Tale of a Pentium IV 3.20 GHz

[Chris Nahr]

Seeing how there are people wondering if they should build their own system, I'd like to share a little tale...

When I recently upgraded my system I installed the Intel "boxed" fan for the new P4 3.20 GHz CPU. Big mistake, as it turned out.

The fan worked but it was incredibly noisy, like a vacuum cleaner. First I tried the "Q-Fan" function of my Asus motherboard which would keep the fan turned off until the CPU reached a threshold temperature, 50-55°C, and then run the fan for a few seconds to cool it down again.

That worked... until I packed a large archive with WinRAR. Apparently this particular algorithm heated up the CPU so quickly that the fan couldn't kick in fast enough. Result: repeatable system lockup whenever Q-Fan was activated. So I had to turn this function off and decided to get a silent Verax P16Cu cooler instead.

Well, now I had to remove the Intel cooler. Apart from the usual mechanical challenges, there was a nasty surprise: Intel had smeared the contact surface with about a ton of heat transfer paste, and the damn thing stuck to the CPU like superglue!

I used a hairdryer as recommended in the assembly instruction to heat up the paste so that I could pull out cooler and heat sink. It didn't work, the whole CPU was ripped out of the locked ZIF socket along with the heat sink.

Then I had to continually heat up the whole thing with the hairdryer for about five minutes, all the while inching the CPU off the heat sink with a screwdriver. Then I had to wipe the whole mess off the CPU using petrol and a soft cloth, naturally.

It's a wonder that the CPU still works after all this torture, but it does. The rest of the operation was uneventful.

Morale: Removing CPUs and cooling systems is much more difficult than installing them, so make sure you get the ones you want right away! And think twice about upgrading your existing CPU...

[Jason McCullough]

Adhesive remover (Goof Off or Goo Gone here in the US; yes, the names are that stupid) gets it off easily without scratching up the heat sink.

You'll probably want to replace it now, as all the scratches screw with the heat transfer.

[SpoofyChop]

I was fairly annoyed with the process of transferring my Athlon 2200+ from the original (fried) MB to the new MB.

Although getting the heatsink fan combo off the MB was easy, and getting the heatsink off the CP was easy, the thermal pad left a ton of tacky stuff all over the die and I had to sit there scraping it off with alcohol and qtips.

:(

Again, not for the faint of heart.

[Chris Nahr]

Adhesive remover (Goof Off or Goo Gone here in the US; yes, the names are that stupid) gets it off easily without scratching up the heat sink.

I'll have to look around if any such special chemicals are available over here. There aren't actually any visible scratches on the heatsink, I've used a very soft cloth to avoid that. The process was rather lengthy and tedious, though...


You'll probably want to replace it now, as all the scratches screw with the heat transfer.

Looks good enough to me, another drop of heat transfer liquid should be enough to fill any invisible scratches. But I don't intend to re-install it anyway because it's so damn noisy.

[John Reynolds]

I just bought a 3.06 P4 a few weeks ago and I immediately noticed that the new fan was definitely louder than my 2.4's fan.

[Desslock]

I just bought a 3.06 P4 a few weeks ago and I immediately noticed that the new fan was definitely louder than my 2.4's fan.

Agree. Noticed the same thing. I just moved my PC on the floor though, so it doesn't really bother me. I'd rather have a cool processor than risk mucking around, especially after Chris's tale.

[Chris Nahr]

A wise decision. I just tried that WinRAR operation again, now with the Verax cooler running and Q-Fan disabled, and guess what? The system crashed. :roll:

Then I activated another one of those Asus motherboard features, a heat guard that slows down the CPU when it runs too hot. That seems to help -- right now I've completed the RAR operation. Even though I haven't actually seen the clock speed reduced... hmmm.

Anyway, I noticed even before that experiment that the Verax cooler lets the CPU run quite a bit hotter than the Intel cooler. It was 30-35°C with the Intel cooler but always 40+ degrees with the Verax cooler in idle mode. So I guess it's smart to keep the Intel vacuum as long as you don't mind the noise.

Edit: Apparently that "thermal guard" didn't do anything after all. Instead, a "Turbo" setting that was enabled by default seems to have been the culprit. God, I hate this crap. If I wanted a CPU with heat problems I might as well buy an AMD, and save money too!

[Chris Nahr]

Jesus Christ on a hyperthread. Looks like this problem wasn't heat-related after all. The system is completely stable in all benchmarks and RAM tests but WinRAR still locks up the computer, sometimes after running only for a few seconds on a cool CPU.

So I guess RAR's solid archive algorithm somehow trips over the CPU's hyperthreading feature. That's my best guess at this point. Other packing algorithms with the same data work fine. The crashes with Q-Fan and lack of crashes with the Intel cooler that I noticed initially must have been random bad luck.

Oh well, that's annoying but better than a hardware problem. Hope I don't run into any other programs that don't work on HT CPUs. :cry:

[Kevin Grey]

So I guess RAR's solid archive algorithm somehow trips over the CPU's hyperthreading feature. That's my best guess at this point. Other packing algorithms with the same data work fine. The crashes with Q-Fan and lack of crashes with the Intel cooler that I noticed initially must have been random bad luck.



FWIW you can disable hyperthreading in the BIOS if necessary.

[Jason McCullough]

It's not HT, I use WinRar on my 2.4 all the time and it works fine.

[Chris Nahr]

Well, I've isolated the issue further. WinRAR works fine with "fast" compression (-m2) but always crashes with "normal" or better compression (-m3) when working on that specific data set.

All other packers that I've tested pack the same data set without crashing (tar/gzip, bzip2, zip-implode, zip-deflate, rar32, also WinRAR in zip mode). This includes packers that heat up the system quite a bit. I've re-run some benchmarks as well, and they're still stable.

I'm pretty sure by now that it's not (only?) a heat issue since -m2 works just fine (repeatedly!) even when the CPU is already hot but -m3 will crash a cool system in the first few seconds.

Further tweaking BIOS settings didn't change things. However, when I tried WinRAR with -m3 and higher on different data sets I didn't get a crash.

It seems that only the combination of these specific files (very large and sparse data files used by a stock market program) with this specific compression algorithm and setting, running on this specific system, is causing the crash.

Hyperthreading is the only significant difference between this system and my old one that I can think of, hence my guess that it must be involved somehow. Disabling isn't really an option since XP would then have to be re-installed AFAIK (because "multi"-processor systems use a different kernel).

Anyway, as long as this operation is the only one to cause crashes it's probably not worth the effort trying to hunt it down any further. I'll just use a lower compression level for those files.

[DennyA]

I've never had a crash with Hyperthreading on using WinRar, and I've had my P4 3.06 for a year now.

You can disable Hyperthreading in the BIOS without reinstalling XP. I've done it a couple of times. It doesn't even do any device discovery upon reboot -- you sure about that Kernal thing?

[Chris Nahr]

You can disable Hyperthreading in the BIOS without reinstalling XP. I've done it a couple of times. It doesn't even do any device discovery upon reboot -- you sure about that Kernal thing?

I thought I'd read it back when HT was new... but if you say it works I'll try it and see what happens.

[Chris Nahr]

Okay, I tried disabling hyperthreading in the BIOS. Kevin and Denny were right -- Windows did not explode. Unfortunately, that WinRAR operation still crashed. :cry:

I tried packing a couple of other big files with -m3 or better. No problem at all. It's just that particular fileset.

My best guess at this point is that WinRAR has a really obscure bug related to threading (not HT) that only appears at this speed and with this data; or that I've found a really obscure bug in the current P4 production run.

[Jason Cross]

Intel boxed CPU fans aren't really THAT noisy under normal circumstances, but it's really easy to end up installing them such that they make a ton of noise.

I've installed several where the wires running between the heat sink fins aren't tucked in neatly and rub up against the bottom of the fan blades, that's pretty loud. I've also had times where the fan blades actually physically scraped (ever so slightly) the top of the heat sink, which made for a ton of noise as well. It's nothing a thin nail file or emery board can't fix, but it shouldn't have to.

Long story short: Intel boxed CPU fans aren't the greatest in the world, but they shouldn't be so loud that you go looking for alternatives unless something is wrong. You might want to take a close look at your intel HSF, Christoph.

[Chris Nahr]

No, I don't think anything's wrong with the Intel fan. I know the sound of wire scraping on the fan blades, and I'm positive they were out of the way. I just looked at the fan again, moved it manually, and don't see it making contact with the heatsink. Previous Intel fans I had were fairly noisy as well, by the way, just not as loud as this one.

Apparently I'm extremely sensitive to noise, compared most computer users anyway, and I've carefully silenced my computer system. The noise of an Intel fan may seem normal to you, but it made a huge difference in my otherwise very quiet system.

Have you ever used a silenced system, with Verax or similar fans, rubber suspension for the hard drive, and anti-noise padding? It's a world of difference compared to standard parts. Mine makes actually less noise than either of my consoles (PS2 and GameCube) -- except when I use an Intel fan.

[Jason McCullough]

Yeah, same here. The intel fans make a disgusting amount of noise under load.