[Update] AMD Released Radeon RX 480 Driver Fix with 3% faster performance And/or Reduced power consumption

[Trixanity]

33 minutes ago, typographie said:

snip

It wouldn't on its own. The solution with an 8-pin would have been the same as with the 6-pin: route more power through the external connector. Since they can both easily do 150W it's a non-issue.

[M4st4M1nd]

21 hours ago, Windows7ge said:

I remember seeing a video online where someone was running tests and his rig would shutoff durring bench marking testing the 480. Kinda shows that AMD doesn't bother to test their cards with older hardware to see how the card might affect a broader scope of equipment.

What do you expect them to do exactly? Test every motherboard known to man? That's just silly.

[typographie]

23 minutes ago, Trixanity said:

It wouldn't on its own. The solution with an 8-pin would have been the same as with the 6-pin: route more power through the external connector. Since they can both easily do 150W it's a non-issue.

I don't know how far out of spec the 6-pin connector can "easily" go, but it's actually rated for 75 Watts. The RX 480 was already drawing 80–100 Watts through the 6-pin cable in the original pre-driver coverage. I'm aware most PSUs can tolerate somewhat more, but the proper solution would've been an 8-pin connector (which is actually rated for 150 Watts) so AMD doesn't have to hope the average user's hardware can run far enough out of spec to power their card.

[Sintezza]

22 hours ago, ace_cheaply said:

The problem is that three of the vrm traces go to the pci slot and 3 went to the cables, so it was drawing evenly.  If AMD had ran four traces to the 6 pin, or hell, ALL of them since the 6 pin was overbuilt and can easily handle it, it would have never been a problem.

 

Well thats a statement covered with a pcb picture that i read allot about.

However thats already the second or third change they made to the picture of how the vrm is build up on the card.

But i higly doubt that they have any clue what they exaly talk about wenn it comes to vrm cirquitry uphere.

What i can tell you is that there are 2 sepperate vrm cirquitries on the board most likely, one for the VCCGT and one for the Vram.

The Vram is most likely powered by the peg, atleast that is what i´m assuming.

Because if i remember correctly AMD said something in their statement that the power overdraw had also something to do with the memory.

Which would support this.

 

However how the 6 phases on the VCCGT are build up and routed, that is something i would like to take a closer look at.

I have a clear idea of how this vrm might be build up.

But to be sure i need to know which particular pwm controller they used, and which particular power stages.

 

[Trixanity]

49 minutes ago, typographie said:

I don't know how far out of spec the 6-pin connector can "easily" go, but it's actually rated for 75 Watts. The RX 480 was already drawing 80–100 Watts through the 6-pin cable in the original pre-driver coverage. I'm aware most PSUs can tolerate somewhat more, but the proper solution would've been an 8-pin connector (which is actually rated for 150 Watts) so AMD doesn't have to hope the average user's hardware can run far enough out of spec to power their card.

The 6 and 8 pin cables are the same. The 8-pin is an added ground and sense pin. That's two pins. One 12V is usually not utilized on a 6-pin but is on an 8-pin. AMD's 6-pin connector threw the thing completely out of spec. It's 3 12V pins just like an 8-pin and the rest is ground like it was an 8-pin. They've actually pretty much removed the sense pins from the equation so that means the two sense pins you would find used on a regular 8-pin is gone. So it's 3 12V and 3 ground. It's functionally like an 8-pin without sense pins. It can do 150W easily like an 8-pin can because it's built like one. AMD has built it out of spec but by doing so made a formidable connector that can handle more than a regular one could. The cable itself is fine - it's the same one regardless of what you plug it in to. It can also handle it. So technically this 6-pin is rated for 150W since it has the same pins as you would use to run 150W through it. If I recall correctly, some German site said the 480W would be able to 200W safely over the 6-pin connector on it but I can't remember the details.

[Sintezza]

2 hours ago, Trixanity said:

The 6 and 8 pin cables are the same. The 8-pin is an added ground and sense pin. That's two pins. One 12V is usually not utilized on a 6-pin but is on an 8-pin. AMD's 6-pin connector threw the thing completely out of spec. It's 3 12V pins just like an 8-pin and the rest is ground like it was an 8-pin. They've actually pretty much removed the sense pins from the equation so that means the two sense pins you would find used on a regular 8-pin is gone. So it's 3 12V and 3 ground. It's functionally like an 8-pin without sense pins. It can do 150W easily like an 8-pin can because it's built like one. AMD has built it out of spec but by doing so made a formidable connector that can handle more than a regular one could. The cable itself is fine - it's the same one regardless of what you plug it in to. It can also handle it. So technically this 6-pin is rated for 150W since it has the same pins as you would use to run 150W through it. If I recall correctly, some German site said the 480W would be able to 200W safely over the 6-pin connector on it but I can't remember the details.

 

pci-e 8 pin vs pci-e 6 pin diffrence is 1 additional gnd wire, and a senseB pin.

Technically the 6 pin and 8 pin could provide the same amount of power.

But this does depends on the psu.

[ace_cheaply]

21 minutes ago, Sintezza said:

 

Well thats a statement covered with a pcb picture that i read allot about.

However thats already the second or third change they made to the picture of how the vrm is build up on the card.

But i higly doubt that they have any clue what they exaly talk about wenn it comes to vrm cirquitry uphere.

What i can tell you is that there are 2 sepperate vrm cirquitries on the board most likely, one for the VCCGT and one for the Vram.

The Vram is most likely powered by the peg, atleast that is what i´m assuming.

 

However how the 6 phases on the VCCGT are build up that is something i would like to take a closer look at.

I have a clear idea of how this vrm might be build up.

But to be sure i need to know which particular pwm controller they used, and which particular power stages.

 

 

https://youtu.be/plC7tOYIqBw?t=3243

 

This is where i got that from, fyi.  Sorry no embed, I just can't ever get that to work for some reason. 

 

2 minutes ago, Sintezza said:

 

PCI-e pin vs pci-e 6 pin diffrence is 2 additional gnd wires.

And the 6pin on the 480 is actually over spec. 

[typographie]

6 minutes ago, Trixanity said:

snip

I'm not arguing that the 6-pin connector can't handle this wattage the vast majority of the time. I'll take your word for it. But I think the fact that AMD themselves are delivering a fix that includes a "compatibility mode" that theoretically ensures both the slot and the cable are run within spec is an admission that AMD believes this has the potential to be a problem, at least for a minority of users. I'd be interested to see which mode AMD chooses to be the default in the new driver.

[RagnarokDel]

1 hour ago, typographie said:

I don't know how far out of spec the 6-pin connector can "easily" go, but it's actually rated for 75 Watts. The RX 480 was already drawing 80–100 Watts through the 6-pin cable in the original pre-driver coverage. I'm aware most PSUs can tolerate somewhat more, but the proper solution would've been an 8-pin connector (which is actually rated for 150 Watts) so AMD doesn't have to hope the average user's hardware can run far enough out of spec to power their card.

there's 3 pins that can deliver power over 12V at x amp for a maximum of either 60 or 75 watts per pin if I remember well. the 6 pin has 3 of those, so does the 8 pin, it just also has 2 more grounds.

[Briggsy]

11 minutes ago, RagnarokDel said:

there's 3 pins that can deliver power over 12V at x amp for a maximum of either 60 or 75 watts per pin if I remember well. the 6 pin has 3 of those, so does the 8 pin, it just also has 2 more grounds.

well yeah. ground wires are useless. I ripped the ground pin out of my PC's power cable because the wall socket's third hole doesn't provide any extra power, and therefore because I don't know anything about electrical engineering, I am assuming the ground is completely useless.

 

/s

[Khvarrioiren]

6 minutes ago, RagnarokDel said:

has 2 more grounds.

A ground and a sense. PIN8 is ground, PIN4 is sense.

[Trixanity]

59 minutes ago, Sintezza said:

 

pci-e 8 pin vs pci-e 6 pin diffrence is 2 additional gnd wires.

Technically the 6 pin and 8 pin are pretty much identical.

Only the additional 2 gnd wires are there for that they could handle the increased current better.

But with a decent psu, both the 6 and 8 pin could provide a similar current load.

6-pin power connector (75 W)[14]			8-pin power connector (150 W)[15][16][17]
Pin	Description		Pin	Description
1	+12 V		1	+12 V
2	not connected (usually +12 V as well)		2	+12 V
3	+12 V		3	+12 V
			4	Sense1 (8-pin connected)
4	Ground		5	Ground
5	Sense		6	Sense0 (6-pin or 8-pin connected)
6	Ground		7	Ground
			8	Ground

 

Unless my understanding of electrical engineering is completely off, the sense pins don't ground anything but are signal wires with no load. So that's not two extra ground on an 8-pin but one. 6-pin has one sense pin and 8-pin has an additional one to differentiate the cables. One 12V is not active on 6-pin but is on an 8. If AMD went out of spec and removed the signal wires, we're looking at 2 pins removed. So 3 12V and 3 ground - just like a 6-pin. Of course, you can always provide me with the knowledge that the sense pin double as ground but it doesn't seem to be the case. The videos I've seen of a multimeter on the pins on the out of spec RX 480 and the comparison with an in spec card suggests that the sense pins are NOT ground and that the RX 480 has no sense pins. Then again wouldn't three grounds for three 12V make sense for an 8-pin and two grounds for two 12V make sense on an 6-pin? Seems to me to be fairly logical. But of course, if anyone has a decent knowledge of the spec and of electrical circuits and of PCBs, then I'll be happy to be corrected if I'm wrong here.

27 minutes ago, typographie said:

I'm not arguing that the 6-pin connector can't handle this wattage the vast majority of the time. I'll take your word for it. But I think the fact that AMD themselves are delivering a fix that includes a "compatibility mode" that theoretically ensures both the slot and the cable are run within spec is an admission that AMD believes this has the potential to be a problem, at least for a minority of users. I'd be interested to see which mode AMD chooses to be the default in the new driver.

I think the compatibility mode is there to please the crowd that wants it to be completely kosher. It seems like they only provided any fix to appease everyone but in reality doesn't give a shit about the spec as long as the wiring does its job. Although what they probably forgot in their haste was that people would use the card on budget/old/poor motherboards which could definitely be an issue and that reviewers would found out that it broke spec causing outrage and I guess rightfully so. It could be a danger to some boards. And it was silly of AMD to not lean on the 6-pin instead. It can handle it much better than the motherboard can.

 

Oh and they already stated that the compatibility mode is completely optional. The default will be the rerouting of the power distribution to favor the 6-pin. So that's default and will always be enabled no matter what, so what would happen with the compatibility mode is that the 6-pin connector will be limited too. They do it like that because they'd rather not lower the performance even if it's only by a small amount, so they give you the option to do so instead. By doing it like this, the performance will still match what reviewers came out with sans the detailed power consumption benchmarks which would have to be done again. Both for users' sake so they know this fix works and for AMD's sake, so it doesn't haunt them going forward when people look up reviews and decide against buying the card because they believe the card will destroy their PCIE slot.

[Deletist_Jerk]

No one should be surprised that AMD launched a product with premature driver. They've done that since 6000 series.

 

On the other hand, I'm glad they are quick to fix a problem. 

[Sintezza]

1 hour ago, Trixanity said:

6-pin power connector (75 W)[14]			8-pin power connector (150 W)[15][16][17]
Pin	Description		Pin	Description
1	+12 V		1	+12 V
2	not connected (usually +12 V as well)		2	+12 V
3	+12 V		3	+12 V
			4	Sense1 (8-pin connected)
4	Ground		5	Ground
5	Sense		6	Sense0 (6-pin or 8-pin connected)
6	Ground		7	Ground
			8	Ground

 

Unless my understanding of electrical engineering is completely off, the sense pins don't ground anything but are signal wires with no load. So that's not two extra ground on an 8-pin but one. 6-pin has one sense pin and 8-pin has an additional one to differentiate the cables.

Eh sorry yeah thats correct.

On the 8 pin plug the additional pin 4 is senseB.

And pin 8 is ground.

But both those pins are wired black.

On most decent modern psu´s they do provide 12V to the second pin of the 6 pin connector anyways.

Thats most likely the reasons why sense B is wired black.

On older psu´s the second 12V could not be connected on the 6 pin.

Like you said above, but  supposingly on most psu´s today they are.

[RagnarokDel]

1 hour ago, Briggsy said:

well yeah. ground wires are useless. I ripped the ground pin out of my PC's power cable because the wall socket's third ground doesn't provide any extra power, and therefore because I don't know anything about electrical engineering, I am assuming the ground is completely useless.

 

/s

it still has a ground.

[RagnarokDel]

1 hour ago, Khvarrioiren said:

A ground and a sense. PIN8 is ground, PIN4 is sense.

Fair enough, regardless those 2 pins do not deliver additional power

[Sintezza]

16 minutes ago, RagnarokDel said:

Fair enough, regardless those 2 pins do not deliver additional power

Nope not realy indeed. it just depends on which psu you have.

Most modern psu´s do provide 12v to the second pin of the 6 pin connector.

Thats why they basicly can provide the same amount of power.

If you have an older psu doe, it could be that the second 12V pin on the 6 pin is not connected.

But i dont think that we are still seeing this today.

Maybe on realy cheapo psu´s.

 

 

 

[HalGameGuru]

2 things. 1 on an 8 pin the spec ALLOWS for, another, sense, most connectors do not use sense, they are JUST grounds. Many PSU's simply reuse the existing grounds to fill the 2 extra pins

 

2, we are conflating terms here, the 6-pin is SPEC'd at 75 watts, not rated. The standard requires 75 watts be doable as per spec, rated would be what individual iterations can ACTUALLY push. And even conservatively you could consider them "rated" for 150. And more accurately can probably push well over 200 safely.

 

EDIT: On further reading the "sense" may well simply be ground wires that complete a circuit and let the PSU/GPU know what kind of connector combination is in play. So many are not set up that way I wonder if they just allow unregulated access to the 12v rail, but the sense wires are apparently simpler than I had originally understood and grounds may perform this function just fine

[Castdeath97]

Is it out yet? 

[xAcid9]

Guys if you want to discuss about power connector go open a new thread in PSU subforum.

[zMeul]

the driver is out: https://support.amd.com/en-us/kb-articles/Pages/Radeon-Software-Crimson-Edition-16.7.1-Release-Notes.aspx

Tom's.DE did some testing: http://www.tomshardware.de/amd-radeon-rx-480-polaris-grafikkarte-treiber,testberichte-242146.html

they also lowered GPU TDP:

 

what remains to be tested, how much of a perf hit this translates into

[Trixanity]

24 minutes ago, zMeul said:

snip

Very misleading to say they've lowered TDP. Yes, it's lowered if you go into settings and enable an option. That could also be said if you underclock the card or lower the power target...

By default it remains entirely the same.

[LAwLz]

23 minutes ago, zMeul said:

the driver is out: https://support.amd.com/en-us/kb-articles/Pages/Radeon-Software-Crimson-Edition-16.7.1-Release-Notes.aspx

Tom's.DE did some testing: http://www.tomshardware.de/amd-radeon-rx-480-polaris-grafikkarte-treiber,testberichte-242146.html

 

they also lowered GPU TDP:

<image>

 

what remains to be tested, how much of a perf hit this translates into

Good to see that it got fixed.

Now the card sits at exactly 75 watts from the PCIe slot even without "compatibility mode" being turned on. Seems like they just shifted some of the PCIe slot load to the power connector. Not a fan that it now draws 93 watts from a 6-pin connector, but that's less dangerous than drawing it from the PCIe slot.

 

They really should have just put an 8-pin connector on the damn thing...

 

 

I am surprised nobody has posted a news thread about the 4GB version of the card actually being a rebranded 8GB card, just with a software lock on the other 4GB (which can be removed with a BIOS update).

[Sunako]

5 minutes ago, LAwLz said:

Good to see that it got fixed.

Now the card sits at exactly 75 watts from the PCIe slot even without "compatibility mode" being turned on. Seems like they just shifted some of the PCIe slot load to the power connector. Not a fan that it now draws 93 watts from a 6-pin connector, but that's less dangerous than drawing it from the PCIe slot.

 

They really should have just put an 8-pin connector on the damn thing...

 

 

I am surprised nobody has posted a news thread about the 4GB version of the card actually being a rebranded 8GB card, just with a software lock on the other 4GB (which can be removed with a BIOS update).

over on pcper they are showing that both connectors are drawling less power not no change in the ratio between either.

http://www.pcper.com/reviews/Graphics-Cards/AMD-Radeon-RX-480-Power-Consumption-Concerns-Fixed-1671-Driver

 all the test are showing that the peg connector is sitting right at 5.5 amps now with only metro last light on 4k going to 5.75.

[Briggsy]

PCPer on the ball as usual.
 

Quote

 

Allyn did some simple multimeter trace work to tell us which phases were connected to which sources and the result is seen below.

 

 

The power controller is responsible for pacing the power coming in from the PCI Express slot and the 6-pin power connection to the GPU, in phases. Phases 1-3 come in from the power supply via the 6-pin connection, while phases 4-6 source power from the motherboard directly. At launch, the RX 480 drew nearly identical amounts of power from both the PEG slot and the 6-pin connection, essentially giving each of the 6 phases at work equal time.

 

That might seem okay, but it’s far from the standard of what we have seen in the past. In no other case have we measured a graphics card drawing equal power from the PEG slot as from an external power connector on the card...

 

...As I understand, what AMD is doing now is very similar to what The Stilt in the Overclock.net forums attempted earlier in the week. Power phases 1-3 that source +12V from the 6-pin connection are now given more time than phases 4-6, thus shifting the weight of power draw towards the 6-pin connector.

 

...To be clear: this fix will not affect performance and my testing shows that to be the case.