Stalker CS - CS base mid 30's high 40's
- burnt farm mid 50's mid 80's
Vantage results were taken in the first few seconds of each of the GPU tests, read by eye.
The power is not as big an issue as is being made out (20W extra at 4 hours per day 7 days per week, 52 weeks/year its only about £5.00 annually (15p/unit)), it is louder but only when gaming, possibly quieter at idle, but then the 8800GTX didn't idle by design, the gtx470 does clock down with dual screens (a fuller explanation than some reviews) see table below.
Full dual screen idle single screen idle
Core 1674 1674 135
Shader 1215 810 101
Memory 607 405 51
The low single screen clocks are not a typo, it really does clock that low, I thought the dual screen clocks were ok.
Idle temps drop accordingly, down from 74C/54C (gpu/pcb) to less than 45C/38C. I hope they get the drivers sorted, although then mention in the documentation for the latest drivers that this is unlikley to be fixed, 2-d profiles may have to be created in that case.
Overall i'm a little bit meh! about it, partly because my favourite games are poorly matched to nvidia, i'm not a fanboi but i'd had a GF2,3,4,6600GT,7900GTO,8800GTX and now this... but i'm not too displeased, my 8800 lasted for nearly 30 months and I was still happy near the end, I just got the bug with the 58X0 releases and started to hanker after an upgrade, and having seen teh 5850 and 5870 prices climb 10-20% and not come down after 6 months i thought i'd preorder knowing i'd be ok for 6 months (well thats how i convinced myself)
Unfortunatley the latest drivers don't seem to support the 8800gtx, so its a ballsache to swap them out so I doubt i'll be doing more comparisons.
I'm not equipped to do full benchmarking (and there are 100's out there anyway), but I am able to tell you how it performs from a useability live-with-it point of view compared to a previous monster, I'll try and do a few more, but i get a bit focused on my gaming. BF2 for instance I can tell you caps at 100fps on full everything with about 20-40%gpu utilisation, compared to 80-90's at unknown utilisation.
An additional discovery, when the monitors turn off after n minutes, then the clocks do indeed drop to 135/101/51, regardless of the number of monitors, so teh low power states are not completely knackered.