<- Menü
(please have javascript enabled for this)
May 11th, 2009 Another 10 reworked modules have passed all tests and are under way to their owners.
Or have arrived there already :-) ... only three bare CPUs left now.

If anyone ist still - or again - interested in a 600MHz CPU replacement:
hurry up! the latest price is 160.- Euro plus swap-in of a R5271 CPU module.

September 3rd, 2008 Some minor news: the first of the last three O2@600 CPU module leaves for Slovakia on Monday.
The other two are also thoroughly tested and work fine, one of which Iīll keep for myself as a spare.

Good news: I shot me another 15 brand new bare CPUs at eBay USA.
If anyone ist still - or again - interested in a 600MHz CPU replacement: please ask for a quote.

July 30th, 2008 A year passes like nothing... all CPU modules are sold in the mean time,
and about a dozen replacements were made on trade-in modules.
The repair shop has moved to another location and became even more professional:
Every single module comes with a copy of the x-ray inspection image now,
and no one ever showed up not to work immediately.

Still O2 owners do want their modules upgraded from time to time :-)

January 21st, 2007 Itīs been a long time since the last update was made here...
Not that nothing happened - quite the opposite.

I got me some R5271 CPU modules from china in 2006 and had them converted to 600MHz.
My expectation was that hordes of O2 owners would flock in and want to have their CPUs replaced...
SGI went broke in the mean time, abandoned IRIX - and whoever in the industry had a couple of R14k@600MHz Fuel machines
began to sell them off at eBay for cheap.
I myself got me five of those... on the other hand, still have a crate full of 600MHz O2 CPUs.
So I also began selling them at eBay. Some of them quite under price in order to get at least some of my invest back.

Eleven more 600MHz O2 now exist out there ;-) and I always watch if one shows up at eBay.
None did.

In the mean time, I found out that RM7000C-600T CPUs do not require a heat sink at all.
Under heavy load, the CPU reaches a temperature of only 33°C on its surface.
Which only rises above that if the fan in the PSU is blocked.
Four modules are left - two of them are of the "no-blue-patch-wire" type which obviously cannot be reworked to 600MHz - donīt know why.
Iīll keep the last two working 600MHz CPUs on stock in case someone wants a replacement
(which, btw., I now offer for 265,- Euro, shipment free).

I guess Iīll start to offer replacement sets (CPU & SPROM only) at eBay soon.

March 31st, 2006 The last two CPU modules came back today, both said working.
Both do as far as PROM monitor diagnostics are concerned,
but again one of them only boots into the shutdown message. Sigh.
A first visual inspection revealed that a capacitor broke off...

If anybody out there reads this and has a 030-1493-003 revision B CPU module,
would you please be so kind and flip it over to have a look at C535B on the upper left (7th from top)...
Iīd like to replace it with one of correct value.

And I very much hope that this is the reason for the auto-shutdown behaviour of this CPU module.
I didnīt update the image gallery on this site yet, but it is about time to do so...
now everyone have their CPUs back, right?

None of them feedbacked as to be _not_ working.

Finally :-).

Thanks a lot to you all for your great patience!

I myself am now the proud owner of at least three CPU modules that behave erroneous,
and twelve not-yet-modified ones :-) ... in order to get back a few bucks on my invest,
Iīm forced to go on with this kind of "business" 8-( ...
luckily there are still people willing to participate.

This time, I wonīt collect modules before... Iīll alert you as soon as I have some available,
tested good both in PROM diagnostics _and_ IRIX.
Then Iīll either sell modules or swap them for sent-in ones.
Still no R7000 CPUs here, by now. Time for another... wait.

March 6th, 2006 I have sent out another three boards and Iīm hoping for the last two that are missing.
Unfortunately Iīm pretty busy with different things at the present time
so it is likely to take another while until I can update the image gallery on this site also.

Yīknow, to show you some bent pins and other fine desasters of the like.

March 2nd, 2006 (few) Good news: some modules have been reworked and are ready for delivery.
I have fetched them today but not yet tested. One module that has been working already
but from which the connector came loose now showed a badly bent pin in that connector.
Shit happens...
weīll swap connectors from another module to that one... another wait for some of you, sorry.
There are now only two people left without a reworked and working module, and I fear
that I will be able to compensate only after the next batch will have begun to be processed.

Since I have no idea if it will be possible to find these mysterious PROM-OK-but-no-IRIX errors
this has begun to gnaw into my savings badly. In worst case, Iīll be left with all the defectuous boards...
which I really dislike.

It will take 10-12 weeks until the next batch of CPUs (R7000C@600) arrive.

January 13th, 2006 Good news: half a dozen people have shown up so far for a next batch.
Weīll start as soon as I get new CPUs (will be ordered these days).

Sad news...
two of the CPU modules that already passed all diagnostics on the PROM monitor
arrived back here late this week. Both do not boot into IRIX correctly, either cause
bus errors or panic... I suspect the connectors to the main PCB, since there are
visible deviations relative to its soldering pads - which may cause contact to be weak.

No idea why the PROM diagnostics donīt care a bit about that. Less critical timing?
(only wildly quessing here)

I returned another board to the repair shop because of this type of problem earlier.
For the next batch of modules Iīm going to build a support with a connector fixture,
in order to avoid the connectors to become loose or swim sideways.

January 8th, 2006 Iīve taken the time to improve this site a little.
There is an additional "me, too" page where you can find advice in case you want one,
but are in doubt if you got the type of O2 CPU that is suited for rework.

There now is an "Advantages" page, meant to tell what a 600MHz O2 can do,
and a page called "other Mods" for... (guess what!)
...other moddings that can be done to SGIs O2.
Feel free to contribute!

Last - not least - Iīm working on a glossary, and Iīm about to learn
how to HTML-link its terms to the appearances in the text.

January 7th, 2006 Todayīs news: there are some - but few - R5271 CPU modules on stock, whatever for.

To all of you who where interested in participating in the next batch to process:
the plan is to do this in February.

January 1st, 2006 A happy new year to everybody!

December 29th, 2005 My PFC (Positive Feedback Count) has risen to three now :-) .
Out of 13 there are 3 still pending for the job, 3 left at the shop for further inspection/repair,
2 mysteriously made it through my PROM diagnostics but do not properly boot into IRIX,
1 was working but was left at the shop in order to have a reference at hand
and 4 more modules have been sent to their owners already but still thereīs no feedback.

December 28th, 2005 Today again there is some good news and some bad news (as always - shit happens).
All the CPU boards - even the one bound for italy - reached their owners before christmas.

For only two of the modules I have positive :-) feedback yet.
One CPU successfully made it back into its native environment:

The other one (mine) - the one that lacks L3 cache - didnīt travel and works flawlessly since.

But it seems that the fact that most modules passed the PROM monitor diagnostics
misled me to celebrate far too early - for 2 modules thereīs negative :-( feedback.
Both of those CPUs refuse to boot into IRIX, one giving a bus error, the other an ECC error.
For error messages please see my additional notes on each module at the gallery.

If anybody reading this has a deep understanding of the IRIX boot process
and/or the pinout and logic of the R7000 CPU, and can tell us eventually
where to look for bugs or mistakes - donīt hesitate to contact me!

The guys that do the soldering do their best but do not have any of that knowledge.

If I had plenty of spare modules to rework Iīd simply exchange them
and return the not-properly-working ones to the repair shop until working fine.
Unfortunately I donīt have - each CPU module has to be returned to its respective owner.
If anyone knows a source for O2 R5271 CPU modules - Iīd like buy some!

December 19th, 2005 Boards KVZ 656, KWJ441, KWJ446, KWJ697, KWJ837, KXA675 and MKP050 have been shipped today.
Recipients located in Germany will receive before christmas, in Netherlands, too (hopefully), in Italy - maybe.

December 16th, 2005 The last CPU module - out of those I have back from soldering - has passed the test suite without any error.
Due to some private duties I didnīt manage to pack & ship today, but I plan to do on monday.

Serial Numbers of modules ready for shipment are
KVZ 656, KWJ441, KWJ446, KWJ697, KWJ837, KXA675 and MKP050.

Have a nice weekend!

(noon on) December 16th, 2005 Isnīt it funny:
after months after months of agony and only a few, mostly sad news once in a while
now updates of this twice a day are becoming insufficient.
I hope you enjoy what is going on here actually
(Yes, I know: some do not - those whose modules did not succeed yet).

I delivered the only CPU module that showed a minor mechanical defect (a loose connector)
back to the repair shop in the morning.
Since this is easy work for them on a module that had proved to have a running CPU already
we decided to leave that one for further use as "the master" for first go/no-go tests.
Therefore I brought the other one that tested good home in exchange.
I can now proceed and put the heat sink back onto that one, too (as soon as I finish typing this),
and then run the full PROM monitor test suite on it tonight.

FYI: some of the images that are contained herewith are larger than they appear.

(early) December 16th, 2005 This night I have tested the 600MHz CPU modules more intensively one by one.

I put each one into my O2 and ran the hardware test suite from the PROM monitor.

I encountered three modules that did not work well from scratch... (no, please donīt worry)
Two of them I could fix myself immediately.

The first one that stayed with a red LED turned out to have fallen off the motherboard.
When I pulled it there simply was no CPU on it, it lay loose on the inside "floor" of the O2 instead.
I put it back into its seat and it worked fine through all the tests.
But a tiny piece of the connector housing broke off. I have to apologize.
That defect doesnīt harm any function, but may be glued back... or omitted as well...
Iīll include it with the board.

A second module didnīt work at all. It turned out that the blue patch wire came off its upper end.
I found it hovering above the PCB... soldered it back and the CPU worked fine, too.

A third one worked fine, then caused a PANIC, worked fine after reboot and got a PANIC again.
A brief visual inspection revealed that one of the connectors had become partially loose.

The SMD contacts work only sporadically. I will have this repaired tomorrow (hopefully) at the shop
because I myself do not have the necessary equipment for that kind of reflow soldering work.
This kind of error - some parts becoming loose on the other side when soldering big components - is pretty common.
In fact, I found that some of the passive components vis-a-vis the CPU had been put back in place manually.
No wonder that one didnīt work for longer than a few minutes.

(late) December 15th, 2005 As you may have noticed I produced a little confusion about the number of modules that tested good.
Here at home I counted only seven, but I could have sworn that there were eight at the repair shop.
The simple solution: I left one of the good ones there, so they have one that works to compare.
Yesterday evening I didnīt know which one it was, but I found out on my visit today.
Thereīs no new gallery photo of that module yet, but at least I updated its status.

For those now unlucky with not-yet-ready boards:
theyīll try to inspect and eventually redo three more next week. Why three?
Well, thereīs a latecomer from an O2 I got me from ebay lately - which arrived yesterday... :-)
At the moment the tested-good modules sit on my table and wait for Loctite 315 to fully harden,
a process which is said to take 24 hours - counted from three oīclock in the morning on thursday.

I will do some more final testing on friday evening and finally pack and ship modules on saturday,
together with a desoldered R5271 CPU as a little souvenir - or to decorate the christmas tree.

(early) December 15th, 2005 Everyone is pretty busy before christmas... when in a crescendo of
just in time delivery is at its best.

So am I.
Last night the heatsinks went back on top of the CPUs:

The first payments are said to be made already :-)
With a little luck some of the 600MHz CPU modules can dangle from a christmas tree.

Btw,. I have just learned that this here is some kind of theme blog...
so then Iīm a blogger and didnīt know about it yet.

Maybe itīs about time to install some adjacent tools...
For anyone interested: my tools used here are xnview, bluefish and gftp.
Everything on a couple of SGI IRIX machines ;-)
only the USB port for the Camedia camera is housed on an Intel crapbox,
which runs Xandros 3.0 (some Debian Woody variant) nicely integrated into the IRIX net.

Does anyone know how to make a USB port PCI card work in a SGI O2?

December 14th, 2005 The CPU gallery is updated now so everyone can see if he is among the lucky ones
that will soon receive an invoice for the rework job :-) ...
and a 600+ MHz O2 CPU module back soon (for christmas eventually).

December 13th, 2005 Now again - finally - there is some good news and there is also some bad news.

For the good news first:
I now have a couple of CPU modules back which we together tested good at the repair shop.
For this I brought one of my O2s there and started it with the already soldered modules one by one.
Have a look at 9 pcs. of R5271-300 and a single R7000-300 CPUs from its soldering side...

First, none of the modules worked :-(
until I found out that a 0 Ohm resistor was in place that shouldnīt have been.
Then, in my O2, the first modules went from red LED to orange immediately, then into green (tatatata!)...

For each one I took a short look at the hinv output of the PROM monitor.
Funny thing - some modules reported a 601MHz CPU and others a 603MHz one.

8 Modules out of 10 tested good on the first attempt.
Please notice the little green point on the PCB right next to "the cube" :-)

The bad news is that not all of them were ready - and two of the ready ones tested bad.
Especially the modules that need the reballed L3 cache still sit on the shelf.
I left them and the bad ones (for re-inspecting with the roentgen device)... now letīs hope for the best.

Iīm again a little short of time at the moment, but as soon as I manage to glue the heatsinks
onto the CPUs (and write some invoices) I will ship the tested good modules back to their owners.

November 8th, 2005 I have delivered all the CPU modules to the repair shop today.

Taking off the CPU and replacing it with the R7000@600 is no problem, but
- this is the case only for those "prototypical" R7000 boards that lack L3 cache -
taking off the L3 cache and tag RAM and especially a process called reballing
will take its time (and dime).
During de-soldering the L3 cache / tag RAMs get their balls ripped off (eek!),
and these very tiny things (hundreds of it) have to be renewed manually one by one.
There mustnīt be a single one missing - that would render any re-soldering useless.

Two 180MHz R5000 boards are at hand as L3-cache-donators,
a third one is missing unfortunately - so my own R7000 module will be reworked
to function without L3 cache (sigh).

Next week the first two (L3 equipped) modules are said to be go/no-go test ready,
I will then supply one of the O2s as test equipment. Will there be the desired music?

Now fingers crossed, everybody!

November 7th, 2005 Finally I found the time to try some SPROM programming...

now the SPROMs behave fine, but the programmer does not. Not quite.
Its rudimentary software error-messages something about a "failed margin verify"
Such a thing has been subject to a hardware patch at,
but my prommer is of the latest version and is already patched. Sigh.

In effect, there are only 4 bytes I have to program, and these are correct,
but the following 12 bytes repeat the first 4 three times.
Donīt know why, I have contacted Roman-Jones about this and already got a response.
No solution, though.

Anyway, I got an appointment at the repair shop tomorrow :-).
I burnt SPROMs enough for all the boards and hope the superfluous 12 bytes
will be ignored by the O2 bootup sequence.

If not - I have another 85 blank SPROMs left.

October 28th, 2005 The SPROMs arrived today!

Unfortunately there is no way to get a date earlier as 11/08 at the repair shop.
A first test with the Roman-Jones programmer showed that this time the SPROMs are blank.

September 24th, 2005 Today a written confirmation of my Xilinx SPROM order arrived
from a company called Avnet Silica, Herten (Germany)
Delivery date is said to be November 3rd, 2005....
I hope I can accelerate this a little!
(phone them up twice a week or so).

September 22nd, 2005 Now - some news again... about these tiny f******* SPROMs.
Iīm now completely tired of US buzniz, no, Iīm absolutely sick of it.

I have now ordered a "unit" of SPROMS from the german distributor,
which is 98 pcs., - that will leave me with plenty of them.
... anyway, this will finally cut the gordian knot.

In the mean time, a 300MHz R5200 CPU Module was on eBay,
but the high bid was 150 Euro... my limit was 123 - so I missed it. Sigh.

September 12th, 2005 Now look at this email I got today:
HI Walther Mathieu,
How are you today?
On our Website we have recently added over 30,000 Stock
parts with prices, which you can order online with a credit card. We accept
all major credit cards.
Also we carry all Major Lines! Some of our lines are: Xilinx, Altera,
PMC-SIERRA, Intel, AMD, Dallas - Maxim , just to name a few.
On you have free access to over 1 million current and
obsolete datasheets.
Some of our customers that use are GE, Motorola,
Solectron, Jabil, Intel and many many more. Please take the time to use our
service and let us know how our sales team can help solve your sourcing

Thank you for your time and have a nice day,
Sierra IC Inc Sales Team
Tel 1-631-325-9097
Fax 1-631-207-8305
16 Bailey Lane
Manorville NY 11949

At least some spambot is still active over there.

September 11th, 2005 Fourth anniversary of the crime of the century...
still no response from Sierra IC Inc., not via email, fax nor phone,
no affirmation of my email & fax order, simply _nothing_. Sigh.

How do these american component traders do business?
The way the Bush administration handles flood desasters, seemingly... announcements only.
Personally, Iīd like to suggest some tar and a pillow full of feathers...

September 1st, 2005 Good news again!
No, there are still _no_ working Xilinx SPROMS here, but guess what arrived...
R7000@600 CPUs!

August 31st, 2005 OK, Iīm back.
There has arrived another quotation for Xilinx SPROMs in the mean time,
this time from Sierra IC Inc., for parts that are guaranteed to work (i. e. be blank)...
I have ordered some immediately but have got no response yet.
Delivery time is said to be a few days only... letīs hope for the best.

August 5th, 2005 The CPUs are ordered now (at least) and said to arrive in week 34.
Eventually then more - good - news is available, esp. about that SPROM 8-(.
If CMB components were located in germany I sure would go sue them...
no chance from here, though.
There will be no news for another two weeks from now on (Iīm out).

August 2nd, 2005 No news today... my thread at nekochan proved what I suspected:
It is "good practice" to sell junky components to first-time customers.

No response from CMB components yet.

August 1st, 2005 Good news and bad news.
The good: today is my cute little son Linusī (nice name, eh?) second birthday...
it is always a great joy for me to watch how happy he can be with very few,
like rearranging icons on a windnose desktop...
Another good news: I finally understood the SPROM thingie...
with a little help from Chicago Joe on nekochan again, of course.
The programming device works perfect now.

But some hint Joe gave led me to find out about the real desaster:
the SPROMS CMB components supplied are not blank!

all of them contain the same junk in the first line of code:

and all I can do eventually is erase them entirely.
No programming (it is 4 bytes only!) of the desired bytes - still no luck.
Iīll complain via emal to CMB about that tomorrow.

July 31st, 2005 After mechanically fixing the broken parallel port jack on the NT4 box (a tiny bolt became loose when untightening the printer cable and the adjacent nut went off into the interior of the box making a connection of another printer cable impossible) I have now installed and running the roman-jones Xilinx SPROM programmer. Driver and program work OK.
Iīm a little confused now about different representations of hex code, but anyway, the programming failed :-( (perhaps I got too long a cable, or the on-board battery is empty [DOA]).
I have put some questions about that into a new thread at
nekochan The guy at the repair shop is very patient and promised to start immediately as soon as all the material is available.
Heīd go check for prices and availability (when earliest?) of the CPU again. And I got him convinced to calculate a price for the L3 cache replacement. That will be taken of some other boards and re-implanted onto R7k ones, and Iīll try to keep the donator boards functional through use of no-L3 PROM code - but I have to make the SPROM programming work first, of course.

July 25th, 2005 The Roman-Jones Xilinx serial PROM programmer arrived on saturday! FedEx handled the customs procedure which took the whole week. Unfortunately I was not at home during the weekend, and there will be plenty of project work this week.
Anyway, I have downloaded the software and set up a NT4 box to use with the programmer (thereīs only SGI boxes here, a single iMac and a Debian Linux box.
The NT4 box is actually a ISDN fax and answering machine, but thereīs an ISDN modem sitting on the shelf waiting to replace it - lucky me I didnīt trash the NT4 box yet. The wrong printer cable on it, though - have to get me a DSUB25 gender changer first.

July 22nd, 2005 The programming device from roman-jones left the very same day I ordered it. Some days passed - while I kept tracking the parcel on the USPS web site. On monday, the 18th it was reported to be delivered into german customs. There - or within the german post customs handling - it seems to be stuck... I tried to find out whom to ask about its whereabouts but to no success. That leaves me (us) to do what I (we) have been doing for months - wait. I have to apologize but fear this kind of news is not very much entertaining.

July 15th, 2005 another - hopefully the last - step towards the beginning of the rework - at least for the part besides the L3 cache thingie (which may also be solved by the help of CMB components) - I ordered a programmer for the Xilinx serial PROMs from, and it is said to have left there already on itīs 3-5 days journey...

that will allow me to program the PROMs myself - and exchange the PROM later for the desired switch to L3 cache when I manage to find some.

July 13th, 2005 today a FedEx parcel arrived, 60 cm x 12 cm x 12 cm in size...
buried within a small plastic envelope 18 cm x 16 cm x 1 cm ...
and in that - voila:

July 8th, 2005 guess what happened...
CMB components mailed me that FedEx shipment to germany will be another 373.50$!
Are these guys crazy? I received parcels of all sizes and weights from the US,
even a complete O2 from california, and never ever someone tried that kind of fraud on me.
I have asked them to use US mail (air mail service), which may eventually take another week
A letter of 1kg airmailed from germany to the US is 20 Euro.
The proposed shipment vice versa is - totally incredible - plus 1394%!

July 7th, 2005 CMB components confirmed my order! 30 Xilinx PROMs are said to arrive here soon... next week!

July 4th, 2005 On July, 1st Joe has supplied some addresses where to ask for Xilinx PROMs, which I all checked - and thereīs _plenty_ of Xilinx PROMs available. Luckily CMB components responded in time... price is 6.95$ a piece, and Iīm going to order 30 of them - so I have some on stock if anybody else wants to upgrade.

June 24th, 2005 A single XC1765ELS08C PROM is only 5.51 Euro... Joe has responded that he has only 12 pcs. left of a batch of 50, most of which he needs by himself. But he promised to try to find us a batch of better suited size.
I have to apologize for the delay!
As soon as I manage to get the costs down below 300 Euro Iīll start the thingie anyway ( for the nine boards that do not need L3 cache).
The other three boards may require some more patience, or these will be reworked without the L3 cache (gimme a note if thatīs the desired option).

June 21th, 2005 A quotation arrived for the rework job from the second contractor mentioned below. Unfortunately it is 340,- Euro per piece - which is a result of the fact that they were only able to allocate the Xilinx PROM in a batch of 100 - out of which we need 13 pcs., but they charge us for all of them...

I have now asked Chicago Joe if he could provide the PROMs eventually.
This would help to keep the price for the rework at 295,- Euro, which is the preferrable option.

Another couple of good news / bad news is: the L3 cache chips are available, even in the package that is required for SGIs O2 CPU module. Now for the bad news: these are also sold in bigger quantities only - and we need only three sets of them. To order and keep the other 97 sets sitting on the shelf is no good solution; costs would be huge (details as soon as I know some).

June 11th, 2005 There is some news now at last (just to explain the duration of wait - while nothing seems to happen actually): The company originally scheduled to do the rework failed to locate the necessary L3 cache chips.

Bad luck, and in the meantime some employee unwillingly leaked to me that they were planning to give the entire rework job "out-of-the-house" - buy and supply the components only (not all, though, as mentioned above).

Since I am some kind of insider on the electronics manufacturing buzznizz here it was not too difficult for me to guess where to...

I now have contacted that other company and asked them for a quotation for the rework job, too
(because Iīm still not content with having three R7k boards go without the L3 cache).
They also do HD-BGA rework jobs frequently and they do them inhouse,
they have their own roentgen equipment to examine the soldering of BGAs etc..
I expect to know more my the end of next week.

Again I have to beg for your patience - I know you all want your precious CPUs back asap.. .
thatīs why I finally set up this little site to let you have at least a look at it ;-)

My hope is that I can upgrade this site from time to time to give you all the required information without having to ping-pong emails. As much as I liked to communicate - Iīm simply lacking time, and of course Iīm driven by one of the strongest emotions mankind can experience - laziness.

Or Iīm out biking eventually, or attending some concert... tonight there is a drum workshop with Chad Wackerman for a hand-selected little audience just around the corner.