GTX 970 Owners Club

[Gofspar]

UPDATE: I will still hit the power limit and start throttling after and 25mins on 3dmark Ultra at 2160p so I intend to lower voltage again as much as possible to reduce power draw while still maintaining the max table clock of 1493mhz

 

 

SIDE NOTE: anyone who has modded their own bios will know that the stock bios has a max table boost clock of around 1500mhz give or take but that the default boost clock is whatever the specs for the card say (in my case 1304mhz) the reason the cards never go much beyond that on their own is due to the stock voltage table (note the voltage at the bottom of the voltage table is higher than your card is allowed to go in my case this was 1.281V) By adjusting any clock state voltages in the table that are higher than the max voltage at the top of table to match that of the top of the table (in my case reducing any clock state voltages that were higher than 1.262 down to 1.262) and by setting the default boost clock in the main table to match the max boost table clock you will allow your cards to reach the maximum boost clock in the boost table possibly with the need for a moderate increase in TDP to remove any throttling.

 

LINK TO THREAD:

http://www.overclock.net/t/1553510/gtx-970-gtx-980-why-bios-modding-is-mandatory-for-most-cards-if-you-want-a-stable-overclock/10

 

Also as I have an effective liquid cooling loop with EK full cover blocks my cards never go over 50C even when benchmarking at 4k it is safe to say that this over-current protection issue that causes the cards to shutdown cannot be over-come simply by reducing temperatures.

 

There are however hard mods that can overcome this but I am not brave enough to try!

yeah the palit cards are meh. I tried to make a decent bios for someone else and I was hitting OCP protection.

[Jaqen]

yeah the palit cards are meh. I tried to make a decent bios for someone else and I was hitting OCP protection.

Asus Strix and some MSI 970 owners are having similar problems, interestingly it looks like Gigabyte must have gone the 'more power, more voltage with the 8-pin +6-pin and the more robust VRM's) for achieving more stable boosts where as others like ASUS went for lower voltage and TDP (Strix has a max of 180W at 106% and is limited to 1.212V) with more throttling to avoid noisy fans and poor efficiency. 

[Gofspar]

Asus Strix and some MSI 970 owners are having similar problems, interestingly it looks like Gigabyte must have gone the 'more power, more voltage with the 8-pin +6-pin and the more robust VRM's) for achieving more stable boosts where as others like ASUS went for lower voltage and TDP (Strix has a max of 180W at 106% and is limited to 1.212V) with more throttling to avoid noisy fans and poor efficiency. 

Yeah, FTW+ cards can do extra voltage and TDP too. sad that manufactors had to do that.

[Jaqen]

On a side note it is also true for many versions of the 970 that even if MSI AB and other OC utilities allow you to increase voltage in the software it often DOESN'T DO ANYTHING

 

Also software OC'ing these cards without modding BIOS voltage table or atleast confirming that you can indeed increase voltage in AB is very ill advised.

 

I will not go into explaining why as many of the people modding their own BIOS will already know why but if you don't and want an explanation follow the link in my above post

[Jaqen]

Yeah, FTW+ cards can do extra voltage and TDP too. sad that manufactors had to do that.

I didn't choose the Palit cards myself, knowing what I do now they would not be my choice but if you are interested I will post my findings with regards to much lower voltages and moderate TDP increases. My initial findings are that you can achieve the same or higher stable clocks with less than 1.26V due to reduced power consumption.

 

I am hoping to be able to help those of us with inferior VRM's and no 8-pins (or single 8-pin i.e. Asus Strixx) to be able to achieve those higher clocks.

[Gofspar]

I didn't choose the Palit cards myself, knowing what I do now they would not be my choice but if you are interested I will post my findings with regards to much lower voltages and moderate TDP increases. My initial findings are that you can achieve the same or higher stable clocks with less than 1.26V due to reduced power consumption.

It's not due to reduced power consumption its the way Maxwell scales with voltage. With Kepler you could throw 1.4v at it on air and it'll scale. after 1.262v you wont experience ANY scaling or little at all on Maxwell. With Kepler you could make up for having a shit card with voltage, with Maxwell you want high ASIC because less leaky cards will scale farther with less voltage.

[Jaqen]

It's not due to reduced power consumption its the way Maxwell scales with voltage. With Kepler you could throw 1.4v at it on air and it'll scale. after 1.262v you wont experience ANY scaling or little at all on Maxwell. With Kepler you could make up for having a shit card with voltage, with Maxwell you want high ASIC because less leaky cards will scale farther with less voltage.

interesting you mention this... one of my cards is 89% and the other is 68% this actually holds me back a lot under water! with just one card enabled i got a stable 1623mhz

[Gofspar]

interesting you mention this... one of my cards is 89% and the other is 68% this actually holds me back a lot under water! with just one card enabled i got a stable 1623mhz

Yep, Maxwell is a pain in the ass.

 

I have a 980 Classified now and honestly regret not just getting two 780 Ti KPE's and a 1080mm radiator from Performance PC's and making an external loop.

[rjfaber91]

Wait I saw you in the LTT video, holy shit you lucky motherfucker.

 

Looks so damn sexy, wow!

 

Thanks! What video though, I'm not aware of any?

 

 

IMO a white backplate would look better.

 

On its own, it probably would, yes. The paintjob on the graphics card is part of a bigger (re)build log however, where blue, white and black form the colour scheme.

[Gofspar]

Thanks! What video though, I'm not aware of any?

 

 

 

On its own, it probably would, yes. The paintjob on the graphics card is part of a bigger (re)build log however, where blue, white and black form the colour scheme.

Did you not get that from Linus's meet?

 

He had a G1 Gaming painted EXACTLY the same.

[rjfaber91]

Did you not get that from Linus's meet?

 

He had a G1 Gaming painted EXACTLY the same.

 

No, I didn't. I painted mine myself. Sort of spoils the fun to know that there is one that looks exactly like it, of course...

[Gofspar]

No, I didn't. I painted mine myself. Sort of spoils the fun to know that there is one that looks exactly like it, of course...

I think the one in the LMG HQ meetup video had a blue front and a white backplate, so don't feel too down.

[RyGuy99]

On a side note it is also true for many versions of the 970 that even if MSI AB and other OC utilities allow you to increase voltage in the software it often DOESN'T DO ANYTHING

Also software OC'ing these cards without modding BIOS voltage table or atleast confirming that you can indeed increase voltage in AB is very ill advised.

I will not go into explaining why as many of the people modding their own BIOS will already know why but if you don't and want an explanation follow the link in my above post

So you are saying that overclocking in Afterburner on its own with no bios alterations is a bad thing?

[arniks8]

So you are saying that overclocking in Afterburner on its own with no bios alterations is a bad thing?

No He's saying that changing voltage in afterburner often is pointless.

[Jaqen]

So you are saying that overclocking in Afterburner on its own with no bios alterations is a bad thing?

Only for some cards, the way to tell is if in GPU-Z or AB you can visibly see the voltage going up when you +mv in AB.

 

If you can't then you shouldn't be overclocking from software like Afterburner, the reason being that when you OC in AB you are OC'ing the BIOS's entire boost table 

 

for example

 

"CLK 59 1.050mV-1.268mV" 1304mhz

"CLK 50 968.8mV-1112.5mV" 1190mhz

 

after 50mhz OC in AB

 

"CLK 59 1.050mV-1.268mV" 1354mhz

"CLK 50 968.8mV-1112.5mV" 1240mhz

 

In a nutshell to get a stable constant boost clock over the default boost clock of a GPU that lacks the function to add voltage externally in AB you have to edit the voltage table in the bios to reflect the new OC'ed boost table

[Lethal Seraph]

Oh yes, it is, but I hope you don't mind me if I find my one-of-a-kind custom-painted 970 G1 Gaming to be about as FTW as it gets.

 

 

 

 

I almost thought it was an hall of fame card. GJ!

[RyGuy99]

Only for some cards, the way to tell is if in GPU-Z or AB you can visibly see the voltage going up when you +mv in AB.

 

If you can't then you shouldn't be overclocking from software like Afterburner, the reason being that when you OC in AB you are OC'ing the BIOS's entire boost table 

 

for example

 

"CLK 59 1.050mV-1.268mV" 1304mhz

"CLK 50 968.8mV-1112.5mV" 1190mhz

 

after 50mhz OC in AB

 

"CLK 59 1.050mV-1.268mV" 1354mhz

"CLK 50 968.8mV-1112.5mV" 1240mhz

 

In a nutshell to get a stable constant boost clock over the default boost clock of a GPU that lacks the function to add voltage externally in AB you have to edit the voltage table in the bios to reflect the new OC'ed boost table

The VDDC correct? I added voltage in afterburner and say an increase in the VDDC when I did on my G1 Gaming. I've had a stable 1580 mhz on the core and 2000mhz on memory without ever adding voltage. Those were confirmed through GPU-Z by the way not just Afterburner. So am I good?

[Jaqen]

You should be fine. g1 gaming cards as far as I understand have pretty high default voltages anyway. However the point to pay attention to is potential instability at lower clock speeds.

When your card is at 1580mhz you are likely to be in clock 63 close to that. When you do something less intensive your card may use clock 50 or 40 and with lower clock state comes less voltage but you have the added software OC which means there is potential for instability. Less so on your brand of card though.

Stock max voltage on my jetstreams are 1.262 but on an asus strixx 970 for example it is 1.212. Significantly less.

[Jaqen]

Just an update. Last night I was able to hit 1514mhz boost clock totally stable at 1.212mV

Relative power consumption dropped by over 30W compared to the same clock speeds at 1.26mV which is back down almost to the cards default power limit.

I can now achieve a stable 1514mhz at the same power consumption as the stock bios at 1304mhz.

[Jaqen]

This shows that ideally the best way to begin overclocking a maxwell gpu is to first find the lowest voltage your card can run its Max boost table clock at. This will give more TDP head room to work with aswell as lower temps and voltage head room when moving forward.

[arniks8]

This shows that ideally the best way to begin overclocking a maxwell gpu is to first find the lowest voltage your card can run its Max boost table clock at. This will give more TDP head room to work with aswell as lower temps and voltage head room when moving forward.

You can always do this ( http://overclocking.guide/increase-the-nvidia-power-limit-all-cards/ ) and forget about TDP problem.

[RyGuy99]

After playing with voltages yesterday I think I got a stable overclock of 1603mhz with +20 voltage. Not really sure if its worth it though performance wise.

[SkilledRebuilds]

After playing with voltages yesterday I think I got a stable overclock of 1603mhz with +20 voltage. Not really sure if its worth it though performance wise.

Do tests before and after applying it.

If the difference means something to you, keep it.

If your not fussed (or getting AMPLE FPS already) then yeah, no need.

[arvark]

I've got a Gainward Phantom GTX 970, runs a bit hot but I love it!

[arniks8]

I've got a Gainward Phantom GTX 970, runs a bit hot but I love it!

Attach a fan to it's heatsink .