[geeks] Mac Pro 1,1 CPU Upgrade - failure?

[Nathan Raymond]

Can't remember if this has been discussed - when you swap in the new CPUs,
do the new CPUs have matched steppings? In the case of the X5355 for
instance, there are three different steppings, and you can determine which
one a CPU has based on the S-Spec code:

http://www.cpu-world.com/sspec/SL/SLAEG.html

The X5355 G0 stepping has improved throttling ability (very close to the
original CPU idle temp) and possibly other differences (in some cases there
can also be microcode changes between steppings). I would check the S-Specs
and match them when using multiple processors in one system and if you can
get steppings where the idle power state closely matches the original CPU
that could be it (there are reports of people having to manually adjust
upward fan speeds to maintain stability otherwise).

- Nate


On Tue, Jan 3, 2017 at 1:15 PM, Mark Benson <md.benson at gmail.com> wrote:

> Update:
>
> I dropped the Mac Pro back to it's stock X5130 CPUs, and reinstalled the
> heatsinks using a better (more modern) thermal paste, which when I used it
> on my games rig i7 dropped the CPU temps by 7-8C over the previous
> Silver-loaded stuff I was using. I also applied sparingly using a thin skim
> applied using an old plastic card.
>
> I found after removing the heatsinks from the previous installation that
> good contact on the IHS surface just required the proper torque (tighten
> the screws until they squeak). Contact was even when they were subsequently
> removed after being tightened down properly, so I reinstalled them using
> the same torque technique.
>
> The problem, however, persists.
>
> Ubuntu Linux 16.04 runs fine, on a GUI Desktop. I left burnP6 running x4
> on all the 4 cores and also played some YouTube Videos, surfed and caused
> it to pretty-much roof out at 100% accross all 4 cores constantly for 10+
> minutes. CPU core temps sat at mid-high 50s C and the CPU heatsink temps
> rose very quickly and then levelled out at low 30s C after about 6-7
> minutes. Main and Exhaust fans were at a constant 650rpm (I think my Linux
> install currently has them hard-wired at that)
>
> Nothing broke or fell over, but burnP6 may be a bit of a dumb process to
> test with in that it probably just spams instructions and doesn't look for
> responses. Any better ideas for actual stress benchmarks under Linux would
> be appreciated.
>
> Mac OS X 10.6.8 lasted a good 10 minutes under duress with Geekbench run
> 2x but, again, eventually face-planted and everything locked up, leading to
> complete loss of GUI responsiveness. I've tried installing a fresh OS X
> 10.6 install and that also locks and dies part way through the install
> process. Rebooting and restarting the install crashed very early in the
> loading stages. This is usually indicative of heat build up?
>
> I'm stumped.
>
> Either it's working right and Mac OS X is screwed somehow (but I remember
> running 10.6 on it and it was fine, plus it worked okay before I dismantled
> it!) or if it's overheating somewhere I don't see or I've broken it. I
> really dunno. Only things left are to see if either the Northbridge or
> Southbridge heatsink are loose or unseated.
>
> I've tried 6 different CPUs (2 pairs of X5355s and 1 pair of X5130s),
> completely different sets of RAM DIMMs... I may even swap the Graphics
> board again and see if that's an issue (it didn't resolve it before,
> though).
>
> As ever, suggestions appreciated.
>
>
>
> --
>
> Mark Benson
> twitter.com/MDBenson
> _______________________________________________
> GEEKS:  http://www.sunhelp.org/mailman/listinfo/geeks