System Builder Marathon Q3 2015: Prosumer PC

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.



They'll remind you, once more, that the SBMs are built/sponsored from Newegg. A smart shopper can definitely do better, as I've mentioned just by hitting up my local computer superstore I could pick up an i7 for roughly $70 cheaper which by itself is enough to net a decent graphics upgrade to something like a GTX960 or R9 380. not really a fair comparison if you're intent on scouting eBay listings. :)
 
I'm going to leave that one in there, if only for the remarks about the GTX750Ti and weak compute. While marginally off-topic, it's good information, much more so than someone simply derping about a "better build" that doesn't follow the rules.
Now that the point(s) are noted though, let's stick to the SBM context. Thanks.
 
You probably shouldn't have gone with Skylake, IMO. Haswell i5's sell for less, plus you can get a cheaper motherboard and cheaper DDR3. If you also cut back a bit on the case and the PSU, you could have gotten a locked Haswell i7. Either way, I'm happy that Skylake is fit in here.
 
I guess my critique of the build is its potential doesn't match the title, that of being a 'prosumer' setup. Consumer SATA, a weak compute card, no backup facility wrt storage, not enough RAM, no SSD, no room to grow re low PSU capacity, etc. Really, for a decent system that one could describe as prosumer (as opposed to something which just gives nice looking benchmark numbers), the budget either needs to be higher for a new build, or look elsewhere, or some combination of both.

Let me put it another way: if normal seller stores still sold older tech like SB and 580s for correspondingly low prices akin to typical used valuations, would one really build a SkyLake setup as described when for the same money one could put together something a lot more potent? The old top-end becoms the new mid-range, then the low-end, etc. It's why a 650 Ti can't even beat an old GTX 460.

In a way this shows how the sliding scale of ageing tech is somewhat at odds with the way tech advances while at the same time older products slip away from availability. SkyLake makes this worse by being kinda expensive for the top-end CPUs. I like the updated chipset, but Intel needs a wakeup call with its pricing given the minimal performance improvements.

Meanwhile, it's just sad that NVIDIA has sort of given up on the pseudo strong side of gamer cards for compute, eg. I was surprised to find that even a good 980 couldn't beat two 580s for CUDA (Arion benchmark).

To me it seems like various aspects of PC tech have sort of stalled. Most sites didn't really emphasise the weak speedup over HW, focusing instead on chipset improvements and the bclk change (understandable, but not the key factor for some, especially anyone who might be in the 'prosumer' space). We really need some competition in this field asap.
 

Proof of the pudding is in the eating: IIRC there were two apps out of a dozen that suffered from the weak GPU Compute and one that suffered from the low RAM. If I could build this any way I wanted, I'd first 1.) Want to satisfy the curiosity of readers with an i5-6600 even if it is overpriced, 2.) Buy an SSD 3.) Double the RAM 4.)Upgrade the graphics and 5.) get a decent case. And then it would be a $1040 PC that I wouldn't have to make any price-based excuses for.

The power supply would have stayed, except now that it's more expensive it might not. You can add a hell of a graphics upgrade for the 200W remaining.
 
Alternate builds that do not follow the SBM rules, such as "all Newegg" (as they're the sponsor) are being deleted as off-topic.
I too am very happy to see measures other than gaming. As my own gaming time has diminished, how well a PC plays games simply diminishes in importance. For this reason, I do like builds that are more weighted to CPU power than many gamers.
This "prosumer" PC is only missing a SSD and maybe a second drive to RAID1 the data, to be pretty much what I'd build myself as a nice daily driver.
What would the Compute numbers look like with a R7 260X or R7 265?
 
i built a 4690k rig last year and I'm still on my Radeon 6950 2 gig for that very reason...well that and usually I'll try to nail down a "nearly" top tier card for either that grey area just under $300 and/or just under $400. there's a whole lot of legacy and overclocked stuff kinda in that range but what i'd need for me to pull the trigger is say...a 'lite' version of the Fury for under $400, or maybe a 390X rebate pushing under $300. both pipedreams i know, at least until they stop making them
 
Buy one. Run it through these benchmarks. Call me when it comes up short.

 


Pretty awful for AE, that's for sure. 😉

This is one problem with some of the benchmarks, certain apps are often hw-specific in what makes the most sense to buy for running them. In the case of AE, a CUDA capable card is essential, right now the most potent & cost effective being a 780 Ti or original Titan (Maxwell V2 not yet supported, so the 900s are useless), followed by multiple 580s if the budget is tight. In such cases, the build spec only really conveys how good or bad the specific selection of parts is, as opposed to what's really possible within the stated budget, where or not that is spent entirely on new parts. Other apps may be better served with a strong OpenCL card, in which case there are all sorts of used & new AMD options. CPU/mbd-wise though, despite the moans of slow gains in recent years, atm Intel is still king, for the time being.

Ian.

 
One Issue: The CPU. Pretty sure a Xeon E3 1231 v3 would have cost about the same, and then you could have got a cheaper Motherboard and a better GPU.

Other than that, Nice Job!
 


Yeh, I wasn't sure how much of an O/C I'd get on BCLK alone.
 


Yeh, I wasn't sure how much of an O/C I'd get on BCLK alone.

While true, wouldn't a Stock clocked 1231v3 beat an overclocked i5 anyway (productivity wise)?

Overclockability (pretty sure that's not a word) isn't necessary in SBM though is it?
 


I'd really expect a 5% overall performance boost in just the timed applications (10% or more in some, none in others) due to its hyperthreading and greater cache. A 10% overclock on the 6600k helps both timed benchmarks and games. It might be a good experiment to try both, since I think I can get a 5% overclock out of the 1231v3 off of BCLK. That is, unless BCLK is locked.
 
Good article, Thomas. These are exactly the type of machines I build for myself and many of my clients; professionally oriented that can get some game on. In that respect, however, I question your choice of core components entirely.

Skylake was definitely the wrong direction to go for a "prosumer" machine. For less than the cost of the 6600k, you could have had a four-core/eight-thread E3-1231v3 @ 3.4-3.8 GHz. Yes it's locked, so no overclocking. But much of the "professional" level software will respond better to more threads than higher frequency. Games don't really care, especially at the 1080/1200p resolution we're talking here. This would also allow you to save money on both the memory and motherboard, allowing you a bump to a GTX 950 and/or an SSD (depending on how cheap you go with the motherboard). This would make a seriously nice "home workstation" for many who also like to game; all at the $800 price-point. It would have fared much better against your competition in this quarter's marathon as well.

Just my 2¢...
 


Well, if you have specialized software:

- You are not on the general case.
- I advice to copy the installers on any optical disk to a hard disk, to protect the installation DVD from scratch and ultraviolet light. If you can read the license without unpacking it, do not even unpack it, search for a downloadable source, and use your license. Most software has trial versions which can be activated with the purchased license.

By the way, because they are so commoditized and cheap, most DVD units are disposable. Even the most expensive ones. They break after some time, even if you don not use them, so you may spend the money, and after some months you try to use it, and is broken, the you need to buy a new one anyway.
 
Somehow I find looking at this kind of article about the same as looking at pre-built PC's at Best Buy or whatever. Yeah, it has all the parts, and it runs, but there are always a bunch of things in there that I'd never put in my own build because it's either unnecessary, overpriced, inadequate, or just not the right part. An ODD is something that some people need, but most people don't. A 750 Ti will be adequate for some, but many would rather find a way to afford the 950 (or better). From 2015 on I see no reason to build a system with a HDD instead of an SSD. Everyone can benefit from the added performance of an SSD, but not everyone needs more than 250 GB of storage. If they do, it's much easier to just add a HDD later on than it is to move the OS over to an SSD.

I realize some of these points are mentioned in the article too, but this is the sense I get when I look at a build like this trying to work on a specific budget. When the budget is that tight, it becomes all the more important to get exactly the right parts to suit your specific needs, so when there is no specific target customer, it's almost inevitable to wind up with a system that just doesn't quite satisfy.

Aside from that, trying to build a system like this would (for me) be a project of collecting the parts over a month or two as they go on sale for a great price. I definitely wouldn't be paying $130 for a 750 Ti, or $70 for a 1TB HDD if my budget was that tight.
 

Agreed. This is why "build for a purpose" is a key mantra of experienced system-builders, especially on a limited budget. There are specific comments in these build articles that suggest that focus may have been softened because these machines will be given away, and the winners will likely have different purposes. I've been satisfied with the explanations, including why they all use an ODD.
I like to see what can be done with tight budgets. This was the most useful SBM series in a while; I wonder if taking another $100 out of the budget would yield similarly interesting results.
 
Status
Not open for further replies.