So if you install a Win7,
Wouldn't do that anyways. I'd stick to Windows XP. Is that affected as well from OS microcode updates or was that a thing of later OS versions only?
So maybe it's just the thermal paste dried out after all these years.
Interesting thought but in this case, I wouldn't bet on it.
I mean this is such a hard wall I'm facing with this CPU and its temperatures were below 55°C all the time, so I don't really think heat is an issue here. To my knowledge, the usual behavior is that you go up in clock speeds and if the voltage is too low, it's slowly becoming unstable with errors facing up after some time of load. It is then that you up the voltage one notch and check again, usually resulting in a stable cpu.
This CPU however is not slowly becoming unstable, it goes from being perfectly prime stable over hours on end to "no, I won't do that!" within the smallest possible increment of just 15 MHz additional clock speed. It's the same wall I've seen many times on water cooling back in the days when I was chasing the perfect P4 CPU back in 2003/2004.
However, one has to be careful with this anyways because as far as I know, Intel used both variants (paste & solder) back in the days.
M0 Cores always were soldered, so every SL6Z3 to SL6Z5 as well as their fully functional EE CPUs were all soldered. Delidding shouldn't make much of a difference.
D1 Cores were available in both variants. Most of the D1 were with paste but later CPUs also sometimes came soldered.
D0/E0/G1 Precott cores were always soldered as well, most likely because temperature was already an issue anyways
There's a post on tecpowerup about this as well:
One more thing. I don't believe it has anything to do with high end or low end. It seems to be all socket 478 CPUs after a certain date got soldered. Around(or slighty prior to) when Prescotts were released. Except the Gallatins(which were soldered before any others). IOW low speed P4s and Celerons got soldered too(so far as I've gathered).
EDIT: Here's
that post.
I've recently been working on a project to figure out how to "decap" integrated circuits using cheap and available tools and chemicals. This has mostly been with smaller chips but I decided to do an old socket 478 Pentium 4 as a surprise to my followers. But I was surprised myself to find that...
www.techpowerup.com
Willamette
1 x Pentium 4 1.8(SL5UK) 1.8GHZ/256/400 D0 = Paste
Northwood
1 x Celeron 2.0(SL6VR) 2.0GHZ/128/400 D1 = Paste
2 x Pentium 4 HT 3.06(SL6PG) 3.06GHZ/512/533 D1 = Paste(on both one), Solder(on the other)
1 x Pentium 4 HT 3.2(SL6WG) 3.2GHZ/512/800 D1 = Solder
Gallatin
1 x Pentium 4 Extreme Edition 3.2(SL7AA) 3.20GHZ/2M/800 M0 = Solder
1 x Pentium 4 Extreme Edition 3.4(SL7CH) 3.40GHZ/2M/800 M0 = Solder
Prescott
1 x Celeron D 335(SL7DM) 2.8GHZ/256/533 D0 = Solder
1 x Pentium 4 HT 3.0E(SL7E4) 3.0GHZ/1M/800 D0 = Solder
1 x Pentium 4 HT 3.0E(SL7PM) 3.0GHZ/1M/800 E0 = Solder
1 x Pentium 4 HT 3.2E(SL7KC) 3.2GHZ/1M/800 D0 = Solder
1 x Mobile Pentium 4 HT 548(SL7X5) 3.33GHZ/1M/533 E0 = Solder
2 x Pentium 4 HT 550(SL7PP) 3.4GHZ/1M/800 E0 = Solder
1 x Pentium 4 HT 550(SL8K4) 3.4GHZ/1M/800 G1 = Solder
Coming soon maybe...eventually...probably not...
Prescott
1 x Pentium 4 HT 3.4E(SL7Q8) 3.4GHZ/1M/800 E0
www.vogons.org
I'd have to check the date of the CPU mentioned in my post stuck @ 3990 MHz to see if it's from around the Prescott days which would then hint on wether it likely has paste or solder in the first place but again, having seen this issue so many times on water no matter what temps and what core, I still think it's sometimes just a general issue on this architecture if there's a hard wall like this and not just the usual gradually decline in stability which can be countered by just a step more voltage.
