Erik Moeller wrote:
I agree with Gabriel. Moreover, we should order the
replacement *now*, as
we do still have more than enough money in the bank, and resolve the
issues with Penguin Computing over the next few weeks.
Well, it's all very strange, but I believe that we are now in
possession of a fixed Geoffrin.
Here's the story, as told to me by Jason, and as told to him by
Penguin. This is my understanding, but it may be flawed in minor ways
of course.
Yesterday, when Jason called, the server was still "in engineering"
according to the Penguin internal tracking system and the person Jason
spoke to indicated that it could not be fixed and that we would be
referred to our salesperson for a refund or replacement.
But, that report was incorrect. The machine will still "in
engineering", but it was being burned in after they changed something.
The machine (allegedly) went through a 90 hour burnin with no memory
errors.
The change was to "turn off interleaved memory access". Of course I
only barely know what that means (mostly from just understanding the
component words), but it certainly sounds like a plausible source of
the problem, and a plausible solution to the problem.
Jason is also FedEx'ing me the keys to the machine, so that I can
access the floppy drive without breaking into it. This will give me
the ability to run memtest86.
So, here's the plan.
When I get the keys, I will install the machine in the colo. I will
run memtest86 for 24 hours. If it passes, we will put the machine in
production. If it fails, then we will (a) order a replacement and (b)
demand a refund from Penguin.
I think that I'll get the machine installed late tomorrow afternoon,
Wednesday afternoon, and have it on the net for everyone on Thursday
afternoon.
--Jimbo