Re: too good to believe

Simon's Mailing List Account (slk@scan.shodor.org)
Mon, 4 May 1998 11:14:30 -0400 (EDT)


> I had this e-mail exchange with a vendor in a hardware newsgroup-- but
> it was a group about ASUS motherboards, not CPUs or linux. Since this
> sounded far too good, I am wondering if anyone can confirm or deny the
> numbers here [`me' is myself speaking, `him' is Rick Lindsay, a vendor
> of decent reputation on the net].
>
> him1: Complete Linux kernel compile times:
> him1: Dual P2-300, 64mb SDRAM, 75 seconds
> him1: Dual PPro166/512, 64mb EDO, 45 seconds
>
> me2- Eh, I am shocked. *45* seconds!! What kind of setup are you using,
> me2- Rick? I ran a compile of the Linux 2.1.98 kernel on a machine withg
> me2- PPro 166/512s in a SuperMicro P6DNE, a SCSI-II 7200rpm disk, and 192MB
> me2- in EDO SIMMs [using BIOS memory settings of 60ns, speculative reads on
> me2- and PCI posting on]. I used make -j bzlilo which took a total of six
> me2- minutes to finish. What version of kernel and which compiler are you
> me2- fielding over there?! And more importantly, are you fielding Cheetah
> me2- RAID-5's on a AHA2940U2W?

(note: I know Rick Lindsay, and have directly and indirectly done
a lot of business with him. He's one of the few that'll build OS/2
workstations, as well as rather 'large' Linux boxes)

The system configs were only a little out of the ordinary... the Asus
P65UP8 motherboard is *fast*. (I know that particular dual pro system
eventually had a DPT RAID board, though I don't know if it had it for
the kernel compile) (I was in the process of helping a family member
order a rather large system from Rick). Considering that a K6/233
can get a kernel compile of around 3 minutes on a 96MB system
with IDE drives (Asus P55T2P4 board), I'd say his times are
quite reasonable.

Again, the Asus boards are a big advantage... they're a lot faster
than Supermicro under Linux, as well as (in my experience) noticably
faster than Gigabyte. (ok, so it helps that Asus tests with Linux)

I can say that your 6 minutes on a P6DNE with dual pros is AWFUL,
as I do better than that by 90 seconds or so on a single K6/166
(Gigabyte board) with 5400rpm SCSI disk.

Oh, and the K6/233 times above are under Redhat 4.2, the K6/166
times and (I believe) Rick's times are all under SuSE 5.1.
I've found that the glibc-based compiler in Debian hamm is around
half as fast as SuSE's libc5 based compiler (both 2.7.2.n) in kernel
compiles on the K6/233.

Simon Karpen slk@shodor.org
Sysadmin, Shodor Education Foundation

"On two occasions I have been asked [by members of Parliament!], `Pray,
Mr. Babbage, if you put into the machine wrong figures, will the right
answers come out?' I am not able rightly to apprehend the kind of
confusion of ideas that could provoke such a question."
-- Charles Babbage

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu