This is a case where the sequel is better than the original. 🙂
I spoke about my initial go with this yesterday.
Another day, another processor. Late in the afternoon, FedEx delivered another 1.6A.
This one was a Malaysian CPU:
Pack Date: 4/23/02
Oh, oh. Went over to our here, but thought I had gone to a World Cup rally site by mistake. It just kept saying “Costa Rica, Costa Rica.” 🙂
What I then did was to look for recent Malaysian entries. In this particular case, searching under “L2” would get me any Malaysian chips made in 2002 (just as “32” would get me recent Costa Rican CPUs).
What that told me was that recent Malaysian chips tended to do 2400 or a bit better at stock voltage or a little better.
That was a good sign.
Something else intrigued me. This chip has a number in that particular batch of 0576. Most of the other PIVs I’ve seen have had numbers like 0250.
Just to restate the relevant components used for this test:
Intel 1.6A PIV, Malaysia, week 8, pack date 4/23/02
Asus P4B533-E, BIOS version 1006, beta 1
512Mb Corsair XMS3000 RAM
Western Digital WD800JB hard drive
Enermax EG451P-VE power supply
SENFU Water Cooler II
Tests were run with the PCI/AGP ratios locked at 33/66.
Put the CPU in, set the voltage to default, tried first at 2400. Booted into Windows, the Asus threw high and gave me about 1.55V.
I ran Prime95 for a while until I was satisfied the machine was having no problem with it.
(People may scoff at my use of it, and I’ll grant you it’s hardly the most strenuous test out there. However, since I had watercooling, CPU temperatures were not a concern. In my personal experience (and I know some of you have had differing experiences), Prime95 blows up on me before other benches like 3DMark. Even better, it usually blows up FAST, which lets me get the range of a processor relatively fast with relative confidence. Once I get into the ballpark, then I use other benchmarks as a check.)
I then went to 2480; same voltage. Again, no problem.
Hmmm, we may have something here.
What I found rather interesting was the 12V readings. They did decline, but not quite as much as they did with the Costa Rican chip.
Next, 2560, still at 1.55V. Got into Windows, no problem. Ran Prime95 for a bit, then died. OK, can’t complain about that.
Time to go for the Holy Grail. Set to 166MHz, increased voltage to 1.65V, which gave me a bit over 1.72V.
It worked. This pleased me greatly. 🙂
Obviously, at least some current Malaysian 1.6A chips can do the deed. It’s not just Costa Rica anymore.
However, I did note that a 166-167, the 12V voltage readings were pretty much down to the levels I had run into at 155 with the Costa Rican CPU.
I predicted to myself, “I bet I’m going to have some problems getting much above this.” And I did.
I didn’t try to electocute the thing, well, at least not yet, since throwing more voltage in all likelihood would just worsen the 12V issue.
This leaves me with a few questions.
Does the 12V dropoff indicate a PSU weakness which a better power supply would take care of (and enable somewhat better speeds)?
Or rather, does the 12V dropoff indicate that a PIV is reaching its limits?
Or is this just a coincidence?
I’d rather it be the first, but even the second would be a good diagnostic tool.
I’ll get a new power supply Monday. In the meantime, I’ll play with the machine within the tested parameters to determine stability and performance,
Something I have to look into more is whether or not the Asus P4B533-E can be induced to stop whining about voltage.
One annoying thing it does is stop the boot process to report an error Hardware Manager found. The “error” is a voltage reading above 1.7V. You can resume the boot process by pressing F1, but a lot of people would likely find that that gets old fast.
I haven’t hooked speakers into this yet, so I can’t report on the Asus literally, audibly whining by beeping a bit (or a lot) as the voltage goes up.
This may just be a matter of telling the mobo to ignore the voltage readings; I just don’t know yet. Just don’t want people who’d go stark raving mad over this buying the board until I find out.
More Barking Up A Tree
Our CPU database asks for part of the code number on the CPU in the case of PIVs. It doesn’t ask for the last part of it, the four digits that indicate which one in the batch it is.
It might be useful to see if that makes any real difference in how well these chips overclock.
If I have learned anything the last couple years, it has been that nothing happens to everyone. There are always exceptions. What I get may well not be what you get, or even typical.
Ergo, for those of you with Northwood PIVs (and only Northwood PIVs; I’m trying to keep this simple), I’d like you to do either or both of the following:
1) Set your PIV to a speed where you know it’s just breezing through with little strain at default voltage or close to it.
2) Get whatever monitoring software you’re using up and running, and note the 12V and 5V while idling.
3) Run Prime95, or something like that, any stressing program that will let you keep an idea on those voltage while running. Note what happens to the voltage.
You will probably find that VCore will drop quite a bit (.05-.10V) while running the test. This is perfectly normal; the PIV was built to handle voltage differently than earlier processors. It’s designed to drop in voltage as it chews up more power.
4) Reset your PIV to a speed you know is stable, but close to the limit. Repeat steps two and three.
5) Send me an email telling me what giving me the details on what you did, the setting you used, and what happened to you. Be sure to indicate what power supply you used.
1) Look and jot down the complete third line of print on the CPU. It will probably start with a “3” or an “L” and look something like 3203A310-0123 or L151A247–0456.
2) Please send me an email with the code number, the maximum stable speed (what you run the CPU at normally), the voltage you use at that speed, your motherboard, your RAM, and the type of cooling you use.
This is not a competition. If you give me a result from a machine barely able to get a WCPUID reading before crashing, that’s only going to mess me up. I need honest numbers; I promise I won’t tell. 🙂
Don’t be afraid to report on a machine where you didn’t push it. That tells me something, too.
I thank you in advance for your emails. I may or may not be barking up wrong trees here, but the only way to find that out is to ask. This could yield some very valuable information.
P.S. If you happen to hang out in a forum where there’s a lot of Northwood overclocking going on, you might want to mention this article in that forum. You might also want to mention that forum to me in case I don’t already know about it.