Speed rankings of Next Gen hardware

[Blizz1220]

Good summary.  I'm sure someone will post a response disagreeing with you in 3...2...  

:rofl:

[spirantho]

Those timing tests are dead-on accurate and reflect all other tests made
by big and professional companies I read ... Those tests can be used to
give you a very clear picture of what you're buying ...

If you showed that timing graph to a real benchmarking forum they'd just laugh at you.

1) What version of the tools benchmarked was being used? If not the same version, then no comparison can be made.
2) What compiler was used for each system? If not the exact same version of the same compiler, then no comparison can be made.
3) What RAM did each computer have - did each platform have the best RAM available? If one platform was using sub-optimal RAM, and another was using fastest available, then no comparison can be made.
4) Do the tasks involve HD access? If so, what HDD was used? Was one using an SSD and another an HDD? Did one have DMA available and another not? Was one SATA, and another PATA?
5) Were each of the benchmarks carried out on the latest versions of the OS on a clean installation? If not, no comparison can be made.
6) Were all background tasks halted before the benchmark was executed? If not, no comparison can be made.
7) Was the network disconnected from each machine? If not, no comparison can be made.
Does one of the systems support SSD instructions and one not? If so, then the benchmark is meaningless except in the area of SSD apps, and should not be used arbitrarily.

Benchmarking between systems isn't just a matter of running a program with the same parameters on two systems and timing the difference. You have to be certain that ALL other parameters are equal.

I remember the thread that spawned those benchmarks. Piru had gone through the forums and found results from other people, regardless of their system. I also know that his benchmarks were using an older version of the tool under AmigaOS 4.

These "benchmarks" were just numbers trawled from forums, and have no meaning whatsoever, as they only show 10% of the full picture.

[som99]

Have you tried the Hyperclock utility to overclock that sam44ep-flex?

Haven't tried yet (quite new on AOS hardware), got the 800MHz sam and I will give it a go and put some extra cooling on it Thanks.

[nicholas]

@Nicholas

Thanx dude!  You're a handy guy to have around!  I love timing tests!

Do you know much about those G4 and G5 machines?

I am thinking "Damn!  I want one of those MorphOS G5 machines!"

But then I am thinking "Wait, I don't want a blacksmith forge in my bedroom that can smelt iron.  I want something coooool so maybe I better get a 2.0Ghz G5 PowerMac instead of the 2.5Ghz version"

But then I am thinking "Wait, maybe they are different versions of the chip or different masks or whatever.  So for all I know the 2.5Ghz one runs cooler than the 2.0 Ghz version.  Like how my 50Mhz 060 is much cooler than my 25Mhz 040."

So do you know anything about their heat dissipation? (watts)


If you never had one or read up on them then I will "just assume" that I should aim for getting a lowly 2.0 Ghz model to keep the heat down.

Is there a list anywhere of TDP for both the CPU itself and the whole computer for all systems listed above?

I'm conservative so electrical efficiency is my prime criteria when selecting any type of computer and has been so since 1993.  I don't like wasting resources.  Using resources is ok but wasting them is not.   And it may seem weird but I also prefer to keep my money in MY bank account instead of the Electric Company's bank account.

I can't take the credit for the benchmarks, I pinched them from Piru.

The 2.7 GHz model runs at around 80 degrees C sat around doing nothing when fan cooled but runs a bit cooler when liquid cooled (the default is liquid cooled, fan cooling needs a mod).

They don't use that much electric when on OSX as it has more power saving features than MorphOS.

I recommend the PowerBook G4 1.67GHz machine with 128MB Radeon 9700 Pro myself. Runs cool and quiet and is very fast.

[Zac67]

Overclocking wastes giant blobs of electricity.  Electrical consumption does not increase linearly.  I think it increases exponentially.

Fortunately, it's not exponential or we wouldn't have 4 GHz machines today.

You can think of an electronic circuit as a capacitor and the clock as an AC source: the current is proportional to the voltage and to the frequency: I = C * U * F

With P = U * I you get P = U * C * U * F thus P ~ F:
20% higher clock, 20% more calculation power, 20% more electrical power

However, if you increase the voltage for better overclocking you'll notice that P increases with U * U: P ~ U²

And this is what starts burning electricity; a 10% increase in voltage uses 21% more power. Then you can possibly raise the clock by another 5% but you'll be using up 27% more power.

I experimented quite a bit with K6-II and -III CPUs and reality is extremely close to these formulas (C is highly constant with a given CPU).

[vox]

The fastest Amiga currently available is the PowerMac G5 2.7GHz Dual Processor machine running MorphOS 3.2.

TRue in raw CPU power. In disk, memory and graphics, X1000 is still first, SAM 460 second, then G4 CPUs (amongst OS4 systems).

[nicholas]

TRue in raw CPU power. In disk, memory and graphics, X1000 is still first, SAM 460 second, then G4 CPUs (amongst OS4 systems).

I feel a challenge ahead!

When you get your X1000 install a fresh Debian and run a benchmark of your choice.

I'll do the same with the G5 and compare.

[minator]

Benchmarking between systems isn't just a matter of running a program with the same parameters on two systems and timing the difference. You have to be certain that ALL other parameters are equal.

If you a attempting a scientific test of a microprocessor running a specific bit of code then the above might be valid, sometimes.  But the result will be completely meaningless because that's not what people buy.

In this case it is systems being compared with neither the same hardware or operating system.  The compilers versions supported are likely to be different and the same versions of the code might not compile.

It is not some small part of the system being compared but the entire thing.  It's how fast can system X produce a result compared to system Y.

Look at the industry standard benchmarks - SPEC and TCP.  They all run on wildly different systems, you are allowed to compile them however you want using whatever compiler you want on whatever OS / hardware combination you want.  There is no requirement other than you can actually buy the system.

Consider the Top500 supercomputer list.  It tests systems where *everything* is different.  You are not only allowed to modify the source code but you are expected to!

[toRus]

Benchmarks are mostly marketing rather than mathematics or computing science.

[ChaosLord]

When I do a benchmark it is entirely science.  I did the benchmark so I know it isn't rigged.  I am not trying to sell anybody anything so it isn't marketing.  I just report the results.


As to the benchmarks in this thread some of them were performed by random forum members.  Are you claiming they lied or cheated in some way?

I have known Piru for many years and he has always been a good scientific Vulcan.  I don't believe for one second that he would rig a timing test. He is like me, he puts the facts first, ahead of anything else.

[Duce]

Without a full breakdown on what hardware was in each machine (for example, same hard disks, same RAM speed/amounts, etc. in all machines?) it isn't worth much to me.  Then there's the common sense approach of the whole matter - I wouldn't be using *any* of these machines to do transcoding, not when a $300 PC can do it much, much faster (and likely more elegantly).

FYI, I use both MOS and OS4.

[Themamboman]

When I originally asked this, I was specifically looking for systems that run Amiga OS4.x.  I knew that MorphOS on fast Apple hardware would be mentioned.  I've run Morphos and it's good.  Amiga OS4.1 is the only next gen I haven't run yet due to the hardware costs.

I'm probably going to break down and get the Sam440ep-flex due to the fact that it's the lowest priced one out there (apart from used hardware which isn't easy to come by and still commands high prices years after being discontinued).  It won't be a speedster but it will do for testing and some code porting.

[magnetic]

TRue in raw CPU power. In disk, memory and graphics, X1000 is still first, SAM 460 second, then G4 CPUs (amongst OS4 systems).

Unfortunately raw cpu power doesnt do anything for you when you want to watch a video or play a game and you have no overlay...

[toRus]

@ChaosLord

I didn't accuse you of rigging any benchmarks. I said the cpu benchmarks are inherently biased and their most valid purpose is for promoting a product, an opinion, etc. Not every optimisation matters the same for every cpu architecture, not everything is robust enough to be used as a metric, and not everything is indicative of general-use real -life performance. You can have a set of benchmarks to suggest an approximate trend regarding performance as easy as you can have a newer/different version of the same benchmark that invalidates the previous findings.

[magnetic]

once again to debate the speed of the cpu is kind of laughable when you have an OS that is optimized and super fast on even slow processors. WIthout any real software to use this horsepower whats the point? And again, without proper video drivers quite useless as many sam people have discovered.

[spirantho]

If you a attempting a scientific test of a microprocessor running a specific bit of code then the above might be valid, sometimes.  But the result will be completely meaningless because that's not what people buy.

It's still meaningless. If you want I can benchmark a Core i7 and have it produce results slower than an Amiga 500.  That's why it's meaningless, you need to know all the facts and there has to be constraints.

If a benchmark isn't the same across machines, then it has to be transparent, and the differences must be documented.

It's kind of like running quake benchmarks. One person runs it at 1600x1200 with a software renderer, the other plays it with Quake and runs it at 320x200.

Which is the better computer? without all the facts, you can't say. Ergo the benchmark is useless.

It is not some small part of the system being compared but the entire thing.  It's how fast can system X produce a result compared to system Y.

Which only makes that benchmark valid for that operation using that software, and shouldn't be used in the computing equivalent of "my dad can beat up your Dad" arguments.  It tells you nothing because it doesn't tell you where the bottleneck is.
You could run the same task now and have a wildly different result, therefore you're testing the software (which can get changed) and applying it to the system. That is not good benchmarking! At the very least it should be made clear the EXACT system being benchmarked, the software used, the compiler used and the compilation options. For all we know he could have used numbers from a debug version!

Consider the Top500 supercomputer list.  It tests systems where *everything* is different.  You are not only allowed to modify the source code but you are expected to!

I suspect those will more tightly regulated. I severely doubt the  Top500 is worked out by trawling the forums for numbers and not even knowing what changes were made and what the systems were, which is how that graph came about