Building a Sossaman rig?

[D_A]

Time to whip out the solder sucker and un-mount it then.

[tiro_uspsss]

Time to whip out the solder sucker and un-mount it then.

if I did that, & it works... I swear I'll be pissed.. very pissed.. cause imo, the slot, with nothing in it, is merely an 'extension' so-to-speak of the traces within the mobo. Its like putting in an extension cord into a power outlet - it shouldnt affect the main circuit.. u get where I'm coming from?

[Nosfer@tu]

if I did that, & it works... I swear I'll be pissed.. very pissed.. cause imo, the slot, with nothing in it, is merely an 'extension' so-to-speak of the traces within the mobo. Its like putting in an extension cord into a power outlet - it shouldnt affect the main circuit.. u get where I'm coming from?

Somthing is detecting the extra slot.
Either electronical or bios.

[fart_plume]

bios code is prolly turning off the board telling it has a current overdraw. It's possible that they drilled an extra pin hole in that slot that is not mean to get a pin, such as, how it knows which slot is which, whether that trace maybe part of a feedback loop feedback for a particular slot and by connecting all the pins you now have two slots saming they are the same one.

[niconoise]

code 2a is in the manual on page 139.
it references dim codes on page 143.

exactly which sspec cpu do you have?
which board rev?
which bios rev?
did it work before the mod?

[D_A]

I was searching for "02a"

Initialize different buses and perform the following functions:
* Reset, Detect, and Disable (function 0). Function 0 disables all device nodes, PCI devices,
and PnP ISA cards. It also assigns PCI bus numbers.
* Static Device Initialization (function 1). Function 1 initializes all static devices that include
manual configured onboard peripherals, memory and I/O decode windows in PCI-PCI
bridges, and noncompliant PCI devices. Static resources are also reserved.
* Boot Output Device Initialization (function 2). Function 2 searches for and initializes any
PnP, PCI, or AGP video devices.

It's looking like that extra slot might be a bad thing after all.

[Jaco]

Bios can't handle the extra slot maybe?
No idea how the bios is programmed , but it could just simply fail if it detects something unusual like the extra pci slot.

[D_A]

Might be as simple as not having enough pci addresses in the pool to cover everything, so it craps out.

[Movieman]

nevermind... someone else just told me that 'pni' is supposed to mean sse3, is that true? so why does -mtune keep defaulting to pentium-m instead of prescott?

the chips are essentially pentium-M's not prescotts..
they are Yonahs, a dual core pentium M, sucessor to the Dothans..

[rcofell]

nevermind... someone else just told me that 'pni' is supposed to mean sse3, is that true? so why does -mtune keep defaulting to pentium-m instead of prescott?

Well, PNI stands for "Prescott New Instructions" and is SSE3, much like SSE was originally KNI (Katmai New Instructions).

The reason it's probably picking the Prescott for your mtune is because the instruction set matches with it, as beyond that I don't think the compiler care all that much. I assume it's just trying to optimize around the available instruction set and not really assuming too much about the cpu's organization, but I may be wrong to a degree on that latter part, who knows.

I would suggest maybe trying "-mtune=pentium-m" with "-msse3" to see if there are any advantages, though I'm not sure how much it takes advantage of SIMD instructions unless you use "-ftree-vectorize" as well.

Personally, I don't really expect very big gains out of compiler optimizations(beyond "-O2" or "-O3"), unless you change to a completely different compiler, as from my experience it didn't really help on something like GNUradio (a signal processing framework), where implementing a routine in assembly using SIMD instructions made a night and day difference between any attempts with compiler optimizations.

[tiro_uspsss]

oh, and tiro, what did you end up doing about the board?

I putting it down as a 'bricked' mobo.. so I'm going to have to buy another once I get some spare cash

[D_A]

... and if those aren't as rare as rocking horse ... shavings ... I don't know what is!

[tiro_uspsss]

when ppl were BSEL modding their C2D's, they were modding the ones that had a lower FSB, yet same amount of cache etc as another C2D model in the Intel model line-up.. with these Yonah CPUs, there is no higher FSB than 667 - maybe thats why BSEL modding aint working? - if there were 800FSB Yonah, then it might work.. anyone follow me?


BTW, where the hell is the info, esp stepping for that ultra rare 2.13Ghz CPU?

[Martijn]

All sossaman systems are rare... I bellieve my CPU stepping wasn't even to be found on the Intel CPU finder.

EDIT: Just checked, none of them are.

[Mats]

when ppl were BSEL modding their C2D's, they were modding the ones that had a lower FSB, yet same amount of cache etc as another C2D model in the Intel model line-up.. with these Yonah CPUs, there is no higher FSB than 667 - maybe thats why BSEL modding aint working? - if there were 800FSB Yonah, then it might work.. anyone follow me?

AFAIK, the important thing is that the NB have support for 800 in this case. But I'm not sure if there are working BSEL pins for that on this mobo.
You're right about the corresponding mobile chipsets, the 945 series, they run at either 533 or 667.

[Martijn]

oh, i know what i have to do to get the overclock, i have to pull pin 51 low on the ics932s208, but my soldering iron is too large and clumsy. there's not much room for error since the pin pitch is so small. i'm not sure that a hot air rework station would even work since there are a lot of little components that might get pushed around in the process. i'll be getting a thin tip for my iron in the next little bit (probably next month when i have money to spend again) and then i'll warm the bottom of the board with a hot air blower, and reflow just the pin in question with the iron so i can pull it up with a hobby knife.

i'm determined to get to 2.4ghz on this board. buwahahaha.

Be sure to let us know when you get there

[Movieman]

check this out: http://www.intel.com/assets/pdf/gene...s_m_060830.pdf
it looks like they once offered a model with 4 intel boards (8 cpus/16 cores) and 2.33GHz sammys!

Now thats a Sammy spelled SAMMY!

So I improvised using a little hot air blower I found in the basement with all of the random craft stuff:


Once it was hot enough, I pried up the pin with a hobby knife, but unfortunately the pin just snapped off! No turning back now!

According to the datasheet this should now be un-bootable as the outputs will all be reference for test, but oddly enough, it still boots at 166 FSB. When I connect FS_A to GND using a 10k resistor, it doesn't boot. It's looking like the FSB setting is set via the internal fuses on the device... No overclock for you!

Sorry to hear this but my Thanks for making the effort!

From memory this is a 7520 chipset, the same as on my old Dual xeon Irwindale(3600/2mb/800) so in theory 800FSB is supported but I asked Franck(CPU-Z) back 2 years ago about it and he said the Supermicro board had 2 PLL's to back each other up so no go on making changes. Then I tried the pin mod to 200 and it just did 166..
It's there, somewhere, just have to find it..
OOPs, just checked, chipset on my Irwindale rig was a 7525 not 7520 so there is a difference.

[s0lid]

Tried to change that Crystal next to that pll chip?
If you change it little faster it should raise the FSB :]

[ibby]

oh, i know what i have to do to get the overclock, i have to pull pin 51 low on the ics932s208, but my soldering iron is too large and clumsy. there's not much room for error since the pin pitch is so small. i'm not sure that a hot air rework station would even work since there are a lot of little components that might get pushed around in the process. i'll be getting a thin tip for my iron in the next little bit (probably next month when i have money to spend again) and then i'll warm the bottom of the board with a hot air blower, and reflow just the pin in question with the iron so i can pull it up with a hobby knife.

i'm determined to get to 2.4ghz on this board. buwahahaha.

Hell guys,

Alot of you might not know me but some old and evil people do.

can you ALL

Please check on your boards and confirm if the clockgen is

ics932s208
or
CK409

[NaeKuh]

Hell guys,

Alot of you might not know me but some old and evil people do.

can you ALL

Please check on your boards and confirm if the clockgen is

ics932s208
or
CK409

lol i can get that info for you when i get home.

Whats the easiest way to check for it? I havent touched a clockgen since Pentium 1 days.

[Movieman]

The ICS chip is a CK409, it's just a general term.

You'll find the chip on the Intel board in one of the three circled locations (thanks to tiro for the image). It's most likely the one in the middle.

edit:
I should also mention that the Intel board most likely has the same ICS chip for a clockgen since the Tyan i7520SD also uses the same chip (you can see it in the pics on Phoronix). I'd be surprised if Intel used a different chip being that Supermicro and Tyan probably used whatever was in the reference design from Intel.

ibby is one sharp dude when it comes to this stuff.
he and I started talking about this last year but then he moved and it sort of got forgotten.
I spoke with him tonight, he has chips but no board.
my thinking is I send him 2 boards,mem and heatsinks.
He pays for all but one of the boards.
That one is on me for him to try mods on.
Make sense to you guys?

[Bobsama]

Just a question to those guys wanting to OC with a clockgen--why won't physically breaking one or even two of those clockgens work? As I understand, the entire purpose of redundant clockgens is to ensure that if one dies, there's another one to make sure it keeps running. And heck--most of us have a desktop with a single clockgen--and it's rare that one of those die.

[NaeKuh]

BRB i'll go take a look at what those numbers say.

The white sticker says FBB_07

i have no idea what the other chips say because i stupidly put an enzo over it for the hell of it. And i cant peal them off.

:T

My sammy is definitely the blingist out of them all. Ask dave. :X

[ibby]

No doubt he is. My apologies, ibby if it sounded rude.

Well, that's actually my next move, but I'm not finding the twin clockgen so easily. It really feels like the internal pull ups are getting in the way right now. The other problem is that my connection straight to GND is also violating the limit of 5 uA (per the spec sheet) for the FS_A input pin. The good news is that my sammy has been happily crunching for 2 days now without any issue despite the missing pin. It seems to only check the input states when it initially establishes the phase lock on boot. After that, we have to fiddle with the i2c interface to get the chip to change frequencies.

No need to apologies you have made some very good points and are miles infront of me

[ibby]

Hello guys,

I’ve been looking at datasheets and acquired some advise and help from an proper guru persivore.

These are my findings

To reach the magic 200 MHZ maybe a lot harder then I initially anticipated.
But here goes!

FS_A = 0
FS_B = 0
And
B6b5 = 1

0 = Ground
1 = some voltage of some sort with a value


Lets talk about this value B6b5

Me and my guru came to a conclusion that this value is acquired either from the BIOS or PLL.

So unless some one can reprogram the PLL, I don’t think applying extra voltage to any of the pins, FS_A or FS_B will give us the desired results we needed.

From my previous experiences with ViL ViH, it made very little difference apart from sometimes increasing stability and lowering cpu life.

Admittedly the easiest solution would be to use clock gen, but I think the BIOS is also playing a roll in not letting you get to 200FSB.

Can some one give me a link to the current BIOS, and tell me what version / revision and what brand it is?