got all parts for AMD dually - questions

[wquiles]

Thanks to cmcquistion I am ready to start building my AMD dually this weekend with the K7D Master L.

As I have the Thermalright SKL-800U heat sinks that attach to the motherboard, I can't very easily change the multipliers without removing the mother board.

I was thinking about trying a multiplier of 15 to start (which would give me 2250Mhz. I can probably go higher, but I want something that will run, and since I can always set the FSB to 100 or 133 I should be able to get the board running even with a multiplier of 15.

So, since I have those really good 1800+ that overclosk so well (XP1800+ Tbred B DLT3C:JIUHB:0311MPMW) and since I will be lapping the heat sinks and using artic silver 3 and I have 92mm fans, and since I want to get to FSB 150Mhz, what is a reasonable (and close to max) multiplier that I should try for my system? Is the 15 multiplier to small? Should I use 16 or 17?

William

 

[Blanks]

Those are the chips I had for mine too...

I was going to start at 15*, but first I wanted to be sure it all worked fine so I ran them for a few months at 12.5 using the BIOS options. Once I had them in and had found that they worked I couldn't really be bothered to take them out again and set the multiplier to 15.

My temps at 12.5*150 (1875Mhz) were around 40-50C at only 1.25v, which made me wonder just how hot it would run at 15*150. Bearing in mind that they could "probebly" do that speed easily, but there are two of them in there-thats a lot of heat to disapate.

In the end I sucessfully managed to break them by using the battery trick..... so thats the end of that...

If I were you I'd test them first at normal speeds and although its a pain to take the board out and change the chips it would definatly be worth the effort to make sure nothnig is wrong now rather than later.

Good Luck

 

[cccary]

I started out with 14X and 150 FSB and have never looked back. Had not opened my case since assembly.

 

[nodark]

I'd test the chips individually in a single cpu board that allows for some good overclocking like an nforce2 board
the reason being, it has multiplier, fsb, and voltage adjustments but the pci and agp speeds are locked and seperate from the fsb

just cause those chips are supposedly from a certain batch, it doesn't mean they're gonna overclock good and if you physically mod a chip that isn't a good overclocker, you'll be having all sorts of headaches

also, I know others overclock their fsb with mpx boards but it can cause probs as well cause the it doesn't have a locked divider

on my setup I have two overclocked (multiplier only) XP1800s as MP2600s with a vcore of 1.6v being cooled by two 800U (gonna be going back to w/c) with 92mm high output Panaflo fans at 7v and sucking air rather than blowing down onto the 'sinks (the temps dropped about 3C after switching the fan direction) and my temps right now are about 23C-28C over case temp, I imagine I could get my temps down lower but I prefer a low noise level

another thing is that the 800U is not a hard sink to remove
when you put 'em on, crazy glue or locktite the mounting screws so they don't come undone when removing the heatsink

 

[cmcquistion]

Why not run the system outside of the case while you check things out? That way, if you need to re-mod, it isn't hard to do and if something doesn't work quite right, it is easier to troubleshoot.

I always recommend assembling, at least a bare bones system, outside the case. Do testing and modding, then put everything in the case. In the end, this will really save time and effort.

 

[Deathknight]

Why not run the system outside of the case while you check things out? That way, if you need to re-mod, it isn't hard to do and if something doesn't work quite right, it is easier to troubleshoot.

I always recommend assembling, at least a bare bones system, outside the case. Do testing and modding, then put everything in the case. In the end, this will really save time and effort.

Sage advice here. It is unbelievably cramped in most cases with a dually rig. On my rig I have a paper thin gap between my 2 heatsinks. Then with the tall mosfet heatsinks all around that reduces room even further. Heh its hard enough to work on the thing outside of the case, no need to drive yourself crazy

I actually had my Lian-Li case sitting unused when I assembled my dually and I just attached the motherboard to the slide out mobo tray. That made it nice and stable to work with, but I woulda done it on a table even if I did not have the tray.

 

[wquiles]

cmcquistion and deathnight,

Yes, I am assembling the whole system outside the case first, as you suggested. It would have been way too hard to attempt this inside the case as I had to figure out how to modify the head sinks by trial and error.

In fact, I have been busy during the day and:
- was able to verify each individual chip coming up as MP (the L5 mod works great) !!!
- system also recognizes them as MP's when I installed both of them. They are defaulting at something like 1500Mhz or so (don't remember right now).

I have been busy recently modifying the heat sinks to fit since this motherboard came with larger caps than in the pictures that I have been and the Thermalright 800U's needed a lot more work to fit (they DO fit).

As to my original multiplier question, I was not clear enough, sorry. What I want to know is what is a "decent" multiplier that it is high enough to start testing at? I am still inclined at starting at 15 (first at 100Mhz, then 133Mhz, and eventually 150MHz), unless I get some other suggestions.

Thanks again to all of you for helping me with this "project". I have been taking many pictures and I hope to post about it soon

William

 

[cmcquistion]

Go ahead and try 15X, but realize that it might take a Vcore pin-mod to get the CPU's to post at 15x133 (you might have to change the default Vcore to a higher number.) This depends, of course, on how well the CPU's overclock and the only way to know is to try.

All in all, though, I think 15X is a good place to start.

 

[wquiles]

OK, I got both CPU's working at 133Mhz with the 15 multiplier (about 2000Mhz) under Win2K. Motherboard monitor shows about 36/39 at iddle which sounds OK for now. I did set the DDR to the best settings (per the sticky) and everything boots OK.

I got two instances of Prime95 running, but all it is happening is masive hard drive usage on the torture test and almost no CPU utilization (Win 2K does show both CPU's properly). As a result, CPU temps are about 1-2 deg higher than iddle. I sed the 9876 password and set priority to 10, but it had no impact.

When I run Prime95 in my old PIII system I never had this hard drive usage and temps do go up as a result.

What might be going on?

William

 

[cmcquistion]

OK, I got both CPU's working at 133Mhz with the 15 multiplier (about 2000Mhz) under Win2K. Motherboard monitor shows about 36/39 at iddle which sounds OK for now. I did set the DDR to the best settings (per the sticky) and everything boots OK.

I got two instances of Prime95 running, but all it is happening is masive hard drive usage on the torture test and almost no CPU utilization (Win 2K does show both CPU's properly). As a result, CPU temps are about 1-2 deg higher than iddle. I sed the 9876 password and set priority to 10, but it had no impact.

When I run Prime95 in my old PIII system I never had this hard drive usage and temps do go up as a result.

What might be going on?

William

You're not doing it, right.

Did you read my instuctions on how to setup and run two instances of Prime95?

 

[wquiles]

darn! - right on dude. I messed up!. Although I did properly created the second instance, I did not read your note as to how to change the memory setting to prevent the EXACT same problem that I had. I need to follow instructions better :-(

I now re-started the two instances of P95 and I got 100% percent utilization on both processors and none of that hard drive madness !!!

I am running stock voltage (1.49 or so) at 15x133 = 2000Mhz - will see if both processors last the rest of the night. If it does, how exactly do I start raising the FSB to get it to 150 while leaving the PCI buss "safe"? I have Mushkin 3500 Level2 CAS2 Memory, so I know the memory will work way past 150 (and propably at default voltage, right?).

Thanks again cmcquistion!

William

 

[cmcquistion]

150 is the max FSB the board supports. You won't have to worry about the PCI bus, at all. Even at 150, it is only running 37.5 MHz, which is well on the safe side.

Bump up one FSB increment at a time, and run a short Prime95 test. If it fails, bump up the Vcore and try again. Keep doing this, until you get to 150. Take notes, along the way, noting how much Vcore is required for stability at each speed. This can help, later, when you're fine-tuning speed vs Vcore vs heat vs noise.

 

[wquiles]

cool. I just followed your advice and have my system running at 150MHz. It took several tries but it has been running two instances of P95 at Vcore=1.725 @ 2250Mhz. The temps at load are 49/52 (ambient/case = 34) so I am not likely to try anything higher as I go not want to deal with even hotter temperatures at load.

I will let it run for the rest of the day, and if stable, I will try upping the voltage a little bit more to get some burn-in done as well.

Thanks much cmcquistion and Deathnight for helping me get to this point - two T-bred's 1800 at 2250MHz is pretty awesome !!!

William