Overclocking the Celeron 400MHz & 366MHz chips
Top
Bottom
Top

Feature Article
Overclocking the Celeron 400MHz & 366MHz chips

May 1, 1999
By Scott Lewis

If you have been following this column for the last couple of months you are aware that I have built a computer from scratch. All new parts including a Celeron 400MHz CPU, 128MB RAM, 18 GB 7200 RPM hard drive, RIVA TNT 2D/3D graphics accelerator, CD-Writer, 17" Sony Trinitron Monitor, etc.

Overclocking Part I

I have been reading a lot of information on overclocking lately. It seems that the Celeron processors are easy to overclock, especially the early ones. In fact, I have read rumors that the 400MHz Celeron is just 300MHz Celeron with its clock multiplier set at 6.0X (the 300 has its multiplier set at 4.5X).

I wanted to try overclocking. This was one of the reasons I chose a Celeron processor over a Pentium II. I went with the 400MHz chip thinking it would overclock the fastest, since it was the fastest rated by Intel. I later learned that the 300A (with the 128K cache) is the best chip for overclocking. Regardless of which chip is best; I have the 400 Celeron and have to live with it.

If you already know about bus speeds and multiplier speeds, skip this paragraph. For those that don’t know, two things determine the processor’s speed. The speed of the bus the CPU is sitting in and the multiplier speed inside the chip. The 400MHz Celeron uses a bus speed of 66MHz (as do all Celeron processors), and uses a multiplier of 6.0X. In other words the CPU steps up the incoming speed 6 times. 66MHz X 6.0 = 400MHz. I know, there is a little rounding there, but the true speed of the bus (in my motherboard) is 66.8 (66.8 X 6.0 = 400.8). The Celeron 366 uses a 5.5X multiplier, the 333 uses 5.0X, etc. You can do the math. Pentium II CPUs starting with the 350MHz chip all use a 100MHz-bus speed. So the PII 400 is running a multiplier of 4.0X (100 X 4.0 = 400). You can do the math for the other chips.

Time to see how fast this Celeron can really go. Remember, do this at your own risk. This voids the warranty. However, the chances of permanent damage are slim, and the worst you probably could do is blow the cost of another chip.

Since we always take things slowly when overclocking, the first step was to benchmark the CPU at its correct 400MHz speed. I used ZDBench’s CPUmark 99 for all tests. The 400MHz had a baseline score of 30.4. (Note: I must put a slight caveat here. Although I used the CPUMark 99 test, I didn’t do anything special to the computer as far as other software. CPUmark listed some programs that were running that could influence the tests. I left this alone because I wanted the tests to reflect the way the computer would be as I use it, not some stripped operating system only setup.)

The database that comes with WinBench 99 has three computers listed with 400MHz Pentium IIs. The first one was a Dell, and it scored 30.4. Cool, I tied a PII with a Celeron for about $200 less. They also list two PII 400MHz Xeon computers, one from Dell and one from Gateway. (The Xeon is Intel’s server version of the PII with its cache running at full speed) These machines had CPUmark scores of 30.8 and 31.6 respectfully. Excellent, my scores are just a shade below the best PII 400s.

I first wanted to know if it was true that the multiplier in the Celeron processors was locked. I took the conservative root and stepped the processor down to a multiplier of 5.5X. This should have me running at 366MHz. Power up and run CPUMark. I get virtually the same score. (30.6 if memory serves, but I didn’t write this one down.)

Next I set the multiplier for 4.0X. This will eliminate any possibility that things were close. It should have run at 267MHz. I got the same CPUmark score, and noticed the machine still read 400MHz when it booted. Yes, the multipliers on Celeron processors are locked. In the case of the 400MHz Celeron, it is locked at 6.0X. This disappointed me. I was hoping to run it at 100 X 5.0 for 500MHz. That way I could maximize my 100MHz memory. I knew there is no way in hell that the CPU would survive running 600MHz (100 X 6). That only left me two other options, a bus speed of 75MHz or 83MHz.

I reset the jumpers for a 6.0X multiplier. Time to get serious. I set the machine for 75MHz. Sure enough, the BIOS displayed a Celeron 450MHz Processor on startup. Cool. I run CPUMark and get a score of 34.2. Even better. That is a 12.5% improvement in performance for FREE!

I set the motherboard for 83MHz. The BIOS reports a Celeron 500MHz Processor, and starts to load Win 98. About a second into the splash screen (with the clouds and the scrolling blue line) the machine locks up tight as a drum. I have to pull the plug, it won’t even turn off by the power switch.

I tried setting the AGP slot for the video card to 2/3 speed. This will have the AGP running at 55MHz instead of overclocking it to 83MHz. It is designed to run at 66MHz. I also removed the sound card in case it is having a problem running at 41.5MHz (the speed the PCI bus runs when setting the CPU/Memory bus to 83MHz). It still hangs at the splash screen.

I reset it back to 450MHz, and leave it that way. I don’t have the extra fan for the case yet, and don’t have enough time to do anymore troubleshooting.

The machine runs flawlessly at 450MHz. I install all my applications, and played Starcraft over Battle.net. Everything is running great, and fast.

Overclocking Part II

I read that if the machine gets through its memory tests, reports the correct (for the level of overclocking) CPU speed, and locks in Windows, things are getting close. I also read that it is a good idea to lower the PIO Mode of the hard drives. This is a performance mode part of the EIDE specifications. Since the highest is PIO Mode 4, I set the hard drive controller to PIO Mode 3. I set the bus to 83Mhz, and the AGP to 2/3. Windows starts loading. The scrolling blue bar gets across the screen about 3 times, then I get a memory error.

This bothered me since I bought 100MHz PC100 compatible memory from PNY. Not cheap, and it shouldn’t have any trouble running at a slower 83MHz. I go into the BIOS and see that the default was not to use the ECC error correction. I turn that on and boot up again. This time Win 98 starts loading then suddenly drops to a blue screen to run a registry utility and locks up. This is scary. I never like it when the registry is messed up. I set everything back to normal speed (actually overclocked to only 450MHz) and boot up. Everything is running fine.

I try again. This time Windows 98 locks on the splash screen, but it scrolls the blue bar about three times across the screen. This is an improvement over Part I, since the scrolling got all of one inch before. Now I decide to boot to a diskette. I wanted to see if the hard drives were running all right at the PIO Mode 3.

Everything runs perfectly with a Windows 95 (OSR2) boot diskette. Unfortunately I don’t have any DOS stuff to test with.

My motherboard has a jumper for setting the voltage to a test mode. This voltage is designated for the RAM, AGP and something else I can’t remember. I tried this, but no change. It boots great to DOS, but Windows crashes. I have not removed the sound card, but decided that is not an option. If it runs without the Sound Blaster Live, then I might as well not run it at that speed. A gaming machine needs good sound.

Overclocking Part III

My sister-in-law and her husband wanted to buy my old computer for $500. (Pentium 100, 24MB RAM, 4GB drive, CD-ROM, sound card, 15" Trinitron monitor.) I told them I could probably build them a system for under $1,000 that would blow away my old machine. My wife and I wanted to give the computer to her parents as payback for them saving us $350-400 a month in daycare expenses. (They take care of my youngest son during the week.)

Once I told them I got my machine working the first day I put it together they agreed to let me build them a system. I mention this to you for two reasons. 1) I am going to show you next month the system I built for them. They ended up setting a budget of $1,100 without shipping or tax. (Basically $1,200 with tax and shipping where necessary.) I will detail their system next month, and show you where I put the extra money. I will be sure to tell you how to build a decent machine with name brand parts for about $1,000. 2) The processor I ordered for them was the 366MHz Celeron. This was $70 cheaper than my 400MHz Celeron, and it sacrifices only 8% CPU power. Definitely more bang for the buck. However, I will try to overclock the 366C in my computer.

I have heard it is possible to run the 366C with a 100MHz-bus speed. This is 550MHz. If this is true, I will keep the 366 and let my sister-in-law have the 400.

The 366C showed up at my doorstep from UPS. I put it into my computer and started running tests. Running at 412MHz (75 X 5.5) the computer ran fine. It got a CPUmark score of 31.2. I little better than my 400C at 400MHz. All is as expected. I ran the 366C at 457MHz (83 X 5.5). I did this with the AGP set at 1:1 for 83MHz. The machine ran fine. I had it running for about two hours. It scored 36.2, followed by a second score of 35.7. I was quite impressed. This is better than my 400C processor running at 450MHz. I attribute this to the memory running at 83MHz instead of 75MHz.

Time for the real test. I bump up the bus speed to 100MHz. The computer boots, reports a 550MHz Celeron Processor, counts its memory, and continues the boot process. Then it hangs tight as a drum. I try booting from a floppy. It hangs just as the computer tries to load the operating system from the floppy.

So it looks like 450MHz (roughly) is as fast as I will be able to go. Although my computer ran slightly faster with the 366MHz Celeron than it did with the 400MHz version of Intel’s chip, I decided to keep the 400. This would put less stress on the PCI bus than running it at 41.5MHz.

Overclocking Conclusion (for the moment)

From what I have read, and various e-mail correspondences I have had throughout this project, I have come to the conclusion that the Celeron Processors are topped out at approximately 450MHz without serious cooling measures. I have only heard of rare cases were a 300A was run at 504MHz (112MHz bus). I don’t think this was done without changing the voltage to the chip from the stock 2.0. If you have, or know of others that have, gotten a Celeron to run over 450MHz, please let me know.

My most recent inquiries have led me to believe that it may be possible to get the 400C to 500MHz if the stock heat sink/fan is replaced with a much better aftermarket unit. The next time I find myself in a computer store that sells these, I will get one and try one last time to get Windows 98 to run at 500MHz. I'll keep you informed.

As for next month... For those of you on a tight budget, I will cover the system I built for my sister-in-law. The total cost came to $1,135 including all shipping charges and sales tax. I will detail were they went over $1000, and even explain how you can build a real good gaming machine for around $1,500, maybe less.

Bottom