thats just bitching right now and sounds so fuckin political. 6 pin can pull a lot of juice. this bullshit about it draws another 10w is not going to change the world. Man I hate when there is a solution people find another thing to bitch about. from what I have read RX480 6 pin is wired in the way that it doesn't need 8 pin to deliver more power. So you can call it out of spec, it sort of avoids you having to put 8 pin. Now you are going to complain if it draws 5 more watts? How did you think all the overclocking was done back in the days? 6 pin is more than capable of handling little more juice. If they can patch it so the card doesn't draw any of power to gpu from pci-e and more from 6pin Its all good. At that point it just becomes witch hunt when we all know the cards draw more out of 6 pin and they have been doing it for ages.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
RX 480 is apparently killing pcie slots
- Thread starter Ieldra
- Start date
thats just bitching right now and sounds so fuckin political. 6 pin can pull a lot of juice. this bullshit about it draws another 10w is not going to change the world. Man I hate when there is a solution people find another thing to bitch about. from what I have read RX480 6 pin is wired in the way that it doesn't need 8 pin to deliver more power. So you can call it out of spec, it sort of avoids you having to put 8 pin. Now you are going to complain if it draws 5 more watts? How did you think all the overclocking was done back in the days? 6 pin is more than capable of handling little more juice. If they can patch it so the card doesn't draw any of power to gpu from pci-e and more from 6pin Its all good. At that point it just becomes witch hunt when we all know the cards draw more out of 6 pin and they have been doing it for ages.
Not its not, Its AMD's problem but at least they could have a temporary solution to the problem until they figure it out how to reduce the power usage so that both PCI-E and 6-pin fall inside the specs.
That would give some end users a piece of mind.
Sounds like propaganda talk. Cards have pulled more from 6 pin forever, not anything new bro. This card is around 160's stock in current stage if they don't do anything else and draw another 10-15w from 6 pin you are really think that is going to blow things up? Common now, seen cards pull extra 100 out of it from old days and back then people had more shitty PSUs then now due to lack of competition. Now you can get GOLD certified PSU's for like 65 bucks not before.
I have never seen people ever worrry about their 6 pin when they overclocked the shit out of their cards, the power draw from the board is the main concern here. If you draw too much from 6 pin while overclocking atleast it doesn't blow up your board, you ever heard of failsafe? If you try to pull too much from your psu it will just shut down, that usually doens't happen with 10-15w though.
Daniel_Chang
[H]ard|Gawd
- Joined
- Jan 4, 2016
- Messages
- 1,313
I couldn't find this. Please show me one instance of that card drawing 800W while overclocked, for science
To be fair, he has a point. As I stated in another post. It's ok if the user goes out of spec while overclocking. That's their choice. A card should never be out of spec out of the box, unless it's an ultra enthusiast product with warnings. The 6-pin PCI-E connector is rated for 75W. Yes, you, me, and most here know that, with the proper wiring, it can go WAY beyond that. But a cheaper yet still adequate PSU may decide to adhere to that spec. And if that occurs, you have a 6-pin connector that only delivers 75W + some degree of tolerance. You're going to have problems with an out of box card routing over 100W to that cable.
So, like it or not, Nenu is correct. The card needs to be designed to be within spec out of the box, ESPECIALLY a reference card. Overclocking is up to the user, including any pitfalls that come with it.
The easiest way to fix this is to cap the card's power draw at 150W, with 75/75 being the split, and if power level is increased by the user, draw from the 6-pin as this causes the least harm. Drawing direct from the PSU won't damage anything as over-current protection will cause a shutdown/reboot before any harm could be done. A motherboard isn't design for power delivery. It's a conduit. So no such protection is afforded. A mother board delivering power is like a person without the sense of touch putting their hand on the stove. You don't know there's a problem until, well, you get the idea.
If there's someone that needs to be blamed, its most likely AMD's marketing team, yet again. They wanted a card which can be powered through 6-pin connector and engineering team had to design the card so that it draws pretty much 50/50 from both power inputs to get that PCI-E certification.
If this thing had a 8-pin connector, they could have powered the chip from it and memory, mem subsystem and display controllers from PCI-E slot, like how they did before with earlier series.
If this thing had a 8-pin connector, they could have powered the chip from it and memory, mem subsystem and display controllers from PCI-E slot, like how they did before with earlier series.
I couldn't find this. Please show me one instance of that card drawing 800W while overclocked, for science
To be fair, he has a point. As I stated in another post. It's ok if the user goes out of spec while overclocking. That's their choice. A card should never be out of spec out of the box, unless it's an ultra enthusiast product with warnings. The 6-pin PCI-E connector is rated for 75W. Yes, you, me, and most here know that, with the proper wiring, it can go WAY beyond that. But a cheaper yet still adequate PSU may decide to adhere to that spec. And if that occurs, you have a 6-pin connector that only delivers 75W + some degree of tolerance. You're going to have problems with an out of box card routing over 100W to that cable.
So, like it or not, Nenu is correct. The card needs to be designed to be within spec out of the box, ESPECIALLY A reference card. Overclocking is up to the user, including any pitfalls that come with it.
True but its not drawing a whole lot over spec though thats what I am trying to get it. 10-15 watt over 6 pin connector really won't harm anything. Unless you really have a crappy psu that can't even handle the graphics card in the first place. Thats my point. Most PSUs will be just fine. Out of box yea I am sure they will throttle it enough to keep it around 75w, but the way the card is wired it will pull much much more, from what I have read its wired in a way that 6 pin connector will act like an 8 pin because its grounded already like an 8 pin.
I couldn't find this. Please show me one instance of that card drawing 800W while overclocked, for science
To be fair, he has a point. As I stated in another post. It's ok if the user goes out of spec while overclocking. That's their choice. A card should never be out of spec out of the box, unless it's an ultra enthusiast product with warnings. The 6-pin PCI-E connector is rated for 75W. Yes, you, me, and most here know that, with the proper wiring, it can go WAY beyond that. But a cheaper yet still adequate PSU may decide to adhere to that spec. And if that occurs, you have a 6-pin connector that only delivers 75W + some degree of tolerance. You're going to have problems with an out of box card routing over 100W to that cable.
So, like it or not, Nenu is correct. The card needs to be designed to be within spec out of the box, ESPECIALLY a reference card. Overclocking is up to the user, including any pitfalls that come with it.
The easiest way to fix this is to cap the card's power draw at 150W, with 75/75 being the split, and if power level is increased by the user, draw from the 6-pin as this causes the least harm. Drawing direct from the PSU won't damage anything as over-current protection will cause a shutdown/reboot before any harm could be done. A motherboard isn't design for power delivery. It's a conduit. So no such protection is afforded. A mother board delivering power is like a person without the sense of touch putting their hand on the stove. You don't know there's a problem until, well, you get the idea.
yea I think that is what will happen. They will cap it at 150 and if the people want to take the risk of overclocking well they can run it out of spec all they want. Probably the best solution.
thesmokingman
Supreme [H]ardness
- Joined
- Nov 22, 2008
- Messages
- 6,617
I couldn't find this. Please show me one instance of that card drawing 800W while overclocked, for science
[INFO] What power is required for extreme overclock. Answer is here. - kingpincooling.com
Btw this stuff is common in this world. I can draw 1600w at the wall with my 3930K and two 290x, easily. Need proof for that too?
primetime
Supreme [H]ardness
- Joined
- Aug 17, 2005
- Messages
- 7,338
I agree with this......the 6 pin being overloaded is much easier to deal with. I personally dont care how oems or system builders deal or dont deal with it.....my interest is how well the card works out for me lol. Starting to look like the cards with 8 pin connectors are going to be good to go.Do tell me, which one you prefer, 85w from PCI-E slot with puny pins or 100w from 6-pin connector which can handle +250w, at least on RX480 because it has all six pins there
Daniel_Chang
[H]ard|Gawd
- Joined
- Jan 4, 2016
- Messages
- 1,313
[INFO] What power is required for extreme overclock. Answer is here. - kingpincooling.com
Btw this stuff is common in this world. I can draw 1600w at the wall with my 3930K and two 290x, easily. Need proof for that too?
Nah, you gave me what I needed, thank you.
Daniel_Chang
[H]ard|Gawd
- Joined
- Jan 4, 2016
- Messages
- 1,313
Here's the problem. If a PSU can only handle 75W over 6-pin, it's not "crappy." It's "spec compliant." This is what you're failing to understand (and I TRULY do not mean that as any kind of insult). A GPU SHOULD try to stay within this spec out of box so that it works on as many possible configurations as possible, again, out of the box, especially for a budget card sold at Best Buy. This is being sold to the tech-illiterate, not us. It needs to work.
Oh I won't argue about out of box. As I said they will probably make that happen with update. But on the other part all electronics have variance built in. If it is designed to handle 75w it does have some wiggle room, I know thats not the issue here. And yes out of the box it should not hit over 150. They should cap it at that.
Its not political, IT IS A MUST FOR THEM TO DO, they can't get pci-e cert if they are out of spec how hard is that for any of you guys to understand. There is some flexibility in the specs but not so much that AMD can run rampant with what ever they want to do. The specs tend to be outdated as well, but there is nothing AMD can do about that right now (with the 6 pin and 8 pin), unless there is something there I'm missing which I don't think I am.
The bios solution is good solution for the time being, but AMD still has to fix the root cause of the spec violation for ANY OEM or SYSTEM BUILDERS to pick them up.
THIS MEANS APPLE TOO. Who knows maybe APPLE will make AMD drop their pants more and ask for a discount so they will cover that warranty costs.....
The bios solution is good solution for the time being, but AMD still has to fix the root cause of the spec violation for ANY OEM or SYSTEM BUILDERS to pick them up.
THIS MEANS APPLE TOO. Who knows maybe APPLE will make AMD drop their pants more and ask for a discount so they will cover that warranty costs.....
Not in games. Furmark + Prime and its doable on water. I had the same setup clocked at 4.7ghz and 1250/1600mhz and I saw 1250w as a max power draw at the wall when playing BF4 on 64p server. Fire up the power virus and cpu burner and my gold rated 1200w PSU shut down instantly.
I am not arguing the in-spec part. Yea that must be done! I am just more generally talking about the extra 5-10w drawn from 6 pin won't hurt and its still better than it coming from the motherboard in general. Yea it has to consume 150 or less they gotta make sure it stays around there. Now if its doing 151-152 that is till with in margin of error.
I'm not just talking to you, eveyone in the last page and half, ITS NOT A NEGOTIABLE OPTION to be certified lol. Its either you get it or you don't
The rx480 has more than one issue, the PCI-e power overdraw added to this current overdraw, and then the over all wattage of the card pulling from the pci-e connectors. All must but in spec.
And right now they are all out of spec lol.
So for a temp fix yeah pushing it over the to PCI-e connector is a GREAT way to go. At least it will put peoples minds at ease to not screw around with their motherboards. But AMD can't stop there, as NKD just stated it has to consumer 150 watts or less overall with the other problems fixed as well. So this is what AMD is working on.
The rx480 has more than one issue, the PCI-e power overdraw added to this current overdraw, and then the over all wattage of the card pulling from the pci-e connectors. All must but in spec.
And right now they are all out of spec lol.
So for a temp fix yeah pushing it over the to PCI-e connector is a GREAT way to go. At least it will put peoples minds at ease to not screw around with their motherboards. But AMD can't stop there, as NKD just stated it has to consumer 150 watts or less overall with the other problems fixed as well. So this is what AMD is working on.
Will keep that in mind
Nenu
[H]ardened
- Joined
- Apr 28, 2007
- Messages
- 20,315
Jesus guys.
I'm a realist, I tell it how it is.
When NVidia conned everyone with the 3.5GB 970 and then went back on their promises, I gave them a lot of shit.
edit: at that time I owned a R9 290x, no agenda for either side. But I correct fanboy statements on either side because, well, they arent realistic.
Fanboyism makes people crazy!
I'm a realist, I tell it how it is.
When NVidia conned everyone with the 3.5GB 970 and then went back on their promises, I gave them a lot of shit.
edit: at that time I owned a R9 290x, no agenda for either side. But I correct fanboy statements on either side because, well, they arent realistic.
Fanboyism makes people crazy!
Last edited:
sutyi
n00b
- Joined
- Oct 23, 2010
- Messages
- 37
Well... it's a digital PWM controller, supposedly that is in charge of the power delivery on the card. You don't have a lot of current going trough that particular part.
thesmokingman
Supreme [H]ardness
- Joined
- Nov 22, 2008
- Messages
- 6,617
I didn't need to run furmark/prime to hit that instead it was in firestrike ultra iirc. At that time I was running quad 290x at 1300/1600mhz and a 5ghz 3930K to give you an idea. I ran dual psu and ran two gpu+cpu on one psu and two gpu on the other psu.
lolfail9001
[H]ard|Gawd
- Joined
- May 31, 2016
- Messages
- 1,496
Well, AMD are binded in WSA chains.
I just read one of the undervolt reports and unfortunately there is one critical factor missing in the article, improving fan profile.
I doubt anyone keeps reference cards to the default fan profiles, this was an issue with the 1080 (lesser extent 1070) that the fan profile was just not aggressive enough at say 80c, it just needed a tweak and looks like the 480 does as well.
Reading the following undervolt test: AMD Radeon RX 480 Undervolting Performance - Legit Reviews
Well they note most of the time the card is not holding its boost clock speed in default setup, primary reason would be throttling as it is the sort of behaviour one sees with the more dynamic boost algorithms that calculate clocks from voltage-temp to frequency.
So really the test needs to be done with the fan profile set to be a bit more aggressive IMO, and tbh every consumer should do that anyway with reference cards.
Worth noting legitreviews BF4 default behaviour is very similar to that of here:
However there are games that have a greater impact such as The Division and RoTR shown on the legitreviews undervolt test.
Point is, the undervolt may not actually improve performance if one sets the fan profile more aggressively to reduce boost algorithm from 'throttling'.
But then that is not its primary purpose, and unfortunately still not known how much undervolt is required to hit safe margin for the x16 PCIe slot.
Cheers
I doubt anyone keeps reference cards to the default fan profiles, this was an issue with the 1080 (lesser extent 1070) that the fan profile was just not aggressive enough at say 80c, it just needed a tweak and looks like the 480 does as well.
Reading the following undervolt test: AMD Radeon RX 480 Undervolting Performance - Legit Reviews
Well they note most of the time the card is not holding its boost clock speed in default setup, primary reason would be throttling as it is the sort of behaviour one sees with the more dynamic boost algorithms that calculate clocks from voltage-temp to frequency.
So really the test needs to be done with the fan profile set to be a bit more aggressive IMO, and tbh every consumer should do that anyway with reference cards.
Worth noting legitreviews BF4 default behaviour is very similar to that of here:
However there are games that have a greater impact such as The Division and RoTR shown on the legitreviews undervolt test.
Point is, the undervolt may not actually improve performance if one sets the fan profile more aggressively to reduce boost algorithm from 'throttling'.
But then that is not its primary purpose, and unfortunately still not known how much undervolt is required to hit safe margin for the x16 PCIe slot.
Cheers
Its a possibility or its a combination of that specific architecture on that node as well which is more likely.
I know how much those cards can take when benchmarking, I ran one at 1320/1600 with ocp limiters disabled. Power draw increased exponentially after I pushed more than 1.3v through the chip. But I'm 99% sure that 1300/1600 wasn't game stable
thesmokingman
Supreme [H]ardness
- Joined
- Nov 22, 2008
- Messages
- 6,617
I know how much those cards can take when benchmarking, I ran one at 1320/1600 with ocp limiters disabled. Power draw increased exponentially after I pushed more than 1.3v through the chip. But I'm 99% sure that 1300/1600 wasn't game stable
I've run up to 1.55v on the reference card and over 1.6 on lightnings. And yes, at 1300/1600 its on the edge with watercooling. If I had phase it would drop temps and be stable for gaming. In the past I had a 7970 reference that did 1400w on water with very low ambient. I sent it to a friend in norway and with his stupid low ambient he was gaming on it at 1350/1800mhz lol. Running phase though is not my style, wasting another 1Kw lol you know...
AMD Radeon HD 7970 video card benchmark result - Intel Core i7-3930K Processor,ASUSTeK COMPUTER INC. RAMPAGE IV FORMULA
^^using a tight memory timing bios, 1700mhz on memory here is equivalent to roughly 1900mhz.
Aah good old times when some radeons actually OC'd like beasts. My 7970 under water:
It was game stable at around 1350mhz with WC (in Finland
). But lets drop the OT here before someone gets angry Even without bridging, since half of the core voltage VRMs are connected to the PCI-E, wouldn't it pop if you don't have a 6 pin in whether you bridge or not? Especially during POST, since at that time every component is using full power plus no drivers to keep things in check.
Also if they're gonna change the power balance, that's gonna put a lot more stress on the 6 pin VRMs. I wonder if those are capable of handling the extra load...
Nenu
[H]ardened
- Joined
- Apr 28, 2007
- Messages
- 20,315
I posed this question assuming they must have some method of detecting a power connection on the 6 pin.
Someone must have performed this test outside of AMD by now so its probably safe to assume they are detecting a voltage level on the 6 pin power side.
My point too.
The 6 pin power connection is already out of spec before this change.
If the supply connection is solid the wires can handle up to 3 times the load, the wires wont be a problem.
But one of the potential fail conditions is when there is a bad connection.
This is worse than losing a wire entirely because a bad connection is a tiny connection point (by definition) that is trying to pass high current for its contact size.
This will heat up rapidly and if it doesnt disconnect due to burning up, it will continue to get hotter and risks more physical damage or fire.
In one respect, higher current flow could be safer due to faster oxidation of the connection. But this assumes the metal parts are not being forced together which can still be the case, even when there is a bad connection.
If so, the oxidation would help keep the contact area small and prolong the burning event which could set fire to the connector.
This makes it a none tolerable situation.
Thats not to say the same type of problem cant occur when its running to spec, but the spec exists for a reason, to minimise risk.
With less current flow, less heat is generated and degeneration is slower.
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
Hopefully this works but try this video.
Explains in great detail the power delivery on the card with comparisons to the Fury.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
Tis the same guy, except this is an older video
Palladium@SG
Limp Gawd
- Joined
- Feb 8, 2015
- Messages
- 283
It would be a much safer practical solution to shift load from the slot to the 6-pin, but it's still technically even worse out-of-spec operation on the 6-pin and completely valid grounds to recall or refund this POS no matter how many AMD diehards claim or show how much multiples of 75W max power can be carried by the 6-pin.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
I can't think of many cards that come out of spec on the 6pin at stock. It's nowhere near as big a deal as the PCIe problem, but still technically not allowed and more importantly to me, I don't understand how they ended up with this situation lol
Palladium@SG
Limp Gawd
- Joined
- Feb 8, 2015
- Messages
- 283
AMD fanboys' favorite punching bag the 960 Strix did very slightly and briefing exceeded 75W 6-pin out of the box by PCPer. Not saying it should get a free pass, but at least it is advertised as a factory OCed card and faaaaaaar from a PCIE slot disaster in the making.
If this fiasco is going to make card makers paying attention to slot power and putting 8-pin everywhere even on <150W cards then every consumer wins. If the users insist on being cheapskates using adapter cables then instead of a real 8-pin cable/PSU then they can only blame themselves for being dumb if shit happens.
No card that I know of was out spec on the pci-e bus, amps or wattage draw. PCI-e connects it has happened once the 6990 I think, all other top end cards nope and this is why nV and AMD never lets AIB's make their own versions cause there isn't much room to play with on those usually.
Across the a 6 pin PCI-e connector is not much of an issue at all, the specs really have to be redone for the connectors. In this case though, AMD just f'ed up, two if and buts about it.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
what about the 295x2, titan z, pro duo etc ?
all those had enough power connectors for pci-e compliance.
Power Consumption: Gaming - Radeon R9 295X2 8 GB Review: Project Hydra Gets Liquid Cooling
toms hardware power draw, from bus the 295x2 is drawing 30 watts or so
actually they did go over the pci-e connect limits on this card as well.
The NVIDIA GeForce GTX TITAN Z Review | 3DMark, Power, Sound and Conclusions
they didn't show us the details but they wrote about it
Power Consumption: Gaming - Radeon R9 295X2 8 GB Review: Project Hydra Gets Liquid Cooling
toms hardware power draw, from bus the 295x2 is drawing 30 watts or so
actually they did go over the pci-e connect limits on this card as well.
The NVIDIA GeForce GTX TITAN Z Review | 3DMark, Power, Sound and Conclusions
they didn't show us the details but they wrote about it
Last edited:
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
Yeah the connectors have been overdrawn before its no big deal really, especially on a dual gpu card its pretty forgivable in my book, so long as it's not something ridiculous like +80% at stock lol. PCIe? Never
yeah on these super high end cards I don't see people using cheapo components anyways so there should be no issue.