In my divine and neverending intelligence I seem to have acquired a non-undervoltable motherboard.

All I can do in BIOS is change "CPU Frequency", which is a value ranging from 100 to 255. This is tied, through a ratio, to "DRAM Frequency". Does it mean I'm altering the FSB or something? Altering the FSB doesn't help the heat output, does it?

There is a nice "hardware monitor" page, which lets me see the Vcore at 1.328. I can also inspect CPU temperatures and CPU fan speeds (BTW -- stock HSF is a 72mm running @ ~5000 RPM). But, I cannnot change the Vcore.

The BIOS in question is an American Megatrends, v 02.54, copyright 1985-2003.

(FWIW -- the CPU is an Athlon XP 2600+)

What now? Am I screwed? I heard of BIOS updates, maybe it would be possible to "update" to a different BIOS that supports undervolting?

"Help!"


EDIT: topic changed to reflect developments later on in the thread.

ebay dude, ebay ----------------->

Then buy one you can undervolt. Beats 2 months of research.

http://www.silentpcreview.com/modules.p ... =69&page=1

Pick it from one of these.

Sorry to sound cheap, but are there any solutions that do not involve purchasing a new motherboard?

Well I made the same mistake as you and I ended up returning the board. I bought it from newegg and I returned it bought a refurb Abint NF7-S and so far so good.

It was only about $40 when I got it too, so might want to look into their refurb section.

CPU heat output is roughly linearly proportional to CPU speed, and roughly exponentially proportional to CPU voltage. So reducing voltage will help a lot more, but reducing CPU speed also helps.

I think you are cooked by way of BIOS, most AM BIOSs are useless.

Ok then.

I'll try to manage the noise via FSB. At 133 MHz (1530 MHz clock speed, still very apt), the stock fan @ 5V (reporting ~2700 RPM) keeps the CPU at 44 C when idle. I know it probably won't be enough to keep it at sane temperatures under load, but we'll deal with things as they come.

Not optimal, but not terrible in terms of noise. Easily drowned out by my current PSU (which I'll be replacing ASAP).

I'm willing to get it down to 100 MHz FSB if necessary.


One interesting thing is that the BIOS reports a Vcore of 1.328 V. Now, I am confident I don't have a 90nm chip in here, because they simply don't exist for socket A. Since I've been told my CPU should be running at 1.65 V, I'm tempted to suspect the BIOS is bullshitting me. Or maybe for some strange reason it really is running the CPU at 1.3V? Is there any way to tell?

yeah your vcore should be higher than 1.3V, this is consistent with the mobile CPUs not standard desktop ones.

you could always go for the more difficult option of opening and closing bridges on your processor

http://fab51.com/cpu/barton/athlon-e23.html

isnt P=IV. so Power is directly proportional to Voltage? or in this instance Power and Heat are not so linked with CPUs ?

Hmm. Are there any mobiles that go into Socket A?

The reason for me asking all this is I don't really know what processor is in this board. The board came with CPU and HSF installed, and the HSF mounts are protected by green rubbery goop so the store knows if I took it off. And as I haven't gotten around to running a full OS on it yet, and the BIOS is crap and doesn't tell me anything about the processor, I don't really know what I have.



So even though BIOS doesn't support Vcore changing, I can still change it manually on the CPU? Nice, I might try to do that when I get bored in the summer.

*Slaps forehead**

For two years I used a product called CPUidle on an 1800+XP. It works *very* well so long as your mobo is supported. I had an nForce1 and it took 60W off my power consumption measured through my solar inverter. We're talking savage power reduction down from 11A(12V) for the whole box, to 6A. Reducing the clock speed does almost nothing. Those CPUs just suck power for no good reason. The Barton is 2A(24W) worse because of the large cache. It's the most inefficient chip in existence. I made one up for my dad an was horrified. Later had to underclock it, just to stop it beeping in summer. Thank AMD for Winchester and Venice.

The reason reducing voltage works so well, is that it is not P=IV, because I decreases as V decreases. Plus transistors also have a min voltage above 0, but not sure how high it is in these technologies.

Reducing clock only works if you are trying to get the heat down during gaming. When idle it make no difference.

The only drawback with CPUidle is that it can sometimes muck up games.

Of couse, P=IV still applies, but as you note I is not constant. Roughly speaking, the load has a particular average resistance. The amperage is equal to the voltage divided by that average resistance--in other words, I is proportional to V.

Thus, P is proportional to V squared.

Roughly speaking, of course.

OK, here's yet another twist.

I managed to get Windows 98SE to boot into safe mode, and it's Control Panel -> System identifies this CPU as...

AMD Athlon MP 1900+

According to Processor Electrical Specifications, this is a 1.6GHz Palomino chip, TDP 58.9W, MP 66W. As this is just slightly lower than 62W/68.3W my XP 2600+ was supposed to put out, I feel additionally ripped off.

Interestingly enough, the above page idenfifies MP's Vcore as 1.7~1.8V, which means BIOS is bullshitting me about the 1.328V.

I'll be going back to the place that sold me this thing tomorrow or Tuesday and I'll have a talk with them.

Any recommendations as to what I should get instead? I cannot push it too hard, but I don't think asking for something within CAD 200 (edit: about USD 160) instead of this weirdo I paid CAD 160 for will be unreasonable. Maybe a futureproof socket 754 with a Sempron? This will be a SFF, so the less heat the better.

One of the undervoltable motherboards from the list, that's for sure. One more question -- just because some boards with, say, VIA K8T800 chipset are undervoltable doesn't mean all of them are, right?

plenty, they are very popular on this board, since are veary good underclockers and run much cooler to start off with.

AFAIK, my Gigabyte K8NXP-9 is also non-undervoltable, at least with my Venice 3500+

Tried CrystalCPUID and RMclock, can't get it below 1.1V. I can select any lower voltage, but it reports 1.1V back (also other tools report Vcore as 1.1V)

And even more tricky: if you set Vcore in the BIOS to anything other than normal, no software Vcore changing possible AT ALL. Ugh!

So why is this bloody board in the undervoltable list

Peter

Because it goes to 1.1V, i believe the standard voltage for Venice is 1.35V.

Yes, but that is the standard C'n'Q value that the MB is supposed to support. Undervolting (for me) is going below 1.1V like others have posted on the forum.

I've posted a service request with Gigabyte, we'll see what they say about this - I hope a bios update will deal with this...

Peter