Yeah, 2-300 mhz was that max most people got out of 423 systems.

The recoded for the 1.6ghz 423 p4 OC is 2.272ghz
hwbot.org/submission/852837_cegosf_cpu_ ... )_2272_mhz
With the max stable being around 2ghz.

Also there is a powerleap adapter on eBay right now. New for 160$
https://m.ebay.com/itm/PowerLeap-CPU-Upgrade- … 8%257Ciid%253A2

Yeas its over priced but thought id share it non the less.

I always look at hwbot for guidance and a rough guide on what to expect 😀 Iguess your "max stable" claim comes from superpi and wprime results?

No way I'll touch that adapter. It's cheaper to find a 478 board with RDRAM support to "spiritually" continue this project than buying that.

2. Modding and overclocking, pt.2

A second experiment with the voltage regulator. This time instead of just getting rid of the resistor going into the FB pin, I put it back in place and created and setup a voltage divider with a 50 kOhm potentiometer I had laying around. The idea: report less voltage/current than is actually being fed to the CPU to fool the voltage regulator into giving us more. Here you see a 1.8 kOhm resistor in the pads where there used to be a 2 kOhm resistor (I didn't have 603 handy, so I used the slightly bigger 805), and a couple of wires going into pin 7 (Feedback) and into pin 9 (Ground).

It worked alright as expected, with the caveat that the IC has a safety feature which disallows outputing more than 2 V or so. I was able to boot into Windows at 2080 MHz, but 2133 is still a no go…

The machine seems stable, but the overclocked PCI bus gives the SATA controller a rough time. I will try using a SATA to IDE converter and see if it behaves better.

I could install a similar voltage divider to fool the regulator into thinking its outputting less than 2 V, but I think I'd rather try another CPU first. 2 GHz seems to be where the voltage wall for this one is, I could boot up at 1.77 V and the machine was stable, whereas any higher requires quite a bit more of voltage and is unstable.

More later.

Nice work. Not many go this far to OC SK423 systems. Really neat to see such work. Hope you can push well past 2ghz.
I wonder if the NB could be holding you back? did you try messing with its voltage?
You may be able to push around 2.2-2.4ghz with a 1.9ghz or 2ghz P4. to me 2.4ghz would be a nice speed to hit for such a system.

chinny22 wrote:

P4, RDRAM, Modern? about 3 differ levels of obsolete in that title 😜
But I like what your trying to achieve , and will be following this thread 😀

Yeah, the irony of the title is not lost on me 😀 Thanks for the interest!

Jade Falcon wrote:

Nice work. Not many go this far to OC SK423 systems. Really neat to see such work. Hope you can push well past 2ghz.
I wonder if the NB could be holding you back? did you try messing with its voltage?
You may be able to push around 2.2-2.4ghz with a 1.9ghz or 2ghz P4. to me 2.4ghz would be a nice speed to hit for such a system.

Thanks for the continuous encouragement!

I haven't really looked into the i850 itself. I figured the same chip is used on this board's younger sibling (TH7II) and that board supports Northwoods at 533 MHz, so it shouldn't be a problem.

Almost all the functions are disabled, too. I'm only keeping USB on at the moment, but I'll try a PS2 keyboard and disabling it to see if it helps (don't think so but hey, I've seen weirder things while overclocking).

I'm glad you find it interesting. I'm having a blast, there's something about RDRAM that just makes me happy 😀

Possibly. Wile the chipset may support 533mhz on newer boards, the older boards may not have the same setup that allows it to hit higher FSB at stock voltage.
Take x38 and x48, they are the exact same chipsets, but x48 and the boards are binned so it can hit 1600mhz at stock voltages wile the x38 can hit 1600mhz, but it needs more volts most of the time.

I'd not rule it out so quick.
A good way to tell if you hit a FSB wall or something other then the cpu is holding you back is to dump the cpu multi as low as possible and up the FSB. If your stuck at say 142mhz FSB and still can't get over 142mhz FSB with the cpu running well below stock speeds its not the cpu holding you back.

Also disabling anything that is not needed can indeed help when your really pushing a system with overclocking.

EDIT:
I can't recall, are 423 CPU's locked like the Piii? I could be wrong but I think you can lower the multi on 423 P4's.

Jade Falcon wrote:

I can't recall, are 423 CPU's locked like the Piii? I could be wrong but I think you can lower the multi on 423 P4's.

Locked.

Unlocked downwards would make troubleshooting much easier, like you say.

I'll keep playing with the machine on the weekend and see if I can squeeze some more out of it.

3. Birthday splurge

Small update: as a part of my birthday celebration I spent some cash on a few items for this machine. A used 1 Gb Ethernet card, a brand new USB3 card…

…and a PL-P4/W adapter.

This adapter was sold with a bundled 1.6 GHz Northwood, but it only works with s478 Willamette CPUs. Indeed, I tried it out and it would not work out of the box. I bought it after examining the datasheets for Willamette and Northwood CPUs in 423 and 478 sockets, to see if I could mod it to run Northwood CPUs. Also they are on the rare side, which I get a kick out of. And finally, I also appreciate that it adds to the parallels between this system and my previous build: both are in essence MaximumPC Dream Machine builds (though I didn't intend them to be that), and both use—hopefully, we'll have to see about this one here—adapters to run more powerful CPUs.

Attempting to make this adapter work will be my end of year retro-project—fun! 😀

---
As an update, currently in the machine I have the 1.6 Willamette overclocked to 1920 (16*120) which leaves the PCI bus running at 30 MHz. At these speeds the SATA card behaves properly. The LAN card worked without any trouble, but I've yet to try the USB3 card. I tried both Windows7 Starter and Windows7 Pro and to my surprise W7 Pro feels snappier out of the box, something I have no explanation for. Disabling Windows Search made it even better, too! This is good news because I have a legit W7 Pro license that I can use for the machine.

Also, reading around I found out that the GeForce 6800 I am currently using in the build does not have hardware decoding. It was advertised, and the PCIe version has it, but the AGP card never could do it. I will try a GeForce 7800GS and will read about Radeons to see what's up.

I'll update later when I try the USB3 card and maybe post some benchmarks of it.

You may need a bios update for Northwood cpus to work.
I know that was the case with the abit bw7