RX 480 PCIE Problem

[Curufinwe_wins]

4 hours ago, awesomeness10120 said:

Someone on Pcper brought up a great point. The GTX 960 had the same problem, but no talked about that: http://www.tomshardware.com/reviews/nvidia-geforce-gtx-960,4038-8.html

 

Only the strix. And it was redone. 

 

Plus this according to Tom's Hardware and others is much worse than both the 750ti and 960 (averaging as bad as 80W stock,  100W oc through pcie with spikes another 100W higher) 

 

Also the 750ti is much less of a concern since it can't be used in a dual gpu solution (not that I'd recommend doing it with the 480, but one gpu overspec is much different than running 2-4 over like some of the people hoping to use the 480 as a compute card or CF). 

 

BTW this entire entry is a repost and should be merged with the other 480 pcie topic. 

[byalexandr]

8 hours ago, Kloaked said:

75 watts is the maximum a graphics card should be drawing from a PCIe slot. 

66W on the 12V rail. 75W across all the voltage rails but the card runs off of 12V.

[Hunter259]

9 hours ago, byalexandr said:

Basically the 750 Ti had this issue but has it far worse. The RX 480 'potentially' exceeds PCI-E compliance while the 750 Ti in most cases did. By far. Like it says in that podcast, the 12V rail supplies 66W for the PCI-E power. The 750 Ti, without any sort of overclock (even without a boost clock), draws up to 70W at peak load, around 65W at gaming load, again with no OC.

 

Now to me that seems a bit ridiculous when you compare it to this RX 480 situation. The 480 'might' exceed it, AMD is still testing, but the 750 Ti clearly does, sold a lot more, and is a lot older. And no one even batted an eye, because they have nothing to hate on because it's not from AMD and everyone has to hate AMD for some reason.

I want you to prove this. Not the milisecond spikes that people want to use as "Proof". Small spikes are fine but prove to me you can find a reviewer that had a 750ti use more than 75w.

[IGJoe2192]

Saying other cards have had this issue and the world didn't blow up doesn't change the fact the RX480 has shown to have this higher sustained power draw. I purchased two for crossfire and currently I am waiting for them to arrive, if CF didn't work out my fallback plan was to put the 2nd card into a mini-ATX build that I plan on putting together for my living room. I am now worried about doing both as CF would definitely tax the PCIe bus and I am unsure the mini-ATX mobo I have has quality components that can withstand this sort of load. I am waiting to see how AMD handles this issue i.e. a recall or some other form of action or non-action. Whatever it is causing this issue, I hope for their sake that it is taken care of swiftly. 

[byalexandr]

11 minutes ago, Hunter259 said:

I want you to prove this. Not the milisecond spikes that people want to use as "Proof". Small spikes are fine but prove to me you can find a reviewer that had a 750ti use more than 75w.

66W.... Not 75W. ATX allows a maximum of 66W on the 12V rail, the rail that the card uses. All of the other rails combined allows for 75W.

 

http://www.tomshardware.com/reviews/geforce-gtx-750-ti-review,3750-21.html

 

You can see here in that article that the 'gaming' load is 62W, and the peak load is 68W. That's without any sort of overclock. And the card doesn't have any additional PCI-E power connections so overclocking in the slightest would exponentially increase the chances of breaking your mobo. Not to mention that the card naturally boosts on it's own with GPU Boost.

[Misanthrope]

9 hours ago, byalexandr said:

Basically the 750 Ti had this issue but has it far worse. The RX 480 'potentially' exceeds PCI-E compliance while the 750 Ti in most cases did. By far. Like it says in that podcast, the 12V rail supplies 66W for the PCI-E power. The 750 Ti, without any sort of overclock (even without a boost clock), draws up to 70W at peak load, around 65W at gaming load, again with no OC.

 

Now to me that seems a bit ridiculous when you compare it to this RX 480 situation. The 480 'might' exceed it, AMD is still testing, but the 750 Ti clearly does, sold a lot more, and is a lot older. And no one even batted an eye, because they have nothing to hate on because it's not from AMD and everyone has to hate AMD for some reason.

Initial testing shows a gaming 480 consistently over 75 is almost all the time above 75 watts.

 

I concede that the 750ti probably means that the problem is overstated and it won't immediately go kaboom! on your boards (except if you decide to do maybe overclocked crossfired 480s) so that's a relevant point. But the issue seems to be worst not better for the 480.

[AnnoyedShelf]

I'm so sick and tired of seeing this issue blown way out of proportion.  The 750Ti was way worse, and a good mobo can withstand up to 300W on the slot.  End of story.  Quit your bitching.

[Hunter259]

5 minutes ago, byalexandr said:

66W.... Not 75W. ATX allows a maximum of 66W on the 12V rail, the rail that the card uses. All of the other rails combined allows for 75W.

 

http://www.tomshardware.com/reviews/geforce-gtx-750-ti-review,3750-21.html

 

You can see here in that article that the 'gaming' load is 62W, and the peak load is 68W. That's without any sort of overclock. And the card doesn't have any additional PCI-E power connections so overclocking in the slightest would exponentially increase the chances of breaking your mobo. Not to mention that the card naturally boosts on it's own with GPU Boost.

Since when? It's a max 75W. The 750ti did not go above it. 480 is always above it.

[Hunter259]

3 minutes ago, AnnoyedShelf said:

I'm so sick and tired of seeing this issue blown way out of proportion.  The 750Ti was way worse, and a good mobo can withstand up to 300W on the slot.  End of story.  Quit your bitching.

300W? Are you crazy? No a good mobo can't do that. Those poor traces would die very quickly.

[awesomeness10120]

9 hours ago, rattacko123 said:

 

it's probably fine if you run 1 or two of these graphics cards, as it doesn't exceed the maximum possible power output of the entire motherboard. If you fill up all the pcie slots with more power hungry devices (such as more GPUs) this could potentially be a problem. Besides, people who buy this graphics card wont be using that many pcie cards in the first place, so it is most likely safe.

Wouldn't it be a bigger problem for whoever makes AMD's reference cards, unless of course AMD manufactures the cards?

[Remixt]

13 hours ago, DarkBlade2117 said:

As much as I hate this ego-tistic bastard, he has a good point in this vid.

Again, about the 750Ti issue. I don't think it is as big as a deal as everyone says. I had my 750Ti OCd quite high probably pushing it near the 100w mark and ran it on a 350w HP PC and a shitty HP mobo. Yeah, it is bad but almost for certain AIB will make it a 8 pin connector and I am 99% sure that 6+2 pins can supply 150w of power.

What about Jay is ego-tistic? He bashes himself all the time

[byalexandr]

4 hours ago, Hunter259 said:

Since when? It's a max 75W. The 750ti did not go above it. 480 is always above it.

I literally explained it in my post, 66W. Combined voltages and you get 75W but the 12V rail only supplies 66W.

[Misanthrope]

4 hours ago, AnnoyedShelf said:

I'm so sick and tired of seeing this issue blown way out of proportion.  The 750Ti was way worse, and a good mobo can withstand up to 300W on the slot.  End of story.  Quit your bitching.

Demonstrably wrong on both accounts: The 480 draws way more power way more consistently above the limit, how is the 750ti "worst" then?

 

Also 300w was already debunked on the big thread.

[AnnoyedShelf]

1 minute ago, Misanthrope said:

Demonstrably wrong on both accounts: The 480 draws way more power way more consistently above the limit, how is the 750ti "worst" then?

 

Also 300w was already debunked on the big thread.

The other thing is, only a couple hundred cards are affected by this.  ONLY 2 of 20 reviewers experienced the problem at all.  I just feel that people are blowing this way out of proportion.

[Misanthrope]

4 minutes ago, AnnoyedShelf said:

The other thing is, only a couple hundred cards are affected by this.  ONLY 2 of 20 reviewers experienced the problem at all.  I just feel that people are blowing this way out of proportion.

Only 2 reviewers measured it at first, but it has been confirmed and replicated by all, it just didn't blow up their motherboards. I can concede that most motherboards are able to withstand it but again most reviewers have beefy mobos on their test benches and consumers could potentially have crap boards or try crossfire.

 

So it's not an "omg people will die in fires caused by the 480!" problem but I say is akin to the 3.5 vram back in the day: doesn't affects too many people but it's there and "instert Company" flatout lied by omission about it.

[KillDog]

It won't affect people with mobos having PCI-E 2.0 or higher.

Just when AMD were heading into the light, a small hurdle crossed them. Hope this hurdle doesn't bash them into darkness.

[Hunter259]

18 minutes ago, byalexandr said:

I literally explained it in my post, 66W. Combined voltages and you get 75W but the 12V rail only supplies 66W.

I read it wrong my bad but that still means the 750ti is within spec.

[Hunter259]

4 minutes ago, KillDog said:

It won't affect people with mobos having PCI-E 2.0 or higher.

Just when AMD were heading into the light, a small hurdle crossed them.

 

[grayperview]

5 minutes ago, Hunter259 said:

 

I watched this video too, seems the motherboard he had is a pretty old cheapo foxconn board....more research is needed into this, and AMD need to come out and clarify the situation. even if its a recall, potentially this could be bad of AMD, either way i suspect they are working on a revision 2.0 of this GPU already. 

[ttam]

I don't know where this kids googling his info at but most pci-e slots have been able to safely put out 75w for YEARS now. Any card drawing 75w and under is safe, no if's ands or buts about it.

 

Also brought the 480 into the shop with me earlier and was finding 80w spikes. Nothing scary, but it's quick random bursts. OCing is showing 85w, which is a bit of a concern. AMD designed the card to only draw 55ish from the motherboard, so IMO a recall would be best to move the card over to an 8pin power.

 

Edit: Source: A very nice Asus rep

[Hunter259]

1 minute ago, super_skank said:

I watched this video too, seems the motherboard he had is a pretty old cheapo foxconn board....more research is needed into this, and AMD need to come out and clarify the situation. even if its a recall, potentially this could be bad of AMD, either way i suspect they are working on a revision 2.0 of this GPU already. 

Well it is new enough it has PCIE 2 so his argument is invalid.

[AceTaker]

This whole problem reminds me of the  Xbox 360's "red ring of death".. weird...

[Doobeedoo]

Really don't think moderate OC specially non OC will have problems. Though hopefully we see test on some super OC card and maybe on cheap mobo.
Also custom ones are on their way, so that's great