Looks like AMD issued a statement and will be issuing a driver fix :
Much ado about nothing IMHO.
Much ado about nothing IMHO.
As you know, we continuously tune our GPUs in order to maximize their performance within their given power envelopes and the speed of the memory interface, which in this case is an unprecedented 8Gbps for GDDR5. Recently, we identified select scenarios where the tuning of some RX 480 boards was not optimal. Fortunately, we can adjust the GPUs tuning via software in order to resolve this issue. We are already testing a driver that implements a fix, and we will provide an update to the community on our progress on Tuesday (July 5, 2016).
Wccftech has an article about this dilemma :
http://wccftech.com/amd-rx-480-pcie-power-issue-detailed-overclocking-investigated/
The VRM power delivery circuitry on the reference RX 480 printed circuit board is beyond impressive. More so when compared to the GTX 1080s 5+1 power phase design which simply pales in comparison. The RX 480s on-board VRM voltage regulator on the high-side is capable of delivering 40 amps from each phase at 125c, for a total of 240 amperes.
At room temperature each phase can deliver 66 amps for a total of 396 amps on the high-side and up to 600 amps on the low-side. The GTX 1080 Founders Edition voltage regulator can deliver 30 amps at 125c from each of its five phases, for a total of 150 amps on the low side, 246 amps less than the RX 480. At room temperature it goes up to 50 amps per phase for at total of 250 amps, 350 amps less than what the RX 480 VRMs can deliver at the same temperature on the low side.
if amd made that boardWill AMD/AIBs take some responsability here?
Will AMD/AIBs take some responsability here?
not likely. there have been tons of pictures like that before rx 480. Even with single slot cards. Just because it happened with the rx 480 does not mean its responsible.
I wonder though, because recently we saw that a few cards do actually have much higher spikes on the pci-e slot. Would those fry the motherboard more frequently if you put 3 of them in the system? Seems they should. That's a massive load if the spikes line up.
At least in civil engineering the connector is overspeced compared to the line and breakers.
I think in electronics it is exactly the same.
Anyway. I'm curious if they can change the VRM controller load balancing with software.
The thing is, you need to generate HEAT to cause problems, peaks are not gona cause that, its the constant current draw that gona cause connectors and pins to heat up and burn, 1ms peaks are not just not enoght.
Peaks may blow up components, but i dont think there is much to blow up there, also im petty sure MB traces and components can handle the peaks, the connectors, mainly the ATX one, are the weak points, and even then a single RX480 its unlikely to cause damage, at least at stock, CF, and specially overcloked CF, may do.
They will lower the clocks or voltage and their is no other logically way.
High spikes are fine, and keep in mind that some of these spikes can be measurement artifacts.
Don't these micro-spikes get filtered out by capacitors anyway?
No, the capacitor are indeed a cause for those spikes since charging them after they got discharged a little by an abrupt current rush is what cause thoses peaks more than the GPU itself, at those points the PC PSU must supply both the demanding GPU while charging the capacitors..
950 (no power connector) uses more than 75W:
Don't these micro-spikes get filtered out by capacitors anyway?
Umm NO... Capacitors are used to filter voltage ripple, they are used as "bypass" capacitors for this reason and it is why you see lots of capacitors around ICs as they are sensitive to voltage ripple and need a locally stable supply as other ICs around them will cause voltage ripple on the same rail.
You are referring to inrush current, which is when large capacitors on power rails are charged on device turn on... That is totally different.
That doesnt work like this, the capacitors that are in the PC PSU are the one providing the necessary current peaks to charge (while supplying the GPU) the capacitors of the card that are slightly discharged by a sudden demand of the GPU.
We are talking of current peaks while you are thinking instantaneous voltage variation of the voltage, you could have huge current peaks without the voltage moving much if the PSU and routings have low impedances.
Nice article with some technical infos :
I'm not talking about the capacitors in the PSU, I'm talking about the capacitors on the board, the SMD and Solid Capacitors around ICs and the like... They are bypass capacitors and they smooth the voltage ripple on the rails that power the ICs, the ripple being caused by other ICs on the same rail, NOT the ripple caused by the Switchmode Power supply in the PSU.
The poster you replied to was asking if the capacitors smooth the ripple caused by ICs drawing too much power, and they (the bypass capacitors) do filter this ripple (to a limit).
The capacitors in the PSU are used to smooth the ripple from the high frequency DC coming from the secondary side of the transformer + rectifier. They only provide a baseline smooth DC supply for everything else (usually it's like 120mV max ripple for 12V and 50mV max for 5V and 3.3V).
We are talking of current peaks while you are thinking instantaneous voltage variation, you could have huge current peaks without the voltage moving much if the PSU and routings have low impedances.
I'm not talking about the capacitors in the PSU, I'm talking about the capacitors on the circuit boards, the SMD and Solid Capacitors around ICs and the like... They are bypass capacitors and they smooth the voltage ripple on the rails that power the ICs, the ripple being caused by other ICs on the same rail, NOT the ripple caused by the Switchmode Power supply in the PSU.
The poster you replied to was asking if the capacitors smooth the ripple caused by ICs drawing too much power, and they (the bypass capacitors) do filter this ripple (to a limit).
The capacitors in the PSU are used to smooth the ripple from the high frequency DC coming from the secondary side of the transformer + rectifier. They only provide a baseline smooth DC supply for everything else (usually it's like 120mV max ripple for 12V and 50mV max for 5V and 3.3V).
However, local rails on the motherboard and on add on cards have sources of ripple from the ICs on those boards/cards and they need local bypass capacitors to smooth this additional ripple.
The RX 480 drawing excessive current from the local PCI-E power rails on the motherboard will cause excessive ripple on that rail as a consequence, but the standard max has been derated anyway as any good design does, so it should be safe regardless (as long as you are not OCing or using > 3 cards at once).
Don't these micro-spikes get filtered out by capacitors anyway?
When i read it, it seems they have the data sheets of the high side and low side mosfets. I wonder if they have the inductor datasheets as well. Because that is a lot of saturation current that the inductors need to handle for such a relatively small size.
Basicaly the 1080 has a sub 200W PSU in worse conditions, but this is not a problem apparently...The GTX 1080 Founder’s Edition voltage regulator can deliver 30 amps at 125c from each of its five phases, for a total of 150 amps on the low side, 246 amps less than the RX 480. At room temperature it goes up to 50 amps per phase for at total of 250 amps, 350 amps less than what the RX 480 VRMs can deliver at the same temperature on the low side.
Without those capacitors peaks would be much lower and voltage would sag, they povide the low impedance at high fequencies (short peaks)..