Well as a wild guess, thinking a week or two for a driver update and then a month or so for a bios update. If the root cause is over-volting then this may be a easy fix with clear benefits of less power really needed and potential better OC.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
AMD in Trouble? RX 480 Powergate
- Thread starter Ocellaris
- Start date
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
I think they should release a fix of some sort, even if minimal, soon like yesterday. Attempts are gonna go over well but waiting too long is gonna look real bad.
WorldExclusive
[H]F Junkie
- Joined
- Apr 26, 2009
- Messages
- 11,548
Put an 8pin on it, don't talk about heat and power, and see how far it can OC to annoy a 980.
They will need it because Nvidia is capable of releasing a 1060 with 980 performance for $250.
I don't think Polaris will have to much problem OCing once cooled and powered correctly and exceeding the 980 performance. Just a hunch plus Kyle indicated that as well who I believe. The question is will the 1060 at 1700mhz really benefit much with OC to 2000 to 2100? 6gb ram and 192 bit bus? At 1080p Rise of the Tomb Raider at max settings was already at 6gb. We will see but from the performance standpoint looks like Polaris has some room to grow.
Caffeinated
[H]ard|Gawd
- Joined
- Oct 16, 2002
- Messages
- 1,451
Is it too dang much to ask for an affordable upgrade these days? I had high hopes for the RX 480, and since my system has worked like a charm with overclocked GTX 460's in SLI, I'm pretty sure it can handle one card.
Except ... I don't want to ruin my motherboard. This really concerns me. I am obviously waiting for the AIB varieties, but damn, I don't want to cook my MB.
Apparently the options are expensive(GTX 1070) or dodgy (RX 480) right now. Meanwhile, I'm stuck in lag-ville
.
Except ... I don't want to ruin my motherboard. This really concerns me. I am obviously waiting for the AIB varieties, but damn, I don't want to cook my MB.
Apparently the options are expensive(GTX 1070) or dodgy (RX 480) right now. Meanwhile, I'm stuck in lag-ville
.
Ocellaris
Fully [H]
- Joined
- Jan 1, 2008
- Messages
- 19,077
Is it too dang much to ask for an affordable upgrade these days? I had high hopes for the RX 480, and since my system has worked like a charm with overclocked GTX 460's in SLI, I'm pretty sure it can handle one card.
Except ... I don't want to ruin my motherboard. This really concerns me. I am obviously waiting for the AIB varieties, but damn, I don't want to cook my MB.
Apparently the options are expensive(GTX 1070) or dodgy (RX 480) right now. Meanwhile, I'm stuck in lag-ville
Pickup a used 970 or wait for the 1060.
Caffeinated
[H]ard|Gawd
- Joined
- Oct 16, 2002
- Messages
- 1,451
I guess. My niece just went to college though, and I'm a tad strapped for cash. She's a wonderful human being and I don't resent helping her with tuition, it just sucks that the prices being commanded are nuts for nVidia, and AMD might fry the shit out of your motherboard.
Ocellaris
Fully [H]
- Joined
- Jan 1, 2008
- Messages
- 19,077
I don't see the motherboard frying to be an issue that AMD will let go unresolved. If I had ~$200 I would likely be purchasing a 480 even knowing the power draw is over spec. They should fix the issue through new drivers or a BIOS update.
Caffeinated
[H]ard|Gawd
- Joined
- Oct 16, 2002
- Messages
- 1,451
I sure hope so. I just don't currently have the financial flexibility for a 1070. I've survived this long, though, so I guess I should just count my blessings and keep on waiting.
TaintedSquirrel
[H]F Junkie
- Joined
- Aug 5, 2013
- Messages
- 12,691
Or just buy a nonref card which wont have any problem...
Is it too dang much to ask for an affordable upgrade these days? I had high hopes for the RX 480, and since my system has worked like a charm with overclocked GTX 460's in SLI, I'm pretty sure it can handle one card.
Except ... I don't want to ruin my motherboard. This really concerns me. I am obviously waiting for the AIB varieties, but damn, I don't want to cook my MB.
Apparently the options are expensive(GTX 1070) or dodgy (RX 480) right now. Meanwhile, I'm stuck in lag-ville
Well if you can wait a little, compare the 1060 and 480 at the time of launch. Should have AIB 480's out there as well and pick up the best bang/$. The 480 power issues should at least be resolved by the AIBs and maybe by AMD cards as well. I do believe the 1060 will be lower in power overall. Now you won't be able to SLI the 1060 like the 480 which gives you an extra option later if you need to or if a killer deal comes up. My bet is the AIB OC versions of the 480 will be faster then the 1060 but until they are released who knows. I will also be more interested in DX12 stuff as well since that is the future.
I don't recommend a 970/980 - less ram, so far performs worst in DX12 for Nvidia no better in power use (if that is important to you) and still more expensive. As for the 390/390x - they give off a lot of heat but do perform well in DX12. Maybe best to wait a few weeks.
Caffeinated
[H]ard|Gawd
- Joined
- Oct 16, 2002
- Messages
- 1,451
GTX 1060 doesn't have sli connections?
Nvidia usually do not allow lower tier cards to SLI since it would eat into sells on the higher end. If two 480's clip the 1080 and beats the 1070 when CFX with games that scale - what will a two 1060s do? Still maybe best to wait to see the full and real specs of the 1060. As for performance I actually believe the 480 will be higher but with a significant amount more power usage as in 50w-75w more maybe.
I believe you can also run more then two 480 in CFX while Nvidia limits Pascal to 2 if supported.
I believe you can also run more then two 480 in CFX while Nvidia limits Pascal to 2 if supported.
Last edited:
Caffeinated
[H]ard|Gawd
- Joined
- Oct 16, 2002
- Messages
- 1,451
It looks like I need to wait a bit. Oh well, I can squirrel away my pennies for now.
Wait. . . 1060
the image in WCCTech shows SLI connections on the image with the cover removed:
NVIDIA GeForce GTX 1060 Pictures and Specifications Leaked
Well that will be a cool option if it shows up that way on the street. I say definitely wait to see the 1060.
the image in WCCTech shows SLI connections on the image with the cover removed:
NVIDIA GeForce GTX 1060 Pictures and Specifications Leaked
Well that will be a cool option if it shows up that way on the street. I say definitely wait to see the 1060.
they might be using the same pcb to save cost. I think it is pretty much confirmed it wont support sli, simply put I believe it. Nvidia will find one or another to gimp the gtx 1060. No sli means people have no other choice but to buy their high cards, that simple.
lolfail9001
[H]ard|Gawd
- Joined
- May 31, 2016
- Messages
- 1,496
You only need to slap 1060 in same performance tier as rx480 to make slid 1060s completely unappealing.
SamiiRoss
Gawd
- Joined
- Jan 8, 2012
- Messages
- 709
A driver update will be release after Tuesday next week. They are going to fix the "problem" via software.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
No.
They are going to "fix" the problem via software.
It's the word 'fix' that necessitates quotation marks, not the word 'problem'
Any driver fix will lower the power limit, so this will lower performance, so the performance in reviews is not representative of the final product. Drivers do not determine power draw distribution
SamiiRoss
Gawd
- Joined
- Jan 8, 2012
- Messages
- 709
I'm currently mining with 4 of them no "problems" so far. dont try to be a smart ass with me. Go a play your doom u haven't play it for a while :v
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
That's great, but your not having problems says nothing about AMD's non-compliance with the PCI-Express specification.
Arcygenical
Fully [H]
- Joined
- Jun 10, 2005
- Messages
- 25,066
The entire PCIe specification for power is out of whack IMHO.
6 pin connectors are rated for only, what, 75w? With 3 12v rails? 2.17a assuming 11.5v (worst case scenario) is well, well below the limit for the connector design. Almost by a factor of 4. The 8 pin connector is rated for twice the wattage, with no extra physical ampacity (though, I suppose splitting one of the pins into 2, with a small 20g wire, typically, counts for an amp or 2).
I wish there was a way to limit PCIe slot power to the minimum signal level (3.3v, and 12v) and pull everything from the connector.
This isn't true at all. Depending on the type of BIOS written for the card. PCB shots of the RX480 show digital PLL and power delivery control. It is conceivably possible that AMD can adjust power delivery in BIOS.
6 pin connectors are rated for only, what, 75w? With 3 12v rails? 2.17a assuming 11.5v (worst case scenario) is well, well below the limit for the connector design. Almost by a factor of 4. The 8 pin connector is rated for twice the wattage, with no extra physical ampacity (though, I suppose splitting one of the pins into 2, with a small 20g wire, typically, counts for an amp or 2).
I wish there was a way to limit PCIe slot power to the minimum signal level (3.3v, and 12v) and pull everything from the connector.
This isn't true at all. Depending on the type of BIOS written for the card. PCB shots of the RX480 show digital PLL and power delivery control. It is conceivably possible that AMD can adjust power delivery in BIOS.
Last edited:
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
Are you saying it's possible to alter the power distribution over pcie / atx just like you can do in maxwell bios, but through drivers?
Arcygenical
Fully [H]
- Joined
- Jun 10, 2005
- Messages
- 25,066
Nope, I'm not saying anything. I'm just assuming the card has this capability, considering the degree of PLL control it seems to possess. I have no idea, as I don't plan on purchasing one, don't have one, and I'm not entirely interested in this segment.
And I suppose I consider a VBIOS mod Software, as well. If that's the point of contention here.
Why? Do you expect AMD to do the right thing? Promptly or fall all over themselves and wait until it is to late?
The longer they wait the worse it becomes, More poeple will be aware of it the more damage comes to AMD by word of mouth.
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
True and at least they released a statement that they could/may have a fix Tuesday. Probably a bandaid with a true fix forthcoming later but at least a fix.
The dominant element in CMOS power dissipation is the charging and discharging of gate and wire capacitance. This element obeys the equation P = k * V^2 * C * F, where V is the voltage, C is the combined circuit capacitance being charged and discharged, F is the frequency, and k is a constant that depends on the design. k is fixed by the design of the hardware, as is C. V and F however can be controlled by software, such as by using an overclocking utility. Maximum F tends to scale linearly with the voltage BTW.
So the firmware on the RX 480 card controls the clock speeds - that's how the card base and boost clocking works, and how you overclock. Lower the clock speed and the power goes down as well. Cap the clock at a lower value, and the maximum power draw gets capped at a lower value too. In addition, capping the clock frequency at a lower value means you can lower the supply voltage without causing the circuit to malfunction, saving even more power.
So one fix AMD is certainly capable of implementing is to change the GPU firmware to that the clocks on some or all of it, such as the clocks driving the memory subsystem, are capped at a lower value than they currently operate at. Power draw will go down as a result. And almost certainly, performance will go down too, or AMD wouldn't have bothered running the clocks that fast in the first place.
I believe with three 12 v rails it's just under 200watts, four would put it at 300.
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
He probably is speaking to the pins. 6pin connectors would only have 2 12 wires though most have a third in place. I think the third gets used when the 2nd sensor wire in the 8pin is used. At any rate the PCI-e specs on cables is far too conservative for we actually have. However the specs for the MoBo are far more accurate and likely need be adhering to.
Agreed
We don't want them rushing a solution because sometimes a rushed bad solution is worse than doing nothing at all.
Other than that, yeah, I agree with you.
bboynitrous
2[H]4U
- Joined
- Feb 29, 2004
- Messages
- 2,514
Yeah I'm kind of thankful now. A friend of mine was about to purchase one and has an older board so now he'll have to way until maybe a two six pin solution or 8 pin board comes out. I don't think there's any other fix for this other than that. I even have an older PCIE 2.0 board and now I'm not going to risk it. Crazy.
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
yeah, agree totally. I was just stressing how important it is AMD gets some sort of fix out in a timely manner.
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
honestly guys like us can fix the problem ourselves quite easily. Undervolt and maybe reduce clocks as an immediate solution. Seriously any member on an enthusiast forum that fries a board after having the information we now have should be banned from ever posting again.
I am not saying waiting is a bad or wimpy solution, if you can then it is advisable. But any one privy to the facts and posting on knowledge forums should have at least the minimal knowledge necessary to facilitate an adequate solution till a more permanent solution has been discovered.
So true! I think some people just try to make it worse. "I know whats going on and I am well informed but I am going to push the limits anyways see if I blow shit up lol". I have seen people trying to undervolt and then overclocking I am like shaking my head lol. It's like everyone is asking for it.
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
Love the bold part.
and you know the guy who blew up his board in amd community. Had his card overclocked to 1350. I think what people don't realize is that the card is not voltage limited. As soon as you push that slider up to 50% power you are asking for it it will suck as much power it can. I think that is the problem here. AMD should just disable that next go around on driver update on reference cards. He has it overclocked and he most likely had the power slider up and he games 7 hours straight. Well that will do it. I blame amd for not educating people enough but since his post clearly says 1350 OC well he pretty much lost any lawsuit he could have had. Remember when you fuck with overclocking tools you do agree that amd is not responsible for any damage that may cause.
thesmokingman
Supreme [H]ardness
- Joined
- Nov 22, 2008
- Messages
- 6,617