meant that I didn't need to pick the graphics card, they did it for me.

 

[RedJaron]

Well, if we're all at that magical $600 range, then it could be interesting to see who can find the best balance. Granted, at that price range, are we each going to end up with near identical builds due to us chasing the same sales? I wouldn't mind going after something a little more simple.

 

[Crashman]

Well, if we're all at that magical $600 range, then it could be interesting to see who can find the best balance. Granted, at that price range, are we each going to end up with near identical builds due to us chasing the same sales? I wouldn't mind going after something a little more simple.

TBH, I find anything less than a $1200 PC repulsive I mean, if I added the HDD and upgraded the case, my $1055 machine would be there.

 

[iam2thecrowe]

This was an interesting article, good job, a bit different for a change. Something I'd like to see in the future and some common questions that are always in the forums:
1. "I have $xxxx budget, is this a good build?". Everyone knows an i5 + gtx970 or r9 390 is the bread and butter combo. Maybe try a budget above and below, and try some different things, like socket 2011 and overclocked i3......
I'd also like to see
2. "What is the lowest possible budget to get something that is capable of gaming?" Set a bottom target of say 30fps medium details, 1080p res, (if your not going to get at least console level of detail then why bother right?) and try do it for as cheap as possible. 3 builds, one has to have an AMD cpu and no dual core (without HT) cpu's as I don't believe they are really a good gaming cpu, especially with large multiplayer maps which don't get bench-marked.
3. Premium builds, 144hz max fps target and play-ability @ 1440p and 4k targets.

 

[Onus]

I'd like to see #2 done myself, perhaps going for 40FPS average though (30FPS minimum would be good too).
#1 is a little like past SBM cycles, but there were often lessons there too.
#3 strikes me as mental masturbation; reminiscent of some of the troll threads we shut down from time to time.

 

[iam2thecrowe]

I'd like to see #2 done myself, perhaps going for 40FPS average though (30FPS minimum would be good too).
#1 is a little like past SBM cycles, but there were often lessons there too.
#3 strikes me as mental masturbation; reminiscent of some of the troll threads we shut down from time to time.

Yes, I was thinking a 30fps minimum target, not an average. 40-45fps average is probably a decent target at medium detail settings. Sure you may get random dips in some games down to 20-25fps, but so long as its not like this for extended periods. You could take that 144hz idea, maybe turn it into "competitive gaming build", and forgo the 1440/4k builds, it's something that has never really been looked into before. But yeah its a narrower target audience. Just trying to think outside the box a little, put some more fresh ideas out there.

 

[RedJaron]

If we're doing a gaming centered theme, forcing one builder to use an AMD CPU puts them at a handicap. Arma likes Intel's uArch too much and Grid wants memory bandwidth, and AMD is very behind in that regard. I wouldn't mind doing a barrel bottom gaming theme. So long as I don't need to build ITX for a while, I'm up for just about anything. I wonder if people would be interested to see something maybe more on the professional side. Eight threads and 16GB RAM minimum to go for tops in the real-world app benches and not worry about games so much ( or maybe completely drop them altogether for a quarter ).

 

[cknobman]

Count me in for wanting to see a shootout of iam2thecrowe #2 option.
That is something I bet a lot of people would like to see.
I just had a discussion with my little sister about the same topic as she is trying to build a 1080p gaming rig for under $800.

 

[iam2thecrowe]

If we're doing a gaming centered theme, forcing one builder to use an AMD CPU puts them at a handicap. Arma likes Intel's uArch too much and Grid wants memory bandwidth, and AMD is very behind in that regard. I wouldn't mind doing a barrel bottom gaming theme. So long as I don't need to build ITX for a while, I'm up for just about anything. I wonder if people would be interested to see something maybe more on the professional side. Eight threads and 16GB RAM minimum to go for tops in the real-world app benches and not worry about games so much ( or maybe completely drop them altogether for a quarter ).

That's a good idea. A "Tools of the trade" build. Maybe purpose build rigs, one for video editing/rendering, 3d rendering, CAD, Adobe suite etc. Maybe add a Mac into the testing as that is one of staples of people in these industries.

 

[RedJaron]

I just had a discussion with my little sister about the same topic as she is trying to build a 1080p gaming rig for under $800.

You can easily do a 1440p rig for $800, if you're only worried about gaming. A reasonable 1080p machine can be done for $600 as long as med-high details are okay. $700 will get you ultra details.

That's a good idea. A "Tools of the trade" build. Maybe purpose build rigs, one for video editing/rendering, 3d rendering, CAD, Adobe suite etc. Maybe add a Mac into the testing as that is one of staples of people in these industries.

A Mac runs off mostly the same hardware as a Windows machine, but the software and OS are different, which renders direct bench comparisons dubious. Most people who work on Macs today do so out of tradition or company policy, not because it's the only platform available for their workload.

 

[Onus]

A SBM cycle that forgoes gaming would be interesting. You might toss in a gaming benchmark or two just as a viability test,, but not have it affect the final score in any way.

 

[cmiconi]

At this point, I think that chasing the value benchmark every quarter is getting at bit monotonous. By now, I think we’ve pretty much demonstrated where the sweet spot is for getting the best performance-per-dollar value for a given build theme. The only time that might change significantly is when new hardware comes out.

I’d like to see more of a shift towards more experimental builds, like a true workstation computer with a Xeon CPU, Quadro GPU, and some sort of RAID setup. It would be nice to see how that set of hardware performs in our benchmarks compared to the other builds.

Thoughts?

 

[Onus]

Experiments are good; that's where new "data points" are generated and lessons learned.

 

[RedJaron]

I'd like to think my last few builds have been rather experimental. The response to them has been all over the place.

 

[cmiconi]

I'd like to think my last few builds have been rather experimental. The response to them has been all over the place.

Indeed, and they've also provided some very useful data.

 

[Onus]

Yes, and I think Munchkin 3.0 settles the APU vs. i3 debate until AMD issues something new.