Motorola 68060

[Apollo68080]

Originally Posted by Karlos

MIP: Meaningless Indicator of Performance

This is why I prefer benchmarks of actual software, i.e. the things you'll actually use.

This makes a lot sense.


But also this can cause confusion.Lets say for example you like mp3 and you use as your yardstick lame (mp3 converter tool). On Aminet you can find not only many different version of lame, but also different compiled versions. I saw one compile properly done with compiler optimization turned on,
and another lame compile done (probably by accident) with all optimizations turned off and with debugger enabled.


Both of the lame exe look the same to the user but the not-optimize versions runs like 2 times slower.



So if people report their results - they can vary by 100% for no apparent reason.

[Karlos]

This is why you benchmark a range of applications.

Like Amiga Shopper always used to when doing accelerator reviews and roundups

[Apollo68080]

Originally Posted by Karlos

This is why you benchmark a range of applications.

Like Amiga Shopper always used to when doing accelerator reviews and roundups

The thing is that before you buy something, looking at benchmarks might make sense.
Before I buy a car I might look at the max speed or max horse power of the differentengines options.
You might even to a testdrive of two different engine options before you decide which one to buy.


With computers you maybe can not easily test drive different options yourself before buying. So you rely on others doing tests.


And sometimes these tests are confusing.
Like for example Sysspeed. A well know Amiga benchmark.
It shows for example MIPS scored of both 68K and PowerPC chips.


Is this nice for comparing them?
Maybe?


But is "test" code run on the 68K and PowerPC actually doing the same test on each or it in reality using two complete different in independent and incompatible algorithms?



So even test results that look genuine for the Amiga users - could be totally nonsense.

[AestheticDebris]

Originally Posted by Apollo68080

So even test results that look genuine for the Amiga users - could be totally nonsense.

Well, yes, but that's true of any consumer test results of any computer. Unless the only reason you want to buy a computer is to run the specific benchmark they're using, it will never be better than a vague guideline about performance.

[Kin Hell]

@ thread

Lets not bastardise MIPS please.....

https://en.wikipedia.org/wiki/Instru...r_second_(MIPS)

Thanks.

[Karlos]

MIPS is a pointless metric though. It's good for willy waving but not much else, except in rare architectures where most of the instructions take the same length of time. Even on the 68060 there's a big variation between the slowest operations and the fastest.

[grelbfarlk]

Originally Posted by Karlos

MIPS is a pointless architecture though. It's good for willy waving but not much else, except in rare architectures where most of the instructions take the same length of time.

I couldn't agree more.

[Karlos]

Originally Posted by grelbfarlk

I couldn't agree more.

Lool!

Aw, I feel bad. MIPS was a nice CPU architecture. Lots of interesting quirks.

[Kin Hell]

Originally Posted by Karlos

MIPS is a pointless metric though. It's good for willy waving but not much else, except in rare architectures where most of the instructions take the same length of time. Even on the 68060 there's a big variation between the slowest operations and the fastest.

Originally Posted by grelbfarlk

I couldn't agree more.