• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

AMD is planning a revised Radeon HD 6900

IronRuler said:
My GTX 480 will not work with just two 6 pins. I have to have those other two pins in or it doesn't display signal to my monitor.
Click to expand...

Thats simply card diagnostics verifying that it is connected properly before going full bore on power draw. The gpu itself would be just fine drawing that amount of power from 6 pin connectors.
 
I think a few people are missing what im saying.....

Take a 6950....cut off the 6-pin connector....solder on an 8pin connector...see if it improves anything in any way....
 
People like to argue instead of think, that's all there is to it.

For the nth time, I ran a ridiculous number of watts through my GTX 470s through their two six-pin connectors. Probably almost twice what a 6970 draws on stock.
 
I never understood the point of the 8 pin connector, especially since the vast majority of PSU companies just do the extra 2 pin splice right at the end of the cable (as the pictured earlier). My TX750 was that way, and my XFX 850w Black Edition is that way.
But hey, what do I know?
 
A few suggestive questions that may help clear up the debate:

- Do 6-pin PCI-express lines have trip-points or active monitoring to prevent them from providing as much current as the 8-pin?
- Does that 12v sense-line and Extra Ground on the "true" 8-pin PCI-E connector allow the PSU to provide any more power? Or, in practice, is such 12v regulation easily performed at the PSU-end?
- Will any of this provide a limitation BEFORE the Video Card PCB's OCP?

- Is it even possible to thermally overload a pair of 6-pin PCI-E connectors with a single video card, considering that it required 3-way and 4-way SLI/X-fire setups for people to begin burning up the motherboard-PSU connector plugs? (Now some motherboards have provided parallel lines to prevent this, and there are mods to accomplish the same)

* I think that the 6-pin connectors can provide the same (or at least a non-limiting) amount of current, and that modern PSUs are capable of regulating w/o the 12v sense line (I do not have the answer, but that linked MSI Lightning GTX 275 on LN2 is overclocked like hell on 6-pin connectors... that's some solid evidence for at least the 1st half of this statement)

* But yes, some of the LN2/DICE guys (maybe at XS) may have the answers... I wonder if those guys wouldn't prefer to buy the 6970 from the start, however, because it may be better binned.
 
There's a better test.

But it requires someone here to have two different PSU's AND a PSU adapter:

But if anyone does, please try this with a *6970*.

1) Use the PSU that has two 6 pin connectors (strongly preferred) OR, if 6+2 pin connector draws the exact same amps as a regular 6 pin, then a 6+2 pin connector (not a real 8 pin connector), but I'm not sure about this, as 6+2 pin connectors may (or may not?) have the +12v sense and ground of a native 8 pin...and use a 6 to 8 pin adapter for the two 6 pin PSU (PC power and cooling has one cheaper than newegg), and test that and find the max overclock you can get without the 6970 VPU recovering.

2) Then, switch to a PSU that has a native 8 pin connector (unless a 6+2 pin is the same as an 8 pin) and retest to find the max overclock. Determine if the max overclock is the same with a native 8 pin connector, or with a 6 pin using a 6 to 8 pin adapter.

That will answer the questions once and for all.

Any takers?
 
Falkentyne said:
There's a better test.

But it requires someone here to have two different PSU's AND a PSU adapter:

But if anyone does, please try this with a *6970*.

1) Use the PSU that has two 6 pin connectors (strongly preferred) OR, if 6+2 pin connector draws the exact same amps as a regular 6 pin, then a 6+2 pin connector (not a real 8 pin connector), but I'm not sure about this, as 6+2 pin connectors may (or may not?) have the +12v sense and ground of a native 8 pin...and use a 6 to 8 pin adapter for the two 6 pin PSU (PC power and cooling has one cheaper than newegg), and test that and find the max overclock you can get without the 6970 VPU recovering.

2) Then, switch to a PSU that has a native 8 pin connector (unless a 6+2 pin is the same as an 8 pin) and retest to find the max overclock. Determine if the max overclock is the same with a native 8 pin connector, or with a 6 pin using a 6 to 8 pin adapter.

That will answer the questions once and for all.

Any takers?
Click to expand...

I would assume that the 6+2 power connectors are the same as a solid 8 pin since I know my 850w PSU is SLI/CF certified and it uses the 6+2 setup. Better yet, get someone to plug their 6970s in using 6 pin connectors only and see how stable it is, if the card will even boot. I suck horribly at overclocking so ill pass, lol.
 
USMCGrunt said:
I would assume that the 6+2 power connectors are the same as a solid 8 pin since I know my 850w PSU is SLI/CF certified and it uses the 6+2 setup. Better yet, get someone to plug their 6970s in using 6 pin connectors only and see how stable it is, if the card will even boot. I suck horribly at overclocking so ill pass, lol.
Click to expand...

Good simple easy reading imformation here. http://www.playtool.com/pages/psuconnectors/connectors.html This will take the quess work out of the numbers in the lines and rails.

8 Pin PCI Express Power Connector

The PCI Express 2.0 specification released in January 2007 added an 8 pin PCI Express power cable. It's just an 8 pin version of the 6 Pin PCI Express power cable. Both are primarily used to provide supplemental power to video cards. The older 6 pin version officially provides a maximum of 75 watts (although unofficially it can usually provide much more) whereas the new 8 pin version provides a maximum of 150 watts. It is very easy to confuse the 8 pin version with the very similar-looking EPS 8 pin 12 volt cable.

The 8 pin PCI Express and the EPS 8 pin 12 volt connectors are polarized differently so you won't be able to plug one kind of cable into the other kind of connector. That is, you won't be able to plug the wrong kind of cable in unless you try really hard. Unfortunately, the Molex Mini-fit Jr. connectors used by both kinds of power cables can sometimes be forced into a differently-polarized connector if they only have a few pins and you push hard enough. If the cable won't slide in easily then you're probably trying to insert the wrong kind of cable. The 8 pin PCI Express connector does have a small plastic bridge which prevents it from being plugged into an EPS 8 pin 12 volt motherboard connector. You can see the bridge in the image above between the rightmost two pins in the top row of the connector. But there's no such protection to prevent EPS 8 pin 12 volt cables from being plugged into an 8 pin PCI Express connector on a video card. That combination may fit if you shove hard enough. And if you plug in the wrong kind of cable then expect fireworks. Some of the grounds and 12 volts wires for an EPS 8 pin 12 volt are reversed compared to an 8 pin PCI Express. Fortunately, most 8 pin PCI Express connectors are labeled "PCI-E" so people won't confuse them with EPS 8 pin 12 volt cables. If the connectors aren't labeled then you can tell an 8 pin PCI Express power cable from an EPS 8 pin 12 volt cable by checking the color of the wires which plug into the clip side of the connector. On the EPS 8 pin cable, the yellow wires (the 12 volt wires) go into the clip side of the connector. On the 8 pin PCI Express cable, the wires on the clip side are all black (grounds). That's the same as it is with the 6 Pin PCI Express power cable. Of course, none of this helps you if your cable uses the trendy all-the-same-color-wires design which is popular with high-fashion power supplies. In that case you'll just have to be very careful or hope the connectors are labeled.
Pinout
Pins 1 through 3 Pins 4 through 6
Description Wire color Pin number Pin number Wire color Description
+12 volts yellow 1 5 black ground
+12 volts yellow 2 6 black ground
+12 volts yellow 3 7 black ground
ground black 4 8 black ground


Official cable/connector maximum wattage delivery
Voltage rail Number of lines Maximum current Maximum wattage
+12 volts 3 4.167 150 watts


6+2 Pin PCI Express Power Connector

Some video cards have 6 Pin PCI Express power connectors and others have 8 Pin PCI Express power connectors. Many power supplies come with a 6+2 PCI Express power cable which is compatible with both kinds of video cards. The 6+2 PCI Express power cable is made up of two pieces: a 6 pin piece, and a 2 pin piece. If you put the two pieces together then you have a full 8 pin PCI Express power cable. But if you split the connector into two parts then you can plug the 6 pin part into the older 6 pin PCI Express connector and leave the 2 pin part unplugged. That way, your power supply only needs to have one 6+2 cable to be compatible with both 6 pin and 8 pin PCI Express connectors.
 
Last edited:
NO, because those can probably unlock into 6970's. The new PCB revision will likely disable that ability.
 
So is anyone going to use a 6 pin PSU with a 6 to 8 pin adapter (from pc power and cooling or newegg) to test the max overclock when using an older 6 pin x2 PSU, compared to a normal 8 pin or 6+2 pcie connector?
 
Falkentyne said:
So is anyone going to use a 6 pin PSU with a 6 to 8 pin adapter (from pc power and cooling or newegg) to test the max overclock when using an older 6 pin x2 PSU, compared to a normal 8 pin or 6+2 pcie connector?
Click to expand...

Say What.:D
 
Ok I'll say it again.

Can someone test the max overclock, using an "older" PSU that has only 6 pin PCIE connectors, by using a 6 pin to 8 pin converter, and compare it to a newer PSU that actually has the the 8 pin (or 6+2 pin) connector on it?
 
Falkentyne said:
Can someone test the max overclock, using an "older" PSU that has only 6 pin PCIE connectors, by using a 6 pin to 8 pin converter, and compare it to a newer PSU that actually has the the 8 pin (or 6+2 pin) connector on it?
Click to expand...
How often are overclocks limited by power?
I was under the impression the roof of a GPU overclock is typically hit when the GPU itself can't muster the higher clocks; independent of power.

The question you should be asking is, how does PowerTune effect an overclocked card on both of those PSU's? The feature kicks in when the card is sucking too much power, but having the extra connector might make a difference. (?)
 
Utopia:

Someone above said that using less amps on a line can limit overclocking due to crashes, and I've also had two 6970's that both max out at 920 mhz game stable (ahem? the same speed on two cards?), therefore I was curious, since I'm using a 6 to 8 pin adapter as well :)
 
Falkentyne said:
Utopia:

Someone above said that using less amps on a line can limit overclocking due to crashes, and I've also had two 6970's that both max out at 920 mhz game stable (ahem? the same speed on two cards?), therefore I was curious, since I'm using a 6 to 8 pin adapter as well :)
Click to expand...

My 6950 does 890mhz on stock 6950 voltage of 1.1v and does 940mhz on 6970 voltage of 1.17v. I can't believe this thread has gone on this long, there is no appreciable difference in overclocking from a 6pin to an 8pin.
 
Well as I said, I had two identical 6970's that both did about 920 mhz.

And the RBE is out and most people are getting 1 ghz easy. However I increased vcore to 1.225 and I only gained 20 mhz (920 to 940; 945 requires about 70% fan). That sounds like a power problem right there. 0.05v for 20 mhz? Doesn't sound right does it?

It's already been proven that the 8 pin connector can draw more amps as the 6 pin. And I'm using a 6 pin to 8 pin adapter, which means it's only pulling out half the amps it could.

Pretty sure its a power problem, don't you think? I HIGHLY doubt I have two of the worst clocking 6970's on the planet...( I only have one card now).

I think I have a valid point. Sure, perhaps it belongs in a completely different thread. But seriously...how HARD is it for someone to TEST it? Is it really as bad as pulling teeth? All I asked is for someone to test a 6970 with a 6 pin and a 6 pin using a 6 to 8 pin adapter (with PSU that doesn't have native 8 pin support), and then compare it to a normal 8 (or 6+2) pin PSU with support for that. I shouldn't get flamed for asking for a simple test. I mean a LOT of people have spare systems and parts available....we are enthuiasts, you know...

Anyway, I asked XFX to send me a Molex to 8 pin adapter. (like the one sapphire has). Then I guess I'll have my answer...
 
And looks like I got my answer....from my support ticket:

"(snip) We do not recommend using any power adapters for eight pin connections. The draw through the plug is 150W which is too high to reliably do through any adapters. You should use a power supply which has a native eight pin connection on it. I will have a mDP to DP adapter sent to you. Thank you. Mark "

So you still claim the 6 and 8 pin plugs are drawing the same power?

XFX just claimed that the 8 pin is drawing 150W. And since I'm using a 6 to 8 pin adapter, the card is only getting 75W, which isn't enough for OC'ing very far. (this is why I wanted someone to test with an adapter). I think I'll go by their word. And could explain why my core isn't scaling very well.

I'll probably grab a dual molex to 8 pin, for testing, since that should give the full 150W....
I'll have the proper PSU for my sandy bridge build anyway...
 
Falkentyne said:
Well as I said, I had two identical 6970's that both did about 920 mhz.

And the RBE is out and most people are getting 1 ghz easy. However I increased vcore to 1.225 and I only gained 20 mhz (920 to 940; 945 requires about 70% fan). That sounds like a power problem right there. 0.05v for 20 mhz? Doesn't sound right does it?

It's already been proven that the 8 pin connector can draw more amps as the 6 pin. And I'm using a 6 pin to 8 pin adapter, which means it's only pulling out half the amps it could.

Pretty sure its a power problem, don't you think? I HIGHLY doubt I have two of the worst clocking 6970's on the planet...( I only have one card now).

I think I have a valid point. Sure, perhaps it belongs in a completely different thread. But seriously...how HARD is it for someone to TEST it? Is it really as bad as pulling teeth? All I asked is for someone to test a 6970 with a 6 pin and a 6 pin using a 6 to 8 pin adapter (with PSU that doesn't have native 8 pin support), and then compare it to a normal 8 (or 6+2) pin PSU with support for that. I shouldn't get flamed for asking for a simple test. I mean a LOT of people have spare systems and parts available....we are enthuiasts, you know...

Anyway, I asked XFX to send me a Molex to 8 pin adapter. (like the one sapphire has). Then I guess I'll have my answer...
Click to expand...
I applaud and support you in your endeavor. At least their are a few people willing to do their small part. Combine all the little pieces and the big picture forms.When you have an answer please pm me, I'm willing to solder the two pins if need be, but I'm not sure if all parts on the 6950 are identical to the 6970.
 
Falkentyne said:
And looks like I got my answer....from my support ticket:

"(snip) We do not recommend using any power adapters for eight pin connections. The draw through the plug is 150W which is too high to reliably do through any adapters. You should use a power supply which has a native eight pin connection on it. I will have a mDP to DP adapter sent to you. Thank you. Mark "

So you still claim the 6 and 8 pin plugs are drawing the same power?

XFX just claimed that the 8 pin is drawing 150W. And since I'm using a 6 to 8 pin adapter, the card is only getting 75W, which isn't enough for OC'ing very far. (this is why I wanted someone to test with an adapter). I think I'll go by their word. And could explain why my core isn't scaling very well.

I'll probably grab a dual molex to 8 pin, for testing, since that should give the full 150W....
I'll have the proper PSU for my sandy bridge build anyway...
Click to expand...
So much FUD....

The 6pin connectors are able to transfer at least 4x their rated power draw. If the 6 pins magically stop providing more power at the 75w level, then overclocked Fermi chips must have micro fusion reactors onboard to provide the extra juice (although I guess that would explain the heat...). Look at the furmark power draw Anandtech saw with a moderately overclocked GTX 470 (790mhz at 1.087v). http://www.anandtech.com/show/3836/msis-geforce-n470gtx-gtx-470-sli/3

At load that system card is drawing 552W with idle power consumption being 173W with the 470 (look on the power consumption page). That puts the delta at 379W plus at LEAST 20W for the idle power draw and you get 399W, not too shabby for a power delivery configuration only rated to provide 250W. We know the cpu wasn't a significant portion of that because that system with a 5750 in it only increased by 66W total at load.

But now consider all the guys running this same 470 design with modded BIOS'es and watercooling at 1.2V and 850mhz+. Anand saw power consumption increase 107W with a .062V bump in voltage and a 115mhz bump in clocks and these guys are running another .113v above even Anand's OC'd card. I am not going to try to extrapolate the power draw, but these GTX 470's are clearly capable of pulling 400-500W with just a pair of 6 pin connectors and that extra juice sure as hell isn't coming from the motherboard.

Cliff notes: the 6 pin theory is bullshit.;)
 
Last edited:
Probably in your case, the GTX 470 was already exceeding the PCI-E spec, but that also means that your PSU manufacturer isn't following the PCI-E 6 Pin spec exactly allowing higher current to flow from the power connector.
 
evolucion8 said:
Probably in your case, the GTX 470 was already exceeding the PCI-E spec, but that also means that your PSU manufacturer isn't following the PCI-E 6 Pin spec exactly allowing higher current to flow from the power connector.
Click to expand...

I haven't heard of any quality PSU's that haven't allowed GTX 470 overclocking, even the overclocked 5970's are very likely over the 375w limit provided by two 8 pin connectors and the pci-e slot. As has been stated many times, the extra pins are only grounds and really don't have an impact if the ground wires that are included in the 6pin are handled well.

The pci-e power spec is just that, its a limit that the manufacturers have to stay under to be compliant and has absolutely no meaning on the actual capabilities of the connector. Based on the cards we've seen ship with clock speed and voltage honed in right at that limit with a nod and wink, it seems everyone is well aware of this except for a few people posting on here who treat the spec limits like a fundamental law of nature.
 
that's because the extra 2 pins route power to part of car that requires it to boot. you could plug in a 6pin to 8pin adapter and the card would still work.
 
BTT:

Any news yet as to when the so-called revision will hhit the markets (as i understand it it´s just being released the way it was meant to in the first place : TI drMOS > Volterra) and how i can/will be identified as the new revision ?
 
R-Type said:
We know the cpu wasn't a significant portion of that because that system with a 5750 in it only increased by 66W total at load.
Click to expand...

Wouldn't having a higher frame rate increase cpu load and thus, power consumption?
 
Sottilde said:
Wouldn't having a higher frame rate increase cpu load and thus, power consumption?
Click to expand...

This is true, but worst case could only account for maybe an additional 50W and certainly not enough to bring the delta down into the realm of the 6-pin spec limitations.
 
You must log in or register to reply here.
Back
Top