GTX 680 - Huge Variance

[Jseams]

I uploaded my results a couple of days ago - scored a 4700+ on the 3D video benchmark using an OC'd i5 (4.5Ghz), 16gb of Corsair Veng and a reference GTX 680 from EVGA... this score was consistant over several runs.

My system, while above average - is nothing really extraordinary when it comes to high end systems.

I've noticed the scores for the GTX 680 have dropped significantly today and when I checked by downloading the stats for those systems running that card I was appalled to notice some of the scores...

How badly can folks have their systems configured, drivers installed, etc to post such subpar scores? I find it troubling that a system with an i7 and a GTX 680 is reporting scores in the 2K range... how is this possible?... and more importantly, if you DID manage to blow your system up to the point that it was performing so poorly, why would you then upload those scores, knowing that something is fishy with them?

My system, if you check the baselines, is listed as the Alienware Aurora R3s (the s was so I could identify it later...)

[David (PassMark)]

There many ways you can stuff up your PC.

Even with the CPU itself, which doesn't really require any configuration or device drivers people still manage to have problems.

Check out the graph here,
http://www.passmark.com/forum/showthread.php?t=3502
Showing the spread of CPU results for a single i5 model.

This post also details the many ways you can end up with sub-par performance,
http://www.passmark.com/forum/showthread.php?t=1057

Many people don't know any better when they upload their results. But it kind of balances out with the enthusiasts. We don't have many results for the 680 GTX yet, so one bad or good result can pull the average around a bit. This effect disappears once we have collected dozens of results for any particular model.

You also need to take into account that a slow CPU can reduce your 3D result. It isn't totally dependent on the video card.

[Jseams]

Thanks for the reply.

... I guess I can understand some of the low benchmarks - but I would think that most people who are interested in this kind of thing would know what a "normal" score is and would recognize a severly subpar run and adjust their systems before uploading a wonky test.

btw - made some adjustments and posted a 5100+ 3d graphics score - significantly higher than the 4K average... so I'm not sure when us "enthusiasts" will start to balance some of these 2-3K puzzles that keep being uploaded.

[David (PassMark)]

The 680 GTX has been back at the top of the chart for some days now.

[Jseams]

The 680 GTX has been back at the top of the chart for some days now.

While true, the scores so far are averaging around 4K - which seems a bit off to me. I posted a 4.7 with an OC'd i5 and with an OC'd i7 easily breached the 5K mark. There is only one driver version available from Nvidia atm - and all the systems so far have been i5-i7's.

I will admit that currently the sample size is very low. I'm interested to see how things look in a few weeks.

Regards,

[BillM.Green]

You also need to take into account that a slow CPU can reduce your 3D result. It isn't totally dependent on the video card.

Yes, CPU can make a huge push. My Radeon 7950 @ 900/1250 scores a mark of 4600 points which is much higher than the average 7950 score. The CPU is a i7-3770k @ 4.5 Ghz.