SGI: Hardware

assyrix wrote: IIRC you modded the new chips onto the old CPU-boards. I would imagine that the changed clock frequency could do some serious damage to any electrolytic capacitors (elca) on those boards and probably on the motherboards as well. Usually the manufacturer uses elcas suited to the environment. If you use a faster processor which also generates more heat (and most elcas sit very close to the CPU for efficiency reasons, causing them to receive a lot of heat) you will probably shorten the lifespan of those elcas considerately.

More info (including a cop-and-robbers-story on faked elcas):
http://www.spectrum.ieee.org/WEBONLY/re ... /ncap.html
http://www.ttiinc.com/MarketEye/Aluminum_Cap_Issue.asp

The problem with faulty Aluminum Electrolytic capacitors extends beyond faulty manufacturing to incorrect engineering practice as well. This is not a new problem at all, however the combo of bad engineering and manufacturing has made this a huge problem in the computer industry. Read up on wax paper capacitors to really learn about unreliable components.

First of all electrolytic capacitors are the most unreliable components currently made (the all time winner being wax paper capacitors), and only have a lifespan of about 10000 hours with proper usage. Proper usage means staying well below the rated max temperature and never leaving them sit more than a few months unused. This means that even properly manufactured electrolytic capacitors won't survive near that long in the real world. In my experience you should never let anything that has been sitting for more than 2-3 years unused (of if it is over 15 years old, sit unused for about a year) without significant inspection of the capacitors otherwise you will have a good chance of the electrolytic capacitors drying out which will cause them to either become leaky, open, or shorted (which will give you an impressive explosion when powered).

The way many motherboard manufactures cut corners by not following common engineering techniques like only using electrolytic capacitors rated 200% to 250% higher in voltage than what will be used in the circuit (going outside this could under or over-use the capacitor and them to dry out). Many motherboards also butt the capacitors right next to the CPU (huge heat source) on the board, and should be instead moving them away as effectivley possible and/or using more durable tantalum capacitors instead.

Anyway, increasing the clock rate while keeping the heat dissipation and current draw nearly the same; which should be the case when replacing the CPU with one properly rated for the new speed, shouldn’t harm the capacitors one way or the other. In fact, overclocking won’t harm them either except for the possibly of increased heat from the nearby processor. Which can be remedied with the use of proper cooling techniques on the CPU to keep things within tolerance of the capacitors.

Shtoink wrote: Let's keep this freindly and not get another thread locked. We don't need any name calling in here, or any other kinds of slander.

Yes, that kind of thing will only degrade further the once friendly and informitave nature of this forum. Once that is gone, I'm sure nekochan.net as we know it won't last much longer.

chicago-joe wrote: warerat wrote

I'd be compelled to rip the R10K out of my Octane and plug a R12K in there instead.

In the O2s, the R10K and R12K (up to 300MHz) cpu chips use a surface array pin system and fit into a socket of sorts, so they are not soldered to the cpu board like the R5K and R7K chips. This would make changing the cpu chips themselves an easy task. The R10K/R12K boards use the same PLL and xtal setup for the external cpu clock but they use a set of 64 tiny resistors tied to the input pins of the cpu to set the initial conditions of the cpu. (the same thing the PROM chip does on the R5K/R7K cpu board) The catch is: the resistors are tiny and not labeled so I am having to map the the individual resistors to the cpu pins so I can figure out which ones to change to set the correct cpu speed. I have taken apart my Octane R12K 360MHz cpu board and it is the same type of setup, just that all the resistors are in different places. :-x I would guess that the I2 cpu board works the same way. So upgrading the O2 or Octane cpu board (at least to an R12K 300MHz) should be a reasonable project. If someone wants to take the cpu board out of their I2, remove the heatsink assy and take some high res pictures, I would be interested to see if it is setup the same way as the O2 and Octane cpu boards.

Joe

I disassembled the R10K module on my Octane and it indeed has the surface array with the socket as you mentioned. I'd assume the R10K in the Indigo2 to be same-- they both have that arrangement of allen head screws that hold the CPU down to the socket and against the heatsink. I'll take my Indigo2 CPU apart and take some hi-res pictures of all the pieces. Whats immediately different about it from the Octane one is that it has this copper tape around the entire perimeter of the heatsink-- probably for RF shielding. I'm just going to have to slice through that stuff to pull the two halves apart.

assyrix wrote

IIRC you modded the new chips onto the old CPU-boards. I would imagine that the changed clock frequency could do some serious damage to any electrolytic capacitors (elca) on those boards and probably on the motherboards as well.

That's a good point, but there no electrolytic caps on the cpu board. If you look closely at the picture http://users.rcn.com/joepage/rm7000_400.jpg all the larger power decoupling / filter caps are tantalum and all the smaller digital signal decoupling / filter caps are mica or ceramic. All these types are good for up to 85C and nothing on the board is more than warm to the touch. The power coming into the cpu board is through a TI uc382td voltage regulator, so there should be no issues with caps on the motherboard or in the power supply.

One thing that is clear, as I have worked more with the SGI equipment on the board level, is the almost obsessive overengineering of the boards. I have little doubt that these boards will easily take the RM7900-900MHz cpu chips (when I get them) and still have a life of years.

akimmet wrote

The problem with faulty Aluminum Electrolytic capacitors extends beyond faulty manufacturing to incorrect engineering practice as well.

Yes. My other hobby is vacuum tube audio, in that group it is almost heresy to use electrolytic caps for anything including the power supply.

Brombear Wrote

you mentioned the prices for R12K processor. What is needed to find out if an R14K could be fited ?

The R12K-360 and 400 chips went back to being soldered to the cpu board. It might be possible to remove the socket and solder the cpu chip directly to the board, but without one to look at, I cannot be sure. I have not seen an R14K cpu board so I don't know how they are put together. The R12K 400MHz chips are about $165 - $185 each and are easily available (they were also made by NEC). I have not been able to find ANY information on the R14K or R16K cips anywhere (or raw chips for that matter). If anyone has data , spec or user sheets, or even knows who made them, I would very much like to see that information.

Joe

chicago-joe wrote: Octane Wrote

I was trying to point out that most people do not mod computers to save money, but rather for the joy of modding.

Sorry, that was my bad attempt at making a joke. That was my point exactly, I did it for the challenge (and for a faster machine).
Joe

No problem. Sometimes jokes are misinterpreted in text, even if we use the same language.

warerat wrote: I disassembled the R10K module on my Octane and it indeed has the surface array with the socket as you mentioned.

Any special tricks for disassembling the Octane CPU?

chicago-joe wrote: The R12K 400MHz chips are about $165 - $185 each and are easily available (they were also made by NEC).

How available are the 300MHz chips? Where did you find the R12ks?

chicago-joe wrote: I have little doubt that these boards will easily take the RM7900-900MHz cpu chips (when I get them) and still have a life of years.

How much is the 900MHz model?

I would be willing to provide you one if you have the board to test with.

I wouldn't mind going through the trouble just to know that I will be able to do the same to my O2.

chicago-joe wrote: akimmet wrote

The problem with faulty Aluminum Electrolytic capacitors extends beyond faulty manufacturing to incorrect engineering practice as well.

Yes. My other hobby is vacuum tube audio, in that group it is almost heresy to use electrolytic caps for anything including the power supply.

You haven't seen horrors until you see someone plug their old childhood favorite AA5 (after several warnings from me not to do so) and see that nasty confetti get all over.

Octane wrote:

warerat wrote: I disassembled the R10K module on my Octane and it indeed has the surface array with the socket as you mentioned.

Any special tricks for disassembling the Octane CPU?

It's really easy to take apart after you pull the CPU module off the motherboard. You just remove the two phillips screws on the voltage regulator, and then remove the 1/4" allen-head screws (make sure they're all the way out). Then the top heatsink will separate from the board-- you might have to gently pry the halves apart as there's thermal compound on the cache chips, it looks like white sticky tape. The CPU will most likely be stuck to the top heatsink because of the thermal grease. You'll notice no pins on it, and the special socket on the board. I was paranoid putting it back together so I plucked the CPU out of the grease and made sure it was in the proper orientation before I screwed down the heatsink again.

I am just concerned about the elcas since I had several of them exploding on me in an VW320 power supply and motherboard after upgrading the box to a single 1GHz CPU (mobo rev. 007B), destroying all PCI-cards. Most elcas shouldn't work in an environment where the temperature ecxceeds 50 degrees celsius continuously, otherwise their lifespan will be drastically reduced. I don't know about the Sierra processors but would imagine that (especially) the 900MHz version dissipates far more heat than the CPU it replaces. Furthermore, I don't know how the power supply deals with the altered electricity demand of the new CPU, putting more strain on the elcas to smoothen out the peaks.

Decent cooling and moderate capacity demands prolong the life of an elca whereas overclocking, use of non-certified CPUs expansion cards and overheating (because of insufficient airflow and/or noise damping foam/mats inside the box) will reduce the lifespan of those components, resulting in malfunctioning, damaged elcas or even a fire inside the box in the worst case scenario (like it happened in my case).

On the other hand, with the price of used sgi machines nowadays (all the way up to the Octane) and the possibility of replacing old elcas one might as well go ahead with the mod to extract one or two more useful years of those machines. I would put another fan into the machine, though, or cut the skins open (with the resulting deterioration in looks and electromagentic shielding) to provide adequate cooling.

If you are interested in working out the potential lifespan of specific elcas you might want to have a look at this:

http://www.bccomponents.com/Uploads/Dat ... c01int.pdf

assyrix wrote: I don't know about the Sierra processors but would imagine that (especially) the 900MHz version dissipates far more heat than the CPU it replaces.

If that was the case, would it be possible to use a chassis from a R10k? It provides the space for more cooling due to heat.

With drives so cheap, I wouldn't have a problem with eliminating the second drive bay.

I am sure there is more to the R10k than just the mobo tray and that it would be a matter of gutting both an R5k and R10K O2 to get it to work but it would be worth the upgrade.

I have a 5k and 12k at work, I'll check tomorrow. I did read somewhere that there were versions of 5k cases that you could remove a screw or two on the bottom, take out the divider and put in a 10k motherboard. I wonder if you can just then use a 10k pci riser to give more space for cooling - or, can you use an 02 without the pci? I seem to remember reading it has to be there.

Rev.Bubba Wrote

How much is the 900MHz model?

It's not a shipping product yet. PMC-Sierra has been sending out a very few pre-production samples to their biggest customers, and that ain't me. I'm not sure I even want to work with a chip in its pre-production state, that would be quite a bit of money and time only to find there were bugs in the early rev. I will most likely wait until it is a shipping product and they have some stability information before I mess with trying to upgrade a cpu board to it. I appreciate the offer though.


assyrix wrote

I am just concerned about the elcas since I had several of them exploding on me in an VW320 power supply and motherboard after upgrading the box to a single 1GHz CPU (mobo rev. 007B), destroying all PCI-cards.

The 7000C-600 chip draws 4W at 1.2V core voltage
The 7900-900 chip draws 5W at 1.2V core voltage
The 5271-300 chip draws 4.5W at 2.5V core voltage
The R12K-300 chip draws 30W at 2.6V core voltage
The R10K-195 chip draws 30W at 3.3V core voltage
I have 3 O2s: a 5K-200, a 10K-250 and a 12K-300 (in their original forms) and they all use the same power supply, the Sony APS-90 Rev. N. Clearly the power supply is built to handle the power draw of any of these cpu chips. The 7K-600 chip is only warm to the touch after 5 days of running, the 12K-300 is much warmer but there is also a much larger space inside the computer without the second drive bay so there is better air circulation and I can see NO reason to cut sheetmetal or plastic. From the cpu power data, the temperatures I have observerd and the premium quality of the O2 componets, I don't believe any of these upgrades are going to cause catastrophic failure of any of the electrolytic caps. (and the Octane is even more overengineered than the O2) On the other hand, O2 and Octane power supplies are on eBay all the time at very reasonable prices, so having a spare might be a good idea even if you don't mod.

Joe

acronym wrote: I have a 5k and 12k at work, I'll check tomorrow. I did read somewhere that there were versions of 5k cases that you could remove a screw or two on the bottom, take out the divider and put in a 10k motherboard. I wonder if you can just then use a 10k pci riser to give more space for cooling - or, can you use an 02 without the pci? I seem to remember reading it has to be there.

The PCI riser has a Dallas serial prom (looks like a little TO-92 transistor) on it with the serial number of the machine encoded on it. I think if it's not there the ethernet won't work or some message comes up at the console.

just checked the motherboards, the are very similar. j75 is missing on the 12k, but the holes are there - also there is a different oscillator, but the frequency is the same 50.000

warerat wrote: The PCI riser has a Dallas serial prom (looks like a little TO-92 transistor) on it with the serial number of the machine encoded on it. I think if it's not there the ethernet won't work or some message comes up at the console.

Neither the ethernet nor any licensed software will work if the PCI riser is not plugged in. However, one could do away with the metal frame, or use a R7k CPU in an R10k Mainboard module, with all the additional space for the R10k/R12k CPUs. I wouldn't think that is neccesary, given theminiscule power consumption of the R7k CPUs.

There are at least two different O2 R5k chassis - one that is upgradable to R10k (has two screws at the bottom that hold the separator between the leftmost drivebay and the mainboard, and the small rectangular black plastic part under the leftmost drivebay is removable), and the other one that isn't (no screws for the separator, one larger plastic part under all drivebays and a/v module).

warerat wrote: The PCI riser has a Dallas serial prom (looks like a little TO-92 transistor) on it with the serial number of the machine encoded on it. I think if it's not there the ethernet won't work or some message comes up at the console.

You can start the machine, but Irix complaints that Ec0 interface doesn't work correctly.

So, you have no network interface.

On top of that, you can't use software like Maya, that requires a FlexLM key, which key is based on the mac adress of the built in network card.

For anyone looking for a company to remove the old cpu chip and install a new one (with a guarantee of their work):

Advanced Electronic Services (336)386-1000 http://www.industrialelectronicrepair.com/
contact person: Joe Hughes (866) 789-7909
contact email: mailto:[email protected]

You will need to tell him you want the same work done (PMC-Sierra cpu chip changed on a pcb) as Joe Page had done. Their equipment requires some unique setting for this chip as the "shell" is a copper/polymer laminate, this way they can look-up the settings they used on mine and save everyone a lot of time.

-they guarantee their work
-they do deal with individuals
- you can pay with a credit card
- they will work internationaly

Joe

Hello,

chicago-joe : would an O2 with an R5K@180Mhz be a "good base" for that mod. Do you have any idea for the 900Mhz version of the cpu to come out?. I have seen an 825Mhz proc on the site of pmc? Any idea/advice?

@+

LoWeN

LoWeN Wrote

would an O2 with an R5K@180Mhz be a "good base" for that mod.

The cpu chips on the 180 and 200MHz O2s are V5000 series chips made by NEC and are completely different, swapping a RM7000C chip for a V5000 chip is out of the question. This was my answer to a question Shtoink had earlier on the same subject:

I don't know much about Indys, but the 180 and 200MHz O2s use the NEC v5000 cpu chip. I'll bet the R5000 series Indys use the same series of cpu chips. NEC made a V5000A series chip (pin and command compatable) that maxed out at 266MHz and I have seen these around so they are available. I downloaded the data and use guides for both and the only problems I see are that the 150, 180, 200 cpus use a 3.3V core and IO voltage and the 250 and 266 cpus use a 2.5V core and 3.3V IO voltage and there is not a 2.5 mutiplier for the core speed. I don't know how hard it would be to modify the voltages on the cpu card, if it could be done at all. I don't know what would happen if you ran a 266 chip with 3.3V on the core, it would get HOT so you would have to cool it well and I would think the chip life would be shorter but it might still last quite a while and should be much better in performance. To get around the core speed problem you would have to change the control resistors on the PLL chip (ICS9159-10) to change the Xtal to output ratio to 93/20 (66MHz X cpu mult of 4) to run the chip at 266MHz, this should be easy enough to do. I might try this with an O2 200MHz cpu board, I would think if it works on an O2 board it could be made to work on an Indy cpu board.

At some point I want to try upgrading one of the 180/200MHz boards to a 266MHz cpu, but the R7K chips are larger in size and have more pins so they cannot be swapped for the V5K series chips.

Joe

Hello,

Hummm "bad".
So what can we take as "base" for that kind of mod.
Any R7K sure?
R5200?
Because I never found any R7K here around only R5K/R10K/R12K and one time in my life I have seen 02+.

Kind regards,

Charles