Best type of overclock for a folding box

[donny_paycheck]

Alright, comrades, here is the story:

I have a 1400 Thunderbird, but it's the old green-cored "B" Athlon with a 200mhz front side bus. It has a locked multiplier but I've used some trace tape to handily unlock it. It is on a MSI K7T Turbo mobo (KT133A chipset) with 1 512mb stick of Kingston PC133 SDRAM. I have a 8045 heat sink with a 50CFM Sunon fan for it. This machine will hopefully soon be folding full time...I will use it for nothing else. What I am wondering is this...

Since this chip will probably run with a 266mhz front side bus, should I lower the multiplier to 10.5x and raise the bus speed to 133mhz to run it at 1400/266? This is what I would do normally for all around better performance, especially on a DDR system, but since F@H relies mainly on the CPUs internal FPU performance I am thinking that clock speed should be my main concern. I could do several things:

1. Leave the multiplier at 14x(100FSB) and overclock using the front side bus.
2. Drop the multiplier to 10.5x, raise the bus speed to 133mhz (yielding 1400mhz) and then see if it goes a little higher (140mhz or so).
3. Raise the bus speed to 133mhz and overclock via the multiplier primarily and a little FSB also - i.e. 11x135 or something.

So basically I'm wondering for FOLDING ONLY, what is the best way to overclock this CPU so we can catch [H]?

 

[Mr B]

Trial and error, just like any overclock, but I'd err slightly towards the direction of stability...

I've got a 1.0 T-Bird (266 bus default) in my "3rd" rig... I left the bus alone, and raised the multi as high as it would go (from 7.5 to 10.5). It's stable as heck at 1400. IF I wanted to push my luck, it will run at 10.5 x 150 (1575MHz), but again, I erred on the side of stability.

Same with "Rig 2", which sports a 1.4 T-Bird (also 266 default). This one won't take nearly as high a percentage overclock as the 1.0. It maxes out at about the same 1575 (so far, though I have had it at 1600 briefly in testing), so I just pushed the multi up to 11.5x, making it a 1533MHz chip.

both have been Folding very stable at these settings, and where both are at 133 (266) FSB, everything else in the boxes are in spec.

Both chips are turning in those 5 point monsters in 13 to 14 hours.....

Conversely, the XP1800+ in my "Main" rig is multi locked (not by choice... ), so that one is overclocked by raising the FSB (to 148, or 1702MHz)

I'd start out by raising the multiplier first, and go from there...then try dropping the multi, and setting the FSB at 133, and raising the multi again. Trial and error. Find the "edge of stability", and back off a couple of notches from that....it'll likely be a combination of multiplier and FSB that you'll wind up with.

Good luck.

 

[res0r9lm]

your right, focus on higher clock if it main a folder. try every combination to see what the highest overclock while maintaining temps staying in the low to mid 40's while folding. I have found my tbird start getting unstable at 44c once I go over 1.6

 

[muddocktor]

Donny,
I'm not familiar with your mb but you might run into the same problems I ran into with the KK266 board I used to have and a 100 fsb 1.2 t-bird last year. After I unlocked it and ran the fsb up to 133, I couldn't warm boot the computer; the only way to boot was from a stone-cold bootup and I mean physically everything cooled to ambient temp. The problem has to do with semi-jumperless boards and the way that the board starts up before the bios takes control over vcore and the multipliers. Anandtech had a real good article on the problem. I ended up selling the 100 fsb 1.2 t-bird to a friend with a KT7R board and getting me a 133 fsb 1.2 t-bird to replace it. A couple of things you can try if you run into a boot problem on a 133 fsb is to pencil all the L7 bridges so the default vcore on your proc is 1.85 v and you can also look into changing the default multiplier by connecting or cutting bridges in L2 or L3. Cutting bridges is hard to do though.

As far as crunching WU's I don't think that fsb speed makes too much difference as much as processor speed itself.

Fold on, man!

 

[Mr B]

As far as crunching WU's I don't think that fsb speed makes too much difference as much as processor speed itself.

That's the way I've heard it, too... If the bus is at 133, and the multi is high, and it's stable...that's fine. Overall cpu speed is the goal. (Stability and temps need be watched, of course...)

Three out of four of my rigs run at 133FSB... I just like the idea of not having to worry about something cutting out due to being out of spec... I'd hate to have my NIC snuff it because I wanted to run at 150FSB, when pushing the multi up another .5x nets the same result, while keeping the bus in spec.

 

[donny_paycheck]

As far as crunching WU's I don't think that fsb speed makes too much difference as much as processor speed itself.

EXCELLENT.

Thanks guys, this is basically what I needed to know. I'll go for the highest clock speed in this case, be it via FSB or multiplier or a combination, but instead of favoring the highest FSB overclock like I normally would, I'll go for the most megahertz.

Fold on!

 

[donny_paycheck]

pushing the multi up another .5x nets the same result, while keeping the bus in spec.

And I don't want a remotely monitored folder crapping out on me and sitting idle because it locked up or rebooted randomly! I'll go to 133FSB and then bump the multiplier up to 11x and see what happens. The 8045 HSF should handle a reasonable oc.

 

[walaka7]

Ive noticed, too that pushing multiplier first then a little fsb creates a higher stable O/C then dropping mult and pushing fsb or locked and pushing fsb.. I ended up doing multi increases on both my rigs to get a more stable O/C after trying to push through fsb

 

[ThePunkGeek]

I've got a 1.0 T-Bird (266 bus default) in my "3rd" rig

"3rd" rig lucky you no wonder i can't get higher on the stats i only have my "main" rig
but it folds 24/7 so its all good