SGI: Hardware

dexter1 wrote: I have finished the translation of the overclock article, and will be contacting Gemm today about the exact publication (his site or my site or...) of the story. As soon as the page is up, i'll post it on the news, so you can all read it...

akimmet wrote: ...

As for the CPU in the socket they are the same and can be freely exchanged between modules of the same CPU type. This sometimes applies to CPUs of the same pinout too. I’ve heard of people successfully putting an r4700 in their r4600 module for example.
...
If you would happen to have a serial EEPROM programmer it could be easily possible to make a 300+MHz r5k Indy by buying a fast r5k (a r5000a might work too) CPU then modifying the contents of the 8pin serial EPROM to change the multiplier to an appropriate value. This technique also applies to at least all other Indy modules too just keep the replacement in the same CPU family.

orionpi wrote: R4700???

I belive the R5000a had a differnt physical package, but if you want to fab a PCB a R5500 might work too.

orionpi wrote:

akimmet wrote: ...

As for the CPU in the socket they are the same and can be freely exchanged between modules of the same CPU type. This sometimes applies to CPUs of the same pinout too. I’ve heard of people successfully putting an r4700 in their r4600 module for example.
...
If you would happen to have a serial EEPROM programmer it could be easily possible to make a 300+MHz r5k Indy by buying a fast r5k (a r5000a might work too) CPU then modifying the contents of the 8pin serial EPROM to change the multiplier to an appropriate value. This technique also applies to at least all other Indy modules too just keep the replacement in the same CPU family.

R4700???

I belive the R5000a had a differnt physical package, but if you want to fab a PCB a R5500 might work too.

Satoru wrote: Given the voltage problem, maybe I can try to put the 200MHz I2 CPU in the 175MHz module swapping the oscillator instead of using the 150MHz...
The bad point is that the 100MHz oscillator in the I2 module is larger than the one on the Indy, so I may need to buy a new oscillator.
Do I need to do anything to the serial eprom in this case?
Anyone can point me to a link that may help me understand the voltage of the R4400 CPU from the CPU marking?

Thanks
Marco/Sat

The 4x00 series CPUs are 5V devices, and get very hot. The R4x00 CPUs will have a huge "porcupine" heat sink on the CPU chip itself, for heat dissipation. The MIPS R4600 CPU was redesigned using a smaller geometry, and a 3.3 V process, which means much less heat dissipation. The 4400-175, and 200 MHz. also use the 3.3V process, but still have a porcupine heatsink. The R4600 CPUs have a much smaller heat sink, no bigger than the ceramic PGA package, and about .3" tall. The R5000 CPU family also is a 3.3V process. SGI produced some modules with just a heatsink on the CPU itself, and other modules with a fan/heatsink combination. Also, some R5000 modules had a 5V / 3.3V DC-DC converter on the CPU module, and some did not. Not sure of the details on this yet.

akimmet wrote: ...

Althought I have found there ARE some r4600 cpus out there (I have one) with the huge heatsink.

orionpi wrote:

akimmet wrote: ...

Althought I have found there ARE some r4600 cpus out there (I have one) with the huge heatsink.

The larger heatsink is not a sure sign of the CPU being 5v. My Indy R4400-175SC says 3.3v on the PCB but has the larger heatsink, I'm assuming that my R4600-100PC is 3.3v even though it has the larger heatsink.

I still can't figure out the voltage requlators on the R5k Indys, mine doesn't have one but I can see where one could be soldered on, maybe it was only needed in early models. To my knowledge all R5000's are 3.3v, the R5000a is the lower voltage model.