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

[Math Geek]

anyone bought a 4 gb reference 480? seems the rumors of it actually having 8gb of vram is true!! http://wccftech.com/amd-rx-480-4gb-cards-8gb-unlockable-bios-update/

a simple bios update and you got an 8gb card!! now where to get the bios is something i have read yet.

 

[Rogue Leader]

anyone bought a 4 gb reference 480? seems the rumors of it actually having 8gb of vram is true!! http://wccftech.com/amd-rx-480-4gb-cards-8gb-unlockable-bios-update/

a simple bios update and you got an 8gb card!! now where to get the bios is something i have read yet.

Thats what I have, I'll wait until there is a lot of confirmation of this before I go risking the card, plus of course the BIOS isn't out there yet.

Also this makes sense as to why the 4GB RX 480 was hard to find (my Microcenter had 2 of them vs 30 8gb cards)..

 

[TechyInAZ]

anyone bought a 4 gb reference 480? seems the rumors of it actually having 8gb of vram is true!! http://wccftech.com/amd-rx-480-4gb-cards-8gb-unlockable-bios-update/

a simple bios update and you got an 8gb card!! now where to get the bios is something i have read yet.

Reminds me of unlocking cores back in the ole days.

 

[jimmysmitty]

The 300 Series didn't even have a reference card.

Because it was the exact same GPU as the R( 290, even the PCB, so they didn't need to redesign the custom cooling. Hell the R9 390X having 8GB VRAM was not new since at launch you could get a R9 290X with 8GB VRAM for almost $100 less than the R9 390X.

Most AiBs were able to just reuse the same PCBs and coolers.

Any proof that pcie was the problem yet? I doubt that personally, because I havent found a source that is accurate to electricity, like reading comedy.

Everything is still pointing to the fact that the GPU is pulling more power than it should from the PCIe slot.

I am pretty sure the guys at Toms are more experienced than you think and know more than you think they do.

 

[Math Geek]

anyone bought a 4 gb reference 480? seems the rumors of it actually having 8gb of vram is true!! http://wccftech.com/amd-rx-480-4gb-cards-8gb-unlockable-bios-update/

a simple bios update and you got an 8gb card!! now where to get the bios is something i have read yet.

Thats what I have, I'll wait until there is a lot of confirmation of this before I go risking the card, plus of course the BIOS isn't out there yet.

Also this makes sense as to why the 4GB RX 480 was hard to find (my Microcenter had 2 of them vs 30 8gb cards)..

yah i thought of you when i read it. you can be our guinea pig when the bios makes its way into the wild. 😀 as noted this is likely only going to be in the first run cards as an effort by amd to keep supplies up. the next run of 4gb cards will likely only have 4 gb under the hood. i have not seen them in any local stores or i'd grab one just for the giggles of it later on

 

[TehPenguin]

Spoiler

The 300 Series didn't even have a reference card.

Because it was the exact same GPU as the R( 290, even the PCB, so they didn't need to redesign the custom cooling. Hell the R9 390X having 8GB VRAM was not new since at launch you could get a R9 290X with 8GB VRAM for almost $100 less than the R9 390X.

Most AiBs were able to just reuse the same PCBs and coolers.

Any proof that pcie was the problem yet? I doubt that personally, because I havent found a source that is accurate to electricity, like reading comedy.

Everything is still pointing to the fact that the GPU is pulling more power than it should from the PCIe slot.

I am pretty sure the guys at Toms are more experienced than you think and know more than you think they do.

Tomshardware? Never heard of it.

😛

 

[Rogue Leader]

anyone bought a 4 gb reference 480? seems the rumors of it actually having 8gb of vram is true!! http://wccftech.com/amd-rx-480-4gb-cards-8gb-unlockable-bios-update/

a simple bios update and you got an 8gb card!! now where to get the bios is something i have read yet.

Thats what I have, I'll wait until there is a lot of confirmation of this before I go risking the card, plus of course the BIOS isn't out there yet.

Also this makes sense as to why the 4GB RX 480 was hard to find (my Microcenter had 2 of them vs 30 8gb cards)..

yah i thought of you when i read it. you can be our guinea pig when the bios makes its way into the wild. 😀 as noted this is likely only going to be in the first run cards as an effort by amd to keep supplies up. the next run of 4gb cards will likely only have 4 gb under the hood. i have not seen them in any local stores or i'd grab one just for the giggles of it later on

My card's serial number ends in 0050000 so its probably first run lol

 

[jimmysmitty]

Spoiler

The 300 Series didn't even have a reference card.

Because it was the exact same GPU as the R( 290, even the PCB, so they didn't need to redesign the custom cooling. Hell the R9 390X having 8GB VRAM was not new since at launch you could get a R9 290X with 8GB VRAM for almost $100 less than the R9 390X.

Most AiBs were able to just reuse the same PCBs and coolers.

Any proof that pcie was the problem yet? I doubt that personally, because I havent found a source that is accurate to electricity, like reading comedy.

Everything is still pointing to the fact that the GPU is pulling more power than it should from the PCIe slot.

I am pretty sure the guys at Toms are more experienced than you think and know more than you think they do.

Tomshardware? Never heard of it.

😛

Not that so much as people think that for some reason the oldest tech publication hires interns who may have passed an A+ cert exam to write articles. Chris has been at Toms forever (I remember when he was the Editor in Chief) and Igor has been doing computer stuffs for a very long time. They are very well versed in this and it annoys me when people try to call them incompetent.

 

[Suzuki@1473780084@1473780163]

The 300 Series didn't even have a reference card.

Because it was the exact same GPU as the R( 290, even the PCB, so they didn't need to redesign the custom cooling. Hell the R9 390X having 8GB VRAM was not new since at launch you could get a R9 290X with 8GB VRAM for almost $100 less than the R9 390X.

Most AiBs were able to just reuse the same PCBs and coolers.

Any proof that pcie was the problem yet? I doubt that personally, because I havent found a source that is accurate to electricity, like reading comedy.

Everything is still pointing to the fact that the GPU is pulling more power than it should from the PCIe slot.

I am pretty sure the guys at Toms are more experienced than you think and know more than you think they do.

What people are throwing around as the cause is normal for GPU's. And I've not liked many Tom's articles due to the lack of information.

When is this going to be corrected?

"Believe it or not, the situation gets even worse. AMD's Radeon RX 480 draws 90W through the motherboard’s PCIe slot during our stress test. This is a full 20 percent above the limit."

I don't like the "authority on tech" jumping to conclusions when there's not enough evidence. Clearly a case of grabbing as many readers as possible.

And what about the initial handshake power draw from a GPU? If it doesn't pass, then the card shouldn't run.

Specifications doesn't tell us much, it's not the end nor proof because electronics aren't limited to a specification, they often pull more than what they can handle looking at the spec sheets.

I'm very interested to see if there will be any legal actions considering AMD may have hidden something to cut corners on production.

I'm also interested whether or not this case is/was bigger than the GTX 970 fiasco.

 

[Rogue Leader]

@Suzuki - Did you not see he article showing Tom's having a mountain of very expensive equipment to test for this very thing? They are not jumping to conclusions, they went out and tested it, and then reconfirmed their results.

 

[jimmysmitty]

The 300 Series didn't even have a reference card.

Because it was the exact same GPU as the R( 290, even the PCB, so they didn't need to redesign the custom cooling. Hell the R9 390X having 8GB VRAM was not new since at launch you could get a R9 290X with 8GB VRAM for almost $100 less than the R9 390X.

Most AiBs were able to just reuse the same PCBs and coolers.

Any proof that pcie was the problem yet? I doubt that personally, because I havent found a source that is accurate to electricity, like reading comedy.

Everything is still pointing to the fact that the GPU is pulling more power than it should from the PCIe slot.

I am pretty sure the guys at Toms are more experienced than you think and know more than you think they do.

What people are throwing around as the cause is normal for GPU's. And I've not liked many Tom's articles due to the lack of information.

When is this going to be corrected?

"Believe it or not, the situation gets even worse. AMD's Radeon RX 480 draws 90W through the motherboard’s PCIe slot during our stress test. This is a full 20 percent above the limit."

I don't like the "authority on tech" jumping to conclusions when there's not enough evidence. Clearly a case of grabbing as many readers as possible.

And what about the initial handshake power draw from a GPU? If it doesn't pass, then the card shouldn't run.

Specifications doesn't tell us much, it's not the end nor proof because electronics aren't limited to a specification, they often pull more than what they can handle looking at the spec sheets.

I'm very interested to see if there will be any legal actions considering AMD may have hidden something to cut corners on production.

I'm also interested whether or not this case is/was bigger than the GTX 970 fiasco.

Ahhh I forget, you are the expert on this.

You can go back to the GTX 960 review and see the same thing when they see the spikes in the Asus Strix GTX 960.

They are not saying spikes are what is bad but more that the consistent draw and above the spec current draw is the problem.

Either way they never jumped to conclusions, they made a note of the issue and they probably retested it while also sending this to AMD for a response but AMD did not get back to them.

They also retested it again, hence the new article, and came back with the same results (within margin of error) even with different settings that they thought might change the results but did not.

So should they not mention this information and hopefully find an issue AMD might have missed and instead recommend people buy a product that when overclocked, or if you are very unlucky even stock clocked, cause damage to your motherboard or mention it and hope for a response from AMD to let us know what the issue is?

And I think the fact that multiple people have had their boards die after switching to this GPU along with multiple other sites getting similar results is enough evidence to bring the issue to light.

You may like/prefer AMD but allowing that to blind anyone to a possible issue is not going to help. The job of a tech publication is to test and report their facts and findings, not to hold off just because. I am glad they brought it to the light. I like the performance of the GPU and was half tempted to buy one to replace my wifes old HD7970GHz, mainly for the lower power draw, but I would rather not risk burning her motherboard out is I can.

The funny thing is that they could have probably avoided all of this if they just put an 8 pin PCIe connection on the GPU as it could easily pull most of its power from that instead.

 

[Suzuki@1473780084@1473780163]

Overqualified or not, the card shouldn't run if it exceeds spec if recognized by the motherboards. The motherboard should prevent the card from starting, that's what I want AMD to clarify. Exactly why that is that their 'defect' cards still run?

 

[turkey3_scratch]

@Suzuki: having built-in protection circuitry on motherboards that detects of too much power is drawn though the PCIe slot would never work, because even cards like the 750Ti, for instance, will draw greater than 75W in instances. The only difference is that the average power draw for the 750Ti is well below 75W, though spikes can exceed it for brief moments. Same with other cards, too, they all have those occasional transients that will jump up and exceed the spec.

The RX480 is different because its average exceeds the spec. If there was protection circuitry to shut off the computer when it exceeds 75W, that would be a huge problem due to transient spikes, and a system that actually monitors the average power draw would be unbelievably expensive since it already requires hundreds of dollars worth of scopes to measure it.

 

[Suzuki@1473780084@1473780163]

@Suzuki: having built-in protection circuitry on motherboards that detects of too much power is drawn though the PCIe slot would never work, because even cards like the 750Ti, for instance, will draw greater than 75W in instances. The only difference is that the average power draw for the 750Ti is well below 75W, though spikes can exceed it for brief moments. Same with other cards, too, they all have those occasional transients that will jump up and exceed the spec.

The RX480 is different because its average exceeds the spec. If there was protection circuitry to shut off the computer when it exceeds 75W, that would be a huge problem due to transient spikes, and a system that actually monitors the average power draw would be unbelievably expensive since it already requires hundreds of dollars worth of scopes to measure it.

Even on PCIe x16, 2.0/3.0? Shouldn't be the way. I reckon this should only be a problem with low budget cards. Does any of this make sense? Might hit up Google translate.....

I wasn't referring to the spikes, but the 75W and 150W CF 'limit'. It just doesn't make sense. The testing only tells us half the story.
-------------

I never said that anyone was incompetent, I was just saying that generally from what I've read in other forum posts and articles. About Tom's I said lack of information, which I don't personally like, that's all it is.

If AMD reached this with their cards without overvolting then that's a huge problem and hopefully we find out what they're hiding soon...

 

[turkey3_scratch]

The 8-pin PCIe cable I've found some interesting information about. Basically, it's final design was never finalized so we are stuck with this 8-pin cable that is practically pointless by nature. Good read http://www.jonnyguru.com/forums/showthread.php?t=13418

GI Joe

No, the PCI Sig intended them from the beginning as dual grounds with sense, no matter how stupid that was. There were only endless mechanical revisions and two companies tried anything to prevent a third company from finalizing the spec for release. Ultimately the second-last revision got final, not the final one. I recall certain mechanical issues at some PSU companies Naturally, the PCI Sig did not want to use 8p EPS connectors either.

...

It is implemented as intended, 12Vs was never under consideration. I was there, I talked to the board of directors of the PCI Sig, was at the heart of the delay/revision war and shipped the first 8p PEG. Which was Rev H, not G, and everybody copied it (Insiderjoke)

The 8p PEG is 99% useless by design

Those 6-pin cables can also handle above 200W easily.

 

[jimmysmitty]

Overqualified or not, the card shouldn't run if it exceeds spec if recognized by the motherboards. The motherboard should prevent the card from starting, that's what I want AMD to clarify. Exactly why that is that their 'defect' cards still run?

The problem is that it is not exceeding spec on POST, it is exceeding spec when under heavy load. A motherboard does not prevent a GPU from spiking over 75W, it cannot as that will happen and happens even on the Fury X that averages 33W power draw from the PCIe slot.

Under normal idle conditions and lower loads the card will be fine. When under heavy gaming or stress load it will overdraw and when overclocked it will get worse.

 

[TJ Hooker]

and a system that actually monitors the average power draw would be unbelievably expensive since it already requires hundreds of dollars worth of scopes to measure it.

I don't think it'd be that expensive to get a decent measurement of it. I'm pretty sure the 12V rail is already being measured. You could measure the current by putting a small resistor in the trace path and measuring the voltage drop across it. Then implement some form of integrator, which would remove brief spikes and would only signal an overload once power has been over spec by a high enough amount and/or enough time.

 

[jimmysmitty]

and a system that actually monitors the average power draw would be unbelievably expensive since it already requires hundreds of dollars worth of scopes to measure it.

I don't think it'd be that expensive to get a decent measurement of it. I'm pretty sure the 12V rail is already being measured. You could measure the current by putting a small resistor in the trace path and measuring the voltage drop across it. Then implement some form of integrator, which would remove brief spikes and would only signal an overload once power has been over spec by a high enough amount and/or enough time.

It is not that it can't be done but you have to remember there is more than just "add a part and it should work". With motherboards you would probably have to add another layer for the traces making the board more complex and cost more. Then you have to worry about any potential drop in current/voltage due to the additional components in the way.

 

[-Fran-]

Standards are meant to be used and *expected* of products saying they follow them.

AMD is at fault here; there is no way around that.

Now they have to put a fix forward and call it a day if it works.

Nothing more, nothing less.

As for the partner cards: if it's only an issue with distribution of power in the PCB, then it's totally "fixable" without a full re-design. So, we should be good.

Cheers!

 

[17seconds]

Anyone wants to flash their 4GB card to a 8GB one, here you go....

AMD Retail Radeon RX 480 4GB to 8GB Memory Unlock Mod Works, We Benchmarked
https://www.techpowerup.com/223913/amd-retail-radeon-rx-480-4gb-to-8gb-memory-unlock-mod-works-we-benchmarked

 

[Ags1]

AMD is due to make a statement today on the software solution for the power draw issue (according to AMD), but so far no sign of it. It's a bit worrying.

 

[Sakkura]

And I think the fact that multiple people have had their boards die after switching to this GPU along with multiple other sites getting similar results is enough evidence to bring the issue to light.

Or so they've claimed. I would tend to believe them when they're like that bitcoin miner running three cards in one board, but that's as much an issue on the motherboard manufacturer and the end user's side as on AMD; if three 85W PCIe cards can overload the 24-pin ATX connector, then surely four compliant 75W PCIe cards could do the same (since 4x75 is more than 3x85). Or similar if we look at current on the 12V rail instead.

And in general, it might be prudent to refer back to the articles Tom's has published. In the launch article, this is what they said:

To be clear, your motherboard isn't going to catch fire. But standards exist for a reason.

After which they cut short the overclocking tests because that was ramping power consumption considerably higher (they didn't explicitly say so, but the "mobo isn't going to catch fire" statement would possibly cease to apply when overclocking).

And indeed, the second article reaffirms that:

We never implied in our launch article that a system made up of solid components might be directly damaged by an AMD Radeon RX 480 graphics card running at stock clock frequencies.

Ie. running stock clocks and with everything else obeying the spec, there should be no damage.

But if AMD starts exceeding the spec, what's to stop someone else from doing the same? Then eventually we might not be able to rely on the specification for assured safe interoperability. This is at least my view on the "standards exist for a reason" argument advanced in the launch article.

 

[Mousemonkey]

AMD is due to make a statement today on the software solution for the power draw issue (according to AMD), but so far no sign of it. It's a bit worrying.

Give it time, they're probably still hung over from yesterday.

 

[Ags1]

You don't think their engineers ae in the USA? Are they?

 

[Sakkura]

Standards are meant to be used and *expected* of products saying they follow them.

AMD is at fault here; there is no way around that.

Now they have to put a fix forward and call it a day if it works.

Nothing more, nothing less.

As for the partner cards: if it's only an issue with distribution of power in the PCB, then it's totally "fixable" without a full re-design. So, we should be good.

Cheers!

The board partners tend to do redesigns anyway, so they can tout overclocking and all that jazz.

But the question for me is whether AMD's fix for the power draw issue will have a negative effect on performance, and if so whether non-reference cards can avoid that negative effect.