Looks like it's hitting a thermal limit (84°) and throttling back to 1140MHz boost, though oddly the fan never goes above 45%/2200RPM. The really bizarre thing is the first run was done after a clean boot, but I just did a run with Palemoon and Chrome both still open (with a few hundred open tabs each) and the minimum framerate had risen to 27.
try afterburner or any other gpu tweak tool and put the fan on manual @ about 60% and re test. Seen this happen on a lot of Nvidia cards where the fan profiles are just plain wrong, ie. card set to throttle @ 80C and the fan profile is set to "max @ 90C" / "50% @ 80C" / "30% @ 50C" & "off at anything under 50C". So the fan never spins up enough to cool and stop throtteling, because you never reach the steep part of the fan curve. This is because the browsers hold a load on the GPU in between benchmark scenes and so the GPU does not pull back as much (boost less/lower clocks) but it should effect your max as it is still an extra load.
Some benchmarks you need to run twice in succession - Heaven is one of them, and so is Catzilla. I always let Heaven run through a full cycle before hitting the F9 key to benchmark it.
I love Catzilla for the music and graphics but it really isn't very reliable. I remember benching various rigs with it and the results were a bit wild to say the least.
I hope it's reliable if they're charging $899 for the full version...! The only times I've run into problems with Catzilla have been when I've actidentally left conflicting settings in Nvidia control panel, like Vsync and Phsyx, etc. As long as it's run twice, I can't fault it. I've run Catzilla (Tiger) on a number of systems and the results are very consistent, and that's just the Beta version of the software. In fact, Catzilla is my go-to benchmark when I upgrade graphics cards, because I've been using it for so long. Speaking of, I should upload my Catzilla results to the OP. I'll do that later today.
Well. The great nVidia benchmark of 2016 has hit an impasse because I'm really unwell... Sorry guys - bear with me and I'll get it underway soon!
Well I just installed it (I bought it too, mostly for the cool music) and it doesn't support Crossfire. Some benchmark, then. I looked into it and apparently 1.4 does not support Crossfire in Windows 8 or later. When I ran it with my Titan Black in SLI I scored around 15,000 running 1080p. When I run it on one Fury X I score about the same. It's a shame because it's definitely one of the better looking benchmarks out there.
I can't really comment on Xfire/SLI compatibility as it worked fine in the beta version; it did require a change of video settings though (alternate frame rendering iirc). Here are my Tiger results from as far back as Dec 2012, all done on the same Beta version of the software and with (essentially) the same system: Gigabyte GTX480 OC.................4790 EVGA GTX670 WTF....................8593 EVGA GTX680 OC......................9373 Gigabyte GTX770 OC.................10012 KFA2 GTX770 OC2....................10467 EVGA GTX780Ti Classified..........14750 EVGA GTX780Ti Classidied OC....15856 @jinksy, not to worry. We can still look forward to the results another time!
Thanks for running the tests Reg216uk. For Firestrike, make sure to report the GPU score rather than the overall 3DMark score, as the latter is a result of combined system performance.
Finally got round to doing / finishing mine. In both cases the CPU is a i7-4770k @ 4.4GHz and the 780 / 780s are reported as running at 1355MHz if you believe the Heaven benchmark or 1175MHz if you believe Precision X (no idea) Single Card: SLI I'm impressed at just how linearly the benchmarks scale. Also nice to know that not ditching the 780s in favour of a 980ti was the right decision to make.
My contribution to the cause - A single Zotac GTX 970 AMP! Extreme Core with a 6600k @ 4.6GHz = 3DMARK FIRE STRIKE 1.1 score of 11044 And Unigine Heaven Benchmark 4.0 - It's fair to say this particular example of a 970 goes like sh*t off a proverbial shovel
I'll try to remember running my 780 tonight, I'll be interested to see how it fairs considering it's paired with a relatively weak cpu (i5 3450 I think).