SGI: Hardware

I thought it was anything over 300. Anyone with a 350 MHz PIMM who can check it out?

Oh! In that case I'll open up a 360 when I can find a moment... though if it's over 300 then by definition a 300 should be ok.

Ian.

I think I might have found your problem with the Octane PM20s upgraded from 250.

It appears that SGI used a 1:1 ratio for the cache for 250 and under, while 300 (at least for Origin) uses 5nS SRAMs for a 200MHz cache clock (2:3). Unless you swap out the setup SEPROM with one that changes the cache timing it seems you're running the 250MHz cache at full processor speed when you bump it up to 350, and I don't think the SRAM has enough tolerance. Note I haven't actually opened up my PM20 yet, so this is mostly conjecture based on documentation and hinvs. 250MHz would require 4nS chips, 300 would require 3.3nS, and 350 2.8nS (for a 1:1 divisor), and my guess is there isn't quite enough headroom in the 250s cache chips. If you have a 2MB dualie and 2MB 300 singlies you can pull the SEPROM and swap that as well and you should be good (a 1MB dual 250 would need a reprogram to deal with the different cache size, as would the chips from the Origin PIMMs w/ 8MB SC).

I so need an EPROM programmer

Two cents worth on the O2 CPU modules.

With the exception of the 400 MHz R12k, I think they were all the removable CPU's. Once when I was messing with a module, I got the machine to start booting at 400 MHz (with a 300 MHz chip) but it didn't get very far. A good R12k-300 chip will usually run at 350 reliably, but not much more as you're running into process limits on the silicon. I've found with the Octane modules that I don't think I've had one that had any cache issues assuming your chip will run at 350, and from what I've seen the O2 modules are similar. 350 is probably the highest safe overclock then for an R12k module. Anything below 250 MHz doesn't have the necessary split-plane power supply and can't really be upgraded, so they're basically called 'test modules' around here.

Of interest is modding an R12k-400 to 450 MHz, but I only have one module and would hate to break it, as they're quite rare these days. Probably be the worlds fastest O2 at that point.

I did a lot of messing around with cache speeds on the Octane. There is a single CPU module - originally a 400 - kicking around out there I did quite some time ago, and think it ran at 470 MHz with the stock cache divider. This is the exception rather than the norm.

Attached is my little cache-speed calculator for Excel, it takes the actual 'resistor divider' number and module crystal frequency, and spits out the final CPU clock and cache speed.

For an Octane PM, it's still resistors, Shown is the 1.5 and 2.0 positions, useful for all but the slowest PM's. The duals are similarly configured, though the BGA dual boards have a slightly different layout.

http://www.nekochan.net/wiki/gallery2/v/SGI_ ... d.jpg.html

Thanks for the replies!!

So what about using the core from an Octane 360 module, putting it into an original O2 R12K/300 module? Could this then be adjusted to give an O2/425 or better?

I could move the resulting spare 300 core into an O2 250 module and either leave it at 300 or maybe oc a little.

I wondered whether using an original O2/300 module to hold a core from an Octane/360 would make changing the clock to 400+ easier compared to trying
to put anything quicker than a 300 into a 250 module.

However, atm I'm stuck anyway because I still don't know what resistors to move in order to attain a 3.5X multiplier on an O2 module.

Ian.

The problem I believe is the 360's and 400's are soldered to the board and are not removable. The R12k-360 is basically a binned R12k-400, so basically if you're overclocking these you are limited to how far you can push the chips.

I have a dead O2 module that has had the 350 hack which surfaced recently, I will try to take a pic tomorrow and post it up for comparison.

Oh!! I thought the soldered ones started at 400. Ah well.




Ok, thanks!

Ian.

Can't believe I actually missed this thread; silly to resurrect it like this, but... has anyone got proper/clear instructions on how to overclock an r12k 300MHz to 350MHz in an O2? I'm reading a multiplier needs to be set at 3.5 from a stock 3.0, but $random_deity knows were or how...

If relevant bits were copied to the wiki, we would have a reference point for this topic.

R.

I still don't know what the resistor positions are for the 350 setting. Got a stack of 250s to mod.

Ian.

(bump)

Does anyone happen to have pics showing the correct resistor positions for setting 350MHz on an O2 R12K CPU module?

Ian.

I thought that the R10K/R12K set its clock frequency (PClk, a multiple of the SysClk signal from the board) based on a 64-bit word written by the chipset at reset time. So I don't understand how moving resistors can affect that. The multiplier is part of the data contained in the serial ROM on the module, I thought.

And I thought a DM 10 was something special

In reality, the resistors change the multiplier just like on a old peecee .... been there, done that.

Yep, but I'm still not sure what the resistor positions should be to set a 350Mhz clock. Does anyone know? Or
preferably have pics? Searching the archive doesn't help, web/image link refs are defunct.

I've done a 300, worked fine, but I'd like to try a 350.

Ian.

Also still interested in this. My O2 already is a 300MHz, so 50 extra wouldn't kill it... but I don't want to risk blowing it up for lack of proper instructions.

(another bump...)

Dr. Dave, did you ever have a chance to take a pic of that dud 350 module which you found?
Are you still here? Or can someone else with a working 350 mod supply a pic showing the
resistor positions?


I've just completed five R12K/300 mods for O2 (ie. moving 300 cores into 250 modules, moving
resistors to give 300MHz), but alas the 6th mod did not work (red light). Digging around, I found
an R10K/225 module. Moving the 300 core over gave a module which boots ok to the PROM,
but an IRIX boot fails just about when it starts to mount file systems (screen goes black, then after
a pause the display changes to the prompt saying it's ok to turn off the system). I can boot into
single user mode, but there's nothing in the SYSLOG.

I tried changing the crystal speed to make the CPU run at 250 instead, but the behaviour persists.
Shows up ok in hinv as R12K/250. As I type I'm running the ide tests (so far it's passed icache1,
icache2, dcache1, fpu, and scache1 to scache3; currently running scache4).

Any ideas anyone?

Ian.