🚀 go-pugleaf

In article <4tdcv0$o6r@apollo.kosone.com>, Dave Rolo
<dave@mmbc.kosone.com> writes
>Hi there!  Here's a good little problem for you guys.  I had a 386dx40 
>motherboard and cpu in a system and got a hold of a 408dx33 motherboard 
>and cpu.  I installed the new motherboard (and I've done this a million 
>times) and the system ran slower than the 386dx40 motehrboard.  All the 
>jumpers are correct, I tried different CPU's and still had the same 
>result.  Norton sysinfo rated the CPU performance at about 23, about 15 
>points below a 386dx33!  I got ahold of another motherboard, different 
>brand (one that I have worked with many times before with great results) 
>and in one case, used a 486dx33, and in another case used an AMDdx2/80. 
>With the dx2/80 cpu, sysinfo rates the CPU performance at about 38, just 
>over a 386dx/33.  I remind you, this is with a completely different 
>motherboard and cpu that I KNOW work!  Fine...maybe there is a bottleneck 
>somewhere so I completely remove the network card and cd-rom controller 
>and put a new Trident video card and a high speed LAVA controller card.  
>Still...the same result.  The CMOS is set at the best possible ratings 
>(I've tried multiple configurations with the cache, DMA, shadow RAM, 
>etc.) and the jumpers are correct according to Seanix tech support (the 
>manufacturer of the board).  The only thing I haven't changed is the 
>power supply and the drives (but I can't justify doing all that work just 
>as a shot in the dark).  Any input would be greatly appreciated...please 
>e-mail the reply to dave@mmbc.kosone.com.
>
>Thanks...Dave Rolo 
>
I have just had a similar experience going from 386 DX/33 to Cyrix 486
DX2/66 on different motherboards, but using same hard drive VGA card,
etc.

Although my microprocessor benchmarks faster, memory (8 Mb EDO) is
slower than the old 8 x 1 Mb simms. Forther testing has narrowed it
down (CACHCHK4.ZIP from PCMAKUG forum on Compuserve) to slow
L2 cache (almost certainly NOT pipline burst mode). This test showed
L1 cache at 19.7 ns/byte, L2 at 67.8 ns/byte and main memory at 87
ns/byte compared with another 486 DX2/66 getting 20.7, ?? and 68.3,
respectively. The ?? represents the fact that L1 and L2 were so close in
speed that the program did not recognise both levels. Most disturbing is
that a RAM write takes me 1082 ns compared with 61 ns on the other
486.

Using an old Windows User magazine benchmark suite my 386 got an
index of 5.5, the Cyrix gets 7.0 and my girlfriends Escom 486 DX2/66
gets 16.3!

Am checking whether my machine supports faster L2 cache. 
-- 
Mick Southgate