Topbench results:

Norton SI results:

Checkit results:

Landmark results:

Vidspeed results:

The results clearly show that even though the NEC V30 is very capable CPU, it is no match for a 286 at the same clock speed. The 286 is just more efficient and has a 2 clock bus cycle time vs 4 for the NEC V30. I have always wanted run these head to head and now my curiosity is satisfied.

This is interesting, thänk you for doing this ~

jasa1063 wrote on 2022-08-02, 23:56:

The results clearly show that even though the NEC V30 is very capable CPU, it is no match for a 286 at the same clock speed. The 286 is just more efficient and has a 2 clock bus cycle time vs 4 for the NEC V30. I have always wanted run these head to head and now my curiosity is satisfied.

Are the multiplexed address/data pins the big limitation on 8086/8088 chips as to why it takes so many cycles to access memory?

jasa1063 wrote on 2022-08-02, 23:42:

They are both 10MHz XT class computers with 4 8-bit ISA slots

XT-class with 286 is weird...
I guess single interrupt controller and single DMA controller, right?
Which protocol does the keyboard use? PC/XT or AT?

Anyway, that 286 looks unusually fast, at least in Landmark...
it's normal for Landmark to score 286 12 MHz 0WS as "performs like a 16 MHz AT"
286 10 MHz 0WS should be scored lower, right?

SI looks correct, though...
1.7 times faster than IBM AT 286/8MHz
in Landmark's words, that would be "performs like a 13.6 MHz AT"

jasa1063 wrote on 2022-08-02, 23:42:

This got me thinking that these computers are very similar. They are both 10MHz XT class computers with 4 8-bit ISA slots, 640K 0 wait state memory with a 16-bit data path.

The comparison of different architectures using the number of wait states is misleading, because:

jasa1063 wrote on 2022-08-02, 23:56:

The 286 is just more efficient and has a 2 clock bus cycle time vs 4 for the NEC V30.

So a 0WS 8086/V30 system has a theroretical maximum bus bandwidth of 5MB/s, whereas a 0WS 80286 system has a theoretical maximum bus bandwidth of 10MB/s. From a 286 point of view, the 8086/V30 operates like a 286 at 2WS.

My whole point here was to do a comparison of the two CPUs under as closely matched conditions as possible. The NEC V30 will almost always be at a disadvantage because the instruction cycle counts are not as optimized as the 286 and it is in a 4o-pin package forcing multiplexing. If anyone would like additional benchmarks, I will be glad to run them. Please let me know.

Thanks!

Grzyb wrote on 2022-08-03, 03:57:

Anyway, that 286 looks unusually fast, at least in Landmark...
it's normal for Landmark to score 286 12 MHz 0WS as "performs like a 16 MHz AT"

My system would disagree

Your 286 is definitely fast for a 10MHz.

For comparison:

My ibm model 30-286 has a 10MHz CPU (no fpu) and in checkit 3 it shows just 1889 dhrystones.

My Unitron 16MHz 286 0ws gets to 3854 dhrystones and 409 whetstones. Landmark 24MHz AT and 11MHz fpu.

That's a pretty poor showing by the IBM. My same 12MHz system as above for comparison:

maxtherabbit wrote on 2022-08-03, 14:21:

Grzyb wrote on 2022-08-03, 03:57:

Anyway, that 286 looks unusually fast, at least in Landmark...
it's normal for Landmark to score 286 12 MHz 0WS as "performs like a 16 MHz AT"

My system would disagree

I'm sure I saw 286 12 MHz 0WS scored as "performs like a 16 MHz AT" by Landmark.
I'm not sure, however, WHICH Landmark it was...
There were two widely used versions: 2.0 and 6.0, and their results could be different.
Can you run 2.0 on that system?

Grzyb wrote on 2022-08-03, 20:31:

I'm sure I saw 286 12 MHz 0WS scored as "performs like a 16 MHz AT" by Landmark. I'm not sure, however, WHICH Landmark it was... […]
Show full quote

maxtherabbit wrote on 2022-08-03, 14:21:

Grzyb wrote on 2022-08-03, 03:57:

Anyway, that 286 looks unusually fast, at least in Landmark...
it's normal for Landmark to score 286 12 MHz 0WS as "performs like a 16 MHz AT"

My system would disagree

I'm sure I saw 286 12 MHz 0WS scored as "performs like a 16 MHz AT" by Landmark.
I'm not sure, however, WHICH Landmark it was...
There were two widely used versions: 2.0 and 6.0, and their results could be different.
Can you run 2.0 on that system?

That would be it

"When you think of Standards... Think LANDMARK!" 🤣 🤣 🤣