Sign in with
Sign up | Sign in
Your question
Solved

"Performance Priorities" or "Where to Put the Money" :

Last response: in Systems
Share
February 26, 2010 4:41:09 PM

Yawn .... Preamble:

Quality, value builds are a moving target and personal styles and philosophies play no small part. Certain components and elements (architechtural choices) are desirable for all builds (i.e. an adequate Power Supply Unit (PSU) ). But, beyond the fundamental infrastructure which is common to ALL builds, personal usage and proclivities and budgetas well as anyr special requirements will surely influence the nature of the finalized build. We attempt to "measure" our computing experience with benchmarks, and by comparing specs. Some limits on budget are almost always a factor. Emerging trends (such as the SSD) can further alter lagacy formulae for cost/performance (according to individual "philosophy").

When it comes to academic debates over storage and graphics subsystems, for a given price and purpose, there should still be some underlying "most correct" logic. Of course, the customer's (owner's) priorities and circumstances ($80K/yr Designer needs $12k WrkStstn) ultimately should govern where the money goes.

So where and what is that central core logic? Is it some perfect truth, which the community uncovers and collectively approves? Well, yes except that the buyer/owner has ultimate veto, as it should be, but I still believe that old trends and persistent misconceptions(?) fail to provide the best overall throughput (and "per monetary unit") cost/performance ... or ... "value" ... or ... "computing/user experience".

While benchmarks are invaluable in bolstering a debate (premise), it is real world (practical and specific) throughput and the experience and satisfaction of the owner and any users.

BOTTLENECKS !!! In a given commercial application (for instance), even an inadequate printer or routing can stifle efficiency and profits. On the flip side, we are all sensitive to overkill, given that times are hard, PCs are expensive, and they depreciate very quickly toward obsolescence.

The identification of (and solutions to) BOTTLENECKS is fundamental and key to any generic "core build logic" if such a thing exists.
Further, understanding and balancing the true scale and effect of one bottleneck (i.e. speed of memory) versus another (speed of storage), is crucial if any sort of OPTIMIZED THROUGHPUT IS TO BE ACHEIVED (THROU A BETTER BUILD).

How, then, should we (the community with buyer's) evaluate and prioritize these, more general and central, bottlenecks and ... How should we apply our monies for the ultimate efficiency and satisfaction of experience?

So, in future posts, on this thread, I will make a case for my own views (thus far) on where monies should be "focused", in terms of general performance.

Gamers, tho their priorities (manly gpu cards) are different, are so numerous that they deserve special consideration but, within even their domain, some core arguments (between build consultants) persist (as in the "one premium gpu or two mainstream gpus" debate). On less tricky issues, such as PSUs, there is less disagreement.

I intend to make the mathematical/financial case that storage is "far and away", THE LARGEST GENERAL BOTTLENECK of all the core components (unless print-jobs are your bread and butter). I will also argue, with benchmarks, costs, and logic, that "value" (on a per component basis) does not trump performance, when it comes to disproportionate bottlenecks and, Babies, if there is only ONE MAJOR BOTTLENECK in terms of general architecture (and reliability too) it is the storage subsystem !!!

I would also like to discuss the whole "more memory" vs, "faster memory" and also try to determine how much perceptable and cumulative advantages might be gained by various memory (and MMU) architechtures ... From DDR2x1CH w/off-die MMU to DDR#+ x3CH w/on-die MMU.

How much does "Socket Pin Count" effect system performance and sustained (in years) system longevity ?

And there is that whole gamer's graphics thing, which we just might be able to settle, if we can corral all the proper benchmarks ... we might be able to answer some of those functions as well as . . .

. . . WHAT EFFECT WILL PCIe3.0 HAVE ON GAMERS GRAPHICS ?? (in terms of SLI/xFire) ?

I'm a little fried, for now ... I'll pick this train of thought up (with or without you) at some later date. If I get around to it.

All's fair in love and war so, by all means, flame away!

= Alvin =




February 26, 2010 5:58:30 PM

There are way to many variables to come up with some magic potion for builds. Every build is different and the parts suggested will be different for nearly every build based on the users intended use, budget, screen resolution, case size limitations, etc...

Gamers will always want to pick the fastest GPU possible for the games they intend to play and the resolution of their monitors. There are tons of websites full of benchmarks comparing different GPU's. There's no need to re-invent that wheel. A gamer will nearly always sacrifice HD speed for a better GPU since FPS are much more important than the game loading three seconds faster with an SSD.

The "one premium gpu or two mainstream gpus" debate is really pretty clear cut. In nearly every case, it's better to get the best single card your budget allows. Doing so leaves you an upgrade path, uses less power, generates less heat, and performs better most of the time since some games don't scale well with multiple cards. Going with multiple cards right off the bat severely limits your upgrade options and is rarely recommended.

Workstation builds are entirely different animals. Depending on what software they will be using, it's usually best to get the fastest quad core CPU, most RAM, and best workstation GPU for these builds. Gaming GPU's pretty much suck at CAD type work. For instance, I recently built a CAD workstation for an engineer that asked me to build the best tower possible for $1,200. That particular build used an i7 860, 8GB of RAM, and an Nvidia Quadro workstation GPU. Essentially all of those components would change if somebody asked me to build them the best gaming tower for $1,200. I would ditch the hyperthreaded CPU, go with 4GB of CAS 7 1600MHz RAM, and a fast single gaming GPU. I would pick the GPU based on budget, what games they were playing, monitor resolution, and upgrade options.

Sorry for the long post, I just think you may be searching for a silver bullet that doesn't exist. Cookie-cutter builds/suggestions rarely work. Every build has to be optimized.
m
0
l

Best solution

February 26, 2010 6:08:23 PM

I completely agree with shortstuff, but hae a couple of exceptions, which is entirely dependent on the budget.

The first exception is the massive budgets. Once you hit a certain budget level (about $2,400 right now), you literally have a computer that will the best at everything. While lower budget gaming builds sacrifice everything for the GPU and workstations sacrifice for the RAM and CPU, a large budget build can rapidly reach the point where the only improvements come from the marginal increases from the least important components.

The second exception is that there are a few budget ranges (though small and rare) where there is essentially one configuration regardless of what you're doing. Right now, this is mainly caused by the lack of a good GPU in the $200-250 range and the lack of good CPUs priced between the i5-750 and i7-920. These gaps in prices lead to builds that are essentially the same regardless of the tasks they need to optimize.

I also disagree on the cookie-cutter builds. Not every build has to be optimized. After all, one $1,000 budget gaming build is the same as any other $1,000 gaming build. However, once the budget or use changes slightly, then the recommendation needs to change.
Share
Related resources
February 26, 2010 6:15:24 PM

I agree on those very high-end builds. Those are the builds where SSD's and multiple cards make sense. You have to spend all that money somewhere! :) 

I can see your point on the cookie-cutter builds, but not all "$1,000 budget gaming builds" are the same. The variables of what games they're playing and monitor resolution still determine the hardware suggestions. Those variables make one $1,000 gaming build go heavy on the CPU and lighter on the GPU, where another $1,000 gaming build will sacrifice CPU speed for a faster GPU. That's nearly always determined by the resolution and the other uses the computer will be used for.
m
0
l
February 26, 2010 7:11:06 PM

Very insightful. Righteous, too. HDDs are still orders of magnitude slower than memory, tho. I think that at least that one general factor holds up (except for the price gap in storage options).

Thank you so much for the world class reporting. No drivel HERE !

That pretty much says it all. I'm still a little fuzzy as to the magnitude of performance diff in various mem architectures, tho. Plenty of analysis out there, somewhere. As usual, Wikipedia might be a good place to "bone-up" .

= Thanks again ... that is really great stuff and I/we very much appreciate it. ! =
m
0
l
March 5, 2010 1:49:58 AM

Best answer selected by Alvin Smith.
m
0
l
!