The collected works of mapesdhs - Page 7

On the other hand, if one allows the user to infer such a thing is possible, then they'll probably try it.

It's a wonder I don't find breadcrumbs inside O2s...

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
Note that if it's not yet sold, IMO it should go to weblacky (Josh) as he asked first.

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
weblacky wrote: ... he was trying to ensure that requests where honored/addressed on a first come first serve basis ...


Yep, that was all, Doug! 8) Nothing else intended/implied.

Glad to hear you're ok, emachine! I know stuff does happen to people sometimes. With so many people on a forum like this,
someone is bound to be having a rotten time of it somewhere...


Hope the sales all goes ok for you guys!! Kinda enva ya, I had a nice time visiting that part of the world when I was there way back. 8)

Ian.
Nope, that was a one-off, during my Doom days.

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
May the 4th be with you! :D

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
My gf caught me out with it (again), so I thought I'd spread the pain. :D

Ian.
I did something similar with an Arctic F8 TC, though I rigged it to run at max speed instead of being temp controlled. Still inaudible though, and
with a Fujitsu MAS3036NC disk (36GB 15K), the system is silent. Spec is R7K/350, 512MB. To reduce power consumption even more (it's just
a gateway machine), the unit has no AV module and no CDROM. Keep meaning to install a Gbit NIC, not gotten round to it yet though.

Ian.
I'm working on a charitable PC build for the Learn Engineering YouTube channel. Please PM/email/call if you'd like to contribute! Donations of items I can sell to provide funds are also welcome.
[email protected]
+44 (0)131 476 0796
+44 (0)7434 635 121
I never found a 'cheap' source, just a reputable seller I trusted to get me the items I wanted reliably (Active Network Systems Ltd, www.ans-ltd.co.uk
call 01480 437997 in the UK): ARS-2160H SCSI/SATA U160 bridge box (SCA) and the 68pin equivalent (ARS-2160). The SCA version I use for
performance testing in various systems while the 68pin version is destined for my Fuel with a 128GB OCZ Vertex2E SSD.

Actually I thought the pricing was quite reasonable overall, 108 + 5 shipping for the 68pin model, 99 + 5 shipping for the SCA (+VAT in each case).

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
skywriter wrote:
well, it's not like i would ever recommend you anyway.


Sheesh, chill out dude!

What IS it with people on these forums these days? This is why I don't post here much now, the simplest use of a word and someone
goes freak-out. So I mention a term you either don't know or aren't familiar with, so what? Who cares? Move on already...

Ian.
hamei writes:
> ... your "engineers" were never that good anyway, not like these whip-sharp young kids that grew up with computers ...

Even those of us who grew up in the 80s home-computer boom (who no doubt thought of themselves for many years
as being tech savvy) are now pretty much old-timers & out of date compared to current trends.


> ... If it were under the voltage at which the cpu operates, then the cpu would not function. "Undervolt" is neither a noun nor a verb. Using it any other
> way than describing a failure to function is incorrect.

Wrong. :D Modern PC CPUs don't have a fixed voltage at which they operate. Surely you know this? At stock settings
there's a baseline VID value which shows up as whatever setting in CPU-Z, but even that is derived from a range of
possible values for each chip type (varies slightly for each individual CPU). In normal practice the actual voltage a chip
receives is constantly changing. If the system is idle, numerous functions shut down and the voltage drops. On relevant
i3/i5/i7 CPUs, if one or more cores are active then Turbo increases the clock and increases the voltage to match. Watch
the details in Core Temp and it's plain to see. AMD now has an equivalent though less effective mechanism. Your
description assumes there's some fixed voltage for a CPU and that's just not true these days at all.

And then there's Vdroop, which causes Vcore to fluctuate under load anyway, especially when a load halts.

Undervolting, as those who use the term would say, is deliberately forcing the BIOS Vcore to be lower than normal in
order to achieve power consumption below that possible with the supplied C-State mechanism (or equivalent function
in AMD chips). Plus, quite often, the BIOS is too generous with the voltage supplied even for default settings. Many
users find they can reduce Vcore somewhat and still operate with default settings, again reducing power consumption,
also referred to as undervolting. One can sometimes do the same for the RAM. Depends on the mbd as to how flexible
one can be - some don't have a particularly wide BIOS range for Vcore/Bclk. Enthusiast boards are obviously better.

My i7 870 has a VID range of 0.65 to 1.4V. At stock settings, the Vcore varies between about 0.9V for the chip running
at 1.1GHz up to about 1.35V for the Max Turbo setting of 3.6GHz. I have it set to 1.41V in the BIOS for 4.27GHz (no
Turbo or C-States) which shows up as 1.36V in CPU-Z, but this changes under load because of vdroop. It'll be a lot
more complex for my 875K which allows one to individually control each Turbo setting for 1, 2 or 4 cores active (ie. no
C-States, but Turbo active); when it's all setup, the Vcore will vary much more than the 870 depending on load. Ditto I
suspect for my 990X, though I've not examined it yet.


> But what the heck, we live in the world of NewSpeak so why not ? ...

People invent or adopt terms all the time, eg. where did "modding" come from?


> When "democracies" really mean "puppet governments run by friendly dictators who willingly take our bribes to distribute

Or just blame the voters, half of whom don't bother voting, while those that do choose for person X because they like their
smile, or that's how their parents have always voted, or some other dumb reason. :\ A true democracy makes decisions
based on rationale and reason; very few nations operate this way. Poor old Epicuras...


> no, it's .... Undervolt !!

Sounds about right given some of the oc'ing rows I've seen on forums. :)

Ian.
skywriter writes:
> ... the childish slags words you folks use. ...

That's a bit harsh. Those on the overclocking forums are just selecting terms to describe what they're doing, and I certainly would never
say their hobby is childish. These days there are few other 'techy' hobbies that a young person can have, given the massive black-box
nature and immense complexity of modern devices. It's not like the 80s when one could at least get to the guts of 8bit/16bit systems, write
code for robotics control or other device interfacing (the BBC Micro was excellent for this), doing assembly language, etc. Numerous
people kickstarted their careers on the back of that boom, including me.

I doubt very much that those who use these terms would in any way intend such use to be demeaning for what you're referring to. Since
there wasn't any existing jargon to describe what they're doing, one can hardly be surprised that words & terms are either created or
combined from others to suit. It is however certainly true that a lot of those who meddle with PC overclocking haven't the slightest idea
what (for example) voltage actually means (perhaps not helped by modern sw/BIOS tools making it easier than ever to do all this stuff,
often in an automated manner), but from all I've read there are definitely those who do delve more deeply to understand what it's all about
so they can achieve better results - this can only be a good thing.

Your comments remind me of someone I used to know in the late 90s, head technician at a university I worked at for a while. He really
shouldn't have been there, grossly overqualified man for the job; he ended up working there for family reasons (caring for his mother).
His original career was defense systems, eg. he helped design the guidance system for the Minuteman missile. He once said how in his
opinion the apparent intelligence of students seemed to be going right down the crapper, and I had to agree from what I often saw. He once
commented that 5X more people now were being awarded top-level degrees but that one could hardly conclude modern graduates were
magically 5X smarter. then when he was at Uni.


> To me it's basically a denigration of the profession I've spend my life on. Sorry if I'm a downer, but I suppose you kids would never
> understand how crappy it makes me feel to have my kind of work trivialized in the end.

I'm almost 42. Hardly a 'kid'. :D And there was me yesterday feeling like a grumpy old git...

I agree that it would certainly be a shame if the activities of those who mess about with PC overclocking never resulted in their taking up
electronics or electrical engineering in a more serious way. I'm sure some do, but what proportion? I've no idea. As long as it's not zero
though then who can say it's not worthwhile? It's hard enough as it is to get children and young people interested in science/etc., so I
say be grateful there's still something which is able to garner such interest. Schools these days sure as hell make no attempt to
encourage them.

When sites like tomshardware publish reviews of new CPUs, there are two types of readers. Most jump straight to the performance results;
this is the crowd who are IMO less likely to get more seriously interested in science, etc. Others though read the entire article, trying to
understand the explanation of how the new chip design works, what it means. These are the people we need to encourage. Don't belittle
them by saying what they do is trivialising your knowledge - they would never intend their hobby to have that effect. Better to help them
learn more about the foundations of what it is they're dealing with.

Heck, you should get your knowledge written down, build a website or something, pass on what you know.


> Oh well, everything is commoditized in the end even intelligence.

It certainly feels like there's a general dumbing-down (anyone in the UK watch Horizon yesterday, about the LHC? Ye gods it was so
slooow!), and numerous parties are to blame for that (media, the state, parents, teachers, etc.) but in a way I suppose it's inevitable that
as tech gets ever more powerful, it becomes easier to manufacture products which simplify how people can interact with technology
(touch sensitivity, voice recognition, facial recognition, movement tracking, etc.), leading to the irony of incredibly sophisticated devices
like the iPad being controlled by the most basic of hand gestures - in an odd sort of way it's like we're going back into the cave and are
scribbling on the walls once again.

The push from OS products like Windows and OSX is for GUIs that hide the guts of a system's functions (eg. Metro in Win8), a change
that seems unstoppable due to the rise of mobiles, smart phones & tablets. Ordinary citizens are able to do amazing things with modern
technology, but they haven't the slightest idea how it works. I agree that's a bad thing, but how to change it for the better without making
functionality more restrictive? I've no idea. The media continues to label people in IT and science/engineering in general as geeks, nerds,
boffins, etc., so no wonder prospective students avoid such subjects. Universities churn out endless lines of graduates in psychology,
media studies, hotel/catering, tourism, etc. - pretty much everything the modern economy doesn't need. Worse, retraining centers for the
unemployed only teach this sort of nonsense aswell.

Btw, my step-Father worked his entire life as a research engineer at Kodak (aswell as the obvious film & photography, he specialised in
speaker design, audio, magnets, etc. - he used to build cinema speakers); I've heard numerous stories of what he was involved with, so I
have a healthy respect for the kind of engineering you're talking about. Sadly, engineering is not a respected discipline in most western
nations now, unless you happen to be lucky and live in Germany (or Mexico, so I'm told).


Nuke writes:
> ... but programming is a whole different thing. ...

Doing it well requires the ability to abstract, something which is not really taught anymore. Schools just want pupils to pass exams, so
encouraging free thinking is not part of the agenda. Pupils who try to progress beyond the schedule are discouraged or actively
penalised for doing so. At 16, I had, "You should not know this equation!", written in my physics textbook because I'd learned something
on my own; thankfully a different teacher had the opposite view - he loaned me his course notes & books on Special/General Relativity.

Talking to a woman on a train more recently, she told me her daughter had been marked down because she had tried to learn
beyond the prescribed material.

There are two types of computer courses I've come across: Computer Science (what I did) and Compu ting . The former normally
includes a lot of theoretical background, an emphasis on the mathematical/logical basis of computer-related topics, strong support for
research-oriented follow-up courses, more avdanced optional topics like AI, etc. The latter is nowhere near as detailed, often a year
shorter, doesn't cover advanced subjects, focuses entirely on practical aspects of computers (eg. "multimedia", a term I found
sickening when working as a sysadmin in academia), avoids general concepts and instead teaches highly specific things such as
one particular application or platform instead of a general language or theory, and is aimed at getting people into non-research jobs
as quickly as possible. Too many edu places (especially colleges, and universities that used to be colleges or polytechnics) teach
the latter kind of course, creating students who are of little use in real industry; they can't abstract and often have no idea about the
real foundations of what they think they know. The former type of course is harder & more demanding, so many avoid them.

Example: I was told of a "Computing" graduate who, on his first day at work, was asked to comment at a meeting about how the company
could better organise/process its data. The poor guy said how about using a bubble sort; everyone else in the meeting literally laughed
at him, but that's the level of material which students are being given along with a bit of paper saying they now have a degree. I feel
sorry for such students, most have no idea they're being intellectually conned.


> Heck, one time I recruited a guy who was in college for computer science that was 19, and he whined all day about doing his
> programs in C++ instead of Visual Basic. ...

QED. Exactly the kind of thing I encountered. I met students who thought multimedia meant only Macromedia Director, because that's
all they'd been taught. They knew nothing about the basics of imaging, video, audio, etc. They had a "degree" in Computing with
Multimedia, yet they'd never heard of PAL or NTSC, couldn't explain fields vs. frames, had no idea what RGB meant, interlacing, etc.
Ditto for equivlent audio concepts. It was depressing talking to them.


> ... gone off topic ...

Yeah, but I'm a chatty mood. :D

Besides, it's related in a way. What do I use my SGIs for? Partly to maintain a website which as much as I can helps people to learn
about SGIs and make best use of them, while slowly merging in data about modern systems in a manner which eventually I hope will
allow people to get the bost of both the old and the new.


And hey, skywriter, it's not all gloom! Last year I had to get rid of a dozen Indigo2s and many low-spec Octanes which were taking
up space, not worth enough to sell, etc. I advertised on the local Freecycle forum, strongest interest came from a group of young people
who spend their spare time doing all sorts of cool things with computer tech, a blend of hardware, Linux and electronics. One of them
was building a general machine which could act as a wide variety of different hw systems (it had multiple mbds for different types of
systems in the same box, with custom hw/sw and front end to drive it all). The guy who came to collect the 20+ systems was only about
19 I think. I found it reassuring that people doing this sort of thing still existed. They were delighted to have a bunch of SGIs with which
to expand their experimental work.

Ian.
Nuke writes:
> I agree with much of this. They really need to stop penalizing students for getting ahead. And if I go to college, I don't think I'll learn
> anything, just because of what I've seen.

I guess it depends on where you live as to what extent you can learn at a faster rate than 'expected', obtain books ahead of time, etc.,
without driving teachers crazy. Helps a lot to have a sympathetic teacher, though in my case I was living on an island so there wasn't
that much the teacher could do to help (small school, few resources).

Also easier if you have an older brother/sister as I did.


> And I don't have an SGI, afterall, I am only 15. Hah.

If you're near where I live (Edinburgh, Scotland), come get a low-spec Octane for free. :D


theinonen writes:
> Overclocking and playing with modern hardware is pretty expensive hobby for average kids or younger people, ...

True in theory, but many youngsters these days have astonishing amounts of money to play with. Look at the smarphones they all seem to have now.


> ... and modern hardware is so complicated there is actually not much to learn from it really. ...

Usually, definitely yes.


> ... To really be able to understand how something works one must start from the basics and some
> old 8-bit systems are much better for that purpose.

True, but coming at tech from overclocking is still a start.


> I would also argue that hardware side is not very interesting these days,

I suppose it depends on one's point of view. eg. for some, it's what the iPad allows one to do which is important, while others are
more interested in how it works, tech specs, etc.


> but on the software side there is still much to do that will benefit the user much more than some extra MHz from the processor.

Someone said that home-brew apps for mobile devices is the closest thing akin to the 80s home-coding scene. Who knows.


> Some days ago I diggged my old MSX2/MSX2+ system out from the closet and must confess spent couple hours just playing Bubble Bobble and Space
> Manbow. ...

:D :D


> ... It really brought good memories from my younger days, and some of these days must go through those cardboxes with unbelievable amount of
> disks and find out what treasures are hidden there.

This is why I intend to expand my site at some point to cover vintage systems. I have a vast collection of machines - Spectrum, Commodore,
Atari, Amstrad, Acorn, Sony, CGL, T.I., Oric, Dragon, Tatung, etc. Fancy that as a bundle? Fuel + ZX81. :D

http://www.sgidepot.co.uk/misc/oldandnew.jpg

I have a domain ( www.vintage-micros.org.uk ) but not yet the time to sort it out.

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

Ian.

_________________
SGI Systems/Parts/Spares/Upgrades For Sale: http://www.sgidepot.co.uk/sgidepot/
[email protected] , [email protected] , +44 (0)131 476 0796, check my auctions on eBid!
If you write your own, and it works, you'll be famous. I'd build a statue. 8)

Ian.

_________________
SGI Systems/Parts/Spares/Upgrades For Sale: http://www.sgidepot.co.uk/sgidepot/
[email protected] , [email protected] , +44 (0)131 476 0796, check my auctions on eBid!
(bump)

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

Ian.
rooprob wrote:
... I may consider learning motif to make a little UI for my doings.


A reliable capture app?? Time to plan that statue... 8)

Ian.
Would it be any different for an R10K/R12K system?

Ian.
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.

_________________
SGI Systems/Parts/Spares/Upgrades For Sale: http://www.sgidepot.co.uk/sgidepot/
[email protected] , [email protected] , +44 (0)131 476 0796, check my auctions on eBid!
I guess the key question is, can one use a 'standard' LVDS cable in place of the original SGI-branded
cables for DM2/VBOB/etc.?

Ian.
I use an O2 as a firewall/gateway, running ipfilter. The onboard ec0 connects to my cable modem. Previously,
a normal SGI 10/100 card then connected to my 'internal' switch (48-port Netgear GS748T; this has four ports
one can use for Gbit transceivers). I've just replaced the SGI 10/100 with a Phobos P1000 optical card, but
I'm having problems getting the link to work. It did briefly, but it was slow and kept stalling, now it's not
connecting at all, keeps reporting the link as severed, but I'm wondering if I have some kernel parameters wrong.

The P1000 is detected ok as interface pge0, I've set it up in the same way as the old ec1 card was (in
/etc/netif.options, etc.) so ifconfig -av gives (ignoring lo0):

Code: Select all

ec0: flags=8410c43<UP,BROADCAST,RUNNING,FILTMULTI,MULTICAST,LINK0,IPALIAS,IPV6>
inet 93.174.90.126 netmask 0xfffffc00 broadcast 93.174.91.255
speed 100.00 Mbit/s full-duplex
pge0: flags=14863<UP,BROADCAST,NOTRAILERS,RUNNING,MULTICAST,DRVRLOCK,LINK0>
inet 192.168.252.1 netmask 0xffffff00 broadcast 192.168.252.255
recvspace 262144 sendspace 262144
speed 1000.00 Mbit/s full-duplex


The recvspace & sendspace used to be lower, so I've tried raising them but to no avail.


/etc/hosts has:

Code: Select all

192.168.252.10  winters winters.domain.org
192.168.252.1   gateway gateway.domain.org



/etc/ethers has:

Code: Select all

08:00:69:0c:af:e3   x1-6-08-00-69-0c-af-e3
00:60:f5:08:2a:b9   gateway


ipnat.conf has:

Code: Select all

map ec0 192.168.252.0/24 -> 0/32



In case anyone wants to know, here's the hinv (it's normally an R7K/350 with 512MB, but since I'm
using the system to post this, I've temporarily put in my R7K/600 board to speed up Seamonkey, but
everything else is the same):

Code: Select all

CPU: QED RM7000 Processor Chip Revision: 5.1
FPU: QED RM7000 Floating Point Coprocessor Revision: 5.0
1 602 MHZ IP32 Processor
Main memory size: 1024 Mbytes
Secondary unified instruction/data cache size: 256 Kbytes on Processor 0
Ternary unified instruction/data cache size: 1 Mbyte on Processor 0
Instruction cache size: 16 Kbytes
Data cache size: 16 Kbytes
FLASH PROM version 4.18
Integral SCSI controller 0: Version ADAPTEC 7880
Disk drive: unit 1 on SCSI controller 0
Integral SCSI controller 1: Version ADAPTEC 7880
On-board serial ports: tty1
On-board serial ports: tty2
On-board EPP/ECP parallel port
CRM graphics installed
Integral Ethernet: ec0, version 1
Video: MVP unit 0 version 1.4
with no AV Card or Camera.
Vice: TRE


Note that I can't put the old 10/100 back in because I've sold it. :D Hence the P1000 upgrade was kinda
forced (I'd been delaying trying until I had the time to test & be sure it was going to work ok).

Odd things...

a) The GigE switch does have a link light for the P1000 link as I type, but if I do something like change a
network parameter and restart the daemons, the light goes off & stays off, with SYSLOG giving:

Code: Select all

routed[X]: Send bcast sendto(pge0, 192.168.252.255.520): Link has been severed


If I unplug the cable at the switch and then plug it back in again, the Console reports:

Code: Select all

NOTICE: pge0: link up, running 1000Mbps full-duplex (forced)


but the link light does not come back on.


b) If I try route get for my main Fuel system, which is 192.168.252.10, it gives this:

Code: Select all

# route get winters
route to: winters
destination: 192.168.252.0
mask: 255.255.255.0
interface: pge0
flags: <UP,DONE,CLONING>
recvpipe  sendpipe  ssthresh  rtt,msec    rttvar  hopcount      mtu     expire
0         0         0         0         0         0         0      -665


Is that normal? Just wondering if it's supposed to show 192.168.252.0 instead of 192.168.252.10.

Cheers! :)

Ian.
[email protected]
+44 (0)131 476 0796
recondas writes:
> Not sure how helpful they might be, but the manual ftp://ftp.sonicwall.com/pub/software/NI ... dified.pdf and the
> IRIX 6.5 driver ftp://ftp.sonicwall.com/pub/software/NI ... 0-v6.5.zip for the P1000 are available via ftp.

Thanks! As you suspected though, I do already have them; they're on my Depot Resources page, the file is p1000.tar.gz I think. I read
through the manual, it has nothing useful about kernel tuning.

Just so odd... everything looks as it should be, but it doesn't work. Guess I'll have to find another normal 10/100 & put it back in. :\

Ian.
diegel writes:
> ... So I decided my P1000 is broken.

Hmm, I suppose that is one possibility, a bad card. I have several, so I'll try another.


> a) Are all optics and cables multimode? If you mix single and multimode you can run in similar problems

I'm not sure. Atm the setup only has one optical cable, the one linking the P1000 to a GBIC in the switch. All other
cabling is copper.


> b) If you have a managed switch: toggle the auto negotiation mode of your switchport

So far I've done nothing more than connect CAT5 cables to the switch. :D Not sure how to do anything beyond
that, will have to look into it, though I did set the P1000 to be forced Gbit instead of autoneg.

Ian.
The switch doesn't quite work that way. :)

The four copper ports marked 45 to 48 are simply 'shared' with the equivalent option ports, so one can connect a
copper line to port 45T (as it's labelled) or an optical line via port 45F, but not both. I don't think it needs any
special configuration (it's just a lowly Netgear, not a Cisco. ;D)

Ian.
The cable is multimode (it says 50/125) and yes the card is an SX.

I'm going to try using the Netgear management system from a PC to configure the switch (it employs a web-based interface).

Ian.
Quick update, got it working! Echoing diegel's comments, turned out the card was faulty. Tried another one this
morning and the connection worked first time, though atm the performance is very poor so I'm still looking into
possible kernel tuning issues. However, it's faster than the old 10/100 which is something, perhaps because
now the link is full duplex (10/100 card for O2 does not support full duplex mode).

If anyone has any ideas about how to tune the kernel for optimal performance of a Gbit card, please post, PM
or email. Perhaps sendspace/recvspace values?

Ian.

PS. To give you some idea how bad the performance is just now, a simple ftp only gave 15MB/sec, while a
ttcp test gave a mere 32MB/sec.
[email protected]
+44 (0)131 476 0796
Deary me, don't panic! :D Sorting this out isn't that difficult. Yes, the won't-close issue means the existing unit is duff,
but replacing it isn't really a problem. And SGI did ship later O2s with DVDROMs instead of CDROMs (Tohsiba SD-M1401
and SD-M1711, which I do have available if required), but if you want to replace with a directly compatible CDROM then
of course the Toshiba models to look for are the 32X 6201 or the 40X 6401. Note that other CDROMs may or may not fit
because of the catch positions required to hold the blue CDROM face plate, eg. I once came across a non-Yamaha CDRW
which I at first thought was a compatible fit, but on closer inspection I realised the CDROM cover had been glued on. :\

As for extracting the skins, removing the CDROM cover is easy. Just power on, press Escape when you see the message,
and let the tray eject if it tries to (if it doesn't, press the eject button to make it come out), then gently roll the O2 onto its side
(whichever is more convenient for you based on your handedness) and use a small flat-bladed screwdriver to lift the CDROM
tray catches just a tiny distance out of the way so that the cover can be moved 'upwards' and come free. Trust me, it's not
difficult at all. Once done, roll the O2 back up again, close the tray, hold it in place firmly with your hand while the unit tries to
eject it again; eventually it will give up, at which point you can press the power button once to turn off the system.

Check the O2 owners guide for removing the top lid, but basically slide a finger underneath at the back (after removing the
PSU that is) until you can feel the top lid's locking tab protruding downwards. Press the tab very gently inwards while at the
same time giving a little upward pressure on the back of the top lid with a fingernail of your other hand (or some other thin item);
at some point, the locking tab will be far enough out of the way of the chassis catch so that the lid can pop upwards. Key thing
is, don't press the tab in too far, and once removed, put it somewhere safe and well out of the way while you're working on the
rest of the system (all too easy to put it on the floor, then get up for some reason like the phone ringing, come back, forget
about the top lid, step on it by mistake).

Now remove the top 2 large screws holding the top cover in place and the cover should simply lift off. Two more large screws
of the same type hold the CDROM base tray in place: remove these, lift the unit upwards at the front, disengage from the chassis,
unplug the power and SCSI data cables.

Four small screws hold the CDROM in the tray. Remove these and the unit will slide free (older tray designs force the module to
be slid in/out horizontally; later versions are more flexible).

Remove the unwanted tray plate and flat facia cover from the replacement CDROM (unless it already has a proper O2 cover),
fit the curved black cover from the old unit (should come off easily). As long as you fit a Toshiba unit, the locking holes for the
black cover will be in the same place. Make sure the SCSI ID of the new unit is 4 (doesn't have to be, but it's the default used
by SGI), parity should be on, termination generally on.

Install the new unit in the base tray - make sure it's the right way round (the end with the catch hooks should be at the back
of the CDROM). Reconnect the power & SCSI data cables, re-engage in the chassis, lock it in place with the two scrws.

Reverse the earlier steps to put the system back together again. Once the top cover and top lid are on, power on the system,
let it fully boot. Login as root. Eject the tray, fit the blue CDROM cover, and that's it! Push the tray back in and you're good to go.

Cheers! :)

Ian.

PS. If you're lazy, I have all four CDROM/DVDROM models available, original SGI. Note that generic Toshiba DVDROM
models will not work quite right as they don't have SGI firmware (they do work, but they can be a bit fussy). By contrast,
generic Toshiba CDROM units will work ok.
I'm working on a charitable PC build for the Learn Engineering YouTube channel. Please PM/email/call if you'd like to contribute! Donations of items I can sell to provide funds are also welcome.
[email protected]
+44 (0)131 476 0796
+44 (0)7434 635 121
Cool!!

What controller setup was this using? One card? Two cards? And in which slots?

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
That's more like it! 8) Btw, what model cards are they? 3442X-R? 3800X?

Any tips on how you setup the RAID in XVM? So far I've only used xlv-make because I've
been focusing mainly on video I/O, but soon I want to experiment with RAID0 for general
use, so I'm assuming XVM is better for that.

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
Thanks!!

I'm a bit stuck for doing such tests atm, I don't have enough suitable drives, though I will do soon once again,
should have 6 x Seagate 600GB 15K SAS.

I'll also be testing with 4 x OCZ 120GB Vertex3 MAX IOPS SSDs, though earlier tests suggest performance
with IRIX is better with the Vertex2E. Here's a diskperf run in my /var/tmp (note the 4K random I/O compared
to your RAID0 test... ;)

Code: Select all

# req_size  fwd_wt  fwd_rd  bwd_wt  bwd_rd  rnd_wt  rnd_rd
#  (bytes)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)  (MB/s)
#---------------------------------------------------------
4096   39.75   33.27   39.94   33.07   39.70   22.56
8192   69.13   57.90   69.70   53.91   69.68   30.46
16384  109.03   92.98  107.86   79.81  110.71   54.90
32768  153.29  131.74  153.74  103.54  153.02   90.66
65536  192.25  170.21  188.90  151.19  193.54  137.95
131072  220.07  209.69  217.79  195.31  218.80  182.74
262144  238.19  233.02  234.89  228.38  238.00  220.60
524288  247.30  251.77  248.45  248.34  246.96  246.07
1048576  252.74  263.24  253.12  261.81  252.85  261.06
2097152  256.56  269.65  255.59  269.41  255.87  269.02
4194304  257.14  274.21  262.54  273.46  257.71  273.34


A 60GB V2E holds the root filesystem, but /var and /usr are on a 120GB V2E via a SAS3442X-R.

I also need to find another 3800X, only have one now, unless I can construct another 2-port breakout plate for a 3080X.

Ian.
I wasn't sure how to handle backups for a while, given the mix of devices in my Fuel (three SSDs, 600GB 15K SAS, 1TB SATA).
The third SSD (another 120GB Vertex2E, the early 3.5" version) holds my main 'ian' user account. The 1TB is my movie archive,
mostly backups of orig DVDs atm (the disk is a Samsung F3); I'll upgrade this later to 2TB when it starts to fill up.

In the end I bought a Startech 4-bay 5.25" enclosure, which can hold four 2.5" SAS or SATA devices. I'm going to fit it with
a 500GB SATA (two rotating backups of overall root), a 120GB V2E (/ian backup), 750GB SATA (backup of the SAS) and a
1TB SATA, all 2.5" devices. The Startech will be in an external 5.25" enclosure, the idea being that it uses minimal power.
Can't fit it inside the Fuel (bays taken up with CDRW & DVDROM) and wouldn't want to anyway; the external link will be
via a SAS cable to the 3442X-R, so it'll only use power when necessary. Nice & compact. And if whatever unit I use can't
handle the 4 devices at once, I have a couple of extra mini power PCBs removed from other units I can employ inside the
same case. Here's a link:

http://www.kikatek.com/P147865/SATSASBA ... e-2-5-inch
http://www.startech.com/HDD/Mobile-Rack ... TABAY425BK

I spent some time trawling 2.5" SATA specs, speeds & power consumption:

http://www.sgidepot.co.uk/misc/hdsummary.txt

The tables are ordered by best power consumption under load, but include ranks based on write performance, so one can judge
the best tradeoff of power vs. speed, and price.

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
Oh! :D

True, the Fuel could hold 8. I don't like the idea though that they'd always be powered on,
thus the external case. Longer device life.

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
Hmm, nah, TBH the delay when doing things would really annoy me. Plus, I don't have enough active systems for such a thing to be
worth doing yet, really only the Fuel atm.

Thanks for the info though! I didn't know that's what DMF was for.

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
For the Fuel, the Startech unit is basically it. The only other often-used SGI I have atm is my gateway O2, and that just has
a 2nd disk for cloning every now & then (very little changes on it, just running ipfilter). All my other 'for keeps' SGIs are not
really setup properly yet (original Effect O2 with all manuals, general R7K/600 O2, quad-1gig Tezro, Octane2).

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
Why would I use the Tezro? :D I'm mostly dealing with emails, web stuff, etc. The Tezro is for video, which atm
I don't have time to do. Since it's often convenient to leave the system on while I'm doing stuff so I can constantly
check things, power consumption is definitely an issue for me.

At one point I'd hoped to use the O2 as a mostly-on main desktop, but it's just too slow running Seamonkry/Firefox
(NB: my Fuel launches Seamonkey in just under 2 seconds now).

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
The only thing wrong with my setup atm is the Phobos P1000 Gbit NIC in the O2 runs really slow, barely
more than 5 to 10MB/sec. I'm sure there must be kernel stuff one can do to fix it, but so far I've gotten nowhere.

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
ShadeOfBlue wrote: Have you tried enabling jumbo frames ? It really helps , especially on CPU-starved machines.


Hmm, perhaps I've edited the wrong file?... netstat still shows 1500 for pge0 (the P1000 interface). In netif.options, if1name is ec0
(the link to the cable modem, DHCP), if2name is pge0 (internal static IP, host name gateway). I thought this meant the file to edit would
be /etc/config/ifconfig-2.options, which currently contains:

Code: Select all

netmask 0xffffff00 sspace 262144 rspace 262144 mtu 9000


Is there some other way of changing the MTU for the pge0 link if this method isn't applicable?

I changed snd/recvspace, but performance is still poor, around 32MB/sec. Real ftp speed is much worse, about 14MB/sec for sending
a file from the O2 to my Fuel, and a measly 4MB/sec for receiving a file to the O2 from the Fuel (ftp session running on the O2, 275MB file).

Note that sending a file from my i7 PC to the Fuel gives more like 95MB/sec.


mia writes:
> Does your site goes through the O2, ...

Heavens no. :D My main site is hosted by daily.co.uk, their download speed is pretty good. Try downloading this (89MB file), you'll see
what I mean. I get 2.3MB/sec, which is basically maxing out my 20Mbit Virgin Media link.

Ian.
jan-jaap wrote: I get 7.48MB/s ...


Holy crap on a cracker. :D

I could upgrade to 30Mbit for free, but that would require a different cable modem, a model about which I've heard bad things.

Virgin offer up to 100Mbit atm. You can have 500? Blimey...

Ian.
I started with 1Mbit 8 years ago; they never changed the package name, just kept increasing the speed, though now I
do wish at least once they'd simply drop the price instead (slow price creep over the years).

Ian.
(07/Mar/2015) FREE! (collection only) 16x Sagitta 12-bay dual-channel U160 SCSI JBOD units.
Email, phone or PM for details, or see my forum post .
[email protected]
+44 (0)131 476 0796
ShadeOfBlue writes:
> Try editing /var/sysgen/master.d/if_pge if it exists. ...

It does exist, but I'm not sure what I should change that's applicable:

Code: Select all

*
* Copyright 1998 Phobos Corporation
* All Rights Reserved.
*
* Phobos P1000 Gigabit Ethernet Driver
*
*FLAG   PREFIX  SOFT    #DEV    DEPENDENCIES
cs      pge_      -       -       bsd
$$$
/*
* Change pfe to et to allow SGI tools like net visualizer to recognize
* the interface
* char *pge_phobos_name = "et";
*/
char *pge_phobos_name = "pge";

/*
* Number of transmit and recieve buffers allocated at attach time
*/
int Phobos_pge_TX_DESC = 768;
int Phobos_pge_RX_DESC = 1024;

/*
* Transmit mapping threshold
*/
int Phobos_pge_MAP_THRESHOLD = 80;



> There's also a possibility that the Phobos card doesn't support jumbo frames at all.

Hmm, good point.


> The biggest performance gain comes from jumbo frames, but it's also possible that the driver is poorly written or the card is badly
> made. You should be able to get at least 50MB/s of raw performance as measured by iperf.

Of course I don't expect to get more than about 35MB/sec for an ftp (limit of O2's UW bus) but it ought to be better than it is atm.

I tried iperf (feel free to comment if I'm not using it correctly), here's the O2 acting as the server end:

Code: Select all

gateway# ./iperf -f M -s
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 0.25 MByte (default)
------------------------------------------------------------
[  4] local 192.168.100.1 port 5001 connected with 192.168.100.10 port 19704
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-10.0 sec  49.2 MBytes  4.91 MBytes/sec


but look at the output from the Fuel, re the write2 error:

Code: Select all

winters# ./iperf -f M -c gateway
------------------------------------------------------------
Client connecting to gateway, TCP port 5001
TCP window size: 0.19 MByte (default)
------------------------------------------------------------
[  3] local 192.168.100.10 port 19704 connected with 192.168.100.1 port 5001
write2 failed: Interrupted function call
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec  49.2 MBytes  4.92 MBytes/sec


What do you make of that? Note the Fuel is still using MTU 1500.


Here's the O2 acting as the client:

Code: Select all

gateway# ./iperf -f M -c winters
------------------------------------------------------------
Client connecting to winters, TCP port 5001
TCP window size: 0.25 MByte (default)
------------------------------------------------------------
[  3] local 192.168.100.1 port 57088 connected with 192.168.100.10 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec    290 MBytes  29.1 MBytes/sec


and the Fuel as the server:

Code: Select all

winters# ./iperf -f M -s
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 0.06 MByte (default)
------------------------------------------------------------
[  4] local 192.168.100.10 port 5001 connected with 192.168.100.1 port 57088
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-10.0 sec    290 MBytes  29.0 MBytes/sec


So it's much quicker sending data to the Fuel than receiving it. Strange...

Should I set the Fuel's MTU to 9000? Never bothered looking into it before since the speed of the Fuel's GigE has
always been really good to/from my other systems.

Ian.
mia wrote: Ian,

I'm sorry, this is a dumb question, but have you tried to disable ipfilter on your O2 when using the gigE card? By disabling I mean unloading the kernel module all together.


How does one unload the kernel module?

If you're thinking ipfilter could be a problem, could I test it merely by shutting down ipfilter from with /etc/init.d?

Ian.