Low BogoMips on Pentium 90MHz

Low BogoMips on Pentium 90MHz

Post by Yavuz Batm » Thu, 10 Nov 1994 01:11:48



Hi,

We have recently installed Slackware 2.0.2 on a 90 MHz Pentium box with 8
Mb Ram and 16 Mb swap and got 36.08 BogoMips. This seems quite low to me.
Any opinions?

-Yavuz
--
/---------------------------------------------------------------------\
Yavuz BATMAZ                 \\\//  Electrical & Electronics Engr. Dept.

\-------------------------ooO-(_)-Ooo---------------------------------/

 
 
 

Low BogoMips on Pentium 90MHz

Post by Anson M » Thu, 10 Nov 1994 02:53:50


: We have recently installed Slackware 2.0.2 on a 90 MHz Pentium box with 8
: Mb Ram and 16 Mb swap and got 36.08 BogoMips. This seems quite low to me.
: Any opinions?

On mine, I get around 36 as well (36.88?), which is in line with the
BogoMips howto/faq.  It states that a DX4 100Mhz maxed out around 50.
Can someone please describe the BogoMips algorithm as to why this is the
case?  Many thanks.
--


 
 
 

Low BogoMips on Pentium 90MHz

Post by Daniel J. Langri » Thu, 10 Nov 1994 11:34:00




>Hi,

>We have recently installed Slackware 2.0.2 on a 90 MHz Pentium box with 8
>Mb Ram and 16 Mb swap and got 36.08 BogoMips. This seems quite low to me.
>Any opinions?

Yeah, I'd think so...However, I've got a Pentim-90, 16MB RAM and I get the
same results...
--

Dan Langrill

 
 
 

Low BogoMips on Pentium 90MHz

Post by Warwick Allis » Thu, 10 Nov 1994 12:27:40



>We have recently installed Slackware 2.0.2 on a 90 MHz Pentium box with 8
>Mb Ram and 16 Mb swap and got 36.08 BogoMips. This seems quite low to me.
>Any opinions?

Looks spot on to me.  What were you expecting, 90.00?  BogoMips is a
completely BOGUS number, that measures how long your computer takes
to execute a pointless loop.  If you want to test the performance of your
machine, get software like iozone and xbench, don't worry about BogoMips!

Better still, run speed tests ont he ACTUAL SOFTWARE YOU INTEND USING.
For me, that means povray, xanim, and mpeg_play.

And of course, Read The Farking Manual.  /usr/doc/faq contains the
answers to 99% of the questions in this group.

--
Warwick
--

 /     * <-- Computer Science Department,  /     Microsoft is the question.
 \_.-._/     University of Queensland,    /
      v      Brisbane, Australia.        /     NO is the answer.

 
 
 

Low BogoMips on Pentium 90MHz

Post by anderson, jose » Thu, 10 Nov 1994 22:51:23



> Hi,

> We have recently installed Slackware 2.0.2 on a 90 MHz Pentium box with 8
> Mb Ram and 16 Mb swap and got 36.08 BogoMips. This seems quite low to me.
> Any opinions?

> -Yavuz

On the 2 Pentium machines (both at 90 MHz) this is the typical BogoMips
I get with them.  I have had it drop to 35.88 on my personal Pentium,
but the one at the office usually hangs in at 36.08.  Granted I think
it should be higher too, but I saw a post a week or so ago talking about
how BogoMips were calculated, and this number seems in line.

> --
> /---------------------------------------------------------------------\
> Yavuz BATMAZ                 \\\//  Electrical & Electronics Engr. Dept.

> \-------------------------ooO-(_)-Ooo---------------------------------/

Joe Anderson
-Until my Linux is on the net
 I don't have a catchy .sig
 :(
 
 
 

Low BogoMips on Pentium 90MHz

Post by BILL PARK » Thu, 10 Nov 1994 20:51:00


Thanks for your comment on Bogomips... I'd wondered what they were
about.

I'm also dead chuffed that my slow 486/33 doubled to 66 runs at
33.something when the guy you replied to got 36.xx on a Pentium!

I agree what you said a) about ignoring the speed rating and judging it
on the speed of the apps. you use and b) RTFM'ing... but can I find out
how to get my modem up and running?

I get confused between agetty, getty, uugetty, mgetty, getty_ps, some of
which don't appear to have man pages let alone FAQs

Sigh

Still, I'm happy enough with Linux in all other respects.. it's a good
learning experience, all I've gotta do is suss out how do I earn some
dosh from it!!!


 
 
 

Low BogoMips on Pentium 90MHz

Post by Warwick Allis » Fri, 11 Nov 1994 13:56:35



>I have had it drop to 35.88 on my personal Pentium,
>but the one at the office usually hangs in at 36.08.

Yes, those are the 2 numbers my machines generates.

Quote:>Granted I think it should be higher too

        if (_586) { bogomips *= 2; }

The only thing I don't like about my BogoMips is that that is about the only
thing that slows down my boot, now that I've rebuilt a kernel that only
supports (and looks for!) hardware that is actually in MY machine.  Worth it.

--
Warwick
--

 /     * <-- Computer Science Department,  /     Microsoft is the question.
 \_.-._/     University of Queensland,    /
      v      Brisbane, Australia.        /     NO is the answer.

 
 
 

Low BogoMips on Pentium 90MHz

Post by vincent w. y » Fri, 11 Nov 1994 11:47:42



>We have recently installed Slackware 2.0.2 on a 90 MHz Pentium box with 8
>Mb Ram and 16 Mb swap and got 36.08 BogoMips. This seems quite low to me.
>Any opinions?

Hi, no suggestion, but I get 32-33 on a 486DX2/66 with four mage of RAM
(though that shouldn't make any differenct) so your number is extraordinarily
low (I think it should be more on the order of 70-80...)

Are you sure you have your memory cache turned on?  And is it working?
(SRAMS are pretty sensitive to static as I'm sure you know...)

 
 
 

Low BogoMips on Pentium 90MHz

Post by Giao H. Ph » Sat, 12 Nov 1994 02:33:17



: : We have recently installed Slackware 2.0.2 on a 90 MHz Pentium box with 8
: : Mb Ram and 16 Mb swap and got 36.08 BogoMips. This seems quite low to me.
: : Any opinions?

: On mine, I get around 36 as well (36.88?), which is in line with the
: BogoMips howto/faq.  It states that a DX4 100Mhz maxed out around 50.
: Can someone please describe the BogoMips algorithm as to why this is the
: case?  Many thanks.
: --

36.27 over here.

 
 
 

Low BogoMips on Pentium 90MHz

Post by Markus Schod » Sat, 12 Nov 1994 00:47:46


The reason why the BogoMIPS rating for pentium processors is so low, has to do
with the loop that is used to measure it. It is a simple loop consisting of
two assembler instructions a dec and a jump instruction. Because of the
superscalar architecture of the pentium processor this leads to a stall in the
pipe because the jump instruction is in the v-pipe and attempts to jump to the
dec instruction which is still in the u-pipe. Jumping to an instruction that
is currently in one of the pipes seems to get the pentium out of synch the
result beeing that one run through the loop takes 5 cycles (as opposed to 4
on a i486). If an additional nop instruction is added to the loop the
execution time drops down to 2 cycles (I have tested this myself). In fact you
could add two "simple" instructions that don't interfere with each other and
it would still take 2 cycles. That is something like

loop:   movl %eax,%ebx
        addl %edx,%ecx
        decl %eax
        jne  loop

takes only 2 cycles but the loop used to measure BogoMIPS

loop:   decl %eax
        jne  loop

takes 5 because of pipe stalling.
Though I think we can safely say that the only thing you can do faster with
a i486 than with a pentium is a two instruction loop. Unfortunately this is
exactly how BogoMIPS are calculated.

Hope this clears up things a bit (and cheers up the pentium owners :-).

Bye
     Markus
----
"A man without religion is like a fish without a bike"

 
 
 

Low BogoMips on Pentium 90MHz

Post by David Hedl » Sat, 12 Nov 1994 18:54:44


[snip]

: loop: movl %eax,%ebx
:       addl %edx,%ecx
:       decl %eax
:       jne  loop

: takes only 2 cycles but the loop used to measure BogoMIPS

: loop: decl %eax
:       jne  loop

: takes 5 because of pipe stalling.
: Though I think we can safely say that the only thing you can do faster with
: a i486 than with a pentium is a two instruction loop. Unfortunately this is
: exactly how BogoMIPS are calculated.

Why does the kernel not use the first example you give? If this is more
accurate, then does using the second example imply that time critical
sections of the kernel which rely on the bogomips count to give a measure of
CPU speed, are being fed incorrect information. I'm sure most, if not all
of such loops are of more than two instructions!

David
--

+--------------------------------------------------------------------------+

| Computer Graphics Group                | and all computers are lousy     |
| University of Bristol                  | actors                          |
| England                                |                       - Anon    |
|          *** All opinions expressed are mine and mine alone ***          |
+--------------------------------------------------------------------------+

 
 
 

Low BogoMips on Pentium 90MHz

Post by roo » Tue, 15 Nov 1994 23:59:14


I get 39.94 on a 100 Mhz Pentium. This sounds about right.

Jered

 
 
 

Low BogoMips on Pentium 90MHz

Post by Hennus Bergm » Tue, 15 Nov 1994 04:25:04





>[snip]

>: loop:     movl %eax,%ebx
>:   addl %edx,%ecx
>:   decl %eax
>:   jne  loop

>: takes only 2 cycles but the loop used to measure BogoMIPS

>: loop:     decl %eax
>:   jne  loop

>: takes 5 because of pipe stalling.
>: Though I think we can safely say that the only thing you can do faster with
>: a i486 than with a pentium is a two instruction loop. Unfortunately this is
>: exactly how BogoMIPS are calculated.

>Why does the kernel not use the first example you give? If this is more
>accurate, then does using the second example imply that time critical

Because the first sequence is much slower on 386SX and 386DX machines and
would result in much less accurate timing on those processors.

Quote:>sections of the kernel which rely on the bogomips count to give a measure of
>CPU speed, are being fed incorrect information. I'm sure most, if not all

No. The BogoMips loop is merely used as a kind of ``delay constant''.
During booting the kernel times how often that loop can be executed in a
fixed time period. Given the number of loop executions per second the
kernel routine udelay(n) can then compute how many times it has to execute
the loop in order to produce an ``n'' micro second delay.

So, the actual time needed by the delay loop itself (__delay() in the
kernel source) is not important, as long as it is constant and short enough
to get a little accuracy. The only ``advantage'' of having a faster loop is
that your BogoMips rating will go up. The disadvantage is of course that
on faster machines the udelay computations may overflow.

The purpose of the udelay() stuff is to provide micro second delays in
device drivers where simple loops like

    (for (i=0; i<10000; i++)
         nop();

vary too much in time over various processors. [386SX is *much* slower
than Pentium for the same loop!]

If the delay loop actually needs more cycles on a Pentium, that would
be a *Good Thing* (TM ;-) because it allows the same delay loop to be used
on both slow 386SX CPUs and faster Pentium CPUs without risking overflows
in the udelay() stuff. Someday the processors will be so fast that the code
will have to be revised in order to avoid overflows on fast machines

Quote:>of such loops are of more than two instructions!

>David
>--

Hennus

BogoMips: n. [Linux] A {bogus} indication of processor performance
   on the ultimate {benchmark}, a dummy loop consisting of only a
   decrement instruction and a branch instruction. It was deliberately
   designed to be slower on faster machines, causing a lot of confusion
   on {Usenet}. (See {MIPS} and {bogon}.)

--

   PGP fingerprint: 76 99 FD 31 91 E1 96 1C  90 BB 22 80 62 F6 BD 63
       "when privacy is outlawed, only outlaws will have privacy"

 
 
 

Low BogoMips on Pentium 90MHz

Post by Jiann-Ming » Mon, 21 Nov 1994 16:11:03



>I get 39.94 on a 100 Mhz Pentium. This sounds about right.

You sure the FDIV bug didn't affect that answer?? ;)
 
 
 

1. Need suggestions on running Linux and X on a 90Mhz Pentium

Howdy!

I'm in a market of buying a Pentium 90Mhz machine and Linux would be
my major application platform. I would have about $5000 to spend...
(Well, it's my boss's money... 8^)).

Below would be my proposed configuration:

Pentium 90Mhz CPU
256K cache
32M RAM
1G SCSI 2 HD
PCI SCSI controller
17" color monitor (1280x1024)
PCI Graphics
Double speed CD ROM drive

My questions would be:

1. What's the best graphics card so that I can run X-Windows (Xfree86?)
   under Linux? I know that Diamond's cards are fast, but they are buggy
   in running X in Linux.

2. What sort of SCSI 2 controller of supported by Linux? I suspect
   those made by Adaptec would be O.K.??? Same question for CD ROM.

3. What would be a reliable vendor? I'm thinking of DELL or Micron.

4. Which distribution of Linux should I get? I'm thinking of buying a
   CD ROM distribution... Downloading tonnes of FD is kind of time-wasting.

5. Is there any problems running Linux under a 90Mhz Pentium?

I would appreciate any opinions or suggestions.

Thanks.

2. AIX 4.3 Interfaces for Windows?

3. Ethernet setup on Pentium 90MHz

4. LiLo does not work: it prints strange messages

5. Does Linux work on any 90MhZ Pentium?

6. USR5637 USB modem setup (2)

7. Need suggestions of running Linux and X on 90Mhz Pentium

8. SYSV support

9. SUCCESS: Pentium 90MHz, Slackware

10. 1.2.8 and Intel ZP triton 90mhz pentium

11. Micronics 90MHz PCI Pentium

12. Overclocking Pentium 90mhz to 100mhz - any flakiness ?

13. Pentium 90MHz?????