• Please review our updated Terms and Rules here

Benchmark Request : IBM PC AT @ 8Mhz vs. IBM PC/XT286 @ 6Mhz

Great Hierophant

Veteran Member
Joined
Mar 22, 2006
Messages
1,928
Location
Massachusetts, USA
I am interested if anyone could post benchmark results from these two systems. Both systems should be unmodified, i.e. no processor upgrades (although 287s are fine). Benchmark programs appropriate would be MIPS, Norton SI.

Both machines use an 80286 processor. Obviously, the 8MHz part should execute instructions and register operations faster than a 6MHz part. However, the 6MHz part should have faster memory operations due to its zero wait state RAM. Which is faster overall?

On the 286, a bus cycle takes two clock cycles. The ATs all have one wait state to RAM, so that is three bus cycles to access RAM. On the 6MHz 286, a clock cycle is 167 nanoseconds and 125 nanoseconds for an 8MHz part.

Therefore, with the wait state, the 6MHz AT takes 501 nanoseconds and the 8MHz AT 375 nanoseconds. The 6MHz XT/286 takes 334 nanoseconds. In light of the balance between clock cycles and bus cycles, wouldn't the machines be pretty evenly matched?

Also, if IBM says, as it does in its AT tech references, that the RAM has 150 nanosecond access time and 275 nanosecond cycle time, why the wait state in a 6Mhz AT? The RAM should be more than able to keep up with the CPU.
 
Last edited:
Well I have both at 6MHz and can tell you without a doubt Norton SI rates the AT at 5.9 and the XT-286 at 6.7 I know this doesn't really answer the question but it does prove that 0 wait states do make a difference. It wouldn't surprise me if a XT-286 at 6 would be able to beat a 8Mhz AT as well. Just my 2 cents
 
When the XT-286 was introduced, it was well known that its performance placed it between the 6 and 8Mhz models of the AT and was priced accordingly.
 
Just a small note; The XT/286 uses twice as much bus-time as the AT during memory refresh due to the memory system.
 
Back
Top