Atomic I/O letters column #20
Originally published in Atomic: Maximum Power Computing Reprinted here 24-Apr-2003.Last modified 16-Jan-2015.
Tweaking and tuning
I found the section in my BIOS where I can change the FSB, multiplier, voltage... you know.... da good stuff. I tried to play around with this stuff but it wont let me change anything. I did the pencil thing to unlock my CPU, but still no joy. Why?
My mobo is the really cheep ECS K7S5A, and I'm running a Duron 1.2GHz.
Also, I have a "TV Excel" TV tuner card in my PC and it worked great until I changed to WinXP, which doesn't like it. I go back to WinME and it's all good, then back to XP and it doesn't work. I know I need an XP driver, but where to get it? The TV card has the Conexant BT878 chipset on it and a Philips tuner.
Tiny
Answer:
Your K7S5A lets you see the FSB, multiplier and voltage settings,
but the only one you can change is the FSB, and then only from 100
to 133MHz. You can try that if you like, but since there's no CPU voltage
adjustment, a 33% overclock attempt is very unlikely to work.
If you're really hot to try overclocking on this motherboard, there are some tweaky home-made BIOSes for it available, which are free to download and possibly worth what you pay. Have a look here for downloads and information. All the BIOS-twiddling in the world won't turn this board into an overclockers' special, though; if you put a more than trivial value on your time, you'll do better to just save up for a new motherboard.
Your TV tuner card is a Pixelview 878TV. You can find drivers for it here.
The voltage mystery
I decided to have a go at a 12V/off/7V switch as a fan control. Big problem though, after wiring everything up I'm only getting 12V outputs. I checked everything out with my multimeter, and the voltage between the +12V and the 5V ground is still 12V! Any ideas why this has happened to me?
Alex
Red is 5V, yellow is 12V, and the black wires are identical earths.
Answer:
If you measure the voltage between a 12V (yellow) wire and a 5V (red) wire,
you'll get 7V. Well, unless the lead from your PSU is miswired.
If you measure the voltage between a yellow wire and a black wire, though, you'll get 12V. I think you might be doing that, because you mention "5V ground". Ground is ground; both of the black wires are electrically connected to the same ground back at the PSU. The difference between either of them and the yellow wire is 12V.
Note that the "seven volt hack", where you wire a fan between +12V and +5V to get a 7 volt potential across it and, thus, roughly one-third power, is very simple but not particularly elegant. If you short 12V to 5V (by accident, or because of a fan defect) you'll end up with 12V on the 5V rail, and probably blow up various components.
Budget computing
I've just gotten my hands on an old Cyrix II 300 (runs at 233MHz), and was wondering if there would be much chance of getting it to around 300MHz with adequate cooling?
The mobo is a Socket 7 PC100 SystemBoard M571 and have a manual showing how to do with voltage, FSB and multiplier adjustments.
Also, on the same mobo, there is an onboard video card (don't die of shock) but I believe I can disable it. At the moment, it is greyed out in the BIOS and I can't touch it. Do you think I would have to insert a different video card to be able to get to it?
Simon
Answer:
It's possible to overclock a Cyrix II, but not by much; they run pretty
close to the redline at their stock speed. 266MHz is likely to be the best
you can manage. Even at that speed, the II will be a very unexciting performer;
perfectly adequate for many applications, but not fast compared with an
antique Celeron 300A, much less any current CPU.
The M571's only got the old 66/75/83MHz CPU FSB speeds, and 83MHz is unlikely to work. But it's got multiplier jumpers, and you don't have to do anything to unlock the multiplier of any Socket 7 CPU. The most you're meant to be able to wring out of any chip on this board is 5.5 (the highest available multiplier) times 75MHz, or 412.5MHz core speed. If you hit the auction sites and pick up a cheap AMD K6-2 or K6-III, you should at least be able to achieve that - or more, with a K6-2, since they map the 2X multiplier to 6X. If you can get 83MHz FSB to work stably, this'll let you get all the way to 500MHz core speed.
Those processors want 2.2 or 2.4V core voltage, mind you. Older M571s only offer 2.5V as their minimum core voltage, which is on the upper side of acceptable for a 2.4V chip but rather too high for a 2.2V one. Newer M571s have 2.2V as well.
Assuming your M571 offers a useable core voltage, you should definitely be able to run a dirt cheap K6-2 400 at a smidge more than its rated speed and, as mentioned above, you might manage 500MHz. Even at 412.5MHz, a K6-2 will give you the thick end of twice the CPU performance of a 266MHz Cyrix II.
It's also possible to upgrade the M571's video; you use jumper JP3 to turn off the integrated video adapter, and then you can install a video card.
The M571 doesn't have an AGP slot, though, so you'll need a PCI video card. Hunt around for an old PCI Voodoo 3 or Kyro board; that'll give you worthwhile 3D performance on the cheap.
The M571 is unlike a lot of PC Chips/Systemboard branded products, in that there's actually some worthwhile info about it online; check it out here.
XP or not XP?
I recently upgraded my Celeron 400 to a 2GHz P4 with 512Mb DDR RAM and a GeForce2 MX (I plan on upgrading that later). I am currently running Win98 Second Edition and I am thinking about upgrading my OS to WinXP.
What would be the advantages of me upgrading to XP? Would my PC run a lot better with a newer OS?
Alexander
Answer:
With a fast CPU and a pile of RAM - 256Mb or more - the current NT-series
OSes (Win2000 and WinXP) are both preferable to any of the Win95-series
OSes, including Win98 and WinME. The NT-series OSes are giant resource hogs;
they must have a lot of RAM for decent performance, and it really
helps to have a recent CPU as well.
If you've got those, though - and you do - then the NT-series OSes are better just because they're stable.
Yes, Win2000 and WinXP can crash, and crash hard, but failures certainly aren't the daily event they are with Win98 machines. You also don't have to reinstall the NT-series OSes nearly as often, if ever.
For my money, that's about it for the major advantages of 2000 and XP. There are tons of other twiddly bits in 2000 and XP which may turn your crank, but for most home users, I really don't think they're worth spending money on. Stability is.
512Mb of disappointment
I've got an MSI KT2 Ultra ARU-RAID motherboard. I started out using 256MB of PC2100 266MHz CAS 2 RAM, which worked fine.
Last week I sold that RAM and bought an AMD/Intel certified 512MB PC2700 333MHz CAS 2 module. I'd heard months before from Web sites and magazines that if you run 333MHz RAM at CAS 2 on this mainboard, the system will become unstable.
Thinking that the RAM is AMD/Intel certified, I thought it would be OK to run. When I set it to CAS 2, though, the system becomes really unstable. So I'm forced to run the RAM at CAS 2.5, which is much slower than what it is capable of.
Do you know what the problem is, and do you have any solutions?
Ethan
If you must buy RAM with decorations, try getting the heat-sunk type,
like this, not the type that just has misleading
stickers on it.
Answer:
Um... I wasn't aware that AMD or Intel actually did certify RAM, in general
terms. I don't know what sort of certification they'd provide for the memory
you bought, anyway; they make chipsets to work with RAM that meets
JEDEC standards, but when I first wrote
this reply for Atomic magazine,
neither company even made a chipset that supported PC2700 (333MHz
DDR) RAM. It's my considered opinion that someone just put a sticker on
that RAM module that said "Certified".
Intel do validate memory - see their list of validated DDR400 modules here, for instance, which went up after this answer appeared in Atomic, but before it hit the Web. That page just lists some RAM that Intel say will be fine for use with motherboards using their DDR400 chipsets, though; your MSI board uses the Via Apollo KT333 chipset.
In any case, it doesn't matter how "certified" the memory is, if you're winding stuff up to the point where it's known to go flaky, as CAS 2 at 333MHz is known to do with your board. Extreme RAM tweaking is difficult on lots of boards; changing the RAM you use may or may not help.
The good news is that, in any case, CAS 2.5 is not "much slower".
CAS latency is how many clock cycles it takes for the memory to start a read operation. It doesn't have anything to do with how fast the read operation itself proceeds; that'll be the same, for a given RAM clock speed, regardless of the CAS latency. Lots and lots of read operations happen every second, of course, but the impact a CAS 2 to CAS 2.5 change will have on performance for all but the weirdest of PC tasks will be well under 5%.
For tasks that don't lean on the RAM very hard - and most PC tasks don't - the difference will be barely measurable, and certainly not noticeable.
Medic!
Do you guys know any ways of re-pinning a CPU? I have a few chips that through travelling to and from LAN matches have lost or damaged a pin or two.
Also, I'm trying to find a version of Linux that looks and acts like Windows.
Zsebastian
Answer:
Um... how, exactly do you travel to and from LAN parties? By pogo stick?
Never mind.
Regardless of how exactly you've managed to damage CPU pins without re-kitting the whole computer, it seems likely that the processors are going to stay damaged. It's possible, if you're lucky, to unbend slightly bent CPU pins, but replacing them doesn't seem to be something you can do for less than the price of a new processor, unless you're good at fine soldering and the busted pin is on the edge of the package.
As regards your Linux question - try Lindows, or Lycoris.
UPDATE: After I put this column up on the Web, a number of readers sent me CPU re-pinning solutions. You can read them in this column!
Clear computing
If one were to make a case entirely out of perspex/acrylic, what problems could this cause? I thought that the case acted as earth to all the parts inside - isn't it a very good idea to keep these parts earthed? I'm probably wrong, because the earth on the power connectors etc probably takes care of this, but I thought I should ask, your 1337 knowledge is far superior to mine ;p
Steven
Some windowed cases, like this Lian Li PC-6099,
now come with an aluminium plate installed behind the window. Without it,
they're likely to break EMR rules.
Answer:
There's nothing wrong with making a plastic case, beyond the fact that it
won't provide any RF shielding. Earthing isn't the issue; as you say, the
power connectors take care of that. But metal cases don't let electromagnetic
radiation (EMR) out of, or into, the PC. Anywhere there's no metal, EMR
can get in and out, provided its wavelength is short enough that it fits
through the hole. Grilles aren't an issue; great big holes cut in the side
for windows are.
A clear plastic case would work just fine, provided nothing nearby was bothered by interference from the PC, or broadcast enough RF that the PC had a problem (which is very unlikely, in home and office environments).
Clear plastic can be made conductive by "metallising" - applying a thin conductive layer to the surface. A really thin film of metal is conductive enough to block RF, but won't darken the plastic much. It's not cheap, though.
If you make un-metallised cases and sell them then you could be in trouble, especially if you sell cases with components already in them. But running a PC in a plastic case is like running a PC that's just a bunch of components sitting on a newspaper on your kitchen table; the amount of interference generated isn't likely to be sufficient to annoy anyone, so nobody cares.
Secondary choice
Are you the same guy who I used to read in Australian Commodore and Amiga Review, back in the pre-Pentium days when Amigas were real computers and PCs were just business machines?
Reminiscing aside, I have a problem I hope you can solve. My computer runs Windows XP, and has an Athlon XP 1800+ and an Asus A7V266-E motherboard. Recently I have acquired an All-In-Wonder 7500 video card, a great card for a great price if, like me, you aren't a hardened gamer. Photoshop, Director, Dreamweaver and Premiere are where I get my jollies.
For this I need heaps of screen real estate, so the All-in-Wonder is for my 19 inch monitor to hold my work area; I now need a PCI video card for the 15 inch monitor to hold all my tools.
My problem is which PCI video card will do the trick, without costing as much money as I spent on my primary AGP card. There are not many PCI video cards out there, and those that are make no claim to be XP compliant. Could you help me please?
Jonathan
Answer:
Yes, I used to be half of the crack ACAR editorial team. You know, that
150 watt AT PSU with the hacked-on cable that's sitting on the shelf over
there probably isn't ever going to power an A500 again. I should probably
put it in the cupboard, or something.
Getting back to the present day - you can't buy any old PCI card to use as a secondary, but lots of them will work, and there are WinXP drivers for various popular boards. Any Nvidia-chipset board (an old TNT or TNT2, for instance) will be OK, as will any recent ATI card (a Rage 128 would work), and any Matrox card from the G200 onwards.
A lot of computer stores don't stock many, or any, PCI video cards any more, but they're not hard to find on auction sites and at computer markets.
Handy hint
Hi, just thought everyone should be reminded of a handy site if they are having trouble finding drivers for obscure hardware.
Most products have an FCC ID number on them somewhere. It's a simple matter to enter this number into the page here.
This will tell you the name of the manufacturer of the hardware. Do a Google search and it's a piece of cake to find their website, if one exists.
Of course, most of the more l33t amongst us would probably already know this.
Iain
