Nvidia GeForce GTX 1000 Series (Pascal) MegaThread: FAQ and Resources

[mr91]

I had a friend with 2x 1070's Strix cards and one card was booting higher than the other. He found out that they're using different bios's.
After flashing to the new bios both cards were boosting to 2100 mhz in sli.

Do both of your cards have the same bios?

 

[azzazel_99]

Good question. They both run the same speeds but when in sli they don't clock as high as they do individually. I've read ALOT of people having the same problem out of pascal cards specifically

 

[mr91]

Maybe the VRM or the VRAM temperatures are higher when in SLI. I'm not sure why you can't get your max OC in SLI when your temps are good.

The guy I know with the 1070's is running both over 2100 in SLI after the bios update...

 

[azzazel_99]

I may look into a bios update. See if Asus put anything out

 

[mr91]

Cool, I informed the guy that was successful with the 1070 Overclocking and invited him to this tread.
Hopefully he will give you some tips on how to increase your SLI OC with the Pascal cards.

 

[mr91]

The guy said he copied the bios from the higher performing card and flashed the second 1070 strix card.
The guy also said forcing steady voltage was helpful to achieve max clocks with the pascal cards...

 

[azzazel_99]

How did he force voltage? Its locked on pascal cards........

 

[mr91]

Go to General -- Unlock voltage & then force constant voltage...

I didn't increase my voltage on my card however it's isn't lock on my founders card.


If unlock voltage doesn't work than you need to use GPU Tweak...

 

[renz496]

How did he force voltage? Its locked on pascal cards........

AFAIK it is more about forcing constant voltage value. it did not give free control over what voltage we want to apply. with kepler and above nvidia make the power saving feature even more aggressive.

 

[opio]

EVGA.

[video="https://www.youtube.com/watch?v=WAbl0fLY06U"][/video]

Come on EVGA, this is the kind of crap that ruins reputations.

Damn I glad that I have the FTW Hybrid and I don't have to worry about this

 

[opio]

The point is is that everything is getting smaller, and contrary to what vtarmiricmi says Moore's Law has not stagnated. It has slowed down sure, and companies like Intel and IBM have stated that they can get the lithography on their CPU's down to around 7nm-5nm before quantum tunneling becomes a problem, but technology will find a way.

Technological Singularity FTW!

Yes it is. So called Moore's Law (it actually isn't physics law, moreover it isn't any law) puts number of transistors and time into relationship. There are no more doubling of the CPUs number of transistors every two years. 1st gen. i7 quad (2010) had cca. 750 mil. of them, 6th gen Skylake i7 quad (2015) has cca. double of that. Even Intel's best CPU to date, 10core i7-6950x has a 3,4 bil. of transistors.
According to Moore's Law, starting from 750 mil. transistors in i7 2010., 6th gen Skylake should have 48 bil.
Even if we neglect Intel's own generations of CPUs and take in the equation only number of years (what states ML), 2010 750 mil., 2012. 1,5 bil., 2014. 3 bil., 2016 6 bil.

So what we can conclude here:
1. Moore's Law is not valid anymore,
2. Intel's so-called CPU generations are actually not generations because their gradual performance (=number of transistors) is very low (around 10% in each 'generation'), which says itself that CPU development has stalled, explaining that the term 'generations' is market-term only.

Now I am not into conspiracy theories at all, but somehow I think that Intel back in the 2010 could produce CPU that could have had a 50% performance advancement, that is, to go from the core2quad to the ivy bridge cpus, and from that to go to the skylake-like cpu. That mean they could actually have had a 2010 ivy-bridge performance level, and for instance, 2014, skylake performance line. Instead 2 real generations (by 'real' I mean, significant CPU performance advancement), they had 6 'generations'. I think that is bad, even from the marketing point-of-view, because customers are not stupid; they realize that new CPU generation does not bring revolutionary advance (even not evolutionary), and are skipping and postponing upgrade (from 1,2 to 5,6 gen for instance).

Completely another story is that even if the Moore's Law is right, that would not bring advancement into the whole PC industry, because software is not following. 20 years ago, I've thought that in the 2020, we would have have 'real' and functioning AI, that the voice recognition systems would be usual thing in the OSs, followed with the virtual reality and other stuff. But that is not happening, or very very slowly. So those issues cannot be solved with the 100-core CPU.

I'm aware that Moore's law is not a physical law.

When Intel was founded the number of transistors was doubling every year for a long time, then in the late 80's all the way to the early 2000's it went to around every 18 months, then it went to around every 2 years to where we are today. Sandy Bridge was released in 2011 with a 32nm process, 2013 saw Haswell at 22nm, 2014 saw Broadwell with 14nm. Then we have to take in to account the development of FinFet, integrated graphics and other technologies as well. Moores Law has definitely slowed down, but it hasn't stopped. Or maybe it would be more accurate to say progress hasn't stopped. And I agree, software still has a lot of catching up to do relative to hardware.

One thing that is still true when it comes to microprocessors is that about every 18 months the cost to produce any given chip is effectively cut in half. So Skylake was introduced in August of 2015. So by Q2 2017 the cost for Intel to produce say an i7 6700k will be around half of what it was when it was first released.

Sure, progress hasn't stopped but stalled As of Intel internal costs, why should consumers care about that, when retail prices are not cut in half

Agreed, about a year+ after any given architecture is released whether it's Intel, IBM, AMD or Qualcomm they could lower their prices and they don't. The point is is that in about 10 years, you're going to see what right now is sophisticated processors in EVERYTHING. Think of the best consumer processor you can buy right now. In 10-15 years it's going to be pennies to manufacture.

 

[mr91]

EVGA.

[video="https://www.youtube.com/watch?v=WAbl0fLY06U"][/video]

Come on EVGA, this is the kind of crap that ruins reputations.

Damn I glad that I have the FTW Hybrid and I don't have to worry about this

How does your card OC & is it quiet when overclocked?

 

[larsv8]

So I ended up with $600 in Dell gift cards and I want to pick up a 1080, however their selection seems very limited.

http://www.dell.com/en-us/search?q=1080&~ck=mn

I see a PNY 1080, but I can't seem to find it sold anywhere else, so I am skeptical. Anyone know about this card?

http://www.dell.com/en-us/shop/pny-geforce-gtx-1080-graphics-card-gf-gtx-1080-8-gb-gddr5x-pcie-3-0-x16-dvi-hdmi-3-x-displayport/apd/a9148624/graphic-video-cards#configurations

 

[azzazel_99]

Hmmm you can only purchase through dell huh? Bummer

 

[Math Geek]

the pny cards are just founder's edition reference cards. they sell a lot to oem's which is why dell has them for sale. nothing wrong with them other than the normal FE stuff. it'll run a touch warmer and a bit louder that a custom card do to the blower style cooling. other than that, no reason to not get it if you have to buy one from dell. good airflow in your case and it won't overheat or anything like that.

 

[azzazel_99]

Think pny offers lifetime warranty on their gpu's too

 

[opio]

EVGA.

[video="https://www.youtube.com/watch?v=WAbl0fLY06U"][/video]

Come on EVGA, this is the kind of crap that ruins reputations.

Damn I glad that I have the FTW Hybrid and I don't have to worry about this

How does your card OC & is it quiet when overclocked?

To answer your question in a short manner, it overclocks great and is extremely quiet. But I've also done a lot of things to my rig to ensure it's as quiet as possible. I hate loud fans. Detailed description on my experience with my EVGA FTW Hybrid below!

As far as the card goes in general at full load it maybe gets up to 45c (stock), I have the radiator in a rather unique push/pull configuration as well that keeps it about 3c cooler overall compared to just one fan on the radiator. I don't use the stock fan that came with it for the radiator though either, all my fans are Noctua fans. The one fan that cools the VRAM power on the other side of the PCB is REALLY quiet at 100% though. Just note that for some reason, the three pin coming out of the card won't power certain fans. It wouldn't power the static pressure fan I have in the pull position but it powers the airflow fan in the push position. I think it's cause the static pressure fan is a PWM fan though, either way no big deal. Also the pump is silent and the card has no coil whine. I did a lot of research between this GPU, the msi Seahawk, and the Gigabyte Waterforce and the EVGA FTW Hybrid I decided was the best, for how much it costs plus just how it cools the custom PCB (the Seahawk uses the reference design), down to how it comes clocked from the factory. I have no buyer's remorse. One thing to add about how it cools the card, is that the cold plate is flat (as opposed to the Seahawk, it's just using a Corsair H55, which do not have flat cold plates as H55's are optimized for CPU's not GPU's). Anyway, there is a huge copper piece that comes down that is fastened by the cold plate to cover and cool the VRAM chips instead of relying on air alone to cool them. I may even have gotten the Waterforce but it didn't offer anything that the FTW Hybrid didn't (aside from being completely watercooled) as the Waterforce is $40 more (the Hybrid also has a dual BIOS as well, the Waterforce doesn't).

So far I've been able to get my card up to +158mhz on the core clock and +382mhz on the memory with power limit set at 115% and the power allowance* set to 100% and run it through Time Spy, Firestrike 1080p, Furmark 4k (for 30 minutes) all stable. I know I can go higher, but at this point I really only overclock when benchmarking, I just don't need the extra horsepower for games, at least not yet. At these settings the GPU got about 3c hotter on average under load.

The weird thing is is that I got a lower graphics score by about 200 points on Time Spy with the overclock settings I mentioned above than when I ran through Time Spy with it clocked at +75mhz core and +200mhz memory with power limit and allowance set to 115% and 100% respectively. Anybody know why this might be? Lower clock = higher score does not compute.

P.S. I really wish precision X was as good as afterburner considering I have an EVGA card, and the fact EVGA stopped using rivatuner is beyond me.


* On the FTW Hybrid it doesn't let you unlock voltage control and control how many extra mV you want to put in to your GPU, instead the slider is marked "Core Voltage (%)" I was confused about what this slider was at first but then found out that when you set it at 100% you're allowing the card to pull up to 1.2V if it needs to, usually it sticks around 1.095V

 

[mr91]

I would raise the power limit to max when overclocking. On the founders card max is 120%.

An extra 158 seems like it's too much because you get an OC out of the box. I would stick with the most stable OC.
158 might be too much, I would try 100 or 125 however at the end of the day you won't get much performance out of 50 - 75 mhz.

 

[renz496]

So I ended up with $600 in Dell gift cards and I want to pick up a 1080, however their selection seems very limited.

http://www.dell.com/en-us/search?q=1080&~ck=mn

I see a PNY 1080, but I can't seem to find it sold anywhere else, so I am skeptical. Anyone know about this card?

http://www.dell.com/en-us/shop/pny-geforce-gtx-1080-graphics-card-gf-gtx-1080-8-gb-gddr5x-pcie-3-0-x16-dvi-hdmi-3-x-displayport/apd/a9148624/graphic-video-cards#configurations

should be fine. they probably less popular than some other brands but they usually sell nvidia professional line up of gpu. AFAIK almost all quadro that sold on the market was made by PNY.

 

[opio]

I would raise the power limit to max when overclocking. On the founders card max is 120%.

An extra 158 seems like it's too much because you get an OC out of the box. I would stick with the most stable OC.
158 might be too much, I would try 100 or 125 however at the end of the day you won't get much performance out of 50 - 75 mhz.

Ya on my card the power limit (not to be confused with power allowance) only goes to 115% and the temperature limit to 95c, the highest I ever see the power at is 98% anyway. I just want to know why I got lower graphics scores in some benchmarking programs when my card was clocked higher.

 

[mr91]

I would raise the power limit to max when overclocking. On the founders card max is 120%.

An extra 158 seems like it's too much because you get an OC out of the box. I would stick with the most stable OC.
158 might be too much, I would try 100 or 125 however at the end of the day you won't get much performance out of 50 - 75 mhz.

Ya on my card the power limit (not to be confused with power allowance) only goes to 115% and the temperature limit to 95c, the highest I ever see the power at is 98% anyway. I just want to know why I got lower graphics scores in some benchmarking programs when my card was clocked higher.

Can you add some voltage?

Basically if your overclock is too high it can negatively impact performance unless you find a way to tweak it.

 

[opio]

I would raise the power limit to max when overclocking. On the founders card max is 120%.

An extra 158 seems like it's too much because you get an OC out of the box. I would stick with the most stable OC.
158 might be too much, I would try 100 or 125 however at the end of the day you won't get much performance out of 50 - 75 mhz.

Ya on my card the power limit (not to be confused with power allowance) only goes to 115% and the temperature limit to 95c, the highest I ever see the power at is 98% anyway. I just want to know why I got lower graphics scores in some benchmarking programs when my card was clocked higher.

Can you add some voltage?

Basically if your overclock is too high it can negatively impact performance unless you find a way to tweak it.

I can flash overclock settings onto the BIOS, and I have a dual BIOS just in case, but I've never done that, would like to learn how though.

Ya all you can do is put power allowance at 100% and let the card do the rest .The card CAN pull up to 1.2v, but like I said, I never see it go above 1.095v

 

[Math Geek]

apparently Nvidia added some telemetry services to the newest GPU driver. as is normal with such things they are not saying what they are collecting or what it is for. so naturally we assume the worst and it's part of some sinister plot to steal your Nanna's secret family recipe for pumpkin spice stuffed eggplant fritters.

if you've read about it and/or are concerned about it, here is a simple way to stop the data collection. https://www.overclock3d.net/reviews/software/how_to_disable_nvidia_s_telemetry_process/1

basically, run autoruns program, uncheck the 3 telemetry services and then close autoruns program. now your secret family recipe is safe as is your not so secret fondness for that fat feet food squashing fetish video you've been into since you saw it on Tosh.o........... (don't worry your secret is safe with me)

 

[mr91]

I would raise the power limit to max when overclocking. On the founders card max is 120%.

An extra 158 seems like it's too much because you get an OC out of the box. I would stick with the most stable OC.
158 might be too much, I would try 100 or 125 however at the end of the day you won't get much performance out of 50 - 75 mhz.

Ya on my card the power limit (not to be confused with power allowance) only goes to 115% and the temperature limit to 95c, the highest I ever see the power at is 98% anyway. I just want to know why I got lower graphics scores in some benchmarking programs when my card was clocked higher.

Can you add some voltage?

Basically if your overclock is too high it can negatively impact performance unless you find a way to tweak it.

I can flash overclock settings onto the BIOS, and I have a dual BIOS just in case, but I've never done that, would like to learn how though.

Ya all you can do is put power allowance at 100% and let the card do the rest .The card CAN pull up to 1.2v, but like I said, I never see it go above 1.095v

I didn't get a chance to play with your card & I don't know any good BIOS's that are available. I suggest you start a thread and maybe somebody can help you. There is some risk involved with using aftermarket BIOS's...

 

[opio]

I would raise the power limit to max when overclocking. On the founders card max is 120%.

An extra 158 seems like it's too much because you get an OC out of the box. I would stick with the most stable OC.
158 might be too much, I would try 100 or 125 however at the end of the day you won't get much performance out of 50 - 75 mhz.

Ya on my card the power limit (not to be confused with power allowance) only goes to 115% and the temperature limit to 95c, the highest I ever see the power at is 98% anyway. I just want to know why I got lower graphics scores in some benchmarking programs when my card was clocked higher.

Can you add some voltage?

Basically if your overclock is too high it can negatively impact performance unless you find a way to tweak it.

I can flash overclock settings onto the BIOS, and I have a dual BIOS just in case, but I've never done that, would like to learn how though.

Ya all you can do is put power allowance at 100% and let the card do the rest .The card CAN pull up to 1.2v, but like I said, I never see it go above 1.095v

I didn't get a chance to play with your card & I don't know any good BIOS's that are available. I suggest you start a thread and maybe somebody can help you. There is some risk involved with using aftermarket BIOS's...

Ya I read that it is possible to brick your card doing this, luckily I have a dual BIOS just for that purpose haha. Like said before though, for right now I don't overclock it when gaming as it's stock overclock is more than enough, I'm sure I will when I finally save up enough to get a 1440p monitor, even though on a lot of games I run that with DSR anyway. It sounds like a good, stable BIOS for overclocking a card takes awhile after said card's launch date. One additional question I have is when you flash a BIOS on to your GPU do you use any software? Do I need to use integrated graphics while I do it?