[Princess Garnet]

Well, to those who already saw my comment before in another thread, here it is. For those who didn't, I got a Pentium 4 641 (3.2GHz, Cedar Mill, 2MB L2 cache, 800MHz FSB, Hyper-threading, LGA775, 64-bit, 65nm) in return for trading my parents my old E2160. It's no Core 2 or Phenom II or Core i7, but obviously I didn't get it as a replacement. I've always wanted to try and mess with one for the sake of it, and now I get to, so that's what this is for. I hoped I get pretty high with my motherboard and RAM.

I just put it in, so I'm just now starting, but here's the results so far (so far is the key phrase).

At stock speed at idle, I'm getting an unbelievable (no, really, I'm hardly believing it) 23C idle. If my experience with my E8600 told me anything, it was that after the first load it was put under, it'd end up settling a few degrees warmer. Still, that's unbelievably low, especially considering this is the "hot running Pentium 4" here.

Under Prime95 load (this was just to check load temperatures, not stability, obviously), I'm getting 37C-38C, with it touching 39C a few times. That's a mere ~100F. When's the last time you've seen a Pentium 4 that low? Not often.

For my first attempt, stupid me just went into the BIOS and set the voltage to 1.4125V (Intel lists 1.x-1.4V, so I assume the same ~1.5V for the 65nm Core 2 Duos is about the maximum, or would it be a bit less since there's just one core?), and set the FSB from 200 to 266 without caring what it'd result in, and booted. It was successful. It was at 4.256GHz.

I checked temperatures, and again, couldn't believe the idle temperature of 24C-25C. A 1.25GHz increase raised the idle temperatures ~2C. Are we sure this a Pentium 4 here? I put it under another prime95 load, and I'm getting ~49C-51C. I'm taking a guess right now that at this rate, the temperatures will under full load (idle seems unbelievably fine) will become an issue before the CPU, voltage, motherboard, RAM, or anything else does. If I can still get it to boot into Windows for a screenshot at 5GHz, I'll be happy. I tested it for ~15 minutes. After the test, it dropped down to ~27C, not the ~24C it was at before. As I suspected, I saw the same thing when I first tested my E8600.

Note that I didn't use the method of inching up the FSB slowly to find the maximum voltage for each range (if this was a CPU I was trying to "properly" test and to use 24/7, I would), so for all I know, the current 1.4125V is good for even more, but we'll have to see.

And try to see, I did. This time, before just cranking up the FSB, I tested to see what I'd need for 4.5GHz, and that was a FSB of 281, so I did that, and it booted successfully at 4.5GHz. Idle temperatures are up to ~27C-29C. I' currently testing it under load now (~12 minutes so far), and temperatures seemed to scale up just a few degrees more with the small frequency change (~53C-55C). It's starting to go up more exponentially and less linearly now.

Speaking of which, how will I know if the CPU is throttling? I assume once I start to see that happen, I'm too high on the temperatures. I read it happens around ~65C or so.

Also, for some reason, my motherboard isn't giving me control of lowering the multiplier from the default of 16. I'm not sure if this will matter, but I was able to alter it for my E8400/E8600.

 

[partypooper]

p4 like to thermal throttle....

 

[Phil]

test it using PCSX2

 

[bkwegoharder]

Nice you got a Cedar Mill P4. Those are really good overclockers, they hold overclocking world records. Also your temperatures will be good for way past 65C, It is around 100C that most Intel cpus start to throttle. 70-80 is where you want to keep them for longevity though. So people like 65C as a standard for longevity as well I noticed since that is what I hear a lot.

 

[nonoitall]

From what I hear, P4's notorious temperature issues were often at least partially due to the coolers and thermal compound that shipped with them.

So where are the performance benchmarks?

 

[Princess Garnet]

test it using PCSX2

I did. At 4.5GHz, in Dragonball Z Budokai Tenkaichi 3, it get's between roughly 1/3 and 2/3 full speed in a 1v1 battle.

Also your temperatures will be good for way past 65C, It is around 100C that most Intel cpus start to throttle. 70-80 is where you want to keep them for longevity though. So people like 65C as a standard for longevity as well I noticed since that is what I hear a lot.

I don't know about it being so high.

Maximum CPU Temperature | Hardware Secrets

My CPU isn't listed there, but if that list is anything to go by, it's 70C-80C. You might be thinking of the older Athlons that had a rated limit of ~90C or so.

So where are the performance benchmarks?

I'm overclocking it just to see how high I can get it, not test it's performance any.

 

[ElijahTW]

Pretty impressive.

 

[nonoitall]

I'm overclocking it just to see how high I can get it, not test it's performance any.

I know.

 

[Princess Garnet]

Well, I'm having problems trying to get 4.8GHz, or even 4.75GHz, to boot. I can get it to boot to the BIOS, but Windows won't load. Either it gets past the BIOS to where it would show the loading screen and instantly restarts back at the beginning of the BIOS (i.e., it'll boot loop), or it gives a message about how Windows failed to load, and if I try to continue loading, it fails and reboots again.

I found this.

pentium 4 631 d0 overclocking

It seems I've been using slightly more voltage than needed (maybe, maybe mine needs more). I've been using ~1.4V for 4.5GHz so far, but I couldn't even get 4.75GHz to boot with 1.6V, and if that picture is anything to go by, that should be enough for 5.0GHz, let alone 4.8GHz, so I'm fairly sure it's not a voltage issue. It's not RAM or MCH/FSB. They have the same voltage when I was using them at 400MHz for my E8600, yet they're running far below that speed, so it's not them. I tried upping the FSB Termination and PLL voltage only slightly, and nothing. I can't figure out why it won't boot above this speed no matter what. I have to be missing something.

 

[Phil]

Loosen your timings and see if that would help.

 

[bkwegoharder]

I don't know about it being so high.

Maximum CPU Temperature | Hardware Secrets

My CPU isn't listed there, but if that list is anything to go by, it's 70C-80C. You might be thinking of the older Athlons that had a rated limit of ~90C or so.

That list is very inaccurate. It lists the q6600 in the 60s even though I can do a quick and easy google search and see plenty of people sitting with much higher temperatures. The same can be said for a lot of other cpus on the list. It seems like they are using the specific heat rating that Intel gives to the cpus. That isn't the maximum temperature for whatever cpu you have though. The max temperature is the tj.Max of the cpu. You can check it with coretemp or realtemp.

I hope that you can get to 5GHz though that would be nice. Maybe you can compare the cpu you have at 4.5GHz to your Core 2 Duo at 4.5GHz so we can see how much faster clock for clock a core 2 duo is. Disabling 1 core in the core 2 and disabling hyperthreading in the pentium could give a real close and accurate comparison. I'm guessing the Core 2 is 2x faster clock for clock.

 

[StriderVM]

As far as I know, Cedar Mill processor have the highest overclock records, at least in regards to pure clock speed alone. Good luck.

 

[Princess Garnet]

Either way, temperatures aren't the thing holding me back, so I guess it won't matter. RAM timing shouldn't be an issue. If it can do 5-5-5-18 at 400MHz, it can do the same timings at 300MHz. I got it to boot at 4.8GHz, and I can get the BIOS to boot at 5008MHz, but it won't boot to Windows. I think voltage is the problem. It took a little under 1.62V to boot at 5.0GHz to the BIOS. I probably could get it there at ~1.7V, but if people were getting 8GHz+ at ~1.9V, I don't feel it's worth it to go for 5GHz at 1.7V when it does 4.5GHz at less than 1.4V.

So, I'm impressed at how easy it was able to get up there as high as the Core 2s with equivalent hardware, but I'm disappointed at the same time that I hit that huge exponential wall higher than ~4.6GHz. My E8600 did the same thing around the same frequency, but it wasn't as pronounced. I was hoping to get some ~5.2GHz Windows screenshots, even if it wasn't 100% stable under load.

I think the E8x00 might even be faster than just 2x this at the same speed. This is just one example, but we all how well PCSX2 takes into account the CPU. At 4.5GHz on this CPU, with all the speed hacks on (strangely, all hacks on gave very little boost versus all off, in three separate games), it got between 1/3 and 2/3 full speed. The E8600 gets full speed 99% of the time at 4.0GHz in single core mode. Final Fantasy X was "almost" playable. I got about 2/3 speed average (between 1/2 and full typically). Considering my old 2.8GHz Northwood got ~15FPS last I tried it, that surprised me. However, I assume in other things, it may be less. It all depends on how the application is coded. Some may benefit more from the Core 2, and some may only a bit. At either way, I'd guess the ~2x is a good rough middle ground average (I think Intel stated something like 40%-150%+ when the Conroe Core 2s came out).

 

[Mad]

See if the temps are the same using Realtemp. Sometimes everest doesn't report the right temps.

 

[Princess Garnet]

Temperatures aren't the issue, and RealTemp doesn't work on Pentium 4s anyway.