Discussion: Polaris, AMD's 4th Gen GCN Architecture

[Turb0Yoda]

Max PCI wattage is still 75 watts right?

 

[-Fran-]

Max PCI wattage is still 75 watts right?

Be a bit more pedantic/precise and use the little "e" at the end: PCIe. Remember there *is* another standard (old now) called PCI. Yes, I am being pedantic, but it's for the better 😛

And yes, the spec says "75W of power draw through PCIe". This is not an explicit quote, of course and it is a bit more complicated, but that is the geist of it.

Cheers!

 

[Turb0Yoda]

Max PCI wattage is still 75 watts right?

Be a bit more pedantic/precise and use the little "e" at the end: PCIe. Remember there *is* another standard (old now) called PCI. Yes, I am being pedantic, but it's for the better 😛

And yes, the spec says "75W of power draw through PCIe". This is not an explicit quote, of course and it is a bit more complicated, but that is the geist of it.

Cheers!

Ergh, I forgot that... my bad...

 

[ern88]

A RX 480 and a cheap OEM motherboard=RIP motherboard!!!

 

[Rogue Leader]

Ok so I just tried another test again.

Something is definitely going on on the PCIe bus. I am running a Sound Blaster X-Fi Titanium HD sound card, to a pair of M-Audio Monitors, the whole setup is pretty sensitive, however I have never heard any sort of feedback before using my old cards.

So I ran 3DMark again after making some tweaks and I hear noise in my speakers if I listen up closely during each of the tests.

My computer at idle draws 156w from the wall, it spikes to 400 during the tests, now I know my processor can account for part of that, but even 50w more thats still something else drawing 200w. 99% of the test its round 350w.

 

[-Fran-]

Ok so I just tried another test again.

Something is definitely going on on the PCIe bus. I am running a Sound Blaster X-Fi Titanium HD sound card, to a pair of M-Audio Monitors, the whole setup is pretty sensitive, however I have never heard any sort of feedback before using my old cards.

So I ran 3DMark again after making some tweaks and I hear noise in my speakers if I listen up closely during each of the tests.

My computer at idle draws 156w from the wall, it spikes to 400 during the tests, now I know my processor can account for part of that, but even 50w more thats still something else drawing 200w. 99% of the test its round 350w.

Well, for the safety of your MoBo, remove the card and go back to the 280s. At least, for the time being.

Get in touch with Microcenter and see if you can get a refund before this becomes more widespread and they start putting barriers.

Cheers!

 

[BurgerandChips66]

What does GPU-Z say regarding power consumption of the graphics card during the test? As in Max power?

I read somewhere once that excessive power draw through any PCIe slot can lead to an audible sound through the speakers - if the sound card is also using a PCIe slot. Anyone else read this? I cannot find a reference atm.

 

[-Fran-]

You can usually hear any noise or ripple using speakers (self powered speakers or studio monitors) that amplify the analog audio signal that goes through the MoBo DAC. The higher quality the speakers, the most noticeable it becomes.

I can attest to that, because I have 2 of them next to my TV and are sensible as hell to signal noise.

Cheers!

 

[Rogue Leader]

Downloaded DOOM demo, ran between 60 and 130 fps with the average in the 70ish range on Ultra with everything turned up. Smooth as silk. I should have tried with my old setup, but even my "ancient" system has no problems with it using the RX. Very happy with it

Was this with the Nightmare shadow setting and Nightmare Virtual Texturing Page Size setting?

Wouldnt let me turn them on, says required 5gb or more VRAM. I got the 4gb version

What are your temps and % GPU usage, if I my ask?

In DOOM I just tried again at ultra and it hits 100% GPU useage, 95-98% processor useage. I did nto see one slowdown or hiccup during the game. Max temp was 81 on the GPU.

 

[Rogue Leader]

Ok so I just tried another test again.

Something is definitely going on on the PCIe bus. I am running a Sound Blaster X-Fi Titanium HD sound card, to a pair of M-Audio Monitors, the whole setup is pretty sensitive, however I have never heard any sort of feedback before using my old cards.

So I ran 3DMark again after making some tweaks and I hear noise in my speakers if I listen up closely during each of the tests.

My computer at idle draws 156w from the wall, it spikes to 400 during the tests, now I know my processor can account for part of that, but even 50w more thats still something else drawing 200w. 99% of the test its round 350w.

Well, for the safety of your MoBo, remove the card and go back to the 280s. At least, for the time being.

Get in touch with Microcenter and see if you can get a refund before this becomes more widespread and they start putting barriers.

Cheers!

I actually don't care if it eats my motherboard, its just an excuse to upgrade, as you can see my system is old. I was waiting fro Zen but whatever. Also in theory I can blame it on the card if AMD starts handing out settlements. Probably more than Ic an get selling it afterwards lol.

You can usually hear any noise or ripple using speakers (self powered speakers or studio monitors) that amplify the analog audio signal that goes through the MoBo DAC. The higher quality the speakers, the most noticeable it becomes.

I can attest to that, because I have 2 of them next to my TV and are sensible as hell to signal noise.

Cheers!

I'm using a PCIe soundcard and I hear the GPU as I mentioned, I'm sure its probably worse with onboard sound.

 

[BurgerandChips66]

Are you able to try under-volting it? Does the OC software allow that? Might be interesting to see what effect it has.

 

[Mousemonkey]

Are you able to try under-volting it? Does the OC software allow that? Might be interesting to see what effect it has.

Why should one need to under volt a card? That alone would suggest that it doesn't do what it says on the box and is therefore pants does it not?

 

[Ags1]

I don't want to buy into the fear hype but I'm thinking of returning the RX480 unopened when it arrives and waiting for the custom cards. Also worth waiting for the 1060 to push the RX480 prices down a bit. The rumors are the 1060 comes in mid July with 15% more performance than the RX480 according to a leaked nVidia slide (which means its slower than the RX by a long way in DX12 ).

 

[Bem-xxx]

The RX480 doesn't burn motherboards. Wccftech deleted the article.
Nvidia's astroturfers said that Crimson drivers burned GPUs. False.
Nvidia's astroturfers said that the RX480 burns mobos. False.
The 960 has higher peaks than the 480.
Nvidia's cheating benchmark since 2003.
http://www.geek.com/games/futuremark-confirms-nvidia-is-cheating-in-benchmark-553361/
[video="https://www.youtube.com/watch?v=kFuYc2FHgjw&feature=youtu.be"]https://www.youtube.com/watch?v=kFuYc2FHgjw&feature=youtu.be[/video]

 

[Rogue Leader]

Are you able to try under-volting it? Does the OC software allow that? Might be interesting to see what effect it has.

I am not sure if undervolting it will help much, however you can drop the power limit to negative. That would probably choke the card back and prevent it from drawing so much.

I really don't care to try it, at that point I would just return it. As I mentioned if it eats my motherboard its on them, even if they don't I got my moneys worth out of the system I don't really care I'd just return the card and thats that.

 

[Baralis]

AdoredTV did a video that shows this issue is not limited to the RX480 but multiple other cards including some by the green team. Is this problem being blown out of proportion?

https://youtu.be/kFuYc2FHgjw

 

[Bem-xxx]

I don't want to buy into the fear hype but I'm thinking of returning the RX480 unopened when it arrives and waiting for the custom cards. Also worth waiting for the 1060 to push the RX480 prices down a bit. The rumors are the 1060 comes in mid July with 15% more performance than the RX480 according to a leaked nVidia slide (which means its slower than the RX by a long way in DX12 ).

Are you able to try under-volting it? Does the OC software allow that? Might be interesting to see what effect it has.

I am not sure if undervolting it will help much, however you can drop the power limit to negative. That would probably choke the card back and prevent it from drawing so much.

I really don't care to try it, at that point I would just return it. As I mentioned if it eats my motherboard its on them, even if they don't I got my moneys worth out of the system I don't really care I'd just return the card and thats that.

It's a fake advertising sponsored by Nvidia.
The 960 has higher peaks than the 480.

 

[turkey3_scratch]

Well this is certainly not good. This is why I don't jump onto cards too quickly. I am curious, could somebody explain precisely where the power comes from on the PCIe slot? So you have 16 pins used for transferring data, but where in that slot does the power for the GPU come from (in addition to the 6-pin PCIe cable)? And also, if the motherboard VRMs do not work with the GPU (since the GPU has its own VRM), how precisely does this affect things like audio on the motherboard?

Just wondering for educational purposes, that's all 😛

 

[jimmysmitty]

They need to get those reference PCBs re-designed asap!

I've seen a bit of nonsense about this too. Somebody claiming their motherboard was fried after a 7 hour gaming session. They claimed they had overclocked the card to 1350Mhz with no over-volting (?) and then refused to offer up any evidence when pressed on the subject.

I recall setting my Sapphire Nitro R9 380X to 1200MHz core (+40mV) with the power limit at 120% when running the likes of Valley Benchmark at 1080p Ultra. It never seemed to pull more than 125W according to Trixx. Quite a big overclock (+ 160MHz from the factory overclock of 1040MHz) and it still didn't go anywhere near the 190W TDP.

Some of these RX 480s seem to be pulling much more than their TDP - even at stock frequencies. :-/

If the custom cards solve these issues I'm still happy to get one though to replace a 750Ti in my older rig. Although now the GTX 1060 is rumoured to be releasing in July.... I might have to wait AGAIN for more reviews.

All this waiting is making me crazy 😵

http://www.game-debate.com/news/20679/250-geforce-gtx-1060-could-actually-be-as-powerful-as-gtx-980-according-to-new-slides

New process means they might not need as much power as previous to get higher clock speeds. More testing will be needed.

Max PCI wattage is still 75 watts right?

Actually max is technically 66w (12v @ 5.5A) but there is enough wiggle that 75W is probably fine.

The RX480 doesn't burn motherboards. Wccftech deleted the article.
Nvidia's astroturfers said that Crimson drivers burned GPUs. False.
Nvidia's astroturfers said that the RX480 burns mobos. False.
The 960 has higher peaks than the 480.
Nvidia's cheating benchmark since 2003.
http://www.geek.com/games/futuremark-confirms-nvidia-is-cheating-in-benchmark-553361/
[video="https://www.youtube.com/watch?v=kFuYc2FHgjw&feature=youtu.be"]https://www.youtube.com/watch?v=kFuYc2FHgjw&feature=youtu.be[/video]

WCCFTech is a meh source. There are more reports coming up from other sites.

The Crimson drivers did initially have an issue where the fan speed curve would not go above 20% which could cause damage if the GPU was not cooling properly. It was fixed in a fast beta release though. On the other side of the coin, nVidia also had a driver that cause GPUs to run too hot.

Again there are multiple reports of board no longer working. Are you saying that all those people are lying? Even the ones with AMD 990FX boards and FX 8000 series CPUs?

Higher peaks, which are not the issue. The issue, if you and the guy in the video actually looked at the averages, is that it is doing this over a longer period of time and averaging higher pulls consistently. That is also only ONE GPU, he is looking at the review of multiple GTX 960s and the Asus Strix for some reason is spiking that high.

Of course using one GPU out of a review of multiple that has bad power spikes from the PCIe slot proves nothing and is a strawman argument.

 

[Ags1]

Yes but the transient peaks are apparently not an issue - the issue is the average - 80 for the RX480 and 55 for the 960. I really hope Toms does an in depth article on this issue looking at other cards likely to push the PCIe slot limits (Radeon 7750, GTX 750Ti and GTX960 are most like culprits in my opinion).

Mailed the store to canel my RX480 order. I'm hoping for the custom boards though. Really plan to buy AMD but I was a fool to buy AMD on day zero.

 

[Mousemonkey]

AdoredTV did a video that shows this issue is not limited to the RX480 but multiple other cards including some by the green team. Is this problem being blown out of proportion?

https://youtu.be/kFuYc2FHgjw

I don't want to buy into the fear hype but I'm thinking of returning the RX480 unopened when it arrives and waiting for the custom cards. Also worth waiting for the 1060 to push the RX480 prices down a bit. The rumors are the 1060 comes in mid July with 15% more performance than the RX480 according to a leaked nVidia slide (which means its slower than the RX by a long way in DX12 ).

Are you able to try under-volting it? Does the OC software allow that? Might be interesting to see what effect it has.

I am not sure if undervolting it will help much, however you can drop the power limit to negative. That would probably choke the card back and prevent it from drawing so much.

I really don't care to try it, at that point I would just return it. As I mentioned if it eats my motherboard its on them, even if they don't I got my moneys worth out of the system I don't really care I'd just return the card and thats that.

It's a fake advertising sponsored by Nvidia.
The 960 has higher peaks than the 480.

Your attempts to hide facts are futile, the issue appears to be real and Nvidia have no part in it from what I and others can see. And why are AMD having posts deleted from Reddit? And why have they not made any moves to to dispute any of these claims that you say Nvidia are behind?

 

[turkey3_scratch]

It's safe to say it is indeed a problem that should be addressed by both AMD and board partners before their own cards come out. Even Rogue Leader says he can notice through the speakers. Though with his optimism nothing seems to be able to disappoint him!

This should have been something AMD would have noticed long before these cards were released. It only took Tomshardware one time to test the cards to find this pretty serious problem. PCIe cable rated wattage may be underrated, but it seems the PCIe bus has its rating pretty firmly true.

 

[jimmysmitty]

AdoredTV did a video that shows this issue is not limited to the RX480 but multiple other cards including some by the green team. Is this problem being blown out of proportion?

https://youtu.be/kFuYc2FHgjw

Did you read that GTX 960 review? I don't think you did because if you did you would see that was a review of multiple AiB cards and that only the Asus GTX 960 Strix had that spiking issue, none of the other ones did.

 

[Dark Lord of Tech]

https://community.amd.com/thread/202410
PCI-E slot died with RX 480

Starting to see more of these reports.

 

[Bem-xxx]

The RX 480 is selling a lot. Multiple reports done by who? One is a troll who has been banned. If the issue was real, would be a hundred of reports, not two or three made by the same people with different nicknames.