AMD RX 480 Fails PCI-E Specification

[southernorth]

if you ever want to hear the truth about a product,good or bad..listen to someone who doesn't have a vested interest in that product.

whenever I hear in 1 ear that there are no problems by a salesman,and I hear the opposite in the other ear from an independent party..i know who I believe.

What do you think about all the speculation on why the card behaves how it does? intrinsic design error, or late-stage increases in voltage to ensure performance was in line with the marketing (to a point the card wasn't designed for)

 

[thebennyboy]

What do you think about all the speculation on why the card behaves how it does? intrinsic design error, or late-stage increases in voltage to ensure performance was in line with the marketing (to a point the card wasn't designed for)

The only change i would make to the card would be an 8-pin instead of a 6-pin. With overclocking taken into account they should've had a 8-pin on it.

 

[sqwerty]

The only change i would make to the card would be an 8-pin instead of a 6-pin. With overclocking taken into account they should've had a 8-pin on it.

From that stream with buildzoid the 6 pin is already wired as an 8 pin less the two additional grounds so it's capable of 150w+.

Roman at caseking managed to pull 260w after modding the voltage controller to push 1.35v through naked mosfets! I wonder how much of that 260 was through the slot

https://www.facebook.com/caseking/photos/pcb.10154264173047230/10154264169147230/?type=3&theater

 

[Gerard]

So its an 8 pin in disguise essentially? Wtf are amd playing at?

 

[Rroff]

So its an 8 pin in disguise essentially? Wtf are amd playing at?

One of the reasons it is done that way is so that the GPU can't start up without the 6 pin present due to the lack of normal power distribution hardware. Wish I knew a bit more about it (I have some moderate knowledge of audio power supply design but little working knowledge GPU wise) as its truly perplexing to me what they are playing at there.

 

[sqwerty]

So its an 8 pin in disguise essentially? Wtf are amd playing at?

Yep, usually the centre pin is there as sensor only but this has all three grounds tied.

 

[AlamoX]

So its an 8 pin in disguise essentially? Wtf are amd playing at?

it's probably due to how far they pushed the clock, they probably designed it for lower clock, then when the hype started about it's performance they needed to push it further to not underwhelm ppl performance wise, which i think was a mistake, so you ended up with a card pushed to the maximum of it's cooler ability and power envelope, and less time to tune all this out correctly, they should have left it at 1080mhz base and 1233mhz boost, stay at 970 perf and call it a day, would have been better than grabing those extra 10% performance.

 

[ihatelag]

What do you think about all the speculation on why the card behaves how it does? intrinsic design error, or late-stage increases in voltage to ensure performance was in line with the marketing (to a point the card wasn't designed for)

If they were intending the card to be used on a PCI-E slot, they would need to go through a compliance testing regime that would make it able to get the PCI-E logo for their packaging/boxes. What I don't understand is what happened during that compliance testing for the card that would make them miss the issue? Did AMD skip testing? Was this intentional? For such a large company, you would expect far better.

Shame that AMD don't get a break, we really need them to have some success and provide competition to nVidia, we are getting slammed pretty hard in our wallets by nVidia and I was hoping AMD would successfully make more value oriented cards, forcing the hand of nVidia to reduce prices (starting at the mid to low end).

 

[Yrkoon]

wonder what they can fix with this without a recall, they say driver fix?

been staying off the witcher 3 and project cars just incase and playing dark souls 2 scholar of first sin all night.

don't wanna push the card much untill i see the fix later so i have been going easy on it lol

oh hdmi scaling is broken on crimson, worked perfect on catalyst with my old 270x to tv, any use of hdmi scaling on crimson on the 480 gives massive display lag.

got around that by just telling my tv to fit images to screen on hdmi 2.

i really like this card but this problem has put a dampner on it, hope the amd fix is good then i can go and enjoy my new gpu

 

[southernorth]

If they were intending the card to be used on a PCI-E slot, they would need to go through a compliance testing regime that would make it able to get the PCI-E logo for their packaging/boxes. What I don't understand is what happened during that compliance testing for the card that would make them miss the issue? Did AMD skip testing? Was this intentional? For such a large company, you would expect far better.

Shame that AMD don't get a break, we really need them to have some success and provide competition to nVidia, we are getting slammed pretty hard in our wallets by nVidia and I was hoping AMD would successfully make more value oriented cards, forcing the hand of nVidia to reduce prices (starting at the mid to low end).

It truly is a shame. I usually buy nvidia cards, but thought I would try AMD just this once. As a result of the issues associated with the 480, I regretfully don't trust them as a company anymore, and may just stick with my GTX 660. But who knows, their fix tomorrow might be amazing, but I'm not holding much hope to be honest.

 

[sqwerty]

it's probably due to how far they pushed the clock, they probably designed it for lower clock, then when the hype started about it's performance they needed to push it further to not underwhelm ppl performance wise, which i think was a mistake, so you ended up with a card pushed to the maximum of it's cooler ability and power envelope, and less time to tune all this out correctly, they should have left it at 1080mhz base and 1233mhz boost, stay at 970 perf and call it a day, would have been better than grabing those extra 10% performance.

The problem is that they didn't even make proper use of this potential additional power from the plug as the card is sharing power almost evenly from it's two sources.

Why break spec to make your 6 pin "8 like" in the first place if power was to be shared in such fashion?

Really strange design.

 

[Griffildur]

If they were intending the card to be used on a PCI-E slot, they would need to go through a compliance testing regime that would make it able to get the PCI-E logo for their packaging/boxes. What I don't understand is what happened during that compliance testing for the card that would make them miss the issue? Did AMD skip testing? Was this intentional? For such a large company, you would expect far better.

Shame that AMD don't get a break, we really need them to have some success and provide competition to nVidia, we are getting slammed pretty hard in our wallets by nVidia and I was hoping AMD would successfully make more value oriented cards, forcing the hand of nVidia to reduce prices (starting at the mid to low end).

They are not helping themselves much unfortunately. Self created issue, 100%

 

[AlamoX]

It truly is a shame. I usually buy nvidia cards, but thought I would try AMD just this once. As a result of the issues associated with the 480, I regretfully don't trust them as a company anymore, and may just stick with my GTX 660. But who knows, their fix tomorrow might be amazing, but I'm not holding much hope to be honest.

the pci-e isn't really that much of an issue, it shouldn't have happened, but it's not going to destroy motherboards like ppl claim, the fix will probably be a redistribution of load to the 6pin instead of pci-e, that should help shut up the media about it.
but the value of the card is still great, untill we see what the 1060 will bring and at what price point.

 

[Yrkoon]

They are not helping themselves much unfortunately. Self created issue, 100%

true i bought one because it was awesome performance for a cheapish price, but im thinking about whether i should send it back as it could be a potential bomb in my newish comp

guess ill see what this fix does, i just do not want my new mainboard to start degrading because of a gpu.

plus if it is a 8 pin in disguise why was the power load not more on the pins and not the socket, maybe thats the fix? can that be done via drivers or only a bios update?

hope this issue gets resolved later and i can go back to enjoying an awesome new card

edit:
220 is the most i have ever spent on a gpu only other one the same price was the tnt2 ultra lol

 

[southernorth]

the pci-e isn't really that much of an issue, it shouldn't have happened, but it's not going to destroy motherboards like ppl claim, the fix will probably be a redistribution of load to the 6pin instead of pci-e, that should help shut up the media about it.
but the value of the card is still great, untill we see what the 1060 will bring and at what price point.

Very true, it could all be exaggerated. One concern is that, over-time, the GPU does indeed damage motherboards. This long-term damage won't emerge immediately, and I think perhaps lulls people into a false sense of safety. Another thing is the massive headache this might cause for those selling pre-built pc's with rx 480s. These pre-builts usually don't come with with the best quality motherboards and psu, and so perhaps increases and speeds up damage.

 

[sqwerty]

true i bought one because it was awesome performance for a cheapish price, but im thinking about whether i should send it back as it could be a potential bomb in my newish comp

Nah, it's not really until you start OC'ing them and/or pushing unrealistic workloads (cough 4k Ultra Metro cough) that they are showing over spec.

Besides many are finding them to not only sup less juice undervolted but actually perform better whilst running quieter...http://www.legitreviews.com/amd-radeon-rx-480-undervolting-performance_183699

 

[Yrkoon]

Very true, it could all be exaggerated. One concern is that, over-time, the GPU does indeed damage motherboards. This long-term damage won't emerge immediately, and I think perhaps lulls people into a false sense of safety. Another thing is the massive headache this might cause for those selling pre-built pc's with rx 480s. These pre-builts usually don't come with with the best quality motherboards and psu, and so perhaps increases and speeds up damage.

yup a potential bomb untill later i hope

i tend to hang on to my pc's for a long time nowadays only recently went to a 6600k i5 when my old mother board would be 7 year old this august, so i want my mb to last a good while

 

[Rroff]

Very true, it could all be exaggerated. One concern is that, over-time, the GPU does indeed damage motherboards. This long-term damage won't emerge immediately, and I think perhaps lulls people into a false sense of safety. Another thing is the massive headache this might cause for those selling pre-built pc's with rx 480s. These pre-builts usually don't come with with the best quality motherboards and psu, and so perhaps increases and speeds up damage.

The voltage controller has the capability of balancing the phases but whether that functionality can easily be utilised on cards already released is another matter - in der8auer's video they soldered in an extra card to fully access the IC via I2C and I think MTP configuration requires external hardware as well.

 

[southernorth]

The voltage controller has the capability of balancing the phases but whether that functionality can easily be utilised on cards already released is another matter - in der8auer's video they soldered in an extra card to fully access the IC via I2C and I think MTP configuration requires external hardware as well.

You seem rather knowledgeable on the topic, do you work in the pc hardware industry/electrical engineering?

 

[Rroff]

You seem rather knowledgeable on the topic, do you work in the pc hardware industry/electrical engineering?

I do a bit of audio design for my own amusement - amps, DACs, etc. some principles are shared when it comes to electronics but I lack working knowledge on the GPU side.