Truly great PCs...?

[LongRunner]

I've been thinking about it, and here's what I have in mind:

• First and foremost, of course, are the quality standards - exclusively Japanese electrolytic capacitors are just the beginning, and quietness is not far behind reliability. Removable dust filters, of course, to avoid having to open the case for cleaning.
• Regarding quietness, I would prefer to use 3-wire fans with a true voltage controller, rather than the 4-wire PWM types that sometimes emit strange tones.
• A departure from conventional "value engineering" (where the builders are only concerned with the system lasting for how long a typical user keeps it), to instead make longevity a design priority, so users with lower performance demands can get a second- or even third-hand system (it's been a long time since you actually needed up-to-date hardware to get satisfactory performance for many applications*), saving money and the environment.
  Of course, we'd have to actually convince them that these systems are still worth something, but if that can be done, I think I'd be able to explain it to them.
• I would actually love to have a higher pixel-density monitor. With good old 4:3 aspect ratio. And much dimmer than the eye-wateringly bright screens they make now. But as great as having clearer text would be, no-one is actually ready for the upgrade due to software so far being designed around a fixed pixel density. As for the aspect ratio, the manufacturers abuse the use of diagonal measurement of screen size to take advantage (economically) of the fact that a 16:9 screen only has 89% the surface area of a 4:3 screen with the same diagonal length (less surface area is cheaper). For watching a movie widescreen is good, not so much for PC. If you thought it seemed a bit small when you got it, you're absolutely right.
• Regarding screws, arguably the least interesting part of the system...it's time to say goodbye to Phillips-type screw heads - which have no advantages - and to switch to something stronger, like Torx. If it wasn't for laziness, Torx screws would have become the dominant standard years ago.
• And my case colour of choice is...wait for it...light grey, with dark grey power and reset buttons. I don't have an honest idea why you would want the chassis or PSU to be painted, given that the back panel is seldom viewed and even if it is, it's not exactly a sight to behold, painted or otherwise. I also would like to have no extraneous curves, and suffice to say that gloss is the work of the devil.

*Such as e-mail, web browsing, word processing, etc...so long as the software has any efficiency, of course (I hate bloatware). (But note that an application that seems to use a lot of RAM isn't necessarily bloated - it may be using the space to cache information that would otherwise be either regenerated (consuming processing power) or loaded from HDD. Though if it really does require that much RAM to run...well, for lack of a better explanation, it probably is indeed bloated. But RAM is easy to upgrade.)

I'm not convinced that the way of the future is tablets and storing everything in "the cloud", even if some Badcaps.net members say so (after all, I can fault even them on some of the things they've said). Typing in any volume requires a physical keyboard, and for the forseeable future, online storage will be slower and less secure than local storage.

[LongRunner]

How compact can a desktop PC be? It's a good question. But before I begin, let's make it clear how I define a desktop PC:

• True desktop: Must have reasonable expandability and be straightforward to work on. Should stay cool and quiet under all workloads (with exemption for extreme gamer systems). Must have room inside for 3.5" drives at least (see below for the optical drive size thing), and proprietary components are forbidden.
• Small form factor: May make sacrifices to get the size down - though I would prefer to at least stick to standard components.
• Proprietary: Anything that uses non-standard components falls into this category, and yes, this does mean that "desktop" systems from some of the big names are not true desktops under my definition.

Anyway...

PSU form factors
Yeah, you have to admit that the ATX PSU form is a bit of overkill for small units. As I have a copy of PSU Design Guide 1.1, I know of most if not all of the standard PSU forms. So here's what I think of each of them:

• SFX (60mm front fan) - fine, I think, though the fact that the fan is front-mounted (i.e. with the grille right up against its intake) suggests that it could whine a bit at higher speeds.
• SFX (80mm top fan) - no problem, presumably.
• SFX (80mm top fan) "reduced depth" - swaps width and depth, compared to the other versions. In a desktop system, would most likely make better use of the available space than the other SFX versions (and also has more exhaust area).
• SFX (40mm front fan) - probably has some use, but for a true desktop system, it's simply not a good trade-off.
• CFX - a funny-looking form that's kinda like an ATX unit with a 95mm depth and a section removed, resulting in a sort of "L"-shaped cross-section. Doesn't look to me to be much more than a curiosity for a true desktop system.
• LFX - why would you put one of these in a true desktop system? In case you're wondering, it has an almost square cross-section (72mm by 62mm), and is very deep (210mm!), with a segment removed at the front, and would most likely have a 70mm fan (I'm not certain though).
• TFX - with an 80mm top fan, and dimensions of 85mm wide, 65mm high excluding the fan grille, and 175mm deep. If this review is any indication, it's a thermal failure.
• PS/3 - well, actually the Design Guide says "PS3" - but not to be confused with a certain video game system. Hardly worth mentioning as it's just the ATX PSU form reduced to 4" depth.
• FlexATX - with a 40mm rear fan, dimensions of 81.5mm wide, 40.5mm high and 150mm deep.

Of the group, SFX "reduced depth" looks to me the most useful for desktop systems.

Sub-microATX forms
You can decide for yourself whether the size reduction with a case designed for smaller boards is worth it. I'm currently reluctant to go smaller than microATX, but feel free to say what you're comfortable with.

That said, FlexATX does provide a significant reduction in depth compared to microATX (from 9.6" to 7.5").

Low-profile expansion slots
No doubt like most PC builders, I consider them to be a compatibility breaker not worth the potential size reduction (and well...if you can fit a decent-sized fan between the I/O panel and the side of the case, you can fit full-height expansion slots). Big names love them, of course, but what else do you expect when they're so focussed on money?

Internal optical drives
Should you desire a narrow tower case (I don't feel the need myself), they're rather annoying, as they're quite wide and can't be made much narrower. If you want a narrow tower case so bad (or if you want a very small PC in general), I guess there's always the external option...but then again, making the case much narrower will sacrifice stability, so I'd rather have a desktop case than a very narrow tower. Their length (even for modern, shorter units) is also a major limiting factor in reducing the depth of a tower case, should they be placed with their back end towards the mainboard*. Fortunately, two stacked optical drives (good enough for most users, and overkill for many) are similar in height to an ATX PSU (which of course is shorter than the board), but if you want more than two, or mount the PSU at the bottom of the case... (There's also notebook-type optical drives, but I'm not sure those belong in a true desktop. SFF, fine...)

So what is the answer??? I figure that (with a 92mm rear fan) a 6.75" wide tower would be no problem, and that's a bit narrower than the one I have. Height and depth may be harder, though. I guess they're already about as compact as practical, given the components they're designed around. So shrinking them further would require either using smaller components, or removing them entirely...

*Because "motherboard" is just a cutesy term.

[Wester547]

I don't think you can make desktops any more compact than they already are without causing onerousness. The heat output is already so intense in such a compact space that even Japanese lytics can't take the heat (in other words, fitting desktop-grade components into that which is only a level above a laptop in size). RoHS compliance and BGA packages don't help this at all. I myself have always preferred regular desktops with ample room for airflow and ventilation. They may not be as portable but that's what laptops are for. Noise doesn't really bother me so long as my computer doesn't sound like a lawn mower. I'd take a noisy computer (and a reliable one) over an overheating computer any day (no matter how fast it is).

[LongRunner]

Noise doesn't really bother me so long as my computer doesn't sound like a lawn mower. I'd take a noisy computer (and a reliable one) over an overheating computer any day (no matter how fast it is).

(Don't get me started on lawnmower loudness. It's sickening. It's unforgiveable.)

Sure, noisy is preferable to overheating. But I know cool and quiet is not that hard to achieve - provided you aren't using tiny fans and heatsinks, blocking the airflow path, or dissipating ridiculous amounts of power...

It doesn't help, of course, that the stock coolers for CPUs are designed first and foremost to minimise the heatsink cost. (And, well, better that than to use a low-quality fan.) It may be possible, with good design, to make a relatively small heatsink that performs in line with many larger models, if this old article is any indication (this unit performing almost as well as this screamer). But there are limits to everything. (And the monstrous graphics cards can't have an easy time staying cool at all, which is why I think water cooling would help them even though I don't really recommend it otherwise.)

I never like making anything unnecessarily noisy. Of course, there are any number of noise sources outside of the PC, so there's more to do...

[LongRunner]

The thing about general-purpose electrolytic capacitors, as usually used in low-stress areas, is that they have low endurance ratings - usually 2,000 hours at 85°C or 1,000 hours (for small types) at 105°C, and some tiny ones (e.g. Chemi-con SRA) are only good for 1,000 hours at 85°C. For your convenience:

Capacitor endurance table.png (12.69 KiB) Viewed 34788 times

It's actually a screenshot of one I coded in HTML (though it's not on the net; it's staying on HDD for now, as part of a private project - but I will let you in on the fact that the page containing it (still under construction) is about PSUs, including repairing them). Red cell for unusable (<13K hours), orange for poor (13~25K), yellow for OK (25~50K), green for good (50~100K), and blue for excellent (100K+).

My concern is that if general-purpose capacitors, with their low endurance ratings, continue to be used as mentioned, they could wear out before the low ESR capacitors in PSU filtering. Especially if the manufacturer continues to get GP caps from low-quality brands (in the not unrealistic hope that they will get away with it). The solution is straightforward enough - use low ESR capacitors (with their higher endurance ratings) whether the ESR itself is important or not. Except that low ESR bipolars don't exist, so if you want the endurance, you'll need to put two low ESR caps (of double the value) back-to-back, wasting board space...

85°C (and generally 2,000-hour) capacitors that are usually used on PSU primaries aren't really that great, either, as if you want them to last long they need to be kept quite cool (there are instances of even Panasonic 85°C primaries drying out in hot PSUs). While most 105°C general-purpose capacitors have lower ripple current ratings, Panasonic TS-ED series (3,000 hours at 105°C) is comparable to the 85°C types in this regard. If you want to fit new primary capacitors to an old PSU, I highly recommend them.

[LongRunner]

Oh, yeah, and here are the formulae to calculate the width and height of screens in the common aspect ratios (you can use any unit of measurement):

4:3
Width: sqrt(<diagonal>^2*(16/25))
Height: sqrt(<diagonal>^2*(9/25))

8:5 (a.k.a. 16:10)
Width: sqrt(<diagonal>^2*(64/89))
Height: sqrt(<diagonal>^2*(25/89))

16:9
Width: sqrt(<diagonal>^2*(256/337))
Height: sqrt(<diagonal>^2*(81/337))

where <diagonal> is, well, the diagonal length of the screen in question.

Anyone willing to hazard a guess as to when fine-pitch (>200PPI) screens will become popular on PCs???

[LongRunner]

I have basically discovered, long after the creation of this thread, that the mainboard itself has nothing to do with electrical noise induced into the front audio circuits and that it's all the case's fault for creating an earth loop (despite the fact that "breaking" it properly would require just 1 small resistor (thick film, presumably) and 2 (at most) ceramic capacitors) and, in some cases, using an unshielded cable (which probably would only cause a severe problem on the microphone input but still…). Suffice to say that the more you listen through headphones (like me; it will be heard most prominently with nominal 32Ω (in contrast to the higher impedances) sealed headphones), the worse it will affect you.

My other desires for improvements to the PC audio system are about functionality rather than fidelity, but for what it's worth:

• Dual equalisers: A "main" equaliser used by the rear outputs (which of course are primarily for speakers), and a second "headphone" equaliser that can be applied either on top of or independently of the "main" EQ for the front headphone output (and the rear-mounted stereo/front speaker/headphone output if specifically chosen). (A small thing, but using it well could certainly provide a noticeable, if not dramatic, improvement.)
• (Edit 2014/12/17: It's already implemented.)

Transplanted 2014/12/17: I was wrong on the statement originally here, but I found out something since that is also headphone-related: Small coupling capacitors combined with low-impedance headphones will roll off the bass. (This can sometimes help to tame, to an extent, the usual bass-boost inherent to sealed headphones, but it does impose some extra maths on the discerning audiophile user — if they don't instead use a separate headphone amplifier which for a PC would be a needless expense unless they still use a PC with an ancient sound card that can't drive headphones by itself.)

Anyway: I cannot endorse any proposal that inhibits the freedom of the PC as a platform, in any way, shape, or form. Sadly, some major corporations are apparently trying to do exactly that.

I don't think the extended use of the x86 architecture (or derivatives at any rate) really helps with CPU power efficiency. If nothing else, it would be an interesting experiment to compare one of the RISC designs against a modern PC CPU and see what it is capable of (performing equivalent tasks). (Not that it's going to happen any time soon.)

[LongRunner]

Yes, they did exist. I have one on my desk right now. Only, it's not very useful, because it's from the bygone era of the "traditional" ATX port layout:

• Dual PS/2 (keyboard and mouse)
• 2 USB + optional ethernet (not open on my sample)
• LPT1 and COM1 + either COM2 or VGA
• Optional stereo audio with game port (already open on my sample)

Suffice to say that the last mainboards I've seen with that layout date from 2003.

Nonetheless, compared to typical I/O shields, this unit, with a thick galvanised steel layer riveted to the thin spring-steel layer below, is incredibly strong. A full-colour textured plastic label on the out-facing side completes the deal. What I would like is for I/O shields built like this one (though in suitably modern layouts) to return and become the norm — as I'm no big fan of those flimsy things made of only the single spring-steel layer. Sure, they don't exactly affect the performance or reliability of the PC they are installed in, but they're a quality attribute all the same, just as the construction of the case itself is.

[c_hegge]

Yup. Dell sometimes don't use a pop-in IO shield, and just have the port holes cut into the case. while this makes for a very solid IO shield, it means that the motherboard can only be replaced with another one the same. I have a nice slim Dell Vostro desktop case which I'll have to cut the back out of because if that.

[LongRunner]

As we know, it's a little over 13 years since Seagate's famed Barracuda ATA IV hard disk drive came and took the PC silencing community by storm. Before the fluid-dynamic bearing was developed, permanently quiet HDDs could only be dreamed of. And they've certainly stood up to the test of time — I've yet to hear of a worn-out FDB. But while they've been standard equipment on every new HDD since 2005, fans equipped with true FDBs are still few and far between.

There are a few likely contributing factors:

• Patent on the original FDB: As mentioned in the Hardware Secrets article, Matsushita/Panasonic holds a patent on the real fluid-dynamic bearing design so other manufacturers have to pay considerable licensing fees to use it. To make things worse, there is no entity to govern usage of the term so you don't know what actual type of bearing is used unless the manufacturer explicitly states that it's the Matsushita/Panasonic design.
• Cost: Most of this is the direct consequence of the licensing fees, although the manufacturing cost itself is also a bit higher than for conventional sleeve bearings.
• Not necessitated: But probably the biggest factor is that there is still no requirement for FDBs in fans. HDDs of course couldn't (reliably) use BBs for much longer with ever-increasing areal densities (requiring extremely high precision), but fans have no such demands. And noise from BBs in fans isn't nearly as bad as that of BB-motor HDDs, nor does it usually dominate the overall noise output of the fan except at very low speeds.

It will be noted that the fluid-dynamic bearings in HDDs are considerably more sophisticated than the design for fans — with axial support, and ferromagnetic bearing oil inside a magnetised bearing case to prevent leakage. (Although I would be more than happy myself to pay extra for the advanced design even in fans — not that they're a high-cost item but I still think >100K hours lifespan with no noise increase would be quite fan-tastic. )

Probably the small, high-RPM fans would benefit the most from FDBs. The other "enhanced" bearings (e.g. Delta's Superflo, Panaflo's Hydro Wave, Papst's Sintec) may also provide quite respectable lifespans, but not record-setting.

I have no doubt that the best-quality 2BB fans (i.e. Sanyo Denki) can be very quiet in their own right, at least when fairly new. But the biggest drawback of ball bearings in fans is their vulnerability to mechanical shock. That's the one reason besides price (and lower bearing noise at low speeds) to choose sleeve bearing over ball bearing fans, in lieu of any other option.

I also have to say that the current patent enforcement period (20 years?) is far too long for the high-tech industry. Consider that the Barracuda ATA IV through to at least the 7200.10 all came with two distinctly different motor drivers: STMicroelectronics' SMOOTH family (of which several iterations were used), or conventional six-state drivers from Texas Instruments (SH6950 on Barracuda IV/V and 7200.7, and SH6960 on 7200.8/9/10) that were more reliable, but resulted in a good deal more whine from the drives with them. (The HDD manufacturers themselves have mandatory cross-licensing agreements, but that apparently didn't apply to the motor drivers for them. )