My Galax GTX970 runs at about 1.33 GHz @ 1.05 V and 1.34 GHz @ 1.06 V. I have set its power limit so that it hovers around 140 W. If I would let it use more power, it would obviously clock higher but work a bit less efficiently. Your card uses the same chip in a comparable environment (OC-capable board, solid cooling), so there's no need to specifically look for values of a Strix.

My GTX660Ti is currently also running with "extreme eco-tuning": 1.06 GHz @ 1.037 V at 102 W power target. I'm running both like this because otherwise my system would become too loud with both of them crunching.

MrS
____________
Scanning for our furry friends since Jan 2002

That's pretty normal for a stock factory-overclocked card. To get better performance simply raise the "GPU clock offset" in the usual OC-tool like EVGA precision or MSI Afterburner. You'll hit the power limit, at which point the GPU will reduce clocks & voltage (user a lower boost" state). It will run more efficient due to the lower voltage. I'd expect you to be able to run at about 1.4 GHz for the same power consumption via this method. Or at the same speed with less power, like I do.

You could also try to OC your cards memory. Notice how it's running at 1500 MHz in the sensors page, whereas it's listed as 1750 MHz on the 1st page. It's an issue with the power saving state nVidia chooses for you. I've talked about it over there in the last posts, but was not yet able to solve it. It seems normal for GTX970/980. You can manually OC the memory for GP-GPU using nVidia inspector (the other tools can't), but be warned: my card crashes quickly at 1750 MHz. I expect to be able to run something between 1550 and 1650 MHz if I can't fix this power state issue.

And finally: your card is reported to run at 16x PCIe 2, which would be strange since you're running it in a Haswell i5-4xxx system. The only 16x PCIe slot it should have must be capable of PCIe 3 speeds, as well as the GPU. So I don't know what's going wrong here. Have you switched the system since you made the screen shots?

MrS
____________
Scanning for our furry friends since Jan 2002

There's some variation even running similar WUs with the same config. With +60 MHz at maximum a difference of (1330+60)/1330 - 1 = 4.5% could be expected. Assuming that the resulting clock sped is indeed 60 MHz higher. If you're hitting the power limit it is not, as I described in the previous post.

The memory clock speed is generally not very important at GPU-Grid. But at 40-50% memory bus load, increased memory clocks (i.e. lower bus load) can result in a few % higher throughput.

Similarly for the PCIe bandwidth: GPU-Grid is not particularly susceptible to it, but especially for fast cards it can yield a few % improvement.

BTW: you recenlty got a computation error. Is this because you're trying to find the limits of your hardware? If so: I suggest to first use a testing tool other than GPU-Grid. Determine the rough limits of your hardware there, then you know what you can approximately expect running GPU-Grid. I used the Heaven benchmark this time, but am not convinced it is any better or worse than e.g. 3D Mark.

MrS
____________
Scanning for our furry friends since Jan 2002