Technical support by email?

Resources, guides, discussions for modding your system.

Postby atang1 » Wed Oct 09, 2002 2:27 am

In technical support for memory faults, common mistake is assuming manufacturers are unreliable. That you can only trust Crucial by Micron is as bad an assumption as we can get. Micron in Idaho is run by potato farmers. What would they know, is the same kind of prejudice even if it is true? In the semiconductor industry, Micron is not an inovator in technoogy. Micron is now one of the leader, by buying out their competitors. they use Samsung parts as well as other manufacturer's parts to supplement their inadequate production capacity in Idaho.

Memory incompatibility is the real problem in this would. single sided or double sided, 2 clocks or 4 clocks? Look at the pinouts of Dimms and you can see the D(data bits) and A(address bits). If your motherboard did not have wires in the socket connected then you can not use 512mb modules. So, always check what the motherboard manual had as memory suitable for use.

Memeory faults when you run operating system with internet browser, are mostly synchronization problems of wait state in the software. Most of the older software were not correctly synchronized with the faster cpu frequencies. It will not show up in diagnostic software which runs on the operating system only. Changing IE5.0, 5.5 to IE6.0 should solve the memory faults problem.

Some of the motherboards got bent and the memories would not function. Dimms are made with different thickness of printed circuit boards. The earlier sockets were made to use 0.45" thick modules. The 0.60" thick modules inevitably bend the motherboards and separated the traces from vias. This usually haqppen when you pushed too hard into the motherboard inside the computer case. Never push too hard. Bending the motherboard is a fatal mistake whether you lose the momory function or IDEs or card slots.

Simms have to be installed in pairs, Rdrams or rimms too. Because of the speed of rdram, termination is important for it to work fawlessly.

While DDR does not need to be run in pairs, the speed of FSB is now 533 mhz, termination becomes a good practice in engineering. Fill up all the sockets is terminating the end of the memory traces.

ECC slows down the computer, and can only correct the wait state errors, if you have that problem. So, buy the cheapest memory is still the best investment. Micron is never the cheapest, remember the days when they tried to cartel the simms? For many years. Micron held up the memory price by contracts, until 1998. Enjoy the low prices of today, scale up your memory capacity. You will enjoy your computer more.

Remember always use the fastest speed of the chipset for the speed of your memory. Otherwise your computer will not boot.
Last edited by atang1 on Wed Dec 18, 2002 2:43 pm, edited 2 times in total.

Postby atang1 » Thu Oct 10, 2002 1:01 am

Other than motherboard getting bent, cpu incompatibility probably needs the most technical support. You must buy a suitable fan for your cpu. The motto used to be buy the largest that fits. Today, it is "not too hot, not too cold"; in order to overclock without increasing core voltage but with decreasing core voltage.

Manual of the motherboard tells you what cpu is supported. This may mean the core voltage is limited. Or the cpu socket does not support the newer cpus which requires some wiring alterations. Then in selection of your cpu, the motherboard bios must accomodate the product of multiples and chipset FSB frequency. ie., a motherboard that is limited to 933 mhz can not use a cpu of 900 mhz. the multiples allowed can not do 9.5, but can do 7.0x133=933. Some times we do get mislead, if we don't know the bios limitations.

If those conditions can be satisfied, then a search on the manufacturer's website, on the fixes of the latest bios should reveal whether you can download an updated bios to run the cpu of your choice. You have to use a cpu the motherboard supports to get the bios updated. Or you have to buy a bios from Unicore or Badflash. They are other more desperate methods, such as hot swap on another motherboard(which I invented); but that is another story. Or use another motherboard and use the recovery method to update your bios, assuming the bios does not work in that motherboard. See manufacturer's recovery method before you try it. Both eeprom or flash has to be the recovery type and the same voltage.

Clear cmos, so that the cmos settings does not kill your cpu. The manufacturer may have left the cmos with the cpu defined, which has a different core voltage. Or a different FSB and multiples. Any of that, the PNP bios can change, but you are taking a chance that bios is too long and it may take too long a time resulting in the cpu being damaged. Slot1 or slotA cpu has to be seated down all the way, and click into the retainers.

After that, you may want to invest in a post card(diagnostic card), made by Startech, sold by Proadvantage(least expensive and has the latest post codes). The post card will show the post codes as the motherboard started to boot. If it stalls, immediately shut off the power, to save your cpu. Beep codes or lack of it, due to cpu failure, may be too late to save your cpu. Always remember the post code, which is the failure mode that you have to repair or exchange the failed part.

If your cpu runs, go into the bios and optimize all the variables. You are on you way to become an expert.
Last edited by atang1 on Sat Oct 12, 2002 1:01 am, edited 1 time in total.

Postby atang1 » Fri Oct 11, 2002 12:27 am

You would not believe that floppy drive could have any problems, but they do.

The most frequent is no read problem. It is caused by magnetic powder left on the heads. Common fiber glass disk to clean heads is totally useless, with or wothout alcohol solvent. The Chinon CA-350 green silicon carbide disks are the only one that can remove the ferric oxide or chromic oxide powder, running it dry. It can only be used 15 times, before it clogs, but you can extend its life by using an eraser to clean off the silicon carbide surface. After the reidual powder has been removed, the floppy drive must do a write on disk to erase the residual magnetic field left on the heads. Almost all the floppy drives with no read symptom were rejuvenated this way. Newer 3M floppy disks have aluminum oxide powder imbebed in the chromic oxide, so that the disk cleans the heads while it is being read or write. Avoid using floppy disks made before 1985.

Then there is the problem of overclocking FSB, where the floppy frequency(clock) maybe changed, the only way is to avoid overclocking on your motherboard..

When newer Mocrosoft operating system boots, it needs a disk in the floppy. If you have an emtpy floppy drive, it will exercise the head movement. Eventually worn out the floppy drive itself.

So much for floppies. It should be obsoleted many years ago, most software are too long for the disk capacity.
Last edited by atang1 on Fri Oct 11, 2002 1:16 am, edited 1 time in total.

Postby atang1 » Fri Oct 11, 2002 1:04 am

Mean time between failure on hard disks had been improved to 230,000 hours. And if you keep it on, it should not fail for 26.25 years. Unfortunately, turning it on and off leaves electrical arcing spikes on the magnetic surface. ATX power supplies are much easier on the hard disks. Scandisk can sometimes remove the magntic spikes. If not, you might have to repartition the disk to avoid using the sectors which are no longer usable. Otherwise, you have to replace your hard disk.

Another problem is the data transfer mode. PIO mode 4 can do 16 mb/sec transfer. UDMA started using double date transfer using two clocks sync to leading edge and trailing edge same as DDR memory scheme. But then, UDMA 33 and 66 had many inconsistancies, and you have to check the hard disk manufacturer for bug fixes. When ATA 66 and up started to use higher frequencies, the disk ribbon cable has to be shortened, or parallel double cables had to be used. Round cables are now preferred.

As hard disk capacity increased by 60% each and every year. The data compression used to do most of the work. Today, the heads are much smaller and magnetic disks have special coatings to improve read and write characteristics. What this means is the head gap to disk surface is much tightly controlled. Head expansion of the gap creates no read or write until it cools dowm. The same tempersture problem also extends to the semiconductor ICs(DSP and embebed flash). Cooling for hard disks becomes a necessity.

Hard disk has to be assembled in clean room condition. It dust gets into the hard disk, read and write will be interfered. the error(or dust particals) will move around on the magnetic disks. Room temperature change forces the design of the hard disk to allow breath in and out of the hard disk case. This is done by filters covered with a scotch tape(disk), Or a sootch metal coated paper tape is used around the seam. In any case, another layer of scotch tape will cut down the dust intake further extending the life of the hard disk from dust. If those adhesive tape broke, some dust will get in. Adding scotch tape, then run the hard disk day and night, off and on, will reduce the dust inside the case thru temperature differential breathing. Osmosis of dust if you will.

Good luck with you hard disk, it is $100.00 worth of your computer system.

Postby atang1 » Tue Oct 15, 2002 12:58 am

One of the most important part in the computer is the dial up modem. Earlier modems are straight forward duplex up to 33.6 baud. Then came the pwm(pulse width modualation) scheme, and troubles began with all variations of host modems. In general the modems are depending on the cpu to function, instead of the straight forward modems which only uses a serial port to communicate.

Pwm is audio in nature, and some of the modem uses sound chipsets, such as pctel, stc. Others uses DSP chipsets such as Lucent, etc. Then the cpu of a minimum speed of 166mhz takes over the modem operation. These host modems are now transparent since the slow cpus obsoleted some time ago. Any cpu of speeds above 300 mhz has no problem using host modems. So, technical support has more to do with drivers then speed of cpus.

There are so many modem manufacturers that specific drivers must come with the retail box. But, fortunately the chipset manufacturers do offer genriac drivers which covers most of the modems. On Pctel drivers, they allow configuration drivers to patch different modems. If you want a fastest dial modem, pick a modem which uses the shortest driver. Pctel only uses 700kb of codes. the rest of them modems use 1mb or longer drivers. If you are a collector of modem such as myself, do obtain a CA(computer associates)Bitware for Windows driver software for all modems(copyright2002 is the latest)

Often, people get slow connections. The first concern is noise. Remove all devices attached to the phone lines. Anything noisy will slow down your modem. then add them to see if speed slow down.

Modem card also needs low temperature burn-in. Drive off moisture in the chipset. Cure capacitors which had been dried up. These generally takes two weeks of continuous power on without intensive usage.

One of the component that fails is the relay which turns on the modem to the telephone line and gets the dial tone. No dial tone means you have to change the relay. There are two kinds. One is miniature relay(rectangular boxy type), which has a 50,000 cycle life. The other one is reed relay(long and slender shape), which has a life of 30 million cycles. The erosion on the contacts cause noise, after many cycles of use(on and off cycle). Of course, thunder lightening will destroy them, because the gap of contacts can only stand 200 volts. But reed relay at Radio Shack is only less than $3.00. Miniature relay is slightly more expensive. When they become noises its time to get a new modem or new relay. Its your choice.

If you can not get speeds of 50,667 baud or better; you are living too far away from the central office. The telephone line is too long with or without repeaters. Today, most urban telephone operations are digital, so you are only concerned about how far is the pole with the electronics. Most modems have input filters of inductors and capacitors to compensate for line length. They(line filters in modems) are no longer a consideration(speed) for choice.

Is Kflex or V.92 a choice over v.90? Not necessarily. Kflex has the shortest handshake, and will connect 50,000 instead of 50,667. V.92 has more features in drivers and will slow you down some(by waiting and looping to activate some functions).

The future of dial up modem is not switching to cable modem, which is just an ethernet card connected continously. Hubs and swtchers add up cost quickly. The future is submicron technology with multiple shot guns and collated data(ADSL or better) thru pots(plain old telephone system) by becoming cheaper and better dial up modems.

This technial support will keep you happy until then.

Postby atang1 » Thu Oct 17, 2002 2:27 am

What every one fears most is no video, no beeps. And the problem is never a fault with the video card, or the monitors. Some, video cards and bios will beep if monitor is not connected to the video card.

First video cards are now PNP(plug and play). Meaning it can be hot swapped. Most people like myself would not risk it, and always shut down power before seating or unseating the cards. Monitor cable can be connected while power is on, except when you are using some monitors with automatic functions depending on the video card you use. In which case the monitor should be turn on after the computer is turn on. If you turn on the monitor first, it will just search for the video signal and go into a loop. Older monitors don't care what you do, no automation at all.

No video and no beep of course is a problem of either cpu or motherboard is dead. Some times it is the memory too fast or too slow.

Now to the video cards. Older video cards need at least 2 mb to get decent color and resolution. Because of the lack of video memory, it is fast for single page display. Larger memory only caches multiple pages, which you can store in your large sdram or DDR. If your video card does the rendering on board with large memory, then it is going to be slower. Most of the rendering is now done by the latest cpus. PCI is slower than the AGP by the bus speed. AGP2x is 3.3 volts; and the 4x video cards are incompatible with 2x motherboards and may even be damaged.

Common video card failures are memory defects, and chipsets getting too hot. In technical support, some times, we use our hand to feel the temperature of the chipset. But never wear wool clothing to zap the computer. Always wear cotton. Believe it or not, cold semiconductor is also trouble. Coupled wth no vidoe, cold or hot semiconductor; you have to try a known good video card to isolate your video problem.

All video cards are made of two parts, VGA and more sophiscated video chipset. The VGA bios will work first, then the operating system will recognize your video card and install it. You may need a specific driver to take advantage of all the benefits of your video card. Many of us may have misplaced it, and we have to search on the internet to download the proper driver. When your VGA bios fail, you may be able to download a VGA bios which will reside in the memory.

What about all those newer more complicated video cards? What about them? They are not necessarily faster, depending on the operating system you use. Whether your game needs a minimum of 32mb on the video card or not, you have to comply with the software design.

If we believe all computers are backward compatible then I am still using S3 True64 PCI video cards made for IBM. The AGP video cards are not any faster that I can detect. Most larger memory video cards actually are slower because they load up the entire memory and it takes time to transfer the data.

Next, we will look at the problem of overheating in your computer? Your computer may become flacky. And computer cases do store heat gradually. What to do, what to do?

Postby atang1 » Fri Nov 29, 2002 8:22 pm

Recently, some one complained to me that a brand new Intel D815 mothrboard did not recognize any drives on the secondary IDE channel. The first question was if it had been installed in a computer case. We all know that insertion of cables may bend the motherboard to damage the traces leading to the IDE connector.

But what is most often mistakenly forgotten is that the cmos clear on ATX motherboards must have the power supply plug pulled from the wall, before it will work. Without cmos clear, the bios may have been set with quick boot or quick POST enabled. Quick boot or quick POST does not recognize any new devices, they will stick to the items in the cmos. Many a cpu could have been destroyed without cmos cleared; if you change from a ppga cpu to a fcpga cpu, when cmos supported 2.07 core voltage instead of the 1.65-1.75 volts.

So, if you don't want to clear cmos, when you experiment with new devices or components, change the quick boot or quick POST in the bios to disable. When you try a new motherboard, clear cmos is the only safe thing to do. Do it without the ATX power supply connected. Many motherboards came with some cmos info that you do not want.

Postby atang1 » Wed Dec 18, 2002 2:28 pm

Failure mechanism for hard disks is three folds. Ic failure when they become too hot. Or the air in the hard disk got too hot. Thermal expansion of the head space between the media surface is much trickier. Or bearing wear creates problem with head space to media. You could hear the bearing wear noises though. Heat problem started with 2.5 gb hard drives due to miniatured heads and more advanced analog data compression schemes.

In the setup of WinMe operating system, the hard disk is used fairly intensively. As the files are being decompressed and copied to the disk the heads left the media. Between 10% and 13% done, the transfer stopped, waiting for the temperature to go down. On Maxter drives, I put a cooler over the pcboard on the drive. On Seagate, I put a cooler on the topside of the drive and it cool the air inside the drive. Both waited a few seconds and pickup the speed to transfer the files and finished in good time.

Unfortunately, the coolers are made to fit over the pcboard in the drive. Many drives needed the cooler to be on top of the drive to cool the air inside the drive. You may have to glue it in place?

There are many drives that some parts of the disk is not readable anymore, as you start them up. Is it because they have part of the disk not near the heads anymore. Try a cooler on top and a cooler on the bottom? Maybe you will get luck? This is because, the heads no longer can gouge out any part of the disk media by design. But thermal expansion or bearing wear still play havoc. Can we do anything with bearing wear? Not likely.

Postby atang1 » Mon Dec 23, 2002 6:27 am

Why is visual inspection the number one quality control for the United States government?

It is because most problems started when visual uniformity is missing. Anything that deviates from the usual is a suspect of quality control.

When things go wrong, do a visual inspection.

An Abit motherboard failed, visual inspection shows two leaky electrolytics. Most cpu bypass electrolytics bulge on the top indicating overheating. Miniature electrolytics has a life of 3000 to 5000 hours only between failure.

A Pcchips motherboard failed, visual inspection shows a hole in the mosfet bypass transistor. replacement is needed.

Another motherboard had no keyboard. Visual inspection on the bottom of the motherboard shows the blown trace leading from the keyboard connector. Soldering with a bare wire connecting broken traces solve the problem.

Visual inspection sighting(along the edge) on the motheboard will show the flatness problems in wave soldering of the motherboard. Card board trick may be needed.

Visual inspection will show the contacts in socket or slots touching each other due to abuse. Use a tool to separate them.

Visual inspection also shows IC pins bent and touch each other whuch can be separated by single edge razor blades.

Visual inspection of card shows damaged or missing gold fingers. Use printed circuit repair kit, glue copper foil and tin it.

It goes on and on, all you have to do is visual inspection.

Postby atang1 » Fri Feb 14, 2003 1:06 pm

What to do when WinME or later operating system would not recognize the cards or components you installed on your motherboard?

Clear cmos or disable quick boot or post in the bios? Still it may not recognize even if drivers had been installed? So, you have to remove the uninstalled card or component. Start the computer to clear all the software recognition from the operating system. Then you reinstall the hardware. Then the operating system will search all the new PNP devices and install the drivers already in the hard disk.

The concept is that bios and operating system, both have to be cleared before a new installation will work. If the operating system had used a wrong driver, you have to clear it by removing the card or component from the motherboard first. Operating system register will then recognize the new PNP device.


Return to System Modification

Who is online

Users browsing this forum: No registered users and 1 guest