AMD RX 480 Causing Faults For PC Gamers – PCIe Slots Have Died On High-End Boards

[Starelementpoke]

1 minute ago, klh2000 said:

Give me a minute...

 

Starting the timer.

[Clyne]

5 hours ago, LAwLz said:

I put on my robe and skepticism hat when I opened this thread. Surprisingly enough some of these seem pretty legit.

 

roquen22 used an ASRock 970 Extreme4 motherboard. That's a pretty old (probably 4-5 years by now) motherboard and it was not exactly high end back then. On top of that his RX 480 was overclocked on the core (although not overvolted and stock VRAM).

 

The person on overclock.net (the one with the Rampage IV Extreme motherboard) was using 3-way crossfire.

 

 

I would not be surprised if the majority of reports are fake, but it is worrying that some of them seem legitimate. I was expecting reports like these in 6+ months. Not 2 days.

It's funny with those specs.

Asrock 970 Extreme 4, FX-8350@ 4.5GHZ, 480 OC'd to 1350MHZ core, CX500.

What do you expect, pushing a 8350 on a 4+1 phase board with a OC'd RX 480 at 1350MHZ core.

[Colonel_Gerdauf]

5 hours ago, LAwLz said:

I put on my robe and skepticism hat when I opened this thread.

Hold on a minute here. I get the skepticism hat, but robe? Just how many lemons did you end up buying in the grocery stores?

[Curufinwe_wins]

2 hours ago, RagnarokDel said:

The number of power-phase has nothing to do with this issue oO

The number of time I accidently mined on a overvolted  7990 just to see it reach 105° in seconds(different gaming and mining settings) and not break because it just reboots your pc rather then damage itself. Yeah I highly doubt that's was a thing. as for PCIE slots, they're different. I guess one could go in their motherboard bios and fix a hard limit on how much wattage is available via the PCIE slot. I'm inclined to agree with you that 99% of those claims are bullshit.

Sorry for the delay but it actually can. Each power phase has a specific source. Some are from pcie slot some are from connectors. By adding more connector loading power phases at the same total load the power draw from pcie phases diminishes significantly.  (This is one of the many reasons cards for ln2 have significantly beefier power delivery.)

[Curufinwe_wins]

29 minutes ago, Bubblewhale said:

It's funny with those specs.

Asrock 970 Extreme 4, FX-8350@ 4.5GHZ, 480 OC'd to 1350MHZ core, CX500.

What do you expect, pushing a 8350 on a 4+1 phase board with a OC'd RX 480 at 1350MHZ core.

GPU load is completely independent of cpu power delivery (it goes directly to slot, hence gpu's have their own power delivery components) That said, that is a pretty ambitious oc for a 970 4+1 board.

 

 

[Starelementpoke]

2 minutes ago, Curufinwe_wins said:

GPU load is completely independent of cpu power delivery (it goes directly to slot, hence gpu's have their own power delivery components) That said, that is a pretty ambitious oc for a 970 4+1 board.

 

 

You could say it FXcking ballsy.

[Curufinwe_wins]

1 minute ago, Starelementpoke said:

You could say it FXcking ballsy.

rofl

[Doobeedoo]

Quite interesting. I guess it wasn't much intended for reference card to be OC really. So yeah, for those that plan to OC this GPU custom cards it is.

[klh2000]

58 minutes ago, Starelementpoke said:

Starting the timer.

Sorry this took so long, but here you go:

 

 

And the video (sorry about portrait mode, I was in a hurry):

 

 

[That Norwegian Guy]

19 minutes ago, Curufinwe_wins said:

GPU load is completely independent of cpu power delivery (it goes directly to slot, hence gpu's have their own power delivery components) That said, that is a pretty ambitious oc for a 970 4+1 board.

 

 

It still comes from the EPS 12v connector on the motherboard.

 

If I had an 970FX board I wouldn't even dare put an 8350 in it, let alone overclock it. There are heaps of those dead from just burnt PCB around the socket. Those motherboards are meant for 6300 and below.

 

Hell, even on 990FX I still only trust the ASUS Sabertooth R2.0

[Starelementpoke]

5 minutes ago, klh2000 said:

Sorry this took so long, but here you go:

 

 

And the video (sorry about portrait mode, I was in a hurry):

 

 

Bah, graphs and incomprehensible charts, my weakness!

[HalGameGuru]

The 8-pin is a non-entity. They can pull plenty of power out of a 6-pin, the problem is them not properly limiting the card's ability to draw from the Slot.

[Clyne]

51 minutes ago, Curufinwe_wins said:

GPU load is completely independent of cpu power delivery (it goes directly to slot, hence gpu's have their own power delivery components) That said, that is a pretty ambitious oc for a 970 4+1 board.

 

 

Maybe not power delivery, but i suspect his Northbridge may something to due with it. 

OC'd FX may have caused Northbridge run hotter and the oc'd 480 may have been the final straw, this is what i think, may not be entirely true.  

I doubt the PCIE died is true, as AMD would have done tests to make sure there weren't any faults with it. 

[Curufinwe_wins]

18 minutes ago, Bubblewhale said:

Maybe not power delivery, but i suspect his Northbridge may something to due with it. 

OC'd FX may have caused Northbridge run hotter and the oc'd 480 may have been the final straw, this is what i think, may not be entirely true.  

I doubt the PCIE died is true, as AMD would have done tests to make sure there weren't any faults with it. 

Ehh, I've had plenty of older mobo's lose functionality in particular pcie slots... In fact the one I have my parents media pc in right now (a ddr2/ddr3 970 mobo w an x4 975 black edition) has a non-functional primary pcie slot. Fortunately the second x16 slot works fine so it doesn't matter, but when I was scavenging the pc (before giving my parents the pc) I was surprised to find out that was the issue.

[leadeater]

11 hours ago, larsi239 said:

To me it really does not matter wich claims are true or not. If AMD had just gone with an 8 pin in the first place, many if not all of the genuine complaints would not be here today, wich in turn would certainly get rid of all the fake ones. 

 

Simply put: AMD brought all this bad PR on themselves.

 

Edit: Oops double post

 

It's not quite as simple as saying use an 8-pin rather than a 6 pin. Without redesigning the entire PCB, power delivery/regulation etc an 8-pin will make no difference. The theoretical problem is the design of the card pulls too much power from the PCI-E slot, which I am yet to be convinced is actually happening on a wide scale and that it is to a degree that would cause problems at all.

 

Power or more correctly current will only be delivered as much as is required and if the additional power connectors aren't being asked to supply this current, 6-pin or 8-pin, then they won't. Also there have been many cards over the years that have exceeded the designed power limits of ether or all of the slot, 6-pin and 8-pin. The cables used for 6-pin and 8-pin connectors can handle far more than 75W or 150W.

 

Keep in mind graphics cards are already a special case exception to the PCI-E standard as it is.

Quote

All sizes of ×4 and ×8 PCI Express cards are allowed a maximum power consumption of 25 W. All ×1 cards are initially 10 W; full-height cards may configure themselves as 'high-power' to reach 25 W, while half-height ×1 cards are fixed at 10 W. All sizes of ×16 cards are initially 25 W; like ×1 cards, half-height cards are limited to this number while full-height cards may increase their power after configuration. They can use up to 75 W (3.3 V × 3 A + 12 V × 5.5 A), though the specification demands that the higher-power configuration be used for graphics cards only, while cards of other purposes are to remain at 25 W.

https://en.wikipedia.org/wiki/PCI_Express#Power

 

When it comes to electrical standards and safety there is always significantly large headroom/margin put in as an extra additional safety measure just in case. If something is rated at 75W it can handle a great deal more than that 24/7, but 75W will be safe even under extreme an unusual circumstances.

 

As a hobbyist with a national certificate in electronics, has a parent who is a registered electrician and a family friend who is a certified electrical engineer I have a fairly good understanding of fundamental electrical theories. No where near the level of the people designing these cards and even further away from the people designing the GPU but enough to know none of the evidence supplied and the figures warrant the response we have seen. I also have no doubt someone or many people of this form are more qualified than me on this topic.

 

I'll put my tinfoil hat on and say a large component of this is people specifically looking to find fault in a product in a extremely competitive market ignoring all other cases of the exact same thing from previous products and competitors. It's not the first time an AMD card has exceeded the PCI-E power spec and zero people cared before, why now.

[totjup5]

Sorry, but I'm calling bullshit on this. The 960 did the same thing at launch and no one went crazy about it. Also, any card will exceed 75w through the pcue slot if you overclocked it. Pure bullshit just stacking on top of the minor issue that people have blown out of proportion.

[Albatross]

Yikes...poor AMD man. They just can't get a break.

[Yoinkerman]

If I had to guess there's a thin trace or 0 ohm resistor acting as a fuse for pci-e slots.  I could see burning one out with 90+W somewhat sustained instead of 60W typical max.

[Kobz360]

lol so the OP (roquen) posted pictures of the RX 480 (which he actually has) and a 750 Ti. 

 

Interestingly enough, he shows the 480 inside his motherboard (AsRock 970 E4), but then to prove he also has a 750 Ti, he takes a photo of that.. Of course on close inspection, you can see that the photo showing the 750 Ti and the photo showing the 480 feature completely different motherboards. When questioned, he claims it's still the 970 Extreme4, although clearly the 750 Ti is in a different motherboard.

 

Though once again, this is a lot of effort this guy is putting in just to troll

[Prysin]

12 hours ago, Bubblewhale said:

It's funny with those specs.

Asrock 970 Extreme 4, FX-8350@ 4.5GHZ, 480 OC'd to 1350MHZ core, CX500.

What do you expect, pushing a 8350 on a 4+1 phase board with a OC'd RX 480 at 1350MHZ core.

ON a rippling, shitty CX500

[leadeater]

8 minutes ago, Kobz360 said:

lol so the OP (roquen) posted pictures of the RX 480 (which he actually has) and a 750 Ti. 

 

Interestingly enough, he shows the 480 inside his motherboard (AsRock 970 E4), but then to prove he also has a 750 Ti, he takes a photo of that.. Of course on close inspection, you can see that the photo showing the 750 Ti and the photo showing the 480 feature completely different motherboards. When questioned, he claims it's still the 970 Extreme4, although clearly the 750 Ti is in a different motherboard.

 

Though once again, this is a lot of effort this guy is putting in just to troll

Yea some magic he pulled to have the same motherboard with completely different number, colour and spacing SATA connectors and also PCI-E slots. One minute they are blue and the next black???

 

Well at least it looks the the photos are taken from the same model phone, checked the image EXIF metadata.

Make	samsung
Camera Model Name	SM-G920V
Software	G920VVRU4CPD2

[Kobz360]

2 minutes ago, leadeater said:

Yea some magic he pulled to have the same motherboard with completely different number, colour and spacing SATA connectors and also PCI-E slots. One minute they are blue and the next black???

 

Well at least it looks the the photos are taken from the same model phone, checked the image EXIF metadata.

Make	samsung
Camera Model Name	SM-G920V
Software	G920VVRU4CPD2

This is so bizarre man.. He could've just said he has another motherboard but chose not to. 

[patrickjp93]

10 hours ago, leadeater said:

It's not quite as simple as saying use an 8-pin rather than a 6 pin. Without redesigning the entire PCB, power delivery/regulation etc an 8-pin will make no difference. The theoretical problem is the design of the card pulls too much power from the PCI-E slot, which I am yet to be convinced is actually happening on a wide scale and that it is to a degree that would cause problems at all.

 

Power or more correctly current will only be delivered as much as is required and if the additional power connectors aren't being asked to supply this current, 6-pin or 8-pin, then they won't. Also there have been many cards over the years that have exceeded the designed power limits of ether or all of the slot, 6-pin and 8-pin. The cables used for 6-pin and 8-pin connectors can handle far more than 75W or 150W.

 

Keep in mind graphics cards are already a special case exception to the PCI-E standard as it is.

https://en.wikipedia.org/wiki/PCI_Express#Power

 

When it comes to electrical standards and safety there is always significantly large headroom/margin put in as an extra additional safety measure just in case. If something is rated at 75W it can handle a great deal more than that 24/7, but 75W will be safe even under extreme an unusual circumstances.

 

As a hobbyist with a national certificate in electronics, has a parent who is a registered electrician and a family friend who is a certified electrical engineer I have a fairly good understanding of fundamental electrical theories. No where near the level of the people designing these cards and even further away from the people designing the GPU but enough to know none of the evidence supplied and the figures warrant the response we have seen. I also have no doubt someone or many people of this form are more qualified than me on this topic.

 

I'll put my tinfoil hat on and say a large component of this is people specifically looking to find fault in a product in a extremely competitive market ignoring all other cases of the exact same thing from previous products and competitors. It's not the first time an AMD card has exceeded the PCI-E power spec and zero people cared before, why now.

According to the PCI-SIG standard, there is an 8% overage protection for the average draw. instantaneous spikes can be just about as high as you want. However, AMD averaging 88W+ is far beyond that 8%. That's the problem. The ASUS Strix 960 never exceeds the 75W average threshold.

[othertomperson]

18 hours ago, LAwLz said:

The person on overclock.net (the one with the Rampage IV Extreme motherboard) was using 3-way crossfire.

That's the thing. With all the people saying "people with good motherboards will be fine," I would not have expected this problem with a Rampage IV Extreme. This was a £300 (~$500) motherboard. I don't think spending more on a motherboard than on the GPU is a good idea, nor do I think 3-way crossfire makes sense for high end, let alone low-mid range cards like the 480, but at the same time if this card is killing motherboards of this extreme tier I think we have a problem.

[leadeater]

55 minutes ago, patrickjp93 said:

According to the PCI-SIG standard, there is an 8% overage protection for the average draw. instantaneous spikes can be just about as high as you want. However, AMD averaging 88W+ is far beyond that 8%. That's the problem. The ASUS Strix 960 never exceeds the 75W average threshold.

I have no doubt the majority of motherboards could handle that kind of draw, likely even slightly more as I'm sure even the 8% overage is also on the conservative side but one should always stick within compliance limits. Sure as hell don't get away with flouting proper electrical safety standards. If the RX 480 was any other kind of electrical device it would have failed safety tests and be banned from sale until the issue is rectified, and none of use would have even known.

 

I actually still have an AMD 4670 passively cooled card which with the factory OC it came with exceeds the 75W, had it running in a HTPC for about 5 years before I retired the system for a better solution. GPU load while watching HDTV channels was pretty much 100%. Was an old 775 system with a E8400.

 

Clearly a bit of a design fail on AMD's side but nothing that I'm going to scream bloody murder about or try and return a working part for no reason, not until more evidence is shown to confirm that these cards will damage motherboards.

 

Not an RX 480 owner in case anyone is wondering.

 

Rather disappointing as I was hoping that even though the RX 480 wasn't going to break any performance records or actually be all that interesting that this would at least be a smooth launch with good sales and even better long term sales of custom design boards. AMD flush with cash, high sales numbers and market share is good for everyone.

 

I don't get why people have such strong ties to a brand, pick your budget find the highest performing card at that price point and buy it. Who cares who's logo is on it.