SGI: Hardware

i've done that last year.. w/o problems
just a bit tricky soldering ..with a big common iron solder

can you get us more details?

I didn't do any soldering. I mean I just did the first stage only, ie. swapping the CPU core, which should result
(am I right?) in an R12K/225, but it didn't work, just gave a red light. Tried a second core from an Octane/300
CPU, same result.

Putting the R10K core back in, still a red light. Worked ok before the swap attempt. Something gone fubar...

I can't think of any other details that would be relevant.

Ian.

i swapped an r12k 300 from octane cpumodule to o2 r10k 250 one
just unscrew swap and screw in again
at the first stage i got r12k 250 working
after second stage it bumped to 300

..maybe some bad contact between cpu and board? ..inspected contacts?
..tried the r12k back again in the original module?
..some problem on cpumodule socket? tried a bare r5k on it?

radrob wrotes:
> i swapped an r12k 300 from octane cpumodule to o2 r10k 250 one just unscrew swap and screw in again

That's what I did; messed up somewhere...


> ..maybe some bad contact between cpu and board? ..inspected contacts?

Inspected, nothing visible.


> ..tried the r12k back again in the original module?

Of the 2 Octane modules I tried, with the cores back in again, one works, the other does not.


> ..some problem on cpumodule socket? tried a bare r5k on it?

(R5K? It's an R10K machine)

A separate R10K/225 module works fine.

The one I tried to mod has it's original 225 core back in, but it doesn't work now.

Ian.

Ahh, the red light. Unless the PM is baffed itself, check the compression pins in the connector socket. It's not so much that the pins work loose or misalign, but its fairly easy to gunk up one of the contact surfaces, or have a miscellaneous bit fall into the socket and cause contact problems. The R12k-300 should have worked fine, as should the original R10k-225, they have the same voltage requirements.

Pulled the module apart just now, cleaned, etc. Still a red light. Ah well.

If I have more than one 250 module then I'll try one of the 250s next. Wierd though.

Ian.

strange thing

seems to be something wrong on the cpu module itself
but i can't mind what...

fingers crossed for the 250 one..

p.s.
despite the r5k - r10k machine class identification
you can swap cpu and cpu socket from each others
like try an r10k on r5k mobo and vice versa..

Yes, I know they're compatible; I've often replaced a bad mbd in an R10K machine with one removed from an R5K unit.

I need to check how many 250 modules I have. I'm not going to bother if I only have one of them.

Ian.

Btw, has anyone modded an R10K/195 module? I gather from Joe's original posts that it needs some extra work; is it
worth bothering with? Or too much hassle?

Ian.

Does anyone have a copy of NeZetiC's images/article? Looks like his domain is gone byebye.

Ian.

A snapshot of the site from August 2008 is still available via the Way Back Machine <today, anyway - but only that single day's snapshot - so you probably better not drag your feet>..

How's your French? NeZetiC.info: SGI O2 Overclocking : de R10K à R12K

Thanks!!

I'm also trying to find an image of the Octane dual-PM which shows the other resistors that would need to be changed
for a 250->350 overclock. The images in the dual-300 overclock gallery don't show all the resistors. Naturally, all
of Joe's images are long gone.

EDIT: no need now, sorted it out. Will do some decent scans at some point, add them to my site somewhere.

Ian.

My Website is still running. I've only lost the .com domain name.

http://www.nezetic.net/index.php?id=38

If you have any question, do not hesitate to PM me.

Thanks!! For the moment though I'm focusing on the dual-250 mod; built two modules so far.

Ian.

Sorry to resurrect an old thread...

Does anyone know which resistors to change to set a 3.5X multiplier on an O2 module?

I've built the R12K/300 modded module ok, the pictures in this thread are fine for that (3X multiplier), but I'm
not sure what needs to be moved to change it to 3.5X.

Some benchmark results for the modded 300 module are now on my site (Lynx compilation, RC5 and movie
conversion), performance is identical to an original 300.

Ian.

PS. Just looking at my previous post - the modded Octane modules pretty much worked out ok (dual-R12K/350)
but with one peculiar exception: they passed all the benchmark tests on my site, including the 'heavy' tests like
the Alias render, but trying to clone the system disk onto another drive with xfsdump/xfsretore would almost always
cause a cache error. Wierd...

Ian - what are the cache chip speed differences between 250/1MB and the 300 modules? I'm looking into doing this on my Octane with an O2k/300 PIMM-sourced R12k, but will leave the PIMM intact if it looks like things aren't going to work.

Er, not sure offhand. It was some months ago I was fiddling with the Octane modules.I'm focusing more on O2 now.

Ian.

Today someone asked me about the O2 mod images, so I thought I'd scan in my own diagram I made for changing a 250 module into a 300:

http://www.sgidepot.co.uk/o2/o2modresis ... to_300.gif

Hopefully that's a bit clearer than some other descriptions I've seen.

However, I still don't know what resistors to move in order to have a 3.5X multiplier (to get 350MHz). Anyone know?


Also, does anyone know if the core from an Octane 360 module could be used to get an O2 425? (ie. by moving the core into what was
an original O2/300 module).

Plan:

1. Convert orig O2/300 modules into O2/400 or 425 (whatever's possible if I can use a 360 core from an Octane module).
2. Used the 300 cores removed from O2/300 modules to convert orig O2 250 modules into either 300 or 350 (whatever works best).

I figure one couldn't move a 360 core into a 250 module as the cache would be too different, yes?

I have a couple of 360 modules I can expt with. However, I'd need to know a bit more about the multiplier resistor positions to try it.

Ian.

So are the O2 300 modules' R12ks BGA?

I highly doubt it... I was under the impression from earlier comments that the BGA stuff didn't start until the 400 speeds. Hence why I
wonder about being able to fiddle with a 360. An O2/425 would be nice. Orig 400s are too valuable to stick in hobbyist systems.

Ian.