@ mr_bigmouth_502

Science Studio claims to have used an older AM2 board and the machine shutting itself off. He didn't show a video of this though, that would have made it even more dramatic.

Newer boards are simply more tolerant I guess. Reminds of the AGP power draw issues with V3 cards. Although here it was the motherboards that cut corners.

A driver update is coming, but can you put a current limit on the PCIe slot through a driver? AFAIK what cards do, when you OC them even, is hard limit the PCIe power draw and get anything extra from the 6 pin or 8 pin plug.

Any way, we will know more soon. AMD can be sure that reviewers will re-test this with eagle eyes, so they better fix it properly.

Still, it's huge stuff up. For a chip that was hyped for its energy efficiency to come out and break specifications? Not a good thing 🙁

And of course they knew about the issue! A reviewer even said that during a press talk, AMD straight up said that 150W is not the maximum, but the typical load you will see while playing games.

Still, a great bang for you buck card, if they can sort out the driver. The RX470 might be the safer card to go for...

I haven't bought a non-Nvida graphics card since the Radeon 9800Pro, and this card is doing nothing to change my mind. Pretty sad, I was actually hopeful with the 480. I'll be interested in how AMD addresses this situation and how the 470 fares not only in performance, but power tests.

PhilsComputerLab wrote:

And of course they knew about the issue! A reviewer even said that during a press talk, AMD straight up said that 150W is not the maximum, but the typical load you will see while playing games.

The thing is, if it was pulling the extra power from the 6-pin connector instead of PCIe, it would probably be fine.
It would be exceeding the spec, but most PSUs are over-engineered and will handle more than 75W over the 6-pin connector just fine.
The card really should have had an 8-pin connector (150W) instead of a 6-pin (75W) though, which would have solved all these issues.

There was another followup article yesterday:
http://wccftech.com/article/radeon-rx-480-red … ing-efficiency/

Undervolted RX480 4GB version, consumes about 112W, reaching better clocks too. Optimizing for efficiency might be the answer for RX480. Launch first, optimize later... 😒

At 150W target, RX480 is within the same power envelope as stock 1070--which is going to be the all round perf/watt winner and price/perf winner if prices are coming down to msrp (it already is starting, I'm seeing $450~460 compared to 500+ last week). RX480 should stay strictly within the ~$200 pricepoint.

Deep Thought wrote:

The thing is, if it was pulling the extra power from the 6-pin connector instead of PCIe, it would probably be fine.

Yup, it probably would not have been a big deal... However, now that AMD was 'caught' with going out-of-spec on the slot, they won't be able to get away with going out-of-spec on the 6-pin connector anymore either, because everyone is watching AMD's next move very closely now. They need to make sure the card stays under 75W draw on both connectors.

I think it's also sad that some people started to draw comparisons with some heavily overclocked nVidia cards (eg Asus Strix GTX960), and say: "Look, nVidia does it too!": https://www.pcper.com/reviews/Graphics-Cards/ … S-GTX-960-Strix
Well no, Asus is doing that, and the Strix branding is a pretty damn good warning that this is not a part that operates under the original specs.
The stock version of the card (the one nVidia designed), does not draw as much. And we have stock RX480s here that do. That is the issue.
Besides, the average draw of the Strix on the motherborad was still way below spec. It only had pretty extreme spikes. It drew slightly too much from the external connector only.

I think that the 480 has the potential to be like Chrysler's K platform.

Was it the biggest, fastest, or most powerful? No. It was never intended to be.
Was it affordable? Yes.
Did it offer good enough performance for most use cases? Yes.
Was it extremely boring? Abolutely.

The entire point of the 480 is to offer, at $200, a significant upgrade over other $200 cards. The fact that the 480 can sometimes trade blows with the 390 and 970 is just a nice bonus.

All things considered, I'll probably get one. It's about twice as fast as my 7870 for roughly the same price new.

Extreme TL;DR for the K platform:
By the late '70s, Chrysler was facing bankruptcy due to stiff foreign competition and a rapid shift in consumer interest from expensive, high margin luxury cars to cheap econo-boxes. In order to compete, they developed the K platform, which was mediocre in pretty much every concievable way. The silver lining was that it was cheap to make and met all of the new car market's needs well enough. By finding a sweet spot in performance and cost, Chrysler effectively saved itself.

I don't really expect the 480 to be some magic bullet that completely reverses AMD's position in the GPU market. However, I think that the 480 will sell fairly well once AIB cards start coming out.

meisterister wrote:

I think that the 480 has the potential to be like Chrysler's K platform. […]
Show full quote

I think that the 480 has the potential to be like Chrysler's K platform.

Was it the biggest, fastest, or most powerful? No. It was never intended to be.
Was it affordable? Yes.
Did it offer good enough performance for most use cases? Yes.
Was it extremely boring? Abolutely.

The entire point of the 480 is to offer, at $200, a significant upgrade over other $200 cards. The fact that the 480 can sometimes trade blows with the 390 and 970 is just a nice bonus.

All things considered, I'll probably get one. It's about twice as fast as my 7870 for roughly the same price new.

I wouldn't. The 1050/1060 will likely blow the 480 out of the water in the value department. Sure, against the 390 and 970, the 480 still looks sorta okay, but both are outdated. The 480 doesn't compare favourably to 2016 technology. With its 14 nm it only matches the performance/watt of nVidia's 2-year old 28 nm cards. And feature-wise it is still behind those 2-year old cards even.
Given what nVidia has shown with the 1070 and 1080 in the higher price brackets, it's a no-brainer that their offerings in the 480 pricebracket will be considerably better as well.

More Pascal rumours, GP100/102-related: http://vrworld.com/2016/07/05/nvidia-gp100-ti … r-geforce-1080/

For some reason, even some (many? all?) of the 4 GB cards actually have 8 GB of memory installed on the PCB.
A BIOS flash can enable all 8 GB: https://www.techpowerup.com/223913/amd-retail … -we-benchmarked

Here's a review doing some crossfire tests with RX480: http://arstechnica.com/gadgets/2016/07/amd-rx … gtx-1080-ashes/
No surprise there, AMD cherry-picked the results in their presentation of the RX480, where they 'demonstrated' that two RX480s were faster than a GTX1080.

I just don't understand why people want dual GPU setups. The framerate is meaningless with AFR induced latency. Though I hear that D3D 12 allows developers to use multi-GPU without AFR so maybe there's hope in the future.

Unfortunately DX12 is somewhat snake oil right now.

Here is AMD's solution to the power consumption issue: http://www.anandtech.com/show/10469/amds-tues … y-late-thursday
It seems to be two-fold:
1) Shift the excess power draw from the PCI-e slot to the 6-pin connector. Still out-of-spec, but potentially less risky. (default)
2) Add a 'compatibility' mode, where power draw (and performance) is reduced to get within spec. (optional)

So by default it is still not within spec.

If anyone is in the market for this card, I would suggest looking for a third-party version where the 6-pin connector is replaced with an 8-pin one (these are not yet available afaik, but should arrive in the coming weeks).

Performance and power consumption look really good and the price is more reasonable than the R9 Nano was. Estimates from some journalists put the 480 at a little faster than a 280x or 380x, but with only one six pin connector, that is pretty amazing. It makes me wonder how powerful a low end Polaris variant that runs only off of slot provided power and with passive cooling will be.

sliderider wrote:

but with only one six pin connector, that is pretty amazing.

Not sure if trolling or srs 😀

Scali wrote:

sliderider wrote:

but with only one six pin connector, that is pretty amazing.

Not sure if trolling or srs 😀

What do you mean? All the pics I have seen so far of Rx 480 have only one 6-pin connector and guesstimates so far give the card only about 200 fewer stream processors than a R9 390 that uses a lot more power.

sliderider wrote:

What do you mean? All the pics I have seen so far of Rx 480 have only one 6-pin connector and guesstimates so far give the card a few hundred more shaders then a 380x.

A few posts up we were discussing the fact that the RX480 does not stay within the 150W envelope, and pulls too much power from the PCI-e slot as well as the 6-pin connector, since it is a 50-50 configuration currently.
AMD will release a new driver in the coming days, which should draw less power from the PCI-e slot, and more from the 6-pin connector (which would mean it draws even more out-of-spec from the 6-pin connector).
Alternatively, you can set a 'compatible' mode, where it lowers performance somewhat to try and stay within the specs of the PCI-e slot and 6-pin connector.

In that light, I thought it was rather comical to see someone claim: "Wow, it only needs one 6-pin connector!".
Technically it needs an 8-pin connector.

What's even more sad is that NVidia's 1070 has a 150W TDP as well, and actually stays within it. So in theory it could run on a 6-pin connector, but that would limit any overclocking headroom.
Also sad that the 1070 has *exactly* the same memory configuration: 8 GB GDDR5 on a 256-bit interface, yielding 256 GB/s bandwidth. It does so much more with the same power and bandwidth, it's just ridiculous. Performance and power consumption on the RX480 don't look even remotely good.

Scali wrote:

A few posts up we were discussing the fact that the RX480 does not stay within the 150W envelope, and pulls too much power from […]
Show full quote

sliderider wrote:

What do you mean? All the pics I have seen so far of Rx 480 have only one 6-pin connector and guesstimates so far give the card a few hundred more shaders then a 380x.

A few posts up we were discussing the fact that the RX480 does not stay within the 150W envelope, and pulls too much power from the PCI-e slot as well as the 6-pin connector, since it is a 50-50 configuration currently.
AMD will release a new driver in the coming days, which should draw less power from the PCI-e slot, and more from the 6-pin connector (which would mean it draws even more out-of-spec from the 6-pin connector).
Alternatively, you can set a 'compatible' mode, where it lowers performance somewhat to try and stay within the specs of the PCI-e slot and 6-pin connector.

In that light, I thought it was rather comical to see someone claim: "Wow, it only needs one 6-pin connector!".
Technically it needs an 8-pin connector.

What's even more sad is that NVidia's 1070 has a 150W TDP as well, and actually stays within it. So in theory it could run on a 6-pin connector, but that would limit any overclocking headroom.
Also sad that the 1070 has *exactly* the same memory configuration: 8 GB GDDR5 on a 256-bit interface, yielding 256 GB/s bandwidth. It does so much more with the same power and bandwidth, it's just ridiculous. Performance and power consumption on the RX480 don't look even remotely good.

I was not aware of any power issues and I am not in the habit of reading through 5 pages of old posts before commenting on a topic.

oh, and here's something for you to chew on. Tom's Hardware says the PCIe slot being out of spec isn't the issue that it's being blown up to be anyway.

http://www.tomshardware.com/reviews/amd-radeo … nts,4622-2.html

sliderider wrote:

oh, and here's something for you to chew on. Tom's Hardware says the PCIe slot being out of spec isn't the issue that it's being blown up to be anyway.

http://www.tomshardware.com/reviews/amd-radeo … nts,4622-2.html

For me to chew on?
What do you even mean by that?

Fact of the matter is that AMD had to act on this issue, which they did.

Scali wrote:

For me to chew on? What do you even mean by that? […]
Show full quote

sliderider wrote:

oh, and here's something for you to chew on. Tom's Hardware says the PCIe slot being out of spec isn't the issue that it's being blown up to be anyway.

http://www.tomshardware.com/reviews/amd-radeo … nts,4622-2.html

For me to chew on?
What do you even mean by that?

Fact of the matter is that AMD had to act on this issue, which they did.

If you read the article linked, you would see that they didn't have to act. There was no threat of harm to anyone's computer as a result of it running out of spec. According to Tom's, a healthy ATX PSU will still be able to provide the power draw through the PCIe slot without any risk of burning anything out.