kernel module for math co-processor, MCP78

GlennsPref · 09-23-2008, 08:35 PM

I have this one module that is not recognised, and I wonder if it is contributing too bad kernel re-builds and the in-ability to install nVidia drivers for my graphics setup.

Here is some info...

My tag shows the current system details,

Quote:

Identification
Vendor: ‎nVidia Corporation

Description: ‎MCP78S [GeForce 8200] Co-Processor

Media class: ‎Co-processor

Connection
Bus: ‎PCI

PCI domain: ‎0

Bus PCI #: ‎0

PCI device #: ‎1

PCI function #: ‎3

Vendor ID: ‎0x10de

Device ID: ‎0x0753

Sub vendor ID: ‎0x1043

Sub device ID: ‎0x82f2

Misc
Module: ‎unknown

Quote:

bash-3.2$ uname -a
Linux GamesBox.GlennsPref.net 2.6.27-0.rc6.2.1mdvtmbcustom #1 SMP Sat Sep 20 14:16:26 EST 2008 x86_64 AMD Athlon(tm) 64 X2 Dual Core Processor 5600+ GNU/Linux
bash-3.2$

Any ideas?
Thank you, Regards Glenn.

TalebAtheist · 03-15-2009, 02:53 PM

well Glenn, after the usual annoying registering crap that refrain many from sharing knowledge (an IP address should always be sufficient to sue abusers, but here we go after 8 years of internet freedom attack and security paranoia by the Bush gang), I can answer to you.

That's not a math coprocessor, but a SMU (system management unit). It's a coprocessor that supervises to the SMBus and synchronises the multiple CPUs.
For windows drivers (nforce) can be downloaded from the nvidia site. The same was for linux drivers for previous versions of the hardware, but after the linux kernel developers had reverse-engineered the driver, actually improving it quite a lot, we have the nforce driver compiled into the kernel (either as a module or static). This driver works for various things, the most important of which is S-ATA bus.

It would be great, in terms of performance, if they had done something for this coprocessor too, given that this recurrent error from dmesg is the rule on dual CPU system from AMD64:

Code:

checking TSC synchronization [CPU#0 -> CPU#1]:
Measured 1066575215 cycles TSC warp between CPUs, turning off TSC clock.
Marking TSC unstable due to check_tsc_sync_source failed
Brought up 2 CPUs

It suggest me that we are not using the full potential of these machines and the key would be to actually make the SMU work. Something like that is in PPC systems, although I don't have experience on them:

Code:

/usr/src/linux/drivers/macintosh/smu.c

GlennsPref · 03-15-2009, 08:15 PM

Hi, thanks for the responce.

That explains a lot actually, at least I am not too concerned now.

Next time I rebuild a kernel I'll look out for SMU.

No doubt it is <NOT> there in this 2.6.28.x kernel I am running now.

Code:

glenn@GamesBox:~/bin$ uname -a
Linux GamesBox.GlennsPref.net 2.6.28-Preempt-0.rc8.1mnb #1 SMP PREEMPT Wed Mar 11 14:23:21 EST 2009 x86_64 AMD Athlon(tm) 64 X2Dual Core Processor 5600+ GNU/Linux
glenn@GamesBox:~/bin$                                                                                             (16-03 11:12)

Thanks for the info, I'll keep an eye on it.

cheers, Glenn

btw, Welcome to LQ, try out the search function here at LQ, it may save time waiting for an answer. ;-)

TalebAtheist · 03-17-2009, 07:02 PM

Hi Glenn,
Thank you for your tips on LQ features, sure I'll have a look!

About the coprocessor issue, neither will it be in the new 2.6.29* kernel.
But it's something I'm willing to investigate, by asking to the kernel developers for this new features. They'll clarify if it is really worth to write some code or my interpretation was wrong.

Another interpretation I stumbled into some months ago was that it is a fake GPU, but sounds unconvincing to me because we are using two different pci addresses for this...
I'll keep you informed if any news will come.

cheers, Florenzo

GlennsPref · 03-17-2009, 10:05 PM

Hi, firstly, I meant if you had a question, not meaning to direct you some place else.

Or to assume anything. I was not very clear on that I realise.

About the co-processor, Could it be a spdif connection on/for my graphics card?

The only other thing is physx processor, which neither my motherboard or graphics card has.

I tried earlier (oct last year) to point out to pciid's that the gf8200 is not the chipset I have.

But alas, it is called that anyway. I tried to point out that not all Motherboards have the graphics chip onboard.

Anyhow, It causes no problems as far as I know, and I will probably have new hardware by the time it is corrected(?).

Cheers, and thanks for your input.

Regards Glenn