🚀 go-pugleaf

In article <92252.163135I18BC@CUNYVM.BITNET> I18BC@CUNYVM.BITNET (Spartacus At Cuny) writes:
>From: I18BC@CUNYVM.BITNET (Spartacus At Cuny)
>Subject: Re: Why shouldn't I buy Cyrix
>Date: 8 Sep 92 20:31:35 GMT

>In article <Bu9urG.7zu@nntp-sc.Intel.COM>, dmccart@gomez.intel.com (D. J.
>McCarthy) says:
>>
>>The following are just my opinions and not Intel's.
>>
>>Daniel Moran (daniel.moran@pcb.batpad.lgb.ca.us) wrote:
>>> jas37876@uxa.cso.uiuc.edu (John A. Slagel) writes:
>>>> I can, all things considered, upgrade my non-coprocessed 386-33,
>>>> to the Cyrix 486-33 for ~$150.
>>>> Is there problems with this chip ??
>>> Ditto on that request. I'm extremely interested in this chip. My
>>> understanding is that it does not have an on-board math coprocessor; but
>>> I'm more interested in the cache and pipelining anyway. I'm running a
>>> clone motherboard with a very recent AMI BIOS. Any comments? Anyone
>>> who's done this upgrade?
>>
>>
>>From what I've read, the Cyrix chip:
>>
>>1) Has a hardware multiplier instead of a full-blown math co-processor
>>2) Has a 1K cache instead of the 8K on an Intel 486
>>3) Is pin-compatible with Intel 386's, although you can't just pop out your
>>   i386 and plop in the Cyrix chip; you will have to replace at least the
>>   BIOS and probably the motherboard too.
>>

>     Yikes! A friend from school was thinking of upgrading his
>Opti 386-25DX motherboard bought last November with the
>Cyrix 486DLC chips.  He's running OS/2 on his machine right now.
>Would anyone know if Cyrix offers a money back guarantee?
>-------
>------------------------------+      All The Best,
>Reply To:                     |
>                              |      Edward Galarza,
>     I18BC@CunyVM.Cuny.Edu    |      Brooklyn College Of The
>     I18BC@CunyVM.BitNet      |      City University Of New York
>------------------------------+      "Que viva el OS/2!"
>New accounts above 8-)

Here is some information about the chip from the Microprocessor Report.  The 
Cyrix 486DLC does not include any on-chip floating point hardware.  It does 
feature a hardware integer multiply, which gives it a faster integer 
multiply than even the 486DX.  The Cyrix instruction pipeline is five stages 
and very simliar to the Intel 486.  The Cyrix design does not include a 
seperate address computation ALU, which makes some instructions take one 
cycle longer than an Intel 486, but most other instructions take the same 
number of cycles.  The on-chip cache is 1K, and it suffers slightly because 
of the 386DX pinout.  Intel added some cache control signals to the 486 
pinout, some of which can be provided to the 486DLC in specially modified 
or re-designed systems. 

The Microprocessor Report reported SYSmark92 benchmarks showing the Cyrix 
486DLC to be about 10% slower than the Intel 486SX at the same clock speed.  
SYSmark92 is a good benchmark of overall system speed, so it should be 
respresentative of performance on real-world applications.

Cyrix will also offer 33 MHz and 40 MHz versions of the 486DLC, which may 
give them a competitive advantage relative to the 486SX.  Intel does not 
offer fast clock speed versions of the 486SX to protect the higher margins 
they achieve on 486DX sales.

Cyrix is marketing the 486DLC with their 387 replacement, which is somewhat 
faster than Intel's 387.  They can advertise a 486DLC + floating point 
against a 486DX.  Of course the floating point performance is somewhat 
slower than the 486DX's on-chip FPU, but the price is dramatically lower.  
Intel charges a huge premium for the 486DX and the 487SX (what the market 
will bear), so the much lower incremental cost to add some kind of floating 
point to the 486DLC may be another advantage for Cyrix.

The Microprocessor Report says that the chip can be used to upgrade existing 
systems, but they do not elaborate on the compatability issue.


Mark Gilbert