Microcode for PIII-S 1.4GHz please?

[Th3_uN1Qu3]

I run dual 1.4GHz Tualatins on an Asus CUV4X-DLS. A microcode error appears at every boot and i would like to get rid of it. So does anyone have the microcode for a PIII-S? I know the procedure, i just can't find the files i used when i did it on my old Gigabyte board.

On a side note, is the microcode error purely cosmetic? Or does it affect performance? So far the system runs great so no complaints on this matter, but i'm curious.

[edwin]

run something like CPU-Z to see how they are detected (bus speed/multiplier). I think you should be able to get them from the horse's mouth:
http://downloadcenter.intel.com/SearchR ... crocode%22

[Th3_uN1Qu3]

They are detected fine in CPU-Z. However i'm getting a different issue - the board sometimes gets the VID as 1.28v instead of 1.45v, and as you would expect this results in a certain degree of instability. No major crashes or BSODs, but some application crashes, yes. I know how to do a hardware VID mod ie adding a set of dip switches to manually control the voltage, but it involves too much sensitive work for my taste. Maybe the microcode can help with this.

Thanks for the link, i'll see what i can fish from Intel's site. Edit: No such thing as PIII around there. I'll keep googling.

[Th3_uN1Qu3]

I found microcode, but CTMC said it's already present in the BIOS... I flashed it anyway. Now the error went away and the CPU core voltages are at 1.42v, but this brought another unexpected consequence - the CPUs now use up so much 5v power that my HDD failed. My power supply has 27A on the 5v rail so i don't think i'll be able to find one much more powerful than this.

I powered the hard drive from a separate PSU but the damage was already done. Windows says something about the registry being recovered (recovery my a**, the copy it installed is at least 2 years old so basically half my software is bye bye), and on next restart it hung on "Loading your personal settings". Same thing happens in Safe Mode. I'll run SpinRite now to fix up any damaged sectors that might have occured, then grab my data off the C drive with WinPE, then... reinstall time. It'll take a crapload of time to get all that stuff installed once again though. Crap. This install survived 2 mobo swaps... and now this.

Edit: SpinRite found 2 bad sectors and it said "Unrecovered". Officially my HDD is toast. Hooray!

[edwin]

a $13 power supply has 34A on the +5V rail already:
http://www.newegg.com/Product/Product.a ... 6817165023

maybe, just maybe, it is time for a new one then...

[cp]

is the microcode error purely cosmetic?

yes.

Or does it affect performance?

no. microcode updates come with Windows and are loaded on bootup anyway.

the board sometimes gets the VID as 1.28v instead of 1.45v [...] Maybe the microcode can help with this.

no, the microcode doesn't do anything to the VID pins. it just can't.
how often is sometimes? every other second? undervolting is not a problem for the hardware. it can only ruin your data but won't do permanent damage to your hardware.

Now the error went away and the CPU core voltages are at 1.42v

the cpu core voltage is determined by the VID pins of the CPU. they are hardwired in the CPU. so only mainboards with manual voltage adjustments in the BIOS can alter the CPU core voltage (or let's say: the user can).

Officially my HDD is toast.

HDDs tend to die from time to time.

[Th3_uN1Qu3]

@edwin: Do you really think that PSUs like that ever HONOR THEIR LABEL RATINGS? I have about a dozen of that type that i've scavenged for parts. I've gotten them for (almost) free. At most $1 each. So, conclusion is... NOT. Please come with a more plausible fix. I know enough to build my own switchmode power supply, do you think that a "Linkworld" (aka noname) power supply can impress me? Thing is, building a SMPS is darn complicated, so i'd rather not do it if i can just buy one. But since i can't... i'll build it. I'll build a secondary power supply that takes power from the 12v rail, and supplements the 5v one with the extra amps needed. This seems like the only way to go with this system.

@cp: Thanks for enlightening me, actually the microcode update and the "proper" detection of my CPU VID was just a coincidence. The board will either set 1.42v for ONE of them (which is stable), or 1.28 for both (which is not)... But certainly setting 1.42v for BOTH CPUs is what crashed the drive. On next power up it was back to the "usual" behaviour of setting 1.42 just for one of them, but my drive still did not work. I managed to grab important data and reformat... A subsequent run of SpinRite indicates the drive is actually fine, so it was nothing permanent. The trouble wasn't with undervolting which the board likes to do, the problem was with running the CPUs at their nominal voltage, which asked for so much power that there was not enough juice left to run the hard drive, and it crashed (actually spun down) while trying to load Windows.

And it's not a matter of "not enough power", it's a matter of severe crossloading, which unfortunately is unavoidable in a computer like mine. PC power supplies are designed for (somewhat) equal loads on all the rails. Since my system uses next to nothing on +12v but puts extreme load on +5 and +3.3, the power supply cannot meet the specs. This is similar to people running car amplifiers off a computer PSU (with NO 5v load) and wondering why their 12v rail falls to 9 volts and their car amp (which actually contains another SMPS) sounds like a fart.

Anyway, i think it's about time for manual VID selection via dip switches. Mods ahoy! I'll get to it in the next few days. And no... HDDs don't die from time to time. They *only* die from power failure or physical shock.

Edit: The core voltage is determined by the VID pins *in theory*. Like pretty much everything else, there are a number of things that can go wrong with this detection. At this moment bypassing it entirely seems like the healthiest option IMO.

[edwin]

Some do, not this one, but I was assuming you had an old PSU in there, not a custom-built one. I guess you have to build another +5V rail then.

[Th3_uN1Qu3]

Some do, not this one, but I was assuming you had an old PSU in there, not a custom-built one. I guess you have to build another +5V rail then.

Well it isn't custom... yet. It's a Xilence 420W unit. Certainly enough power for this computer, but it's the distribution that is the problem. So yeah, i will make myself an extra +5v rail.

Anyway, i've VID modded it today. Went the easy way - simply hardwired it to 1.4v as i only had to wire one pin of each voltage regulator chip to ground. I'll see if it's stable this way (the CPUs are 1.45v stock), if it isn't, a second wire will switch it to 1.45v. Will post a pic too.

[edwin]

My apologies for not instantly seeing we're dealing with a electronics engineer. They are rare creatures in these forums...

[Th3_uN1Qu3]

My apologies for not instantly seeing we're dealing with a electronics engineer. They are rare creatures in these forums...

Heh. If i told you my age, you wouldn't believe me. I am a wannabe EE indeed, but i'm just 19. Just starting college this fall, by this time i think you already guessed what i chose. I'm also a DJ on occasion, what i do best is make people enjoy themselves and dance, when i mix no one sits down. This is why i only do random parties, i bet i would become bored if i had DJ'ing as a nightly job. I do it for the music. Oh, and for the pretty girls of course, and for the friends of the pretty girls hehe. My current GF, is a friend of someone i met at a party in June.

This dual PIII is a hobby of mine, it goes back to 8th grade when i barely knew anything, but i had my first dual (accidentally at that time, and it was a dual PII-450). And the bug bit. And it went on. The fun thing is that time goes on, grade school ended, highschool ended, college will start, old friends go away, new ones appear, but this computer is still here. It's simply something i'm used to. And i'm constantly expanding it year by year, pushing the limits on what can be done with this hardware that many may call obsolete. It does what i want it to, that's all i care about.

I was about to post the pic but i forgot where i put the cable of my camera, and my laptop's card reader does not accomodate MS Pro Duo. Will do tomorrow.

[cp]

Xilence 420W

Xilence is a low cost and low quality brand. you should really get something better. if you don't want to spend too much money, i'd recommend the cheaper 80PLUS PSUs from Seasonic or BeQuiet. the brand really doesn't matter. make sure it has a 80PLUS logo (or better).

They *only* die from power failure or physical shock.

there are many more ways to kill a harddrive. like anything manmade they have an MTTF (not MTBF but i guess they write MTBF when they mean MTTF). not enough power on any rail and the resulting voltage drop isn't one of them. anyway, assuming that the hdd died because you observed something on a non-calibrated meter (you didn't really measure the vcore, did you?) before the real incident took place is PURE speculation.

Edit: The core voltage is determined by the VID pins *in theory*.

not only in theory. the core voltage really is determined by the VID pins. take a look at any Intel VRM document. they are available for free all around the web and ofcourse at Intel's page. you can even take a look at one of those VRM datasheets. let's take that well known Intersil HIP6004. there are plently of datasheets and even application notes (read THIS one) with many examples.

and to make one thing clear: the VID pins of the CPU (output) and the VID pins of the VRM (input) are hardwired on the board if there's no voltage adjustment (by jumper or software). what can go wrong with hardwired pins?

Like pretty much everything else, there are a number of things that can go wrong with this detection.

there's nothing that can go wrong. really. maybe you can describe these things that can go wrong. don't hesitate to get deep into electronics since i am a real Electrical Engineer with a real diploma.

[Th3_uN1Qu3]

Xilence 420W

Xilence is a low cost and low quality brand. you should really get something better. if you don't want to spend too much money, i'd recommend the cheaper 80PLUS PSUs from Seasonic or BeQuiet. the brand really doesn't matter. make sure it has a 80PLUS logo (or better).

And what about the 5v rail? 80Plus is 100% ATX 2.2... At least my Xilence isn't a "gutless wonder", it's got PFC (albeit passive) and full protection. I had to choose between this and a FSP. I know you'll say i should have chosen FSP, but this was my decision.

not enough power on any rail and the resulting voltage drop isn't one of them. anyway, assuming that the hdd died because you observed something on a non-calibrated meter (you didn't really measure the vcore, did you?) before the real incident took place is PURE speculation.

Don't tell me you have a Fluke connected to your motherboard at all times... Anyway, my HDD is fine, thanks. It just crashed Windows and made a couple logical bad sectors, nothing permanent.

Edit: The core voltage is determined by the VID pins *in theory*.

not only in theory. the core voltage really is determined by the VID pins. take a look at any Intel VRM document. they are available for free all around the web and ofcourse at Intel's page. you can even take a look at one of those VRM datasheets. let's take that well known Intersil HIP6004. there are plently of datasheets and even application notes (read THIS one) with many examples.

My board actually has a HIP6004 serving CPU2. For CPU1 there's a RichTek (forgot its model but will post pics of the mods so you'll see what it's about anyway), chosen because it has more outputs. That RichTek is probably in charge of the RAM and chipset too.

and to make one thing clear: the VID pins of the CPU (output) and the VID pins of the VRM (input) are hardwired on the board if there's no voltage adjustment (by jumper or software). what can go wrong with hardwired pins? there's nothing that can go wrong. really. maybe you can describe these things that can go wrong.

Such as, improper connection between the pins and the socket? That is a definite possibility in my case. Anyway i measured the resistance of the relevant VID pins with a meter and they read over 7kOhms, clearly reading an "open". I hardwired VID0 and VID1 to ground on each reg and now everything is fine and the CPUs boot at 1.42v all the time.

But of course, it became even more unstable. Why? Well, even with a second PSU feeding the hard drive, there was still not enough power. Luckily the ASUS guys left me an extra power connector on the mobo, with two 3.3v rails and one 5v. And even more luckily, this connector was AT PSU style... So i took one AT PSU cable from my junk box, wired it up to this secondary power supply that also runs the HDD, hit the power and hope for the best. Result: 100% stability at 148MHz FSB (1.55GHz). A pic is worth a million words, but i gotta run now, i'll download 'em from the camera as soon as i come back.

[cp]

80Plus is 100% ATX 2.2

ATX 2.2 doesn't define effiency, 80plus does. just for starters: find out what PFC is for and what it has to do with efficiency. sum up the wattage of the different voltage rails (assuming they are independend of each other) and compare it with 420W.

Such as, improper connection between the pins and the socket?

yes, possible (possibility for just those 2 pins out of 370 pins..you can do the math). there are more than enough pins (data, address, ...) that are way more sensitive to improper connection than those VID pins.

but back to the CPU vcore generation: the two processors consume ~64W (TDP) in the worst case. resulting in ~45A@1,45V. on the 5V rail the same wattage would require 13A (or 5,5A on the 12V rail) assuming that the vcore regulator has an efficiency of 100%. if the vrm has an efficiency of 85% ~15A would be required on the 5V rail.

get a PSU that is at least 80plus certified and delivers at least 20A on the +5V rail (i'd get a Seasonic SS-400ET or Seasonic SS-350ET, both available for 30-40€ each)

[Th3_uN1Qu3]

just for starters: find out what PFC is for and what it has to do with efficiency. sum up the wattage of the different voltage rails (assuming they are independend of each other) and compare it with 420W.

What PFC does is keep current and voltage in phase, lowering the stress on the power grid. I didn't have to look that up btw. IMO it is only important when you are drawing close to the limits of the breaker on your circuit, a power supply with PFC is less likely to trip the breaker than one without. Otherwise... CCFL bulbs are among the worst offenders when it comes to injecting harmonics and noise in the power line, but tell me, you ever seen PFC in one of those? Of course not, because they have to be as cheap as possible (and they're more expensive than the regular incandescents anyway). Oh yeah and they're banning incandescents... I'd rather run a good ol' fluorescent tube rather than a CCFL bulb, at least a regular fluorescent tube actually gives usable light. Perhaps LED lighting will go down in price before all incandescents are banned, anyway, enough off topic.

yes, possible (possibility for just those 2 pins out of 370 pins..you can do the math). there are more than enough pins (data, address, ...) that are way more sensitive to improper connection than those VID pins.

You don't know how many things with <5% probability have happened to me. I don't have to do math for that. But regardless, we are talking of a board that wasn't designed for those CPUs in the first place, AND after the VID mod everything is fine.

get a PSU that is at least 80plus certified and delivers at least 20A on the +5V rail (i'd get a Seasonic SS-400ET or Seasonic SS-350ET, both available for 30-40€ each)

I bet i could put in this system the Seasonic SS-500HT that i have in my Core 2 Duo at home, and it won't run it. I already mentioned before that my Xilence has 27A on the 5v rail and it isn't enough. It's not about the rail capability per se, you forgot to think about crossloading. Since all rails are derived from the same transformer (well except the 3.3 which is regulated by a magamp, but it's regulated from the 5v rail so it still loads that up), the power supply is designed to be loaded with somewhat equal amounts of power on each rail, to stay in spec. Since my system consumes next to nothing on 12v, the 5v rail sags. If enough feedback were applied to keep 5v in check in this situation, 12v would probably go so high that it would blow something up. I've seen it at 12.8v on the old motherboard, this one seems to be a bit more 12v hungry and it stabilizes around 12.5v.

This is only avoidable in the new designs based on DC-DC converters, where there is basically one fat 12v rail and everything else is down-regulated with buck converters. Unfortunately those power supplies tend to be expensive, and i have yet to see one with more than 20A on the 5v rail, they're all 12v oriented for modern hardware. But in essence, this is what i'll be doing, just that the buck converter i will build is intended to supplement the 5v rail coming directly from the transformer, not replacing it entirely.