Benchmarks: what flavour?

[.Wilier.]

Tell me good people of Freddy's, what benchmark results do you prefer to see?

Would you rather see synthetic results, such as 3dMark and Aquamark, semi-synthetic results, like a time-demo, or recording played back in FRAPS or something, or real-world results gained from playing a complete level of a certain game with each different card (or monitor or CPU or whatever is being reviewed)

Also, do you think its a good idea to only see results from the fastest most expensive machine currently available? or do you think there is something to be gained by seeing results from a more......enthusiast system, like a P4 2.8 or an XP3000, something a little older than the very latest?

Or a combination of all?

Or none of the above?

Just wondering is all.

 

[Jonty]

Hey Wilier

Depending on time constraints, perhaps a mixture of tests may be good; 3DMark03/05, a custom time demo and maybe a quick blast yourself like HardOCP. If I had to plump for just one, I'd probably go for the custom time demo, but that's just instinct, nothing very scientific about it

As for the system, having a standard rig to test on would be cool as it would provide a consistent benchmark (or if possible, a high-end, mainstream and budget rig ). You could then do like FiringSquad, time permitting, and analyse Intel chips or AMD chips, an see how they compare to the standard setup, or nVidia cards or ATI cards etc. I don't thnk it matters too much what the system is so long as it's kept roughly the same throughout the various tests (nothing too slow as to create bottlenecks, performance differences with various settings in benchmarks are often more noticable the faster the system).

Kind Regards

 

[smurkin]

I prefer 3d marks for system tuning. But for system/hardware comparisons, I like games. Mainly because I mostly have them, and they are easier for me to relate to (plus its more useful information).

Top end system benchies are ok, but tbh, its often slightly meaningless due to some bottleneck or other.

Pray tell, wht the intrerest (*intrigue*)

 

[.Wilier.]

I have to say that I agree with you Jonts, although practically speaking, it must be very time consuming to do benchies on several different machines, using a suite of say 6 different applications.

Personally, I sway toward the custom time-demo and some gameing.....although its nice to what 3dmark can do, since so many sites use it.

I guess at the end of the day, its purely a comparison that you need. If X card does ***fps in Y machine, and Z card only does ***fps, then there's a clear winner....in a certain application with a certain CPU etc......

The interest smurkin is purely that, Im interested

 

[Danya]

Anything in a game is good with me - a number from a game is generally far more meaningful than whatever new trickery futuremark have decided to implement that no card on the market can even run properly.

 

[Jonty]

True, most things struggle on 3DMark05 That said, two SLI, watercooled, 512MB Gainward 6800 Ultras should fit the bill nicely, although I shudder to think how much a setup like that would cost.



*Waits for Alienware to unveil somekind of a 1GB system with an outrageous price tag to match*

Kind Regards

 

[Clown]

Danya already has that
I'm only running a Barton 2500+ and a 6800GT. I get 13007 in 3DMarks 2003. HAY GUYZ is this good????

 

[PLightstar]

I tend to prefer Games, but its always trickey as all machines are different, and even if two machines have the same hardware set-up with the same device drivers, the make-up of the software will change things, do u run tests with Security Software on or off do u use XP flashy with the shadowed windows and other stuff or basic windows with no xtras, or do u have Desktop X which affects preformance. The list could go on but Games are prolly the best way to tell as most benchmark programs tand to crash on my machines and games don't. and I have a nice machine
AMD 64-FX
1Gb Crucial 3200 Paired RAM
200GB Sata drive
GF 6800 GT with Cooler
Audigy 2
550W PSU.

 

[Clown]

I don't tend to benchmark really. I don't even look at game figures. If it 'feels' alright then I'm happy with that. Go go 60FPS in Quake2.

 

[Jonty]

Hi Clown

That's a good score for that kind of setup. That said, the current record for 3DMark03 is is 26,236, but naturally that's with a very special system

Kind Regards

 

[Clown]

Something like Danya's

 

[Danya]

I have 2 watercooled 6800GTs in SLi, and 3dmark 05 still struggles in many places. I don't think there is a system that can run the whole benchmark smoothly. HL2 and Doom3 are smooth as silk though.

 

[.Wilier.]

I have 2 watercooled 6800GTs in SLi, ....and far too much money which I really should donate to Wilier

....

 

[Danya]

Nice try. Can I donate you my credit card bill instead?