> Date: Sun, 5 Dec 1999 14:09:29 -0800 (PST)
> From: "Mr. James W. Laferriere" <babydr@baby-dragons.com>
> To: Robert Dinse <nanook@eskimo.com>
> Cc: Linux Kernel Maillist <linux-kernel@vger.rutgers.edu>
> Subject: Re: Hyper-Sparc CPU's in Sparc20 chas.(was Re: SS-10 -> Ultra Question.)
>
>
> Hello Robert,
>
> On Sun, 5 Dec 1999, Robert Dinse wrote:
> > On Sun, 5 Dec 1999, Mr. James W. Laferriere wrote:
> > > Hello Robert, I tried an grunt experiment . I put Two Ross
> > > HyperSparc Dual 66Mhz cpu's in my sparc20 chassis & attempted
> > > to install RH-6.1 on the system . After several lock ups ,
> > > and just plain funky behaviour (ie: key presses mis-interpreted) .
> > > I opened the case and removed the cpu's emmediately after power
> > > off , they were almost to hot to handle as a rough guess approx.
> > > 125 Deg. F. Now in my book this is way to hot for a cpu to be
> > > and would definately cause the flaky behaviour I (& maybe you)
> > > were seeing . I will later attempt to add some additional fans
> > > to the case & see if the added airflow helps .
> > > When I popped these out I dropped back in an old set of Ross
> > > 40Mhz Dual cpu's & started the process over , So far nothing
> > > ill has happened . Hth, JimL
>
> > I am sure that's not the issue. The reason I say this is that
> >SunOS 4.1.4
> > will run for months without a crash on the same hardware.
> I am not so sure this is a fair comparison . The reason I am
> saying this is (a while back) someone had mentioned that some
> instructions can have the cpu generate more heat . But, I never
> did hear of any proof of the concept .
While I'll grant you that Linux does allow you to utilize the CPU's more
effectively, the CPU occupancy is not high, nor does this problem seem to be
random as it does with your machine. I.E., I don't see misinterpreted
keystrokes, etc, I see spin_lock problems that always occur at the same
addresses. I don't have just random lock-ups, it's always a spin_lock() issue.
> Now this is in a sparc20 chassis right ?
>
> > My main shell server here is a SunOS 4.1.4 machine on the same hardware
> > except that it's also got a FDDI card in the box generating additional
> >heat and
> > if I don't have to boot it for some other reason it will generally go
> >about six
> > weeks between crashes.
> Hmmm, this still smells of something wrong . & probably with
> the cpu's . IF you have put other cpu types in & the difficulties
> diminished or completely went away or were differant .
Well, those CPU's work fine in a 4/670MP, the same modules and they work
fine under SunOS in ss-10 chassis, it only does this under Linux.
> Fyi, sbus option cards don't put more heat on the cpu's in a
> sparc20 chassis . The airflow is from power supply toward the
> option cards . Thus the power supply gets the coolest air then
> the cpu's then the sbus cards . But they (of course) do draw
> more current . Which is another item I am worried about in the
> sparc20 chassis is max. current draw at rated voltage . Which
> if the voltage drops will cause more heat in the system .
The chassis I have has a 150 watt power supply in it, but the SS-10 had
the exact problem.
I've seen heat problems before, and I've seen power problems before, and
I've never seen them always fail at the same location. Rather, I've seen more
sporadic things like random memory errors, etc. I am not getting any of that.
This is very reproduceable.
> > For what it may or may not be worth, I don't have drives in this box so
> > there isn't that source of heat, just the CPU's, memory, etc.
> I had just the one 4GB IBM sca (dchs04y) , which was warm but
> nowhere near as warm as th cpu's . Mine was configured as such
>
> 2) Ross Hyper-sparc 66Mhz cpu's
> 4) Dram simms totaling 128MB
> 1) Vsimm 8Mb
> 0) sbus options
> 1) 4GB IBM sca (dchs04y) HD.
> 1) sun 1.44 floppy .
>
> Now I have the slower Ross cpu's in the chassis & no more heat
> problems air flowing out is temp nominal (afaict) , & the install
> is continuing un-eventful . Twyl, JimL
I have to wonder if the fans are well on your box, or maybe the 66mhz
modules dissipate a lot more heat, I don't know. If memory serves weren't
those 5-volt CPU's?
To be honest, I've pretty much given up on getting these machines to run
Linux. Trying the SS-20 chassis was pretty much a last ditch effort. I
figured since it was stable on the 4/670MP, maybe the bug was specific to
SS-10, but apparently not. Maybe try again when 2.3/2.4 gets ported to
Sparc-32, if it does, since the locking is all changing again maybe it will be
less buggy. I've still got a few applications that can run OK under SunOS, so
I think I'm just going to try to migrate those applications that can run under
SunOS to these boxes. I have to keep SunOS on a few boxes anyway because there
are no drivers for some FDDI cards I have under Linux.
Now I'm playing with an ultra and if it proves to be stable I guess will
migrate other things to it. If not, perhaps give up and try BSD or something.
I want to be ready for Y2038 anyway.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/