Help needed with 3000+ winchester overclocking

[Kinghunter]

Current Specs:
Gigabyte K8NF-9
a64 3000+ winchester
Kingmax PC3200 2x512mb

In my first venture in overclocking I followed Zebo's guide.
My max numbers seemed somewhat low
Max HTT: 270Mhz
Ram: 230 at 2.5-4-3-7 1 T
Max Cpu speed: 2.15Ghz @ 1.5v and 2.1 @ 1.4v multiplier at 9
(I assume when windows starts up and after login screen only the desktop wallpaper loads means you have gone to far? and the computer is just not acting funny)
HTT Multiplier is set at 4x
The temps never went above 39 degrees during any stress tests under full load


I was hoping for atleast 2.2Ghz without pushing it to far unless im missing a crucial setting or combination any suggestions would be appreciated.

 

[o1die]

I would set your ram at 166 and 2T, and try for higher than 270.

 

[Kinghunter]

When I found all the max numbers the memory was on a 2:1 divider. So if I did that would that help much?

 

[chusteczka]

The 512MB sticks of RAM work well at the 1T setting, this being the advantage of 512MB stick over 1GB sticks of RAM. For this reason, I would recommend leaving the RAM at the 1T setting but adjust the other timings to the factory tested settings. Would these settings be 3-4-4-8 for KingMax model MPXD42D-2K? It may be best to start with these factory tested settings to find the max RAM speed. The RAM timings can be adjusted from there afterwards.

Regarding your latest comment, it would be best to find the highest RAM speed without a memory divider, at the full speed of the RAM at 200MHz, or a 6:6 divider.

The 2.15GHz setting for a Winchester seems low since these chips are often able to reach 2.4GHz. (9 cpu multiplier * 270MHz HTT = 2430MHz cpu) It seems to me that something was not set correctly when finding the max cpu speed but do not worry about that for now. The most important is the max HTT that was found to be 270 and this is sufficient to reach the max typical speed for a winchester cpu. Although this max of 270MHz may also be able to reach higher if other settings are made appropriately.

The next most important setting to find is the memory. This is what will limit you. My patriot RAM had a max speed of 230 and I run it at 225MHz. This provides the following calculations.

HTT MHz * memory divider * cpu multiplier = system speed
225MHz HTT * 6/6 memdiv = 225MHz
225MHZ * 9 = 2025MHz

225MHz HTT * 6/5 memdiv = 270MHz
270MHz * 9 = 2430MHz

Find your max memory setting, at the full speed of the RAM at 200MHz (no memory divider set), to perform your own calculations.

Then adjust the memory divider to 6/5 or 166MHz (5/6 * 200MHz = 166MHz). Then see how far the system can be pushed, by 2 or 3 MHz increments at a time.

I would recommend using Prime95 to test the system. Select Options | Torture Test and let this run. Most instability problems with the memory are found with a miscalculation after 2-3 hours. For me, Prime95 was able to find such instability before memtest was able to. It is for this reason, I recommend testing with Prime95.

 

[Kinghunter]

When I had first found the max memory speed I had no dividers on so that max speed is still in the 225-230 range. The default speeds on the memory seem to be 2.5-4-4-8 according to cpu-z. So I manually set the speed at those and was able to up the htt to around 250 but it was not prime stable.

Currently I am at 240*9=2160Mhz temps still at 35-38 range while running prime.
In order to go any higher do I need to up the CPU voltage it is at 1.425 currently.
Or should this chip be in the 2.2-2.4 range without it?

 

[chusteczka]

My AMD64 3000+ is currently running at 1.525. I think I remember having difficulties similar to yours until I manually set (raised) the voltage to 1.525volts. I recommend raising your voltage and expect that to help.

 

[cubeless]

i need 1.55 to get 2.4ghz out of my 3000... and my kingmax only does around 210 relaibly, so u may need to drop the divider...

 

[Kinghunter]

I just thought of something after failures of prime after raising the voltage to 1.525 attempting 2.2Ghz and failing then at 2.15 and failing after 6 hours.
My motherboard is a Nforce4-4x variant which I believe means its standard HyperTransport speed is 800?
So if thats the case instead of shooting for the Fsb x ht multiplier= around 1000 should it be around 800?
My computer wont boot with a 3x HT multiplier only 4x, 2x, and 1x I found that strange.

 

[chusteczka]

That was one of the first thoughts that stuck in the back of my mind. My HTT is set at 3x since this bus speed is not sensitive, meaning 600MHz will not give any different performance from 800 or 1000MHz.

If your HTT cannot be set to 3x, then try a bios update, and hopefully this bug has been fixed in the latest bios.

An immediate solution would be to try setting it to 2x and see what happens.

I did not think to look up the specs for your motherboard.

 

[DerwenArtos12]

Ok, your max mem speed is 220-230. What dividers do you have available? I would suggest, if available, a ddr166 multiplier or 5/6, set the ram to 3-4-4-8 1T, ram voltage to 2.8, cpu voltage to 1.55, push the fsb as high as you can with htt at 3x if possible or 2 if not, it does make some difference so a bios update would be advised. That should have your ram at ~215. Go and download a program called ClockGen, it'll allow you to adjust the FSB on the fly and test it as you're going. From the data you get off that, shave 10-15%, set your bios to match and work your way from there.

 

[Kinghunter]

Well after updating the bios the issues still remain. I cant get the 3x HT multiplier to boot at any HTT/FSB speed same goes for Cas 3. Currently at 2.2Ghz 245*9 and 245*2.5 for the HT. It is prime stable but is there performance loss for running the hypertransport so low?

 

[jjet67]

I think Zebo posted a thread that hypertransport speed does not really matter. What matters is the speed of CPU. Once setting for 2.2Ghz is stable for prime, I would increase the htt 5 clicks at a time until it gets unstable. Some people gets crazy overclocking results and some don't. So, please don't be upset if you don't get a similar result.
JJET

 

[Kinghunter]

I got up past 2.2Ghz and a lot of strange things started to happen yet prime was still stable after 12+ hours so I reverted back to stock on everything. Now when 2 people are logged on or when the computer is left on overnight and someone signs in a blank screen with a cursor is all that appears. Upon manually restarting it a windows critical error shows up saying that there was a hardware failure (either ram, mb, cpu, or power supply). Is something shot?

 

[Kinghunter]

Now I am getting delayed write failures off my secondary hard drive. Dont know if it is related but something is definetely messed up.

 

[the cobbler]

Kinghunter, something that you might not have thought of that might be causing the problem.

many Winchester with date code 0451 and later have problematic memory controllers. I had a 0503spaw Winchester 3200+ that wouldn't do over 250FSB on certain motherboards, unless the HTT multiplier was dropped to 2x or even 1x. iirc, it's called "cold-boot HTT bug". On my MSI K8N Neo4-F, the bug is @219FSB to this day. on my Chaintech VNF4-Ultra, it used to be 220FSB, then a bios update put it at:
255fsb w/ htt multi of 1
250 fsb w/htt multi of 2
245 fsb w/htt multi of 3
go figure

When the cold-boot htt bug first appeared on some mobos, the Winchesters in question were hitting the wall at 219FSB. Some companies got their bios straightened out (DFI, Epox) many did not (MSI, Chaintech, Abit, probably Gigabyte). but then even the improved bios found wall of 240-255FSB, depending again on the motherboard.

whatever is what that was keeping these chips < 250FSB was fixed with the Venice memory Controller. interestingly, Winchester made before week 0451 do not suffer the cold-boot HTT bug.

just a fwiw, as you may have one of these Winchesters