My PIV Adventure–Part III

It’s rough try to write an article or two a day AND test equipment.

Here’s how life’s been going with this PIV setup.

Better Power, Not Better Results

If you recall, the last time I spoke about this
I was rather concerned about the power drops I was seeing when pushing the PIV to the 2.7GHz, then making it work. The 12V line was dropping to 11.5-11.6V (and the 5V line was sitting at around 5.25V).

So I got an Antec True Power 480, and the voltage irregularities went away. The 12V line does occasional drop a bit, but by “drop,” I mean to 11.9V.

So dropping to 11.6V isn’t “normal.”

To be fair to Enermax, the power supply I was using (an Enermax EG451P-VE) was one of the first to feature a PIV 12V line, long before Northwood. Or it could just be this particular one. Check around to see how others have fared with PIV setups before crossing any more recent models from your list.

One definite disadvantage to the Antec is that it is decidedly noisier than the Enermax.

Was I able to do better with a more solid power supply? Not to any appreciable degree, so if you run into these power drops, but things are otherwise working fine, don’t worry much about it.

The PIV Degrades Slowly

I’ve ended up increasing voltage to essentially 1.75V at 2.67GHz because what initially runs fine doesn’t necessarily keep running that way. I’ve run Prime95 for hours at one voltage setting; tried again another time, and it would blow up after 40 minutes.

Just throwing voltage at it can get you into Windows easily enough; I’ve had this up to 2.8GHz, but Prime95 will crap out almost instantly.

3DMark 2001 has been a somewhat more questionable test; it blows up, too, but adjustments other than CPU memory took care of that.

Temperature has not been a problem, though once you start throwing over 1.7V at the PIV, the temps start mounting even with watercooling.

The Asus P4B533-E Is OK It consistently give you voltages about .05-.08 more than your setting. The voltage reading in BIOS seems to take a reboot or two before becoming more or less accurate.

Asus has been popping beta BIOS revisions left and right, the latest I have is 1007.003.

It’s not bad, though I’m going to have to take a look at the Abit BD7-II to have something to compare it by. I also need to see if some of the minor quirks I’ve had are the fault of the components or the mobo.

A problem many of you are likely to have with this board is:

1) unlike 845G boards, this board only has a 3:4 ratio, and more importantly

2) even that goes out the window at 133MHz or better and goes to a 1:1 ratio.

Not too big a deal for what I’m doing, but it could well be for you, especially if you plan to run, say, a 2.0GHz processor at 133MHz, and memory at 177Mhz or so.

Will do some testing shortly to see how much good that approach does you for real applications shortly.

When all is said and done, I think this board will get high marks for being a solid, reliable board, but it’s not really a tinkerer’s mobo.

Some Questions About The Corsair I need to do some more testing, but at 2.7V, it doesn’t seem to like to run at 2-2-2-5 at 166MHz. 3DMark2001 SE would blow up at that setting, and didn’t when I set it back to SPD. Need to do some more work on that.

The WD800JB Is No RAID Wonder Putting two 8Mb caches doesn’t seem to do much good, either in benchmarks or in real life.

Not that RAID normally give you a whole lot; RAID 0 only gives you a big benefit if you do a lot of long sequential reads. When you see a single WD800JB do better than a dual in some tests, and not dramatically better where it should, that’s not all too good.

Perhaps more importantly, the IBMujitsu 120GXP seems to give roughly equivalent, more consistent performances for less money. Some tests seem to indicate that the JB either does extremely well or extremely badly given the length of the test file: very badly for short ones, very well for long ones.

Not saying the IBM is better, just that you aren’t going to get the sun, moon and stars from a RAIDed JB.

PCMark2002 Is Not Ready For Prime Time One minor quirk that could give you great angst if you don’t know about it is that with the free version, when it runs a hard drive test, it tests ALL your hard drives. If you have a slow clunker in there, you’re going to end up with an amazing low score.

Nor can you change options in the free version, you have to pay for the Pro to do that (I HOPE that’s one of the options).

The far more serious charge against PCMark2002 is that either it sometimes yields absurdly high scores, or absurdly high scores are easy to fake.

By “absurdly high,” I don’t mean “scores high than mine” 🙂 but rather scores that are not even theoretically possible. For instance, sure, you can beat my memory score, but you can’t more than double it. Mad Onion appears to have cleaned up the more absurd CPU scores, but when you see people claim over three times the CPU score of someone you know is legitimate while running at 2/3 the speed; you know something’s wrong. The same looks to hold true for hard disk scores.

MadOnion is trying to clear out the absurd CPU scores, but when you find the “top” 175 memory scores all have the CPU score blanked out, I think they have a problem.

3DMark2001 SE: For Ti4600s Only Right now, I’m using a Radeon 8500. I get a 3DMark2001 SE score of a little over 10,000. Per any competition, I might as well be coloring the polygons by hand with crayon myself. In the 3DMark2001 race, I’m in the Special Olympics division.

The best score from a Radeon, from someone who huffed and puffed and pulled every trick in the book was a little over 12,000.

I looked around comparing scores, and the solution to my “problem” is very simple. Buy a Ti4600. 94 out of the top 100 scores used just that (and the other six were Ti4400s).

No other factors made as big a difference. Not CPU speed, not platform.

The point of this is not to say that nVidia is somehow cheating or whine about the unfairness of life; but: