Cheapo 3000+ Winchester underwhelming OC

A

alan.p

Limp Gawd
Joined
Apr 30, 2006
Messages
409
I read through Eclipse's A64 sticky guide, and attempted to follow it through. My comp is as follows --

3000+ A64 Winchester (failed to grab stepping before putting the HSF down)
Gigabyte K8NF-9 mobo
TwinMOS 2x512MB RAM (nothing spectacular)
Sparkle 6600GT

The mobo is, I suspect, what is really hamstringing my efforts. I'm not a hundred percent positive as to what speeds the RAM can manage before it coughs it's lungs up.

I first went through the "FSB (HTT) OVERCLOCKING" section. The K8NF-9 bios seems to have two modes -- double or nothing. You can change the CPU multi down, and you can wind the voltage up, but nothing else. I couldn't even set a RAM divider.

Heck, half-ass-ery never stopped me before. I put the CPU multi to 5x, and booted into windows. An instance of Prime95, Speedfan, and CG-NVNF4 later, I was gingerly increasing the FSB a few MHz at a time. (Gingerly, since I'd mucked about before and knew that this thing does not like being pushed at all.) It got to 225MHz and Prime95 packed a wobbly. Fair enough, I stepped up slowly from 221MHz, finding 224 to be stable.

Well. The guide says that if it can't do 250, something is wrong. The damn thing couldn't even do /half/ that.

This FSB wouldn't do miracles, but the heck with it. I went on to "CPU OVERCLOCKING". I wound the CPU mult back to 9x, and went for gold (rust, at least) with 224MHz, gaining me a whopping 2.016GHz.

I didn't even /bother/ with RAM overclocking. The mobo wouldn't let me tinker with RAM settings beyond the voltage. (No timings, no nothing.)

I bought the Gigabyte since my previous mobo was a Gigabyte 7ZXE for my 1GHz Morgan core Duron, which would happily run all day at 1.2Ghz. 20%, or double the overclock that this thing will do. Without really tweaking it at all.

So. My final questions are --

Without a doubt, I need to get a new motherboard. My TJ06 has a flipped tray, BTX style, so it can't have any heat pipes (Don't want to take chances of the things not working upside down). SLI is of less than no use to me -- I'm in this OC thing for the savings, not for the play aspect. What would be the best brand? Gigabyte is great for stock speed, but it ain't so hot for overclocking.

Also, is my TwinMOS ram (no fancy labels) possibly the next bottleneck I'll encounter should I get a new motherboard?
 
I think your RAM is actually holding up your overclock in this case. When you set the CPU multi to 5x and pushed the FSB to 224mhz, your RAM was running at DDR448. I doubt generic chips can do much beyond that at stock voltages.

What you'd need is to be able to set a RAM divider. That way you could increase your overall FSB speeds (which overclocks your CPU) but keep your RAM as close to 224mhz as possible.
 
Use Memtest86+ to test the ram for errors as you up the FSB.
Just set it to load from either a floppy or CD before Windoze loads and let it run a loop or two.
If you get any errors at stock speed then the memory is BAD.
If you start getting errors as you overclock then your pushing your memory to fast.

Once you've found the max the memory will run at then its load Windoze and Prime time.

Luck........ :D
 
Thanks -- I'll check the RAM when I get home from work tonight. Well, I think I'm definitely going to get a new mobo now.
 
also, if you have a CBBID stepping winchester, don't bet on getting much more than 2.3-2.4ghz :(
 
If I ever meet the bios guy for Gigabyte, I will be having unpleasant words with him. A hidden, undocumented key combo (ctrl-F1) unlocks all the necessaries, like HT link and memory tweaks.

I went through the guide again, this time following Eclipse's instructions a bit more religiously.

HTT OC'ing. HT link mult of 2x, CPU mult of 5x, and a RAM divider of 100MHz (or 1:2). I stepped it up from 200 through steps of 4MHz (I'm impatient), finding that it crashed at 290MHz. So I'm putting the usable max HTT as 286MHz.

CPU OCing. Stock CPU mult, 2x HTlink, RAM of 1:2. I stepped up from 200MHz in steps of 4MHz, and Prime95 errored at 248MHz. Rather than wait, I just dialed back to 240MHz for "stable". (Quick and dirty.)

I tried RAM OCing, only for the damn memtest disc to not work. Eh. I'll assume that the 224MHz ceiling from pre-hidden-bios is the RAM ceiling.

The results of this? An increase from 1800MHz to 2162MHz, with the RAM going at stock speeds. Weird, but true, as the HTT is 240MHz, with a divider of 5:6, resulting in RAM of 200MHz. 20% increase, not up to scratch for the general 'net, but it rings true with my past OCing experience with my Duron. The K8NF-9 will OC, it seems, but it is still not very happy about *letting* you.

An Asus A8N-SLI will be arriving next week, I'll transplant the CPU to it (and finally find out the stepping!), and find out more about any limiting factors.
 
Hey at least you got a 20% overclock. That's free performance.

CPU-Z will tell you your CPUs stepping.

Did you try other RAM dividers? You can continue to play with the FSB, and with the memory dividers. Try to bump your FSB up a few more MHZ, but use memory dividers to keep your RAM running somewhere between 200mhz and 224
 
Got the new A8N-SLI board today, and promptly disconnected the TJ06 to do a heart transplant on it. A quick test showed that the CPU would still nae go nae faster.

I did find the stepping, though.

ADA3000DIK4BI
LBBID 0609APAW
Z610949 C50876

A quick look with google doesn't turn up any good OCs with LBBID stepping. Luck of the draw, I guess.
 
Had a play around with voltage today, and was quite surprised at what a bit of juice could do.

At stock voltage (1.4V), I managed 240MHz.
At +0.5 voltage (1.45V), I managed 250MHz.
At +0.1 voltage (1.5V), I managed 260MHz, and didn't want to run over +0.1, since I'm running stock cooling. My copy of Clockgen - NVNF4 didn't want to do more than +0.5V, for some reason, so I had to reboot and put +0.1V in the bios.

So. The way I see it, my options are as follows --

1. Run at 2160MHz, 240MHz FSB at stock voltage. Stock mult (9), HTT mult of 4, RAM at 166:200
2. Run at 2250, 250FSB @ 1.45V. CPU x9, HTT x4, RAM at 166:200
3. Run at 2340, 260FSB @ 1.5V. CPU x9 HTT x3 RAM 133:200

... sailing closer to the wind in terms of reliability with each step.

Just thought that it was mildly interesting. The results of increasing voltage surprised me!
 
You must log in or register to reply here.
Back
Top