overclocking gtx 1070 hybrid ftw

culzone

Honorable
Apr 27, 2016
555
1
10,995
So I installed my gtx 1070 hybrid ftw and saw wow 2000mhz out of the box without changing a single thing.
But then... anything above 2050mhz crashes in fire strike..
Just unlucky or is something wrong?
 
Solution
The hybrid isn't bringing anything to the table. Other than the FE cards, no 1070 card is throttling. So making the GPU cooler can in no way affect performance. When a card's performance is impacted starting at 82C, then choosing a hybrid cooler which delivers 52C has no advantage over a cooler that delivers 68C. The issue with the SC and FTW lines is that every one that shipped before November had no thermal pads on the VRM. This has always been a weakness on the SC line but with 10xx it spread tot he FTW.

Perhaps you saw that the SC and FTW designs w/o thermal pads were reaching VRM temps of 14C and some even were catching fire... EVGA has a fix kit that you can get if you are affected. I have not seen anything on how this relates to the hybrid FTW, but it is something I'd look into

https://www.techpowerup.com/227133/evga-gtx-1070-1080-overheating-issues-company-says-thermal-pads-a-solution

If any 1070s non FE 1070s are going to be limited by temps, it's going to be the VRM temp.
 
Graphics card vendors bin their chips and use the better ones in factory overclocked versions
that they can sell for more. You may be able to OC more, but then again, perhaps not.
They try to differentiate by adding cosmetics and fancy coolers which also sell for more.

The FTW is already binned and overclocked. Be happy with what you have.

If you need more performance, register your card with evga and use their 90 day free upgrade option to trade it in for a GTX1080.
 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360


This is categorically untrue. With the Pascal architecture, the cooler the GPU runs, the more efficient it is, and the better it will perform. Which means, at stock voltages and core at 80c, the GPU max clock could be around 2100, but with stock voltage and the core at 25c, the max clock will be 2200+. These "steps" happen every 10c or so. I've seen clock / voltage adjustments made by the GPU with core temps as low as 24c.

In order to get really good core clocks with Pascal, you need to use the voltage / frequency curve to overclock, and not just offsets...AND....keep the GPU as cool as possible. The cooler, the better. Keeping the GPU cool can make for some AWESOME overclocks, even at stock voltage.

Start with something simple on the frequency / voltage curve. Like this:

cHLlI5C.png


 

culzone

Honorable
Apr 27, 2016
555
1
10,995

I did overclock it with the curve, 2063mhz crashes within 2min fire strike :p with maxed out voltage and power limit.
Did only +65 on the offset

 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360


What did the curve look like? Can you post a screenshot of it? They're very particular about how they're setup, and the voltage/clocks/temps they're running at.
 

culzone

Honorable
Apr 27, 2016
555
1
10,995

Is there something I should change to get a better overclock? http://imgur.com/a/S67MF
My gpu gets max 47C during unigine heaven 30min
 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360


Everything after your desired clock and voltage, should be flat. Like the one I posted. The picture you shared, looks like you were trying to hit 2075 @ 1093mv, but there was a jump after that spot...also the two tick marks for 1075mv and 1081mv are too close to your desired clock. Drop them a bit...down to like 2030 or so.

You'll also want to make sure that the voltage slider on the main page is at 100%..or it won't matter what you set in the slider, the voltage will max out at 1050mv, and run at the clock that corresponds with that voltage on the voltage frequency curve.

If you're having trouble getting things started, you can always try a really aggressive curve line to the clocks / voltages you're looking to run. Like this:

Og1PpjK.png


 
Solution

culzone

Honorable
Apr 27, 2016
555
1
10,995

why should everything after my desired voltage be flat when it can't go any higher than 1.093v?
btw will try this when I get home
UPDATE!!!! IT WORKED!!! 2114mhz now holy SSSh. THANK YOU! I just had to put all of the other ones that was over 1.093v to the same clock and it worked!
 


And yet not a single hybrid card has ever shown a higher fps than an quality air cooled AIB card .... meanwhile some have shown worse due to higher VRM temps limiting the OC. Sitting here typing from a custom water cooled system with 5 x 140mm worth of radiator, twin pumps, 6 liquid temp sensors, optical temp sensor and twin water cooled GFX cards.

Running Furmark full tilt, GPU temps are 39C. By lowering pump speed and fan speed I have tested in approximately 5C increments up to 70C ... if what you say is true, how do you explain why performance remains exactly the same end to end ?

This is a simple concept. Throttling kicks in at 82C ... Look at techpowerup reviews. There is no throttling ... ever ... unless temps exceed 82C.

 

culzone

Honorable
Apr 27, 2016
555
1
10,995

How come I throttle at 42C 50C and 55C then? 13mhz less each time

 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360


Thermal throttling starts at 82c, yes. But boost 3.0 will raise / lower voltages and frequencies every 10c or so to keep peak efficiency. At stock clocks, anything below 55c or so, you probably won't notice a difference. Start pushing up over 2150, and you will.

Pascal is all about temps, and those "thermal layers" that boost 3.0 will use to adjust voltage and frequency. On air, both of my GPUs would run at 2202 @ 1.093v with a really steep curve, but...they didn't run like 2202 should be running. Under water, and an normal ambients, they run 2214 @ 1.093v and do it very well. If I lower the ambient temps, they run 2240 @ 1.093v and do it very well. On air, they wouldn't have touched those clocks because the temps were too high....I've got runs done on a single card at 2278 @ 1.093v with coolant temps at 1c and the core temp never going above 22c.

Simple concept? Yes, it is quite simple, and you obviously don't get it. There's a TON of good info in this thread....I suggest you read it.

http://www.overclock.net/t/1601288/official-nvidia-gtx-1080-owners-club

You can believe whatever you want, but......I've seen it. I've done it. Been playing this game for a while, and I'm telling you....you want higher clocks, lower the temps, because you'll never see good scores at higher clocks until you do.

sATNfJu.png
 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360


I've never seen an NVIDIA GPU with a VRM sensor....you can add a sensor. XSPC and a few others make some that you can put next to the VRM, but they're not going to be completely accurate.

 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360




Congrats, man. I'm glad to see you got it figured out. Just keep those temps in check, and run the voltage as low as possible.
 
My VRMs are running at 42c right now, GPUs (nVidia) at 44C per HWiNFO ... Radiator fans are spinning at only 550 rpm. When I have looked at this, I see pretty close to what Guru3D reports with their thermal imaging, tho I haven't compared extensively.

Sensor is CHL8318/IRF ASP1212

Some have gotten it to work with AB ... some not

http://www.overclock.net/t/1398725/unlock-afterburner-limits-on-lots-of-cards-some-with-llc-one-tool-for-all-ab-versions/1190

More here:

http://www.overclock.net/t/1522651/disable-boost-and-bake-in-max-game-stable-clocks-for-maxwell/370
 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360


Interesting. Never seen a VRM sensor on an NVIDIA GPU. Which GPU are you using?
 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360
I'm not seeing anything about VRM temp sensors in either thread....unlocking voltage control in AB, seems to be the topic in the AB thread, and the other is bios modding.

I found some info on it, but it looks like it's very limited with NVIDIA GPUs. I've always used EVGA....can't say I've ever seen a VRM temp sensor on any of their GPUs, even the Classys.
 

TRENDING THREADS