Benchmarking

Hardware: the vital organs of any computer system, the technical genius, the painstaking agony of design, the years of development and not to mention the countless millions of pounds that companies invests in it. Its shit because the other company can make it run Quake 3 Arena at 4 FPS more at 1024x768 in 32bit colour at 100hz with 8xFSAA and 4x Ansitropic filtering. Why did they bother to release such a 'POS' piece of hardware?

Hardware performance is something that has been talked about since the dawn of time and more then likey will go on beyond the end of it too. There seems to be more focus on which is faster than the technical achievements of the item in the first place. The whole for and against argument can and more than likely will continue for years to come, for some reason people don't care about what they need, they care about which is faster.

So why do we need a guide to Hardware Benchmarking when clearly its all about which is faster? surely we just plonk on this months flavour of First Person Shooter (which happens to be PainKiller which is awesome btw ;P) and let two combatants duke it out, jot down the results and harp on for half an hour on how card X compares to card Y. Lets face it, those damn 3 or 4 FPS make all the difference when the system is already churning out 150FPS, yup those extra few make all the difference!

Hell no, what would be the point of that? you can find more fanboy hardware sites on the internet then you could possibly read in one lifetime. Each one debating the 4 FPS margins back and forwards until the next hardware release comes out. This is an int13h guide after all...

Stuff, we need some stuff to do our guide.

Lets see, take a somewhat current PC system :

A 2003..
Intel Pentium 4 - 2.53ghz Cpu
512meg of DDR Ram
Nvidia - Geforce 4 TI4600
Creative Labs - Sound Blaster Audigy 2
Running Windows erm.. thing.. Service pack 14.
Patches:
KB234982a
KB234894a
KB234844b
KB234243a
KB234565c
KB234782a
KB234823a KB234823b KB234823c KB234823f
Nvidia Drivers : Forceware 3.4 Detonator 4.54.7443 The ones that worked.
Sound Blaster Drivers : oh fuck it I'm bored now.

Ok, next we have a old classic :

A 1996..
486 DX4 100mhz Cpu
16meg of SIMM Ram
VESA Local bus graphics (1meg!)
Creative Labs - Sound Blaster 16
Running FreeBSD 4.9-RELEASE (Various Kernel)
XFree86 3.6.2
Patches : yeah probably some when it built the software

and for a laugh lets stick in :

A 1993..
Silicon Graphics ONYX Supercomputer
4x 195mhz MIPS R10000 Cpus
2 Gig of Ram
4x Reality Engine 2 (RM5 graphics pipes)
No Sound =(
Running IRIX 6.5.22m
Patches... hang on.. bugger, cant find any.

That's a nice range of random hardware, now we need some fair and uncompromising software so that we can benchmark these beasties. It's got to be something recent, not more then a few years old. Its got to have full access to the sourcecode (NO not because we are some freak linux hippies!) so we can compile the software across our 3 platforms and to make sure there isn't any funny business going on in the code that would favour one system or the other, well we will take that last bit for granted as we aren't going to spend months looking at someone's messy code to find out where all the :HACK lines are.

What to choose.. hmm.. Ah MAngband. This will suit our purposes just fine.

We will begin by compiling the software by hand on to each of our platforms, this may take a while so why don't you go off and make a coffee, or watch a film, or maybe liberate a small country.

Still here? great, the software has been compiled and has passed the 'That's not an executable file' test. The test environment is ready and its time to get to work. First we will start off with a small benchmark often overlooked by benchmarking websites.

Does it work?

P4

Yay go Nvidia + Intel!

486

Looking good!

Onyx

Awesome!

All 3 machines have loaded the software fine, there was some degree of loading time differences but we didn't think to note them down and quite frankly I can't be arsed to go through it all again to find out. Bring on the next test!

Image Quality

P4

As you can see the Nvidia Cards MipMapping has caused quite a mess when we zoom in so we are going to have to poopoo nvidia and ask DEMAND that they update their drivers now or we will get our readers to boycott all their cards.

486

The BSD system put up a good all round performance but again when we zoomed in we found out something quite scary, the BSD drivers did not support the zoom tool very well and well that was that.

Onyx

The expensive Onyx Looked system looked great but managed to microwave the cat that was sleeping on top of the system during the tests. A worthy note is that the picture looked quite a bit darker then the other screen shots, this is simply due to the fact that all the lights in the house dim while the system is on.

The interesting thing to note here is that none of the tests used anything more then 2d ascii graphics so we couldn't run any of that FSAA nonsense. So this is raw.. erm.. graphics capability showing through. We enlarged a section of the screen for some reason, looks a bit pixely if you ask me. So bare in mind that you should NOT sit with your face pressed against the monitor nor should you use a magnifying glass when you are playing games. Naturally FSAA would be a great help in either of those two situations.

Frame Rate

Whoa its time for the biggy. The one you are all waiting for!



As you can see the [Website's fanboy product] is blazing along and is just leagues ahead of the competition. Of course there is a small chance there could be a margin of error in these tests but with the framerate differences you can still clearly see the winner.

But of course who in their right mind would run their brand new hardware at stock speeds? We have clocked the 486 from its stock speed of 100mhz upto 347mhz using a Liquid Nitrogen home kit. We had doubts about the stability of the system and to be honest there was some worry that the 486 might not survive the test, which it didn't, but we managed to get a score off it before the state of the art system buckled and burnt out.



Hmm, not much change there apart from the smell, though wasn't it exciting? Maybe we should make a keyring out of the cpu and give it away in some kind of compo to increase the hits to our sponsors adverts? Nah.

UPDATE: The drop in frame rate on the P4 turned out to be us moving the mouse during the benchmark.

Interesting test

Seeing as it's the current trend to test all applications to make sure there are no 'optimisations' buried deep in the drivers we tried renaming the mangclient.exe executable on the P4 system to mungclient.exe and re-run some image quality tests. The results speak for themselves.



Looks like the graphics card drivers force some form of ascii filtering when it detects mangclient.exe running. Shocking.

The Results.

Now you have a good set of results its time to upload them to your 'EXTREME' Hardware website (which if you are serious about becoming a hardware website webmaster you will have already registered and posted at least 2 months worth of news posts mostly aimed at begging for hardware and paypal donations). Get those sponsors in now because the fanboys will be piling in by the bucket load! Remember that after you have established your site as being a 'hardcore' hardware review site you can jump ship to other companies and constantly bitch about how crap the previous company is, you have enough loyal fanboys posting in your forum that you can radically change your opinions without anyone pointing out the massive lack of integrity you have.

The Verdict. The Slant. The Rant.

As you can see having the insane framerate levels for the application we are running makes one hell of a difference, I would definitely favour playing this game on a system that can run the game 4 times the level of framerate I need rather then looking at other aspects of the system/platform/environment. There is always a trade off with performance. One system may do better at one thing but dreadful at another, for example the P4 running windows might have the advantage of running a multiplatform application faster but the downside is going to be the constant patching of the host operating system's flaws and the constant worry of overheating and failing components due to the poor build quality.

The grass isn't necessarily greener on the other side either, running a 11 year old $500,000 supercomputer is neither a cheap nor quiet task but the balance of performance, reliability and uptime can outweigh other factors.

The point of this Guide is not to argue for/against which is better/worse. Its about the fact that in the pursuit of better performance people are missing the point of the vast majority of this technology, capability should outweigh performance everytime. Anyone can make a benchmark and jump on a bandwagon, look at the amount of websites that jump from manufacturer to manufacturer everytime a new card ousts the current performance leaders card by a insignificant margin and declares war against the old manufacturer. How can you trust a site to give a fair review of an item when they are just making sure they are in the right fanclub? Goodbye credibillity.

In closing, you might have noticed that the guide has somewhat slipped into a big rant towards the end but this is merely because its very hard to keep the fanboy thing going for too long, I guess that's why its rare to see any kind of review on int13h.com.

We love a good rant every now and again.

Fanboyism is a condition, we are still looking for a cure!

0 Comments



Heavy Engine Console
--------------------------------------------------------------------------------
Loading Page... /894-Benchmarking
Console: