build your own PC parts list and how to
monitor brightness, brands, raid, cmos latch-up and other matters
you can build your own pc or upgrade at quite a reasonable cost and actually get what you need
the trouble with macs is you are usually paying at least twice what an equivalent pc costs and only getting half the function
to build one, you need a static control wriststrap, screwdrivers (but keep any magnetic tipped ones away from a rotating hard disk drive)
the world has changed, incompetence is the norm and its so hard to buy value for money services these days
what follows below are somewhat meandering notes over several builds of a low cost, low power, quiet, reliable and reasonably performing system
ssd's don't like power interruptions so in my view uninterruptable power supplies are essential
the utility of maintaining an OS swappable clone
gaming is a big timewaster tho imo, messageboards are a more subtle, useful and social skill building form of warfare and puzzling
why game which is just a consumer activity when you can be doing something real learning how to write software, web pages or learning the in's and out of the operating system or hardware?
if you don't need gaming keep well away from it, higher end gaming ability can double the hardware cost, triple the power supply requirements, induce software and operating system problems, and reduce reliability
in my view is it's a menace and there are a lot better things to do with your time
“ a little sleep, a little slumber, a little folding of the hands to relax, and your poverty will come like a robber, and your need like an armed man ”
young men are entering the armed services too readily with little appreciation of the reality of battle field injury gennady dobrov
i don't see those sort of injuries on call of duty, the bodies miraculously integrous as they get blown up !
in other words it's all bullshit, make your stand against it and boycott this sort of crap !
wonder if they
beware spyware doctor, 'pc tools' in ireland who make spyware doctor and do stupid things like you have to ring them (061 715900 shannon, clare - iercom white pages) to get your subscription stopped or they will automatically renew your subscription and debit your credit card without asking permission, and i think the account they claim to set up is your name is phoney
article and study on the susceptibility of anti-viral packages to transformed viruses !
investigate dr. web as having static analysis system and able to cope with transformed viruses !
computer hardware repair or diagnosis is just a question of narrowing it down to the likely area and swapping or leaving out a suspected bad item and swapping in known or likely working items, the hard disk/ssd is particularly easy to do this for
that's all they do when you send it away or drop it in for repair
of course the skill is in narrowing it down and not making more problems in the install
see also my computer maintenance page
checking out the bios gives a relatively clear demarcation between the hardware and windows
the likely order of failure is ssd's way first, next, the hard disk and motherboard about equal, followed by the ram
cpu's seem pretty reliable if heatsinked adequately
being able to depower or short the cmos ram/rtc or removing the coin battery is a useful skill
turn the power off to the computer first tho leave the battery out for several minutes
if you are continually needing to do this, the motherboard needs replacing !
also i am not keen on more than one OS per hdd/ssd, better to select which OS/ssd/hdd to boot from in the bios rather than multi-boot !
over the years i also learnt for cold boots to boot first to the bios and let it idle for a couple of minutes so the chips can warm up before doing the heavy work of loading the os and have never had a cold boot problem since !
sleep works much better in this respect because the ram, cpu and bridges? keep warm in this mode !
also i disallow hybrid sleep so that the ssd is not worn unnecessarily with a mirror of the ram being created !
an under appreciated smart parameter is "unexpected power loss count", because when that happens the mapping table is rebuilt so you don't want too many of them
POR on the samsung ssd's?
"I've been looking back at the smart data every day, and I think that 235/EB (ed. POR Recovery Count, samsung ssds;, same as unexpected power loss count on micron/crucial ssds?) is basically unsafe shutdown count. I went and back and looked at the SMART screen pictures, and for the two weeks I was gone at the end of December it was at 18. Then the day I did the FW update it increased to 31 from 21 -- I roughly had to cycle the drive by unplugging it about 10 times that day."
in windows 7 & 8 the hard disk power off setting in advanced power management wants to be "never" to avoid unecessary power offs to the ssd
1. notes for intel upgrade, have used an asus 97M-E with a pentium g3450 and noctua nh-u12s heatsink all working really well! (may 2015)
hdmi and dvi easily interconvertable with a cable, however display port to hdmi needs a proper converter!
“ i buy asrock but their bios support, update rate, and memory support is not on par with my gigabyte boards ”
there is no advantage to msata, and maybe a disadvantage as the smaller form factor forces denser flash chips !
both msata and sata go to the sata controller !
old gigabyte amd GA-MA785GM-US2H motherboard
NQC offers no advantage for desktop performance
i removed the north (aprox 75mm hole to hole, and 40 x 60mm excluding the pins) and south bridge (53.5mm hole to hole) heatsinks, removed the rubber washer heatsink on the southbridge, removed the existing heatsink compound with cotton buds and isopropyl alcohol, and spread a small amount of arctic silver 5 on the bridges with a plastic knife/plastic flosser before putting the heatsinks back
if the arctic silver has not been used for a while, the very tip material may be a bit harder and should be discarded
i took the clear plastic backing and the black fur off the northbridge, but the adhesive holding the clear plastic backing is too difficult to remove, i even tried lithium hypochlorite
the southbridge was worth doing, i had to do the northbridge anyway since i broke the glued joint removing the gigbyte name plate, the northbridge is probably worth doing as well
the great weakness of these heatsinks is they radiate back onto the components on the board under them so any effort to lower the heatsink temperature is very worthwhile
as far as i can see the 780 and 785 chips have no temperature sensor in, temp 1(sb), 2(cpu), 3 (cpu?) (pc wizard, speedfan?) showed no change at all when i sprayed crc freeze on the northbridge
the northbridge gets too hot with the 785G chip doing the windows 7 aero workload at 500mhz (the 780G appears to be ok without the windows 7 aero workload) so have replaced the stock heatsink with a thermalright HR-05 IFX heatsink cooled with a scythe 80mm, 1500 rpm (1638 according to the bios) s-flex SFF80B which is very quiet
with this heatsink set up, the 785G seems to be long term stable at 710 mhz (800 mhz caused editpad to lose a file ?), which is good enough for midrange games and makes this system very viable, best in class for cost and reliability actually
there is a very big mhz gap between long term stability with editpad and graphics glitches
i had used a noctua NF-R8-1800 80mm fan (needs to push air into the heatsink and not suck out as otherwise it is too noisy), but it had a significant low frequency rumble/bearing noise compared to the scythe
without a fan and just the thermalright heatsink on, the computer crashed on the passmark jet planes flying around stone islands in a crystal clear sea scene 3D graphics test (northbridge overclocked to 1030 mhz)
i assume was this was either a thermal cutout or parameter skew on the northbridge
the HR-05 IFX has to be mounted at almost right angles to the mounting hole line which leaves the end over the PCIe sockets and the first PCI socket preventing any cards (graphics!) being put in there
a blog showing mounting on a 780g, bit different from the way i did it, i am not sure straight is best, though i have lost the use of the pci-e and one pci slot through it being at an angle
i used the ends of a longish tie to mark the depth of the hand nuts on the mounting screws so that both sides were tensioned similarly
i doubt the heatpipe designs that gigabyte has on some of its otherboards help much, it really takes fans
the arctic silver 5 heatsink compound requires successive hot/cold thermal cycles for maximum conductivity, using the S3 sleep mode should give that cycling, however it is capacitive though resistively non-conductive, only suitable for well controlled pads and heatsinks which limits it to CPU's on computers (good though for other heatsinking applications like heatsinking a solenoid coil on my washing machine !)
the arctic silver ceramique doesn't have the capacitance problem and may perform better or .8 of a degree worse ! in the short term
the thermalright chill factor heatsink compound is quite fluid and is ok for heatsinks that move (arctic silver eventually sets !) on their pads during their life, (1.15C worse than the arctic silver in the benchmark reviews 80 way test)
xbit labs thermal compound review
the polysynthetic arctic silver 5 needs time and cycling to "set" and so is usually unfairly compared in reviews
its goood stuff but requires extreme care in using; you want to just keep it on the heatsink, i smear it very thin in a square to about within 4mm from the edge of the cpu heatsink so it can spread a bit when the heatsink goes on
this review rates the nano oxia TF-1000 and arctic silver 5 as roughly the same
thick copper power and ground planes very significantly helps with heatsinking and reliability
ASUS H97M-E, may 2015
asus in an act of supreme stupidity have insulated their round PCH heatsinks by putting on a big label with a heat insulating double sided tape to hold it on . .
you can remove the heatsink and then take the label off which i did . .
a scythe 120mm S-FLEX 800RPM SFF21D fan with a screw in the top left hand corner (slightly inboard of the original hole) to catch in the horizontal NB heatsink rail and supported on the bottom by a wooden block holds position ok and seems to work well for the 780G
unfortunately the newer gigabyte boards have a different heatsink which doesn't look to have a suitable mounting point for the fan so will have to be mounted from the floor of the case, unless an old style heatsink is retrofitted !
cut down aluminium extrustion would be best for this i think, tho maybe a mounting hole can be drilled in the heatsink !
the fan has to be able to be removed so that dust can be cleaned off the heatsinks !
this 120mm fan cools both the northbridge and southbridge heatsinks, and may be the best all round solution, but the fan has to be right up close and touching the heatsinks almost to get the needed air movement !
it wouldn't bring the NB down enough if you are doing high overclocks of the 785G, but i have gone off high overclocks of the 785G as you have to bring the clock rate down as the board ages due to parameter shift anyway !
the s-flex fans are amazingly quiet!
pentium g3450 is working well, the g3440 is similarly spec'd with the intel "clear video HD" which eliminates/reduces streaming video jittering by hardare decoding and post processing !
" The funny thing about RAM is that Windows (ed. 8.1?) manages it heavily and sends things off to the pagefile when the limit approaches.
At work, I just upgraded from 8 GB to 16. When compiling, I saw RAM usage peak out around 7 GB before the upgrade. Afterwards, I regularly see utilization upwards of 12GB "
my comment: yeah i am finding because i can have up to 150 chrome windows open that 16GB (kingston hyper x) is worthwhile, keeps the ram standby expanded and so not working the SSD !
basically i think you want about 2GB of free ram in normal use, so "standby" isn't crimped (windows 8.1 uses it as a temporary store) and there's some free gigs obviously available for expansion so less look ahead trimming of "in use" and "standby" is necessary (my speculation !)
a good explanation of why disabling the page file is not a good idea !
more ram may also help with the "rowhammer" problem by spreading the access range !
1866 mhz allows for transistor switching speed degradation over lifetime use for 1600 mhz, though one could let 1600 fall back to 1300 as cpu's also degrade !
gskill more overclockable, kingston more reliable, i am using 2 x 4 gig of 1800 Mhz kingston hyperX !
ram failure rates
Kingston 0,20% (vs 0,20% before)
- Crucial 0,46% (vs 0,39% before)
- G.Skill 0,90% (vs 0,95% before)
- Corsair 1,08% (vs 1,18% before)
kingston is also easily the largest seller @59% in 2014 !
" Frequency is more important than timings when it comes to top-end LGA1150 configurations. DDR3 SDRAM kits currently available don't differ much in their timings but vary greatly in terms of their specified clock rate. And indeed, it is the clock rate of DDR3 SDRAM that affects performance the most "
kingston 1800mhz ddr3 hyperx timings from adia64 engineer beta 19/7/15
@ 933 MHz 13-11-10-30 / 42-243-5-15-8-8-26 (RC-RFC-RRD-WR-WTR-RTP-FAW)
@ 933 MHz 11-11-10-30 / 42-243-5-15-8-8-26
@ 933 MHz 10-11-10-30 (CL-RCD-RP-RAS) / 42-243-5-15-8-8-26
@ 841 MHz 9-10-9-28 (CL-RCD-RP-RAS) / 38-219-5-13-7-7-23
@ 747 MHz 8-9-8-25 (CL-RCD-RP-RAS) / 34-195-4-12-6-6-21
@ 654 MHz 7-8-7-22 (CL-RCD-RP-RAS) / 30-171-4-10-5-5-18
1.65 volt DDR memory is usually cheaper because it wasn't as stable as needed at it's advertised max speed when ran at 1.5 volts. So, higher voltages are needed to make it work without errors at it's rated speed.
Running memory at 1.65 Volts also means it's drawing more power and generating more heat.
" Now, you may find 1.65 Volt memory that can run fine at 1.5 Volts if you don't OC it; and vice-versa; and motherboards may let you run it at other voltages. But, I'd stick the voltage the memory controller in the processor was designed to run at (1.5 volts in the case of the Core i7 3770 you mentioned wanting to go with) "
" I've ran 1.75v for DDR3 31xx through my 4770K while benching with no real issues "
use memtest86 to test ram timings, disconnect the windows boot ssd as it may corrupt !
i think just about all modern ddr3 will fail the row hammer test (#13), the kingston hyper x i use certainly do, but i haven't had a program crash yet !
ddr4 enables a memory controller work around for this problem !
" Source: dram engineer : Just an FYI to everyone running row hammer on their dram, you are permanently damaging it when you do so. If your dram is over clocked/ over volted especially so
the vulnerability does not get easier to exploit on that specific cell each time you hit it
The row you are hammering gets electrons trapped in its gate oxide. It makes the affected transistors more difficult to toggle; slowing access to that row. This is how flash memory works, and why it has a limited life span
Using anything degrades it over time. Uneven extremely hard usage has odd wear patterns. Commodity dram has a 10 year lifetime. This lifetime assumes somewhat even wear. There is no non malicious reason to do a row hammer. When you do it you are opening and closing the same row ~100,000 times a second. Do that for 10 hours and you have toggled the row 3.6e9 times.
This will prematurely age your dram
Row hammer is the easiest way to force channel hot carrier stress on dram. This will slow access to that row primarily, but will have lesser effects to rows that share the same circuitry. If your dram is on the higher voltage side and cold (cold = 0c and hot = 85c) it will accelerate the stress. Do note this is really only an issue if you are running a row hammer overnight. This will not make the row hammer worse "
overclocking frequency plus voltage plus timings
instructions for replacing/upgrading ram:
probably quite a good time to install the new ram is while its raining, if it's not you will need an anti-static wrist strap
the label on the little boards (if there is one) needs peeling off the heatsinks so the heatsink works better - this may void the warranty tho, your choice! i took the label off, but i had tested the ram before doing that
turn the comp off, pull the cables out of the back, taking care to remember which goes into what hole
then put the comp down flat on its back and just replace the memory boards in the computer
they only go one way around, if you look carefully you can see the slot is slightly to one side
and the boards need to be pushed right in so the levers on the side click up and in, the newer sockets with a one sided lever are quite tricky, you need to push down on the non lever ened of the module so it places, then push the lever end down of the module so the lever clicks into place !
removing ram heatsink with ruler and blowtorch
removing heatsink with heatgun
use pliers to handle hot heatsink ?
" From my personal experience (Are RAM coolers really effective?) it depends on the construction of the heatsink, plain plate of metal just closing memory to box is not that much efficient as it more prevents air from getting heat out of board and chips.
Ones used for example by corsair dominator series are much better and can bring temperature down more than 10 degrees.
In my case it was diference betwen ~60 and 40 degrees i have now (and new memory even run on higher frequency) "
4. cpu heatsink
noctua nh-u12s works very well !
Be Quiet! pure rock CPU Cooler, yet to try !
not making the cpu too cold is as important as keeping it from getting too hot since a cold cpu signals will ring more, especially if overclocked, i did buy a zalman 9700 (PWM) but it had an irritating low frequency hum/growl, much worse than the freezer extreme, this must be intrinsic to PWM
the scythe ninja 2 cooler and the noctua NH-U12P don't have PWM, but quiet, low constant speed fans, 1000 rpm for the ninja and 1300, 1100, 900 for the noctua and its two resistor cables
Cougar Spike 5SS7 Mini Tower Case
antec NSK4000 "super mini" = midi ? case. i have used foxconn and aopen but prefer the antec and my latest antec case (NSK 4480 bII) has an "earthwatts" 380W power supply in (elieen's), which is ok and maybe is quiet enough
most importantly the hard disk mounting bays use rubber grommets to free float the hdd's which seems to improve reliability !
6. power supply
Antec Neo Eco 450C-BR $72, actually seasonic, very good, holds up well in a power dropout, japanese capacitors, fan not quite dead quiet
Zalman ZM460B-APS 460W Noiseless ATX Ver2.03 / ATX12V, Ver2.2 20+4pin Power Supply, 120mm internal Fan, Dual +12VDC Output, without power cord
almost identical (some different semiconductors) to the silverstone strider ST50F
FSP Raider 450W 80 Plus Silver Power Supply looks good, 5 year warranty !
on the zalman power supply i have, the mains socket pins are not thick enough so i get intermittent contact (and a sparking noise) when i move the cable.
i have got around this initally by prying the metal socket sides in the power cable closer, but bought a belkin power cable which seems to have solved the problem. (belkin mains plugs and sockets are a tight fit)
this does seem to be a general problem with power supplies and cables, but i still think the mains power pins on the zalman are too thin.
the zalman has performed flawlessly and in fact the oem [ I II ] is FSP which have a good reputation, so maybe it's better to look for FSP aurum or another quality oem ? !
this 460 watt zalman is very quiet with a ball bearing fan and the fan speed is temperature dependent.
Antec TruePower Classic, FSP Aurum S ?
enermax is good, but more expensive, their nominal power ratings are real though which is more that can be said for most brands
also it has a single horizontal ball bearing fan which lasts a lot longer than vertical fans because of the even weight distribution on the bearings
a nominal 350W power supply is not have enough to hold up in dropouts, 350 nominal in practice might be 250 in reality, like 460 nominal might be 350 in reality and has worked well.
in fact this computer probably only draws 150W max, but i like a bit of spare wattage (without being excessive) in a power supply to hold up over momentary dropouts, which this zalman does well
150 as 25% of load capacity = 600 watts, 150 as 30% of load capacity = 500 watts
150 as 40% of load capacity = 375 watts
7. my views on ssds and hdds being complementary !
Samsung 840 pro Series 128gb, samsung is much more reliable than crucial/micron!
i think the micron/intel joint venture flash memory chips have read problems, they use micron on OCZ and that may be a major factor in the ocz reliability problems !
my crucial M4 CT128M4SSD2 ssd has not really been ok, there is a slowly increasing "reported uncorrectable error" rate and a constant raw read error rate of 8 !
no special cables need for sata 6Gb/s
disconnect the sata cables at the motherboard (after the power is off) for any other hdd/ssd drives before the install !
don't reconnect the sata ports of any other drives until you have fully updated windows 8/x !
i think that two 128 gig ssd's are better than one 256 gig, since if one fails you still have the other and can cross back-up important data on each !
lower power dissipation per package too !
20/8/14, i turned the write cache flush back on, possibly helps localize the damage more with a power outage !
" if the write cache flush is disabled, this strips away write-through flags from disk requests and removes flush-cache commands "
19/8/14, lost a file (pp123) from a power outage with the write cache flush turned off, maintained size and name but all text written to 00, i presume the ssd was caught during a write from cache to the nand and the wrote the whole file back to 00 by way of housekeeping ! :o(
17/10/14 hmm ! lost a file (pp126) from a power outage with the write cache flush turned on, maintained size and name but all text written to 00, figure maybe the last file written to the ssd is always vulnerable, the ssd is not guaranteed to have a fully flushed cache despite the write cache flush being turned on ! or is it simply the 85 seconds it might take for the ssd controller to write the cache to the mlc nand memory ?
between losing those two files, there was one power outage with no file loss !
17/10/14 so have turned the write cache flush off again on both the ssd and hard disk, does seem to help with smoothness and speed a bit ? !
21/12/14 two power outages within about 7 minutes of each other, first ok recovery, 2nd (double out?) had file structure? damage on the ssd, windows need to do a restart repair, but i left the write cache flush off ! should have waited 15 mins to see if power flicked off again ?
first bought samsung 840 pro ssd is 16J, 2nd is 25N serial number endings !
dvd-rw drives have three motors in and represent a surprising source of potential power rail glitches and crowbarring during power outs so, since i hardly ever use it , i leave it disconnected, it was connected when the loss of pp123 occurred !
excessive "spin up" time is the likely future failure mode for the current samsung hdd i am using ?
"Spin up time describes amount of time it took to spin the platters up to their rated rotation speed (usually 5400 or 7200 RPM). Values above 80 should be considered good. Values between 70 and 80 are still acceptable"
3.5" seagate, ST1000DM003, 1 terabyte, 2 heads on a single platter, quiet; new egg: early failure rate, really not reliable enough, but same with wd blues and toshiba ! !
ST500DM002; 500 gig half the probability of head failure, since only one head on the platter !
or is that nearly the same as two heads have the same use/shock hsitory and one just fails a bit earlier on average ?
wait for toshiba/hgst to come out with a decent consumer desktop 1T hdd?
backblaze are currently buying the ST4000DM000
"Historically, Seagate drives have performed well at first, and then had higher failure rates later"
19/2/15, still the seagate as the choice as wd arm movement too noisy ? all there is ?
WD better raw read and seek error rates or just different algorithms?
500g green (5400rpm) WD5000AZRX ?
blue and black arm movement too noisy, but red ok as is 5400 rpm, like green ! ?
green has an overaggressive spin down/head parking causing wear, but idle time to head park can be delayed by wdidle3
bad sector problems ! ? new egg opinion is green is unreliable !
wd green no good with linux ?
has a separate 8 pin bios eeprom chip !
defintely not 2.5" which seem to have heads sticking to the platter problems !
i am using a Samsung 320GB 'HD322HJ' HDD - SATA II 7200rpm, 16MB Cache, NoiseGuard, SilentSeek, 3-year warranty, 8.9ms
the pcb on the samsung is likely to be able to be directly swapped without having to transfer a bios chip !
with a Seagate 250GB 'ST3250310AS' SATA 3Gb/s HDD - 7200rpm, 8MB Cache, 5-year warranty (barracuda 7200.10) 8.5/10ms random read write seek times as a back up
samsung offer 3 years warranty compared to seagate offering 1 year for an equivalent 8.9ms product which seems abit anomalous
after 16 months use on two computers without trouble in what has been a damp year, samsung has my vote of confidence ! (13th jan 11)
26th march 2011 (showed up on 7356 hrs reading), the first indication of trouble, the seek error rate dropped to a value of 100, from 253, the threshold is 51 !
it tested ok on the seatools short test so will keep an eye on it !
getting a touch rattly maybe i should replace it ! no the rattle is in the frame i think ! it is quiet doing CrystalDiskMark !
current predicited TEC is 19th aug 2011 (read on 28th march 2011) up from the 15th of august read on the 25th of march
i would guess that the servo mechanism has a bit of wear, the magnetic surface appears to be in good nick having only one reallocated sector that showed up on 665 hrs, according to a google inc. paper reallocated sectors are an indicator for disk failure, presumably the the magnetic surface is going
sequential read and write speeds for this disk on CrystalDiskMark are the same at 97 MB/sec
i am using the passmark disk checkup for the s.m.a.r.t. attributes
a good page explaining s.m.a.r.t.
exporting the s.m.a.r.t attributes from "disk checkup" to my files so i have a record to compare, i was able to see a degradation of performance in the raw values on a samsung HD252HJ and in fact it failed the sea tools short and long self tests, so i swapped the seagate back as the boot disk
the soft read error rate (D) went from 25 to 35 which matches the problems i had with casper with the samsung as the boot disk
the Raw Read Error Rate (1) went from 33 to 44
BB went from 50 to 60
"Hardware ECC Recovered" (C3) went from 66,826 to 734,181 to 2,687,678
i took it apart and am fairly sure the problem was corrosion on the connector leaves on the pcb mating to a connector going into the disk internal area and then to the armature
which was really disappointing as i think i could have fixed it if i had not done unecessary damage during disassembly
the internal area (head disk assembly cavity) is not gas tight, but airtight and dust sealed
so its best not to open it, but remove the pcb and check the connector leaves on the board
otherwise it all looked good
the samsung's are quiet and in fact about 5 % faster than the seagate
they disassemble easily and non-destructively, except for allowing dust in the head disk assembly cavity
delayed boot may be a good early warning sign for hard disk failure
boots should be instantaneous, otherwise the process is waiting on something
full surface scans from the seatools (windows and DOS) and samsung (DOS only) diagnostics are useful/definitive
seagate sound like they have real problems, tho i still have some 7200.10s i use for back up, their last reliable drive series before they switched their manufacturing from thailand to china, tho i think things have improved now ! (feb 2013)
9. Samsung S27B370H 27in Widescreen LED Monitor 1920 x 1080 from pc case gear in melbourne, screen and colours superb, no stuck pixels, only disappointment is the high gloss on the bezel and stand !
however running some 3M/scotch brand 19mm magic tape around the bezel seems to provide an adequate antiglare surface !
i had to drill some holes in the back panel to improve the cooling !
took the pdf manual from the disk and downloaded the latest driver and MagicTune version from their website! (13/02/13)
for 2560 x 1440, display port 1.2 (preferred, but 1.1a likely quite ok, just can't daisy chain monitors and has half the bandwidht compared to 1.2) or hdmi .14 or dual link dvi, ips? lcd, samsung S27A850D, asus ASUS PB278Q (hp, dell?)
display port 1.2 (preferred, but 1.1a on motherboard likely quite ok, just can't daisy chain monitors and has half the bandwidth compared to 1.2, ivy/sandy bridge supports 1.1, haswell 1.2)
for a second monitor i use a samsung S24D300H : HDMI, LED, 1920 x 1080, excellent price/size/performance combination !
LCD Monitors IPS vs TN Panel difference video
a good video showing the trade off of degraded signal versus high frequency suppression for ferrites on computer cables !
10. power surge protected mains and modem plug box, 3500 joules seems to be ok
11. laser mice are better than the ordinary optical, have usefully more screen precision
i have a logtitech LS1, works well, haven't used the logitetch driver and doesn't seem necessary and logitech say their driver has to be uninstalled before installing windows 7, so i am leaving it at this point
i quite like logitech, their pro 9000 web cam is good too
12. the microsoft wired keyboard 600 is good, the genius keyboard keys wobble too much on the keypress throw
i don't use unwired because of the emf
usb mice and keyboards are pretty much replacing PS/2 because of their hot-pluggability
13. a canon MG5150 all-in-one inkjet printer. i think these inkjets need to at least have a test pattern printed off once a week or the ink can dry in the nozzle !
canon, though more expensive to buy use less ink so may not be more expensive in the long run and seems to be better quality so far in both harware and software
i have had an hp printer die on me, not keen on HP
MONITOR BRIGHTNESS, BRANDS, RAID, CMOS LATCH-UP AND OTHER MATTERS
turn the brightness right down on monitors, they last much longer (backlight half life!); but you have to reduce the contrast level as well so that you don't need to depress the brightness so much that you get a hum from the PWM circuitry as the "off" time lengthens
my settings for the viewsonic on the actual monitor are: contrast = 52, brightness = 56, which seems to give a reasonable balance between getting the brightness low enough and minimising hum; a brightness of 49 makes for too much hum
you can go to control panel - display - appearance - advanced - and set the 'desktop' and 'window' to the custom colour: hue 130, sat 185, lum 199; red 177, green 228, blue 245
this seems a good balance between the blue rich light (white is also blue rich) needed for synchronising the circadian rythym and over-bright
in theory software should be well ahead of hardware, in practice however, its well behind
i think this is because software being released in a buggy state is intrinsic to the software development and evolution process and the overall cost benefit of this favours both the developers and users whereas with hardware it doesn't nearly so much because the users lose the value of the hardware
i am very manufacturer and brand concious when buying computer gear; reliability and consistent booting are so important.
gigabyte may have the best combination of features, cost and reliablity, their thick ground planes give a quantum improvement in reliability and stability !
MSI motherboards should be ok, but usually lack a DVI-D port,
asus may be ok, my last board from them which was an early version was bit buggy
basically i am sticking with gigabyte or asus at the moment !
ddr ram needs to be on modules with the chips selected for timings like from 'team group inc.'
all ram and flash memory needs to be from “name” manufacturers/brands because with no-name brands you don't get a random quality sampling from the manufacturers, but a certainty of second rate quality because what you are getting is the leavings from what the name manufacturers/brands have picked through !
trap brands and manufacturers to avoid :
nvidia/geforce are fast and feature-full but buggy and prone to cmos latch-up, however, they may be addressing a few issues now (dec 2008)
“NVIDIA strongly recommends that customers transition to this latest revision of the NB8E-SET GPUs as soon as possible. These latest revision units utilize "Hitachi" underfill packaging material that improves product quality and enhances operating life by improved thermal cycling reliability.”
however i am still dubious about nvidia, away with the fairies a bit
sis chipsets are fast, basic and a good price but the hardware is buggy, unstable with long term use and prone to cmos latch-up
ecs makes unstable motherboards, my experience with the ecs k7s5a which i have seen described as the bad girl of motherboards was enough to make me vow never to touch ecs again
september 09, just browsing on the net, ecs is still scorching newbies to the scene with its attractive price performance ratio and bad girl habits
generic ram isn't fast enough and can't be overclocked reliably
also i think the dram module makers are cherry picking the dram so only the leavings are going into the generic
in australia, two woolworths essentials, ie no name usb sticks died within a fortnight on me !
RAID is counterproductive for home users and small businesses, i copy the working ssd to a hard disk with macrium every couple of months and store the hdd in an area not sharing the same fire or theft risk.
when you do that bootable backup copy, let the disk bearing and magnetic media warm up to normal operating temperature and run for a while before copying
also have the backup disk in the same plane for the copy as it would be if it replaced the normal disk !
i have done copies with the backup disk on its side and this has been ok, but to be doubly sure i now do it in the same plane (horizontal) to eliminate any slight slop
also it may pay to just leave the hard disk connected for a day to dry out the pcb
getting the same the magnetic disk temperature is important for getting the lowest ECC error rate as the manufacturers push the densities to the limit
have at the most one hard disk in the frame during normal operation and this eliminates any vibration resonance issues between two hard disks spinning in the same frame
a single disk runs cooler because it has no hard disks above or below it and arrhenius's law dominates in electronics (failure rate is exponentially proportional to temperature increase)
bascially RAID only suits arrays of hard disks, it doesn't fit a home computer with one or two hard disks, a standard that gives both redundancy and decreased read and write times using only two hard disks with a complete recovery if one fails is wanted if its technically possible
having a bootable backup copy hard disk is important because the computer is not really the cpu and motherboard, but the hard disk which is the permanent memory if you will, analogous to ourselves and the memories that make us who we are
latch up of the cmos ram is a surprisingly common problem, especially with the more poorly designed motherboards and north and south bridge chips
this would show as unresolvable bios problems and the secret is to take the lithium coin battery out and not turn the computer on at all for at least 24 and maybe 48 hours in extreme cases to allow the charge to bleed off whatever cmos transistor its latching up
my computer latched up after being left off overnight with the 2-wire asdl module being left connected and on, left it all turned off for 10 minutes and it restarted ok
when storing motherboards they are best left with the coin batteries out as you can get latchup or something not quite right happen to the bios
my micron ssd latched up from a power flicker and i recovered the use of it after leaving it disconnected for about 20 hours!!!!!!!
mightn't have been a latch up, but some time intensive repair and checking process the ssd goes through when it powers up again after a dicey power failure?
my experience of samsung ssds has been very good to date!
it might be a good idea to keep ssd's on their own power lead and not in series with a hard disk !
i leave a side of the case off so the computer runs cooler and you don't have to have noisy fans pushing air through the case, however i guess this is not an option if you have young children around
some amd cpu's need a higher Vcore setting for stability, in the bios i upped the Vcore on my 89W 5400+ amd cpu from the 'normal' 1.35 volts to 1.425, this is the practical upper limit of vcore i think given the limits of the on board power supply
interestingly when i changed to a version to the motherboard with thicker copper power planes the same cpu was perfectly stable at the factory setting of 200mhz for the master clock without a raised vcore, so it must have been a mother board power rail issue rather than the cpu itself
“on current amd cpu's, there's a master clock, which runs at 200mhz. other clocks (core, hypertransport, and dram) are derived by multiplying the master clock by different values.” (no fsb .ed)
in general if you are leaving your computer on 24 hours a day, overclocking cuts the life of the motherboard by too much
q=cv, upping the voltage a bit increases the charge held on the chips to act as a short term energy source, thus relieving the on-board power supply
more voltage promotes speed of switching and charging, but more ringing
“As silicon gets colder, the internal resistance drops, and output transitions become faster. Signal crosstalk and ringing will be worse at low temperature and high voltage. For a worst case test, increase the operating voltage to the upper tolerance limit, and cool the product to the minimum operating temperature.” article
before decent sleep modes, i used to leave the computer running all the time, if you have quite fast settings the computer may not boot successfully from cold, its best to go into the bios on cold booting and let the comp warm up there for a few minutes so the chips move into a more lossy higher impedance state with less signal line ringing
also leaving the computer running all the time pretty much eliminates thermal cycling stress on the chips, moisture ingress into components and power-up bearing damage to the hard disk
the bios makers should have a time option for staying in boot longer (they don't), boot has much more relaxed timings than after the boot settings are loaded and the ability to set dwell time and the level of cpu and bridge work in those relaxed settings to allow the cpu, dram and north and south bridges to warm would be useful
this section is on timings and voltages is for eileens GA-MA785GM-US2H motherboard
8 gig 2 x (2 x 2) gig team dark (TXDD4096M800HC4DC-D)
press ctrl - f1 before going into the MB tweaker menu to get the ram settings, this also gives the ability to change the "advanced chipset features" as a new menu entry
cpu nb vid + .1v = 1.275V
northbridge + .3 volts = 1.4 volts
above combination seems more stable than anything else, very quick bios start
1T/2T command timing crashes on 1, leave all dram timings as is, except below
4-4-4-13-17 = 4525 MB/s
L1 cache 51640 MB/s, L2 cache 18103 MB/s athlon II 250 3 gig 65 watt
cpu frequency 210 Mhz
vga clock 1030 Mhz, crashes at 1100, unstable at 1050?
6 to 7 mths later eileen had a problem with uncontrolled scrolling, have tamped the clock back to 800 mhz, then again to 710 mhz after editpad lost a file
the chess titans moving aerial view is a good test of the GPU, but windows needs to be in an enhanced visual effects mode in - start - control panel - system and security - system - advanced system settings
this section is on dram timings is for 2 x 2 gig team dark dram (TXDD4096M800HC4DC-D) on the GA-MA780GM-US2H motherboard (current)
press ctrl - f1 before going into the MB tweaker menu to get the ram settings, this also gives the ability to change the "advanced chipset features" as a new menu entry
“I would set more NB Voltage for 8GB for sure maybe 1.25-1.3, and try setting your trfc to 195 or 327Ns, if you have it in auto it may be using 127 which would be too low”
ddr2 voltage = 2.0v (+.2v); northbridge voltage = 1.2v (+.1v)
my notation in stored passmark bench's: cas - trcd - trp - tras - trc - write recovery time - read cas to precharge time
tRCD (Row Address to Column Address Delay) is the number of clock cycles taken between the issuing of the active command and the read/write command. In this time the internal row signal settles enough for the charge sensor to amplify it - a ringing parameter
cas latency (CL) = 4T, ras to cas delay (Trcd) = 4T, row precharge time (Trp) = 4T
minimum ras active time (Tras) = Cl + Trcd + Trp = 12T
memtest86+ gave a bandwidth shift from 3434 down to 3352 going from 12 to 13, whereas from 11 to 12 it stayed the same at 3434, but using passmark there is marginal advantage at 11, however leave at 12 as some other settings can be slightly relaxed on 12
1T/2T command timing, faster (much?) on 1T on 4 gigs, but 1T crashes on the 8 gig board
Twtr command delay or time write to read = 2T ok, 1T crashes, slower at 3T
write recovery time = 4T ok, 3T crashed, 5T slower
read cas to precharge time =2T ok, 1T crashed, down from 3T, but 3T as fast
RAS to RAS delay Trrd = 2T ok, 1T crashed, but still 3434 MB/s at 3T so leave at 3T
Trc bank cycle time = Tras + Trp - 1 = 15T
trfc (refresh to act delay) = 105ns ok, 75ns crashed in just the first or both banks
tried 3-3-3-10 and trc = 14, but it crashed.
3-4-4, 3-4-3 and 4-4-3 (trc=15) all crashed
4-4-4-11-14 has proved fast and stable, but use 4-4-4-12-15
on memtest86+, disabling bank swizzle gave a huge boost to bandwidth from 3434 to 3715 but i think thats an artifact of the systematic addressing of memtest
in fact bank swizzle is supposed to be faster and eset seemed to have a considerably shorter power up with bank swizzle enabled, so i have left it enabled
interestingly the memory bandwidth i could get on memtest86+, 3434 MB/s is in a cooee of the L2 cache @ 4261, tho not the L1 @ 21975
an explanation of DQS as per the DQS training control bios entry, am leaving it disabled as the ram is best warm when the calibration is done !
possibly one could do DQS if the ram timing became marginal !
i went to 8 gigs for windows 7 on this board (4/12/09) and used the same timings as the 785G board, but keeping the northbridge voltage at only + .1 volts = 1.2V
the command delay (Twtr) MUST be 2T for 8 gigs
this dropped the speed from 3434 M/s to 3270 M/s which seems a fair trade off for less hard disk access
having the north bridge power management disabled seems to make the memory more stable at tighter settings
it's how hot the northbridge, southbridge and dram get that's really the upper limit
i test the temperature by putting a finger on, being careful to place it centrally on the north and southbridges so the heatsink is not rocked and the heatsink compound between the heatsink and bridge disrupted
they shouldn't be too hot to hold a finger on for a while and the temperature should even drop as the finger itself acts as heatsink
stability was tested on memtest86+, windows memory diagnostic, and passmark
prime95 and amd overdrive were also used, the amd overdrive a bit unstable?
i don't think theres any absolute answer in terms of metrics, one has to synthesize a view
the windows based metrics don't discount accurately memory sharing with other applications, and the bootable metrics are too regular in stepping through ram
generally the tighter settings either worked, or the board didn't boot into the memory test programs and recovered to the last good bios, but one time memtest86+ had to pick the errors which showed almost immediately
amd overdrive (bit buggy, but functional), passmark (very good!), hd tune, cpu-z, pc wizard 2008, sandra lite (very sensitive to any instability but otherwise not so useful) and primenet V5/prime 95 are helpful overclocking utilities
long term/fine grained stability can only be tested by the comp staying up (not resetting) if left on continually for weeks on end, software programs cannot pick this up
be careful not to cook the ram or northbridge with primenet! i only use it on a minute timing, theres no point in stress testing so the dram gets unnaturally hot
if updating the bios be sure to use Q-Flash and NOT @Bios, as @Bios corrupted my bios (fortunately the dual bios back up works!)
easy tune pro 6 appears have installation issues, not advising the use of easy tune pro at this point, though easy tune 5 may be ok, but imo its better to adjust the settings directly in the bios or via amd overdrive and avoid possible hassles with gigabyte software
award BIOS settings for the ga-ma78gm-us2h motherboard
standard cmos features: if no floppy then disable drive A
advanced bios features: amd k8 cool&quiet control = disabled, enable s.m.a.r.t if you have a smart capable hard drive
advanced chipset features: you have to press ctrl + f1 to get this to come up, disable the NB bridge power management and locate the frame buffer above 4 gig to free up more dram. azalia is the new intel audio codec and should be left enabled
integrated periperals: disable 1394 if not using, disable onboard serial and parallel ports if not using
power mangement setup: this is how i set it because i leave the comp running all the time and want maximum reliability
ACPI suspend type = S1, usb wake up from S3 = disabled, pme event wake up= disabled, hpet support= disabled (for xp only, enable for vista)
because of the bridge heatsinks radiating back onto components underneath the sinks and really need small quiet fans i am not overclocking this board at all
it's a pity to leave the VGA core clock at 500mhz, but a higher frequency needs a fan on the northbridge
24/12/12, on my 780g board (ma78gm), moved the VGA core clock from 510mhz down to 470mhz (8%) then down to 400mhz (24%) because of green screen problems with adobe flash !
back to 450mhz and upped northbridge votlage by .1v, so up .2v total !
"the chip (gpu only?) slowly degrades and requires higher voltage and/or lower clocks to be stable.
lowering GPU voltage does seem to prolong cards lives wonderfully"
the bios cpu temp may read higher than pcwizard
for 64 bit windows & the desk top heap @
original setting 1024,20480,768, upped to 1024,40960,1024, upped again to 1024 x 50 = 51200
next increase to 2048, 81920, 2048, working ok so far 19/5/10 (8 gig ram)
81820 also worked so the numbers don't have to be multiples of 1024, but maybe a few bytes are lost to use?
“The values for the SharedSection substring are described as follows:
The first SharedSection value is the size of the shared heap common to all desktops, in kilobytes.
The second SharedSection value is the size of the desktop heap needed for each desktop that is created in the interactive window station, WinSta0, in kilobytes.
The third SharedSection value is the size of the desktop heap needed for each desktop that is created in a noninteractive window station, in kilobytes.”
for windows XP, if there is only one user account, set the interactive desktop heap size to 8192KB or 12288KB, my current settings are 1024,12288,1024 against the original 1024,3072,512
set eileens comp, two users 1024,8192,1024
the above higher desktop settings for XP enable many more windows to be open
with windows XP, two gig minimium is needed to keep things fast enough. 4 gig is very worthwhile and makes the base for any easy upgrade to 64 bit windows 7 , especially with matched memory modules from 'team group inc'
19th september 2009
updated the chipset, audio and lan drivers following a couple of possible driver induced crashes, possibly from installing .NET over the graphics drivers, but turned out to be a ram module
i uninstalled the ati vga/graphics drivers (via the ati unistall program in the ati folder in the programs files folder) before installing the update (minus .NET 2 and the parental control options in the custom install), seemed to go ok on the reduced screen
also updated the cpu driver which seems to make things go marginally faster 20th september 2009
identified failed ram module with both memtest86+ and windows memory diagnostic
what i should have done and didn't do, is relax some of the latencies, or according to anandtech, the clock speed and see if the module recovered
module function exists in a cornered space of latency, clock speed and voltage
hoovered dust off heatsink and turned around, also did the inside of the heatsink and taped up again
copied the samsung HD to the seagate HD as a back up
23/24th september 2009
installed the 'team' 4 gig kit and settled on 4-4-4-11-14 timing
19th october 2009
replaced the samsung with the seagate as the boot drive as the samsung failed seatools and casper xp couldn't copy a file on each of two passes and they were different files
also booting was delayed which may be a typical sign of a hard disk problem as maureen's computer (nietta) had the same problem and her hard disk was no good
hoovered the front of the extreme heatsink as it was a bit clogged - needs doing every month?
29th october 2009
swapped new HD322HJ samsung over with the seagate
had an intermittent connection with the heatsink fan connector but seemed to be ok when pressed firmly
vacumed exterior and interior if heatsink, interior needed it (every three/four months?) all the holes on the top need taping - signficantly improves airflow
watch fingers in zalman power supply fan!!! draws blood
used plastic tube in vacum cleaner end to get the dust on the bottom of the heatsink
2nd june 2010, copied the windows 7 partition from the samsung to the seagate, upright seagate seemed to work ok, vacuumed the back of the cpu heatsink which needed it, the flow is from the back to the case front at the moment
tested that the seagate booted after the casper partition only copy, which it did
increased the vga clock from 500 to 550 mhz on the 780G, 700 mhz causes glitches in ie and editpad
550 had offset cursor problem in editpad, go to 530 ? maybe ok, just an issue with editpad?
8th july 2010, 550 adjusted down to 510 as editpad may be hanging, 21st nov editpad seems to be glitchless !
eileen vacuumed her cpu and northbridge heatsinks, needed doing, probably needs doing once every three months
21st november 2010, vacuumed the cpu (including interior of split heatsink) and northbridge heatsinks and also the interior of the power supply which need doing, first time i have vacuumed the inside of the power supply !
made bootable hardisk copy !
the long metal vacuum pipe fitting reached into the bottom of the split cpu heatsink, but a close fit !
the northbridge heatsink moved during the cleaning, need to look under it when a suitable opportunity arises!
28th june 2011, put the comp in sleep mode and disconnected all the cabling except the power and pulled the comp out to lie flat on the floor with the power cable on and still attached! it seemed to recover ok !
glued the side flanges on the cpu heatsink !
vacuumed the heatsinks!
used car heater hose shoved into the end of the vacuum cleaner to get down at the bottom of inside the heatsink !
4th december 2011
cloned/made bootable hardisk copy, next clone 2 mths? vacuumed interior of cpu heatsink, 5 mths since heatsink interior last cleaned, about right timing, maybe 4 mths better?
power supply on eileens comp needs vacuuming
16th march 2012
updated ati driver "automatically select and install"
improved? ie 9 and ie 8 stability problems and probably editpad pro as well !
17th june 2012
1.1 "all in one" asus et2201iuks spec 21.5" led fhd i3-2120 3.3g 500g 4gb dvdrw intel uma w7h (non-touch) et2210iuks-b016c
purchased 17/6/12 with 3 year replacement warranty ("product care") and annual check ups
i accidently!!! burnt the cases and COA's of my two windows 7 installs !
so i needed to find the keys!
the microsoft genuine advantage tool gives the last 15 digits and the magical jellybean keyfinder gives the full number and works on the 64 bit windows home premium edition and xp at least
as far as i can see produkey likely contains viruses or malware which is a pity
some web quotes on changing a motherboard with xp, i didn't get the blue screen going from nvidia bridges to amd, but there was some problem and i had to do a new install, maybe the bridges were too different? messy driver software?
had no problems swapping the ga-ma78gm-us2h for the ga-ma78gm-s2h motherboard, perfect compatibility
motherboard upgrade/windows xp: (athlonmb.com)
"From a Microsoft XP tech re. an XP install: "Boot the system from the XP CD. Have your CDKEY ready. Select the install option (Don't select repair!). Setup will find the install of XP that is already there and ask if you wish to repair it. Say yes. Setup will run the upgrade code that will re-enumerate the hardware and set itself to boot from the new controller. BTW, the first repair option only verifies XP files against the XP CD versions and makes no system setting changes."
from the ecs forum on ocworkbench.com
Hybee posted 29 September 2002 02:03 "I've changed motherboard from K7S5A (SIS 735) to K7S6A (SIS 745) 3 months ago. I havent reinstalled anything - I've just switched the boards and booted to WinXP, windows redetected all the devices and everything worked. No reinstall needed. Give it a try..."
Sludge posted 29 September 2002 02:06 "I just did a repair install on a computer that I went from an old Via motherboard with K2-350 to the K7S5A with Athlon XP 1600 processor. You replace the motherboard, then set the bios for CDROM boot. You boot off the Windows XP cd and do not choose the repair option! Instead choose install and select the partion where XP is already installed. It will then ask you if you want to do a repair install, or wipe out the current installation of XP. You then choose to repair your current installation. Hope this helps, Gordon PS: Webshots survived the repair as well as all the other programs. "
the quote below from this article
“Before you swap out the current motherboard go to device manager and select the IDE ATA/ATAPI Controller and select your current storage controller. Right click, select update driver and select install from a list or specific location. Click don't search I will choose the driver to install and select the standard dual channel IDE controller.
This will prevent the inaccessable boot device blue screen. I have done this multiple times on different platforms with no problems.
Also I found that booting the first time with the new motherboard is best done in the Safe mode where XP will install the drivers it needs and at the same time you can install the new motherboard drivers from its CD.
This saved me doing a repair or a reinstall of XP. My whole system came right up and worked great and very stable.”