The collected works of SAQ - Page 2

regan_russell wrote:

SAQ wrote: Not sure if IRIX 4D1-3.3.2 has a bug or something, but the machine slowed down noticeably after going from 32MB to 64MB. Guess I'll back down to 48 and see what happens.

Is it possible that RAM chips are different speed. Does one have -60 on it and one have -70 or -50 or it..?
Chips used to have how many nanosecond they were rated for marked on the chip itself..

Regan

They're all 70 or 80nS, and 4D/25 is supposed to support 80-100nS (IP12 SIMMS are 80nS). Each set of 4 is matched.

porter wrote:
< 1023Meg - yes

1.3Gig - no.

And you can't just write out a disklabel that abandons some blocks?

Didn't the rise of SCSI and other LBA-based schemes (and on-disk managing controllers) drastically change the interface to the point where the physical disk location of data is no longer manageable by the O/S?

Pico/Nano (the GNU Pico clone) are pretty easy text-mode editors (Pico is what the Uni started us out on), but he might be more comfortable with an X-based editor. I do kind of like Nedit under X.

You could always install the CDE packages and have him use DTpad- that's pretty stripped down and easy to learn, but Nedit is much more powerful. I never really got "into" the whole Emacs/XEmacs "thing" because they were so resource-hungry.

qumefox wrote: Even if it does work. SMD rework is not for the feint of heart. I freakin hate doing it even though it's part of the job.

But you must admit the coarser pitch SODIPs are surely the best of all possible worlds involving SMD ICs.

Or you can get a $1.50 used PCI SCSI card and mount the drive from your Debian system (provided it's a 3.0 or 4.0 release), and use your favorite editor to zero out the root password.

Final Cut's pretty nice, but keep in mind I haven't been looking for ~4 years.

TextWrangler is a must in my opinion. Stock systems don't come with much in the way of ASCII editors.

Get a good FTP client.

josehill wrote:

SAQ wrote: TextWrangler is a must in my opinion. Stock systems don't come with much in the way of ASCII editors.

...or make sure that you've installed X11 (it should be on your MacOS disks as an optional install package) and then download nedit for that good, old-fashioned, chewy, IRIX-flavored, text-editing goodness. Now available with special MacOS X Intel binary packaging!

Yeah, I've got a bunch of X11 stuff, but Apple still makes the X progs feel like second-class citizens. Much like Classic apps.

Definitely make sure you install all the stuff from the OS disks- not sure what things are like on the newer releases, but for a while Apple had some essential utilities (for a UNIX-head) as optional installs, including X.

Make sure you put on the dev stuff, too.

I set up my Origin 200, and upon noticing that one of the CPUs had a 2MB cache and the other a 1MB cache, I put the 2MB cache module node as the master, because of course you want CPU0 to be the most powerful, after all it's the most used, and any Sun howto will tell you to put the most powerful CPU as the primary one.

After getting IRIX installed and firing up gr_osview to see how things are going, I noticed that IRIX seems to default to scheduling more things on CPU1 . Is this the usual behavior to keep CPU0 open for system operations, or is this just a temporary thing?

Well, just hit a string of activities biased towards CPU0 load. Looks like it was just luck of the draw.

Like I said, it seems to have evened out later.

I think I recall reading that (in addition to the cache bit) SGI NUMA archs try to keep things local to a node to minimize remote memory access, and since I have only one processor per node, that would necessarily mean keeping it to a single processor.

I just wasn't sure - MP SPARCstations recommend the faster processor be the first one, wasn't sure if SGI had gotten away from this (seems like they have).

On the plus side, it's noticeably faster than the old I2

D-EJ915 wrote: Considering their performance, old PowerMacs do seem to hold their value relatively well compared to x86 computers, really odd but whatever.

(1) Macs have a slower replacement cycle than PCs (want to see some really wild value-holding? Look at RS/6000 )
(2) For modernish Power Macintoshes (fast G4s and especially fast G5s) you're running into the groups of people that need a Mac to run an old Classic/PPC or Classic/68k app that isn't available under OS X. The fastest machines that will run these apps will command a premium for a while (look at the Quadra 840AV back until PPC performance on emulated apps was able to match it). You see the same thing on VAXen and will probably see the same thing on Tezros and the like.

I did it. The R10000 processor at 175 and 195MHz uses the same voltage across all platforms. For the modification you use your Indigo2 PMT5, and the only reason you get a Octane or O2 PM10 (or whatever the O2 name for the module is) is as a processor donor for the R10000/200 rated chip. The resistor mod (on your I2 PMT5) changes the clock multiplier to 4x rather than the 175MHz' 3.5x.

If you want to risk it, you can even overdrive your existing PMT5's R10000 (they're rated at 180MHz or better) and dispense with the Octane or O2 module altogether.

R10000 >= 225MHz are a no-go, though, because of the different voltages.

Mine worked just fine for several years, but after that I decided that driving parts beyond their ratings just wasn't worth it for the minor boost it gave. I didn't get the proper 200MHz - rated CPU, though, so it was more of a risk.

Anyone here know why SGI tends to underdrive their processors (R10000 at 195MHz rather then the max rating of 200MHz, and I've seen a 4D/35 with a full 40MHz-rated R3000)?

been out of country for a while.

The PMT5 is the plug-in CPU and cache daughterboard for the Indigo2 IP28 (it has the heatsink and fan on it). Not sure if you figured this out, but when you get a O2 processor module (or the Octane module, which is a PM10 (single processor) or PM20 (dual-processor module)) you open it up and remove the 200MHz rated R10000 LGA chip and put it in your original Indigo2 PMT5 processor module.

I wouldn't recommend trying to go over 200 on a first-gen R10000. When they boosted the speed they tweaked the chip a bit, I'm not sure if it was a different process or not, but they did have to do something. The fact that they never released chips rated faster than 200 indicates that they didn't get enough running reliably to make it worthwhile. They probably ran into timing errors or thermal problems.

(a) prevent people from monkeying with system IDs.

(b) "You want to change your O3400 config? That must be done by SGI service - we'll have them send you a quote. Have you ever considered a service contract?" .

Hopefully they'll release the codes when the O(3*10^X) series hits end-of-support and they won't make any more money from service.

shyouko wrote: Tiger is faster when given sufficient memory, otherwise Panther is still a good choice but you'll have to put up with older third party software.

What is the problem with Apple software developers? Many Windows apps can still run on 2000, I'd say most Sun apps run back to Solaris 8, AIX is usually back to 5.3 or 5.1 - but Macintosh apps are almost invariably compiled for N-1 (or even current only), and Apple updates ~every 2 years.

nekonoko wrote:

scottE wrote: Thank god for archivists.

Agreed, but here is a danger though. For example, there was a well known archivist of ancient CP/M boot media for just about any CP/M machine you could imagine. People would go to him whenever they came across a rare machine and he'd set them up with appropriate boot media.

Then one day he died, and his entire collection went with him. None of it was stored on the Internet, so there were no mirrors. Lots of one of kind software sets are now apparently lost, perhaps forever.

If you've got a nice collection of something, it's a good idea to have a preplanned agreement between you and someone else (or a couple of someone-elses) stuck somewhere near the will. Make it known that when you're dead, so-and-so will take care of everything related to [fill in blank here]. Most of the time the loss of rare material is because the survivors don't know what to do with it and aren't in an emotional state where they want to deal with the "junk" so to speak.

what would you advise me to go for?

Fast disk subsystem, plenty of memory and a good server-quality network interface. Offload as much off of the CPU as possible. For pushing data out over a network link any of the dual-processor/dual-core systems will work just fine (within reason, a NCR 4x80486/33 won't cut it), you just want to make sure that the CPU has to mess with the data as little as possible.

From a technical standpoint, aren't there still potential performance benefits from the VLIW architecture (provided that Intel has a way to deal with the scalability issues - i.e. not wasting units if a future version is wider and attempts to run current software), or has current OoO technology eliminated the edge?

Zerolapse wrote: Yes, I'm using an Origin 3k R brick from another system, and yes I did have serial mismatch errors, but now I don't.

Say more if possible...

The new x86s have execute disable, and with a well-built firewall there won't be many holes to get into anyway. Indeed, a case could be made favoring a big arch (x86, SPARC, probably AXP), as more people will be looking for the little issues and fixing them, and while very few issues will be platform-dependant, you know the x86 ones will be found and fixed. Just keep on top of things and you'll be fine.

The big downside with many workstations as firewalls (especially old workstations) is finding the second network interface. Indigo/Indy really looses out there. Sbus is pretty common, but it's hard to argue with a well-built PCI PC (emphasis on well-built, junk H/W will make your life miserable with problems).

leaknoil wrote: Server dirt sure is a lot easier to clean off then grease though.

...Unless someone's been smoking heavily around the machine.

Gray Fox wrote: Wont SWAP help end the SSD life span sooner?

Not if you disable it. Although Indy's 256MB memory limit doesn't make it the best machine to run without swap.

For a while they sometimes had a setup similar to the Prestoserves where there was a front "cache" of RAM that handled the rapidly changing data before writing it out to the flash. Not sure what they do now. Of course, there's also the really old SSDs made of entirely battery-backed SRAM.

Aren't there some inst commands to upgrade what is present without adding systems? I don't recall ever needing base disks for an update (although I usually do an "install all" rather than "install default")

As long as you aren't switching streams, wouldn't

work?

Japes speaketh the truth. Moving that O3k put any thought of the rack O2k at RE-PC out of my mind immediately.

Now, cramming the contents of a rack Onyx2 (1 compute module and one graphics module) into two deskside chassis might be doable...

Japes has a really nice O2k deskside though that I wish I had the spending cash, space, and domestic capital to lay my hands on. Immaculate, 8x400 and with the graphics head hack. 95% sure it's cube logo skins, too.

Indyboy wrote: Fujitsu MAM3367MC 36GB 15K U160 80pin
Fujitsu MAP3367NC 36GB SCSI Ultra320

Are these drives quiet? I'm looking for a quiet drive for my I2. I've already found a Seagate Cheetah 15K.4 36GB but it's too expensive

Make sure the MAP3367NC U320 will work on a narrow bus - many disks will, but the U320 standard doesn't require it. Looks like it might not -

8-bit data bus is available only with NP model

Looks like it will work if you get the 68-pin model and a 68->50 pin converter, though.

I haven't the foggiest why Fujitsu decided to not support 8-bit across the board on this drive, since they put it in one of the models.

plain red.

I might need to give that a try. Pity it wasn't standard.