Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
Shaikoten

Budget Micro ATX Options

Recommended Posts

Honestly once you're off of /b/ or any of the porn boards, 4chan is actually not extremely retarded. Some of the boards are nigh impossible to troll too. For example, try trolling the board for "food." You can't do it. They'll turn your bullshit food combination into a serious thread, somehow. They don't get pissed off. It's amazing.

Share this post


Link to post

Instead of worrying about the case size, you should just get a measuring tape and determine the maximum clearance for the case you're about to buy. With my previous suggestion, you get a decent power supply with a reasonably small case.

As for the 4GB, if you ever do any kind of editing, whether Photoshop or sound. You will notice and appreciate the non-virtual memory. Do not skimp on this; I have regularly breached the 4 GB limit when using DB2 and Photoshop CS4 simultaneously.

As for the operating sytem:
http://windows7.digitalriver.com/store/mswpus/en_US/DisplayHomePage?resid=-XlP3goBAkcAAF9MHXcAAAAz&rests=1255896676905
A $30 64-bit Windows 7 for student, need I say more?

To drive the 4GB decision home, Windows 7 will definitely take advantage of that extra room, stretch itself out, and ultimately runs everything extra faster. Don't try to save yourself a few scant bucks that can be recovered later. Remember that you are going to spend a few years with this system.

ADDITIONAL NOTES: Your hard drive, DO NOT BUY SEAGATE OR MAXTOR. They have a very poor reputation in hardware enthusiast forums because they often fail before their warranty. Go for Western Digital drives.

Share this post


Link to post
Doom Marine said:

As for the 4GB, if you ever do any kind of editing, whether Photoshop

I'm going to have to agree here. With 2G of ram in my laptop, Photoshop has crashed many times when run for long periods of time.

Share this post


Link to post
Maes said:

It's mostly because we hit a technological saturation point. Frequencies have stopped rising, and architectural improvements can only get us so far. On a a per-cycle basis, no modern CPU is more than twice as efficient as my Athlon 3200+, which was released some 5 years ago. The only viable thing to do now is going multi-core, but that has its limitations as well as being utterly ineffective for purely serial processing tasks.

Even the crappiest CPU you can buy today will be dual-core, 64-bit (with a few exceptions) and be as clock-cycle-efficient as the best one available on the market, save for stuff such as cache size. That was clearly not the case when "Pentium" could mean anything from 66 to 333 MHz, with or without MMX. Compare this with the enormous power disparity (and requirements gap) that came from releasing Pentium CPUs in a world filled with 486s and 386s: it just stopped being this way from a long time.

But what about the shrinkage of the nm process? This has led to smaller dies, higher transistor density, and better power efficiency.

Intel's hexacore Gulftown processor is set to debut in 2010 with the 32-nm process: http://en.wikipedia.org/wiki/32_nanometer

We still have got a few nanometer processes left before having to resort to other engineering solutions, like 3 dimensional die processes.

Share this post


Link to post

Still, no major breakthrough in frequencies since the Pentium IV and Athlon 64 era (in fact, they are unsurpassed in terms of raw clock speeds). The nm-scale process improvements are used to improve core density rather than clock speeds, which are hindered by the very materials and have almost hit the top limit for reasons other than just thermal efficiency: propagation delays and dielectric effects come into play at frequencies in the GHz range (ask any microwave engineer, especially one specialized in microwave electronics).

Therefore, don't expect any major speed breakthroughs other than improving core densities, ever-cheaper RAM, and more massive parallelism on GPUs. The days of the single, powerful CPU with the ever-increasing frequency are long over. The only possible way to break the slavery of clock speeds is using new materials (carbon nanomolecules, optical microcircuits, or even vacuum microelectronics) or going all the way to asynchronous CPUs.

Share this post


Link to post
Maes said:

Still, no major breakthrough in frequencies since the Pentium IV and Athlon 64 era (in fact, they are unsurpassed in terms of raw clock speeds). The nm-scale process improvements are used to improve core density rather than clock speeds, which are hindered by the very materials and have almost hit the top limit for reasons other than just thermal efficiency: propagation delays and dielectric effects come into play at frequencies in the GHz range (ask any microwave engineer, especially one specialized in microwave electronics).

Therefore, don't expect any major speed breakthroughs other than improving core densities, ever-cheaper RAM, and more massive parallelism on GPUs. The days of the single, powerful CPU with the ever-increasing frequency are long over. The only possible way to break the slavery of clock speeds is using new materials (carbon nanomolecules, optical microcircuits, or even vacuum microelectronics) or going all the way to asynchronous CPUs.

If by speed you mean calculations per cycle, the industry is still keeping up with Moore's law, if you compare clock-for-clock across the different generations, there is a noticeable difference:

http://forums.overclockers.com.au/showthread.php?t=812836

Even if the GHz has reached its physical limits, parallel processing and die shrinkage will still push the limits of cpu speed.

http://en.wikipedia.org/wiki/Moore%27s_law

Share this post


Link to post

Actually, Moore's Law refers to the on-die transistor count, and so far it has been verified by the events (give or take), and it looks as if it will still be verified for as long as traditional microelectronics will be viable.

Calculations per cycle depend on many factors such as: the architecture (CISC or RISC, pipelined, superscalar), predictive branching, pipeline stalling, pre-emptive execution of instructions etc.

So, for example a 68k needed a constant 4 CPU cycles (for the simplest) up to 46 or so CPU cycles (for the most complex) for its instructions, most RISC CPUs can guarantee 1 cycle per instruction, while superscalars can only be defined in terms of maximum theoretical instructions per cycle, usually between 2 and 5, highly code-dependent.

However more components per chip doesn't mean more processing power: e.g. early ARM processors had less transistors than Intel or Motorola 286 and 68000 CPUs, yet kicked their butt cycle per cycle. And slapping two or four on them on a system (there were multi-CPU 68k systems in arcades) was not automatically a guarantee of getting a faster system: the "speed" came from job splitting at the programming level, not from the increased ability to perform a single, inherently serial task quicker.

Since my bachelor thesis had to do with parallel programming, one of the first things everyone is taught in that field is Amdahl's law: a parallel's system efficiency on a given problem is limited by the problem's inherently serial part.

This means that you can throw 100 CPUs working in parallel on something like e.g. adding all elements from an array (inherently or embarassingly parallel problem), but not on something like constructing Fibonnacci's numbers as fast as possible (an inherently serial problem). Only a singularly fast CPU can help you with that.

In the case of the latter kind problems, usually tradeoffs between accuracy and the ability to compute them in parallel are made, or different approaches are tried to the problem (like I had to do on my thesis). The days where you could just wait for the day where a given CPU would be 2x faster just because the clock would be pushed 2x faster, eventually, are definitively over. The speedups are now much more subtle, and certain tasks don't benefit at all (your new quad core may multitask more smoothly than my Athlon 64, but they may still be on par for a purely sequential problem).

Share this post


Link to post
Maes said:

Since my bachelor thesis had to do with parallel programming, one of the first things everyone is taught in that field is Amdahl's law: a parallel's system efficiency on a given problem is limited by the problem's inherently serial part.

This means that you can throw 100 CPUs working in parallel on something like e.g. adding all elements from an array (inherently or embarassingly parallel problem), but not on something like constructing Fibonnacci's numbers as fast as possible (an inherently serial problem). Only a singularly fast CPU can help you with that.

I guess I haven't considered recursive calculations like the Fibonacci sequence when it comes to computing speed and multi-core processing. I can definitely see the limitations of parallel processing when it comes to calculations dependent on the previous calculations.

Currently reading this:
http://www.denbeste.nu/essays/futurecs.shtml

Share this post


Link to post

Interestingly, the serial/parallel dichotomy has ramifications even to something more close to us DW forum dwellers: Doom itelf.

The original DOOM was, by design, like all DOS games, inherently serial (at least the game engine and the majority of its mechanics are). The rendering and audio subsystems can be parallelized to your heart's content, but making parallel monster code that is also 100% vanilla compatible could be impossible, since the processing order can't change (then again, there might be parts that can be parallelized without major breakups, and others that must be applied sequentially in order to keep stuff from blowing)

This means that a port using multi-threaded monster code must take special precautions e.g. a bunch of hitscanners hitting each other will have a deterministic behavior in a serial context (e.g. get hurt in a specific order), but in a threaded/parallel one they may get hurt in a different order, and what's more hard to comprehend is that it might be impossible to record demos of such a context -dependence on actual time, not game tics, becomes critical-.

But I'm digressing here... damn.

Share this post


Link to post

I find GPGPU alot more interesting than just plain multicore processors.

For a relatively new concept, I love that Apple have gotten on board and written OpenAL into Snow Leopard, and that ATI have embraced it as well.

I don't know much about the arithmetic differences between a single complex x86 core and several hundred simple GPU cores, but based on sheer numbers it seems like a promising concept which could provide existing systems visibility into alot of untapped power.

Share this post


Link to post
Jodwin said:

Who said you can't put a normal tower sideways on your desk? :P

Good way to bork disk drives, from what I hear.

Share this post


Link to post

Wrong indeed. All hard disks made after 1990 can work horizontal or vertical or even angled with no problems. Think of external enclosures. Now maybe if you have some very old or cheaply made HD that can't compensate e.g. for asymmetrical gravity on the head arms or makes nasty contact starts..well...that's another story. At least vertical and horizontal are pretty standard and can be coped with, though.

Share this post


Link to post
Jodwin said:

Who said you can't put a normal tower sideways on your desk? :P

Depends on what you're planning to park on top of it. While an LCD monitor shouldn't be a problem my old 20 inch CRT would squash it flat.

Share this post


Link to post
Maes said:

Wrong indeed. All hard disks made after 1990 can work horizontal or vertical or even angled with no problems. Think of external enclosures. Now maybe if you have some very old or cheaply made HD that can't compensate e.g. for asymmetrical gravity on the head arms or makes nasty contact starts..well...that's another story. At least vertical and horizontal are pretty standard and can be coped with, though.

I wasn't referring just to hard disks.

Share this post


Link to post
TheeXile said:

I wasn't referring just to hard disks.


Floppy disks and cassette tape recorders/data streamers work happily sideways too, unless the mechanics really suck ass. However some CD-ROM/DVD-ROM drives may malfunction, even if there are devices with vertical or diagonal disc readers (mostly "arty" sound systems and the such).

Fans may be worth a consideration, depending on their bearing types (etc.) but it will go way OT.

Share this post


Link to post

Apart from the fact almost every computer CD/DVD drive since the old 1x clunkers (with disc caddy!) has been made with features specifically designed to operate on its' side.

Share this post


Link to post
GreyGhost said:

Depends on what you're planning to park on top of it. While an LCD monitor shouldn't be a problem my old 20 inch CRT would squash it flat.

Get a stronger case.

TheeXile said:

I wasn't referring just to hard disks.

Get a case with enough space to put your CD/DVD drive sideways. Or pull it outside of the case for all I care.

Share this post


Link to post

Heh I recall my first 486 had a "dual positioning" case, e.g. you could use it both vertically and horizantally. I started by keeping it vertical (the case was somewhat slim for a full-height, but it had some special removable clamps to help it keep upright).

However as soon as I bought my first CD-ROM drive, I put it horizontally because the 5.25" bays were installed along the broadest side. As for putting a monitor on the case...I've only ever seen it being feasible with CGA monitors or monochrome VGAs, that were comparatively lightweight.

The case wouldn't support the weight of a 14" SVGA monitor ( >12 KG) for a long time without losing its shape, let alone a bigger one. The old XT's top cover might have holded, but that weighed more than 1.5 kg alone :-p

Share this post


Link to post

I built a system with one of these at work. It's one of the few desktop-style cases I could find (although it would also work upright), and it's the perfect size for an office PC. (It does almost exclusively word processing and has no need for any add-on cards, meaning the low-profile expansion slots aren't a problem.) It currently supports a CRT screen which claims to weigh 37 pounds (so it's around 17kg)...

Share this post


Link to post
Doom Marine said:

ADDITIONAL NOTES: Your hard drive, DO NOT BUY SEAGATE OR MAXTOR. They have a very poor reputation in hardware enthusiast forums because they often fail before their warranty. Go for Western Digital drives.

Failure rates are about even for all drives. You're also just as likely to get a WD DOA as you would any other brand. The results are always skewed together from a vocal minority that will rage on the internet when their hard drive shits out regardless of brand and warns people to, "NEVAR EVAR BHUYZ IHT!"

Seagates have fixed the firmware issue which was causing a shitstorm of fail some time back. Yet everyone nowadays jizzes over the WD Caviar Blacks so forget about buying any other brand for your budget PC.

Share this post


Link to post
POTGIESSER said:

Failure rates are about even for all drives. You're also just as likely to get a WD DOA as you would any other brand. The results are always skewed together from a vocal minority that will rage on the internet when their hard drive shits out regardless of brand and warns people to, "NEVAR EVAR BHUYZ IHT!"

Correct.

Share this post


Link to post

So a bit of a bump and an update: I haven't bought it yet, since it's not all too pressing, but I finally got my total down to $400 (minus shipping) without sacrificing anything from my initial build, which was over $450.

Mobo/CPU: http://www.newegg.com/Product/ComboDealDetails.aspx?ItemList=Combo.273368
GPU/RAM: http://www.newegg.com/Product/ComboDealDetails.aspx?ItemList=Combo.285426
Hard Drive: http://www.newegg.com/Product/Product.aspx?Item=N82E16822152100
Case: http://www.newegg.com/Product/Product.aspx?Item=N82E16811154094
DVD Rom: http://www.newegg.com/Product/Product.aspx?Item=N82E16827118031
Power Supply: http://www.newegg.com/Product/Product.aspx?Item=N82E16817709010

If anyone sees anything that's incompatible in this setup, please let me know. And I know the PSU is flaky, but I'm fairly certain I have a decent one on deck.

Share this post


Link to post

Looks good, but be aware that (1) the optical drive doesn't write and (2) the motherboard has on-board video. The video card will still work but it's not required.

Share this post


Link to post

Thanks! Aware of both. I'm thinking of possibly switching to a different mobo with a combo because I don't really need onboard video, but it seemed like a good deal.

Share this post


Link to post

Skip the Samsung hard drive, get Seagate or WD Caviar Black for speed, WD Caviar Green for low noise/heat/power usage.

After years of cooking hard drives in cheap cases, I finally dropped the cash on an Antec P182B and my PC has never run cooler. Antec also have the Mini P180 for mATX. Definitely a good investment which will last you multiple builds.

Share this post


Link to post

I live in the US and already have my system set up underneath my $400 budget, man. See the above posts.

(Although I have switched to a Cavair Green, seems like a good choice for a micro-atx that I want quiet and cooler.)

Share this post


Link to post

Finally got the cash to buy this, and here's how it turned out:

1 x ($24.99) CASE APEX|TX-381-C RT - Retail $24.99
1 x ($71.99) MEM 2Gx2|WINTEC 3AXT6400C5-4096K R - Retail $71.99
2 x ($29.99) GIFT STORMRISE GAME COUPON BUNDLE - Retail $59.98 (lol)
1 x ($69.99) HD 750G|WD SATA2 32M WD7500AADS % - OEM $69.99
1 x ($79.99) MB ASUS M4A785-M 785G R - Retail $79.99
1 x ($69.99) VGA MSI R4670-MD1G DDR3 R - Retail $69.99
1 x ($16.99) PSU LOGISYS|PS480E12 480W RT - Retail $16.99
1 x ($99.00) CPU AMD|PH II X2 550 AM3 3.1G RT - Retail $99.00

Plus all kinds of combo and instant discounts you aren't seeing.

It did nudge over $400, but I got a little bit of extra cash I was instructed to use for computer buying, so I nudged the processor and hard drive up a bit. Apparently the processor I selected is a stupid easy overclock and very, very often unlocks to 4 cores. And whatever that shitty Stormrise game is was free.

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×