Bjorn3D.com Bjorn3d.com – Satisfying Your Daily Tech Cravings Since 1996
Thermals, Noise, & Power
Now we have seen some sites using Furmark or other applications that intentionally drive heat up, which we feel is an unfair method of determining the cooling capacity of a video card. You can get almost any card to max out and throttle using heat driven applications and they just don’t represent real life readings. We ran 3DMark Ultra looping and MSI Kombuster and took the highest reading from the two which was Kombuster using Lakes of Titan.
At idle we got 39°C, which isn’t bad for a card packing 7.2 billion transistors, but we expected a bit lower from a Pascal GPU.
We were pleasantly surprised when the max reading we got after more than 10 minutes on Kombuster was a chilly 58°C. For the nay sayers, here’s a shot of the median reading we got from three runs of Lakes of Titan.
As you can see, we registered the 58°C well after the 10 minute run with a GPU load of 99%.
For noise testing we measured one meter from the open testing station level with the edge of the motherboard and at an idle we got couldn’t register the noise on our equipment which reads at 30dBA and above. The load noise registered (fans auto) 32.6dBA and we had no issues with the low noise operation of the ASUS ROG Strix GTX 1070. Kicking it up to max fan speeds when overclocking, we did get 47dBA, but that’s just an extreme case of manually setting the fans on maximum. At no point did we hear the fans kick to maximum during gaming or benching.
We took total power consumption from the wall using a KillAWatt power reading device and we read each test three times and averaged the readings.
At idle we got 77W with the card installed and since the ASUS Maximus VIII and the 6700K has integrated graphics we pulled the video card and read the voltages without a card installed and we got 62W fluctuating to 64W, so at an idle the card is adding an acceptable 15W-17W. With the 16nm FinFet shrink we expected lower load voltages and we got a reading of 322W which is a total system consumption increase of 245W. Figure the OC to 4.5GHz on the CPU running full-out load we would be close to the Max TDP of 150W.
Why not pish the memory? This card benefits well from memory OC over 9ghz
1) I think its the GTX1080 that has Two power (8+6) connects, whist it is the GTX1070 that has the single 8-pin. Not the other way around (as the File image shows).
2) It seems that the only thing that makes these faster is the clock speeds due to the smaller (16nm) FinFet stranding; because the 1070 has less Stream processors, less Rasterisers, less Tesselation units and less texture units than the stock GTX980 and it also has less Transistors than the 980ti…. So it should be cheaper….
3) Is it really worth the update from a gtx980, if it just cooks quicker?
I`m concerned, the 1070 held back…. and a Ti version will be released with as many, or more:- Stream processors, Rasterisers, Tesselation Units, and Texture Units than the 980Ti.
Hope so…
Darn it… that means I will wait a month or two longer…. Misfiring head.
I have only ever seen ONE 8-Pin power connect on the Asus GTX1070 Strix… and yet many sites report it has Two, as this site claims.
The photos only show One… so, which is it? Is there a `Hidden (secretive) 2nd Power connect`?
Funnily enough, this site also claims the GTX1080 Strix has only ONE 8-pin power connect…. So, it seems the faulty Graph is being shared internationally, without any check or edit.