Date: Sun, 03 Aug 1997 04:19:01 -0500 From: Tony Overfield <tony@dell.com> To: Terry Lambert <terry@lambert.org> Cc: hackers@FreeBSD.ORG Subject: Re: Pentium II? Message-ID: <3.0.2.32.19970803041901.006a69e4@bugs.us.dell.com> In-Reply-To: <199708022310.QAA00461@phaeton.artisoft.com> References: <3.0.2.32.19970802023853.0069c4c4@bugs.us.dell.com>
next in thread | previous in thread | raw e-mail | index | archive | help
At 04:10 PM 8/2/97 -0700, Terry Lambert wrote:
>You guys didn't call them that, but the memory bus was running at 33MHz;
>that's "clock doubled" in my book.
>
>I have a friend who had one until the Pentium math bugs reared their
>ugly heads (he was an IDE weenie, so he didn't care that PCI bus
>mastering failed on the thing).
Can you remember a specific model number? If so, I can reach up onto
my shelf and pull one of them down and measure it. I doubt it, but
maybe I measured it wrong the first time around.
>> Due to that bug, that version of the Saturn chipset does not support
>> write-back L2 caching and it is therefore configured by the BIOS to
>> use the L2 cache in write-through mode, though the L1 cache is always
>> write-back. The missing trace is not used or needed when the L2
>> cache is in write-through mode.
>
>Unless a DMA initiated by a controller rather than the host occurs.
>Check the -hackers list archives for postings on disabling the L1
>and L2 cache on these monstrosities...
I tried searching -hackers ("Dell AND disabl* AND cach*") and only
found one claim (yours, BTW) that this bug exists. Interestingly,
the only reply seemed to disagree with your claim. Perhaps you can
offer some better search keywords.
>Maybe not; maybe the memory was just jumpered for an extra wait state;
>whatever; memory access was half as fast as it should have been in a
>reasonable implementation.
Well, that's quite a bit different, but still apparently wrong. It
is true that the very first Pentium-60 systems had a write-through
L2 cache which makes memory writes appear slower, but it also
eliminates the DMA problem you mentioned.
>> >Just as my 486/50 kicks butt over the same chips.
>>
>> I doubt this, unless you're "stacking the deck" in some perverse
>> way, or you're simply dreaming.
>
>Of course I am, if an I/O intensive benchmark is "stacking the deck"
>and a CPU intensive benchmark is somehow "a good idea".
I have no problem with I/O intensive benchmarks, but improperly
interpreting the results and factoring in bogus assumptions should
be avoided, if possible.
Don't forget than Pentium memory is 64 bits wide and 486/50 memory
is 32 bits wide. Thus, your fancy 486/50 memory bus cannot help to
explain your faster I/O claims, so maybe you've got a "magic I/O bus."
>I'm doing disk I/O, of course, and I am overclocking the EISA bus on
>the machine to 50MHz instead of 2x25. 50MHz EISA transfers data faster
>than 30MHz PCI... or 33MHz.
Wow! You do have a magic bus. Do you expect me to believe that you
run your EISA bus at 600% of the maximum speed of 8.33 MHz!!!???
There's never been an EISA bus that is faster than a PCI bus.
>8-).
Oh, I see, you're just kidding.... ;-)
-
Tony
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3.0.2.32.19970803041901.006a69e4>