About ECS (a fair warning)

[VirtualLarry]

Originally posted by: Peter
No board has ever had six revisions, you say? ROTFL. You know nothing, mate. You're making up facts to support your FUD.

I mean major revs, going from hardware rev 1.0 to 6.0. Most major hardware revs involve layout changes, among other things. I don't mean individual component changes for minor hardware revs. If I'm so wrong on that point, please point out a PC mobo that has had six major hardware revs. I've been putting together PCs for quite some time, and I've not seen any. (Granted, most mobos only have a market lifetime of a few major hardware revs regardless, but the highest that I've ever seen is usually v3.1 or v4.0, most never hit higher than v1.1 or 1.2.)

 

[Peter]

K7S5A had TWO of these major revs, one adding 4-channel sound, the following one adding a VIA USB2 controller on the PCI bus. No more. How does that make six in your book?

 

[CraigRT]

Originally posted by: Peter
Ah, the usual ECS-bashing parrot ... appears every couple of months.

Sorry to say, but one of the only mainborads I've ever had problems with was a K7S5A (and at least its not just me).. I just believe you get what you pay for in this case.... cheapass mainboard = cheapass quality.

 

[Peter]

I'd rather say: Combine any random mainboard with cheap PSU and cheap RAM = trouble. It's just that cheap stuff attracts the cheapasses, who then overdo it ... only to then come back and rant at the most popular target. It's getting boring, folks.

I've seen a handful of K7S5A systems not running right; each and every one was instantly cured by a strong-enough PSU and error-free RAM. If there's a quality issue to REALLY complain about, then it's generic RAM and PSUs. So much over-specified and under-performing stuff out there, it isn't funny anymore.

 

[Creig]

Originally posted by: kogase
It's my hope that ECS will never get another cent from anyone who reads AT forums.

Well, I wouldn't count on your hopes being fulfilled anytime soon. You don't like ECS anymore? Don't buy from them again. That's the strongest statement you can make. Personally, I have nothing bad to say about them. In my computer collection, I currently have both a K7S5A and K7SEM running 24/7 and I couldn't tell you the last time either one rebooted. Nice, cheap, stable boards.

 

[DAPUNISHER]

Originally posted by: VirtualLarry

Originally posted by: DAPUNISHER
Yeah I remember mrathlon2000, he stfu the second ECS threatened to sue him Man those were the days, the controversy surrounding all that was better drama than the current fanboy crap could ever be

I remember the issue with those boards, the RAM incompatibility problems, the mysterious rapidly-draining CMOS batteries, the cold-boot issues, the ZP impedance-matching resistor change causing incompatibilities with certain CPUs. (Which, technically, isn't so much ECS's fault - AMD mandated that change in their design specs. For whatever reason though, it caused problems on that board.)

However, I missed the whole lawsuit thing, apparently. What did mrathlon2000 do/say that triggered ECS's legal response? I would think that the mfg would have to have some serious !@#$# going on, for them to be motivated to sue a customer to shut them up.

I just browsed the old ECS archive@OCWorkbench where it all went down and I think they may have nuked his huge thread about data corruption as I didn't find it during my brief perusal. As I recall, the rep from ECS who had been participating in and monitoring the forum, told him that they would not tolerate his accusations of the product being faulty and requiring the resisitor fix, so STFU or they would bring all their legal resources to bear against those making the accusations. If you don't have the cash for the legal battle you STFU, and he did

 

[imported_OrSin]

All my ECS boards worked great. Now I have gotten 4 chaintec boards 2 was DOA and the third worked for about a month. I thought it was me for the longest. I attched about 4 static guards to me when ever I did anything. 2 ASUS boards later I was filling better.

 

[arswihart]

can we kill this thread?

 

[Tanclearas]

I guess what I find most amusing is the number of times I've read/heard someone claim "I'll never buy from <INSERT BRAND HERE> again, and no else should either!"

If everyone actually paid attention to pointless drivel like this, we wouldn't have any computers today. I've heard this about every brand for every major product type. Every manufacturer has released (and will release again) a less than stellar product. This doesn't necessarily mean a defective product, but rather a product that isn't quite able to operate in a particular way.

As for the K7S5A, I know of two such boards (one XP model, one not) that were purchased in the early days and are still in regular use today.

 

[boatillo]

I feel the same way about Gigabyte stuff (and I've seen others too!)

My Gigabyte radeon 9000pro suffered unreversible who-knows-what that caused 100's of random pixels to flash on screen in only 3d apps. This was after a few months, no overclocking.

Originally paired that video card with a Gigabyte GA-7VAX mb...lets just say it can not, and will not run my new MSI geforce card in 8x AGP (although it says its highly recommended in manual blah) crashing in any 3d app with 8x enabled. And then another who-knows-what one day now makes it necessary to warm the computer up for 5 mins before it will even post.

 

[Peter]

That, very probably, also is a power supply problem dragged along through that history of random misbehavior. See above.

 

[boatillo]

i use an antec tru power 430w on the abominable gigabyte board atm

but yes my ram is cheapo one stick is true ddr400 at like 3-3-3-8 or something, the other is pc2100 in everest at like 2-2-2-6 or something (when I bought it site said it was ddr400, prolly oc'd...)

 

[VirtualLarry]

Originally posted by: Peter
I'd rather say: Combine any random mainboard with cheap PSU and cheap RAM = trouble. It's just that cheap stuff attracts the cheapasses, who then overdo it

That's definately true. The problem comes about, when the vast majority of components are sold on the basis of price, purchased by less-than-fully-technically-informed customers, that likewise pair "cheap" mobos with "cheap" PSUs, "cheap" RAM, and other components. But the problem comes about because the "cheap" designs have far less "engineering margin" in them, to cut costs, so when you combine them, the chances for having an incompatibility are far higher. The fact that you "need" to pair a certain component with otherwise high-quality PSUs and RAM, when other components of a similar nature do not need that special treatment, indicates that those components have very little "engineering margin" built-in. IOW, cheap components are "picky" about what other components that they will work with.

I'm not excusing customer for willfully choosing to go with the cheap-end components, but it's pretty well-known that PCChips makes most of their money in that market segment. (In the past, going so far as to put fake cache chips on the board to cut costs. Hopefully they won't ever stoop that low again.)

Originally posted by: PeterI've seen a handful of K7S5A systems not running right; each and every one was instantly cured by a strong-enough PSU and error-free RAM. If there's a quality issue to REALLY complain about, then it's generic RAM and PSUs. So much over-specified and under-performing stuff out there, it isn't funny anymore.

Well, true, but that applies just as well to motherboards too, or for that matter any "cheap" PC component.
(Or are you trying to imply that the only time that any user has ever had a problem with a K7S5A, that it was due to actually defective RAM or PSUs, rather than just being "picky" about compatibility/interoperability?)

I personally wonder if some of the issues weren't due to the fact that PCChips took the layout of their SiS635T-chipset-based Socket-7 mobo and just did an almost drop-in component swap with the 735 chipset and Socket-A. The layout is virtually identical. I had some very wierd issues with my K6S5AT (?) board back in the day too. (Could just be due to SiS reference designs too, looking like that, I don't really know.)

I'm not saying all PCChips boards are crap, far from it, many users find that they "work". But they are intended for the severely low-cost market segment, and their QC tends not to be the best.

 

[VirtualLarry]

Originally posted by: DAPUNISHER
I just browsed the old ECS archive@OCWorkbench where it all went down and I think they may have nuked his huge thread about data corruption as I didn't find it during my brief perusal. As I recall, the rep from ECS who had been participating in and monitoring the forum, told him that they would not tolerate his accusations of the product being faulty and requiring the resisitor fix, so STFU or they would bring all their legal resources to bear against those making the accusations. If you don't have the cash for the legal battle you STFU, and he did

Wow. That seems really unethical to me. A reputable mfg would have done some engineering analysis and replied with the results when they were all said and done.

I've seen data-corruption issues with a number of hardware components over the years, including the Promise Ultra66 PCI IDE controller cards, even on an i440BX chipset. (Something that was often blamed on the pairing of Promise with Via, as some mobos did back in the day. The Via 686B IDE and PCI-bus data-corruption issue were seperate. But put them all together, and maybe add an SBlive into the mix, and you could have a data-corruption field day!)

OTOH, I've never seen any severe data-corruption from SiS-chipset boards, save for some issues with ancient '496/'497-based mobos when used with AMD 486 CPUs with write-back cache, and trying to use the floppy drive (ISA DMA) at the same time (on an M-Tech board). The PCI and IDE performance from that 635T-chipset based ECS board that I mentioned, were just as good as my i440BX and much better than this MSI Via KT400/8235 board I'm using now. (Which also had IDE data-corruption issues when released, it took eight BIOS revs to work around it. Apparently that issue was more due to board layout than chipset bugs, because other mfg's didn't have that problem with the chipset.)

Sorry, getting a bit OT here.

 

[VirtualLarry]

Originally posted by: Peter
K7S5A had TWO of these major revs, one adding 4-channel sound, the following one adding a VIA USB2 controller on the PCI bus. No more. How does that make six in your book?

Ok, I guess I was wrong there, ECS/PCChips *did* skip revs 2.x and 4.x on that board. For what reason, I have no idea. (Maybe they reserve even-number revs for developmental or OEM purposes? Who knows.)

 

[ChineseDemocracyGNR]

The newer boards designed by ECS are not bad at all. For socket A, the KT600-A, N2U400-A and 741GX-M are very popular motherboards in my country and they're not problematic at all, like the ASUS A7V8X-X is for example. And I own an ASUS board, so this is not biased at all.

 

[Brian23]

I've got a socket A ECS and it runs fine

 

[jiffylube1024]

I remember the old ECS K7S5A; just fixed an older machine with one of those in there. While you can say it wasn't a *buggy* mainboard, it was certainly picky. It did not tolerate crappy or even mediocre PSU's very well at all, for one.

It was definately a good deal for the money, but I usually stick to ASUS and Abit these days...

 

[Peter]

VirtualLarry, the K7S5A's well-known disliking of marginal PSUs and RAM isn't so much about engineering margins. On the PSU front, we must keep in mind it's an old design that doesn't yet use the separate 12V plug for the CPU power circuitry. Hence, it needs a PSU that is adequately strong on the 5V rail ... which happens to be where most of the cheap PSUs skimp. This also causes the "lost CMOS" case - this one is triggered by a 5Vstandby voltage dip in the power on phase, again a PSU weakness. Sure, augmented by the fact that this board runs the CPU off the 5V rail, but primarily the PSU not being able to keep its output steady.

On the RAM end, the board is quite intolerant against underperforming RAM for two reasons: One, the SiS chipset is very very fast particularly on the SDR RAM. Your DIMMs better REALLY be PC133 compliant, or you're in trouble, even if they ran fine on an older board with a slower chipset. Then of course, the four-socket arrangement does, by forces of nature, have an influence on signal integrity, again something that will not influence a good DIMM but break operation with a questionable one.

Nonetheless, I've run mine off a (good) 190W power supply - I had a Duron 600 in it initially - and with (good) PC133 RAM moved over from my socket-7 machine (also a PC-Chips board as it happens). No lost CMOS ever, no RAM problems, no premature death either.

 

[VirtualLarry]

Originally posted by: Peter
VirtualLarry, the K7S5A's well-known disliking of marginal PSUs and RAM isn't so much about engineering margins.

Sure it is. Proper noise-filtration/"smoothing" circuitry costs extra money, both on the mobo and in the PSU. If you leave them off of both, then you're going to have problems. (Unfortunately an all-too-common scenario though, with customers looking to pay as little as possible for components.)

I'm not all that happy that mu current MSI KT4V-L board only has a two-phase VRM for the CPU power either, and is missing a lot of filtering/reserve-power caps between the PCI slots, even though the board layout has spaces for them. :| So I guess I'm not saying that ECS or PCChips is the *only* company in this biz to cut costs/cut corners with components, they all do, to some extent or another.

Originally posted by: Peter
On the PSU front, we must keep in mind it's an old design that doesn't yet use the separate 12V plug for the CPU power circuitry. Hence, it needs a PSU that is adequately strong on the 5V rail ... which happens to be where most of the cheap PSUs skimp.

Considering how many similar system of the time ran fine on those same wattage PSUs in different boards, I can't quite agree that the problem was simply ample power output. More than likely, it was due to the fact that with a much more "beefy" PSU, the power draw from that system was so light in comparison, that it generated very little noise on the PSUs outputs (because that generally increases as power draw approaches maximum capacity, as I understand it), and was required due to the lacking of adequate noise-filtration of the power on the mobo itself.

Originally posted by: Peter
This also causes the "lost CMOS" case - this one is triggered by a 5Vstandby voltage dip in the power on phase, again a PSU weakness. Sure, augmented by the fact that this board runs the CPU off the 5V rail, but primarily the PSU not being able to keep its output steady.

I don't see how a PSU being able to keep a steady output should be a factor at all in losing the battery-backed CMOS data. Usually there's some sort of voltage-comparator/'pwrgood' signal that controls whether or not the CMOS memory is powered by the battery or the PSU. That sounds like an outright defect there that the mobo would be suceptable to that. Btw, considering that CMOS batteries are only 3v, how would dips in the 5v line cause that problem anyways? Hmm.

Actually, speaking about the 5vSB, now I remember some other sort of thing related to that on that board. It had a jumper for controlling whether keyboard/mouse soft power-on was allowed, and that chose whether or not the PS2 port (and USB?) was powered by the 5v or 5vSB lines. (My 635T-based board had that jumper too.) I seem to recall various guides suggesting to have that jumper one way or another to alleviate the issue somehow. Your right, I do vaguely remember some alleged relation between that jumper and the CMOS issue, but I cant recall nor fathom quite how it all fits together. Maybe 5vSB power isn't controlled by the 'pwrgood' signal, whereas the regular 5v power plane is, and that "noisy power" caused the issue, if that jumper was configured improperly? That's just speculation though.

Originally posted by: Peter
On the RAM end, the board is quite intolerant against underperforming RAM for two reasons: One, the SiS chipset is very very fast particularly on the SDR RAM.

That's why they invented BIOS-controllable memory timing parameters like wait-states. I suspect that the problem is probably more related to noise in the power/ground planes or the DIMM power-supply circuitry. Good boards have a seperate VRM for the DRAM array, does that board?

Originally posted by: Peter
Then of course, the four-socket arrangement does, by forces of nature, have an influence on signal integrity, again something that will not influence a good DIMM but break operation with a questionable one.

But you still only would be running with only two of those four filled, that's not that much loading on the memory signal lines. I still suspect board noise issues. Isn't that only a 4-layer board? My 635T-based board with similar layout seemed like it, it had far too much "board flex" for my liking.

Originally posted by: Peter
Nonetheless, I've run mine off a (good) 190W power supply - I had a Duron 600 in it initially - and with (good) PC133 RAM moved over from my socket-7 machine (also a PC-Chips board as it happens). No lost CMOS ever, no RAM problems, no premature death either.

Then again, some of the reports were that some boards were "good" in those regards, and some were not. Perhaps you just got a "good" one?

My view of PCChips is that they are like the "McMotherboard" company. Cheap, popular, but may give you digestive problems from time to time, if that's your only diet. More or less, it won't kill you either, at least in moderation.

 

[Peter]

I won't re-quote, that'd be a mess.

CMOS loss: The one thing to remember is that ATX boards don't run the CMOS off the battery when they're in ATX-off state. You'll be running on 5V-standby rail in that state. Now at poweron, some PSUs show a quick brownout glitch on that rail when the CPU starts sucking power off the main 5V rail, and that is enough to trigger the CMOS RAM chip's battery status flag. When I built systems using the K7S5A, I had quickly identified the one PSU brand that I better not use, and didn't see the problem again.

RAM reliability: The harder the chipset pushes, the less relaxation time do the RAM chips have, and the warmer they do get. Cheap RAM, particularly in the late days of SDR, often was relabelled B-grade stuff from questionable or unidentifyable origin, possibly binned by a major manufacturer because it wasn't up to spec. On a P-III or socket-7 mainboard where everything's slow and easy, people wouldn't notice. I've seen things you wouldn't believe - like a "128-MByte" DIMM that actually had 512 MBytes of RAM on it, with zero-ohm resistor bridge arrays to pick out the working 1/4 of each individual defective RAM chip. Needless to say, K7S5A didn't like it ...

RAM slots: Yes you populate only two. But you still got trace length and parasitic capacitance of all four slots ... and a SDR-DDR combo layout always is a compromise between DDR and SDR requirements. And yes of course it's a four-layer design, like everyone else's on socket-A desktop.

Quality: I didn't have just one. I used one myself, and I configured about a dozen systems for other people that had the same board. Nothing but happy users, zero dead so far (users and boards). After that, I've been using the even cheaper but DDR-only L7S7A2 with equally solid results, and now am warming up to the 755-A2 to get the budget sector off dying socket-A. I'm not buying off questionable retailers, I'm not using $20 case-PSU combos, and I will never again buy off-brand RAM. From my 20 years of experience, these are the rules to stick buy ... there has hardly ever been any mainboard that I'd actually label "unmanageable". I've been using PC-Chips since 1997, so I know what I'm talking about

 

[her34]

Originally posted by: Peter
I won't re-quote, that'd be a mess.

CMOS loss: The one thing to remember is that ATX boards don't run the CMOS off the battery when they're in ATX-off state. You'll be running on 5V-standby rail in that state. Now at poweron, some PSUs show a quick brownout glitch on that rail when the CPU starts sucking power off the main 5V rail, and that is enough to trigger the CMOS RAM chip's battery status flag. When I built systems using the K7S5A, I had quickly identified the one PSU brand that I better not use, and didn't see the problem again.

which psu brand?

 

[Peter]

The one PSU that reproducably wouldn't work with the K7S5A (triggering the "lost CMOS" problem) is the 400W PSU found in "Topphone" cases. Can't remember the brand, I'll look it up next time I see one. Strangely enough, the 300W flavor worked fine (with adequately slow CPUs fitted, of course).
YMMV, if only for the fact that we in Europe get different PSUs than you in the US.

 

[Steve]

Over here it's an Allied 350W PSU in the TU-124 case, and an Allied 300W (usually a 250) in its micro-ATX sibling, the TM-124. Those seem to be just fine. You're not getting Codegen units, are you?

 

[Peter]

No. I never do ... just one of my customers had one already before they came to me. It was unstable, voltages way too low even with the K7S5A and Athlon-1000 they had. Sent them to their original shop for warranty replacement, the 2nd unit was noticeable better, and still runs the K7S5A, now with an XP2400+ and a Radeon 9600. It's fairly marginal though, and in no way lives up to its 350W sticker.

sm8000, that's the case exactly. Over here they ship with 400W units ... I've built a fair few systems, but for the life of me can't remember the PSU brand.