AMD once again violating power specifications? (AMD RX-480)

[zMeul]

3 minutes ago, djdwosk97 said:

FM2+ is PCIE 3.0

by the APU, not by the chipset

or it's because the manufacturer used a PCIe lane hub to aggregate PCIe gen2 lanes into a 16x gen3 slot - I think ASUS tried it

[zMeul]

3 minutes ago, Kobz360 said:

Yeah I'm not denying that it's an issue man, I'm just saying that if the board was relatively modern (I'd say in the last 4-5 years), then it may not affect many users. 

PCie gen2 and gen3 are electrically compatible, leaving to the side the fact it can only provide half the bandwidth oh a gen3

until couple of years ago, even Intel's PCH had PCie gen2 lanes

[Sintezza]

9 minutes ago, zMeul said:

they're most likely in between layers

Yeah but with a thermal laser you can roughly get an idea how hot the traces went.

Older and also cheap boards dont have as much layers as say more expensive and decent boards.

But still that cannot be an excuse, even on cheap boards this card should be capable to function without any issue.

The guy in the video Stated that his Asus Z170-A had not a problem.

But still, like i said a couple of times, cards are getting tested and certified to meat certain standards before they leave the factory.

But if it turns out that they arent tested on pci-e sig, then yeah they will most likely come with a statement.

The video from the guy that has his old board to shut off, is currently the second source there is about the potential problem next to Toms.

We will have to wait and see, if there are more to follow.

Other sources have just tested, and seen that there was a power overshoot, but no failing boards, excepts toms one.

[Bleedingyamato]

4 hours ago, zMeul said:

ADM boards too, since you know .. it's an AMD product

That's reasonable to test those too but most people gaming on their PC are likely to be using Intel CPUs for their improved performance over AMD CPUs.  At least that's my impression from what I've heard of Intel vs AMD CPUs.  So while yes it is important to test AMD boards as well given that Intel presumably has a greater CPU market share it stands to reason that Intel boards be the priority for testing.

[zMeul]

2 minutes ago, Sintezza said:

Older and also cheap boards dont have as much layers as say more expensive and decent boards.

you'd be surprised how thick mobos used be be - power draw of CPUs went down

[Briggsy]

6 minutes ago, Bleedingyamato said:

That's reasonable to test those too but most people gaming on their PC are likely to be using Intel CPUs for their improved performance over AMD CPUs.  At least that's my impression from what I've heard of Intel vs AMD CPUs.  So while yes it is important to test AMD boards as well given that Intel presumably has a greater CPU market share it stands to reason that Intel boards be the priority for testing.

pre-builts using AMD FX cpu's are generally a lot cheaper than the intel pre-builts by good amount. I wouldn't underestimate how many AMD-based PC's are out there, and the motherboards they use are beyond bargain basement level cheap.

[Sintezza]

7 minutes ago, zMeul said:

you'd be surprised how thick mobos used be be - power draw of CPUs went down

 

vrm cirquitry went way up in spec in todays standards.

But it also depends on the quality of the certain board.

With an exception for X58 on which certain motherboards came with a massive 10 phase voltera pwm. VT1165 / 1185.

But those pwm´s are not vrd12 certified, so you cant use them on new motherboards of todays standards anymore unfortunatly.

 

But thats a bit offtopic.

 

[Kobz360]

3 minutes ago, zMeul said:

PCie gen2 and gen3 are electrically compatible, leaving to the side the fact it can only provide half the bandwidth oh a gen3

until couple of years ago, even Intel's PCH had PCie gen2 lanes

I think you're a bit confused with what I'm trying to say, I'm aware PCIe 2/3 are electrically compatible.

 

All I am saying is, is that the newer the board, the better it's going to be built. A cheap $50 AM2 motherboard from 2006 will be nowhere near as good quality wise as a cheap $50 motherboard from 2016. As technology (or really anything) progresses, production techniques and quality tends to get better. But we will see if it really matters when a cheap modern board is tested.

 

Hopefully pcper tests a cheap 760G AM3 motherboard like the 980DE3 as well as that cheap Intel board they're using as well

[zMeul]

2 minutes ago, Sintezza said:

vrm cirquitry went way up in spec in todays standards.

But it also depends on the quality of the certain board.

my Q9550 has a TDP of 95W

my i5 6500, that should be in my hands morrow, has a TDP of 65W

 

7y between them

[Bleedingyamato]

2 minutes ago, Briggsy said:

pre-builts using AMD FX cpu's are generally a lot cheaper than the intel pre-builts by good amount. I wouldn't underestimate how many AMD-based PC's are out there.

Oops..  I hadn't thought about that.  Those are likely to have crap PSUs and cheaper end motherboards so if that Science Studio video is any indication I'd say anyone who owns one of those AMD prebuilts will at least be at an increased risk to have this power problem to worry about.

[Sintezza]

2 minutes ago, zMeul said:

my Q9550 has a TDP of 95W

my i5 6500, that should be in my hands morrow, has a TDP of 65W

 

7y between them

TDP doesnt realy tell a whole lot.

TDP stands for thermal design power, and is a number measured in watts of which your cooling solution should be capable to dissipate from the cpu die.

Its not TPD or total power draw.

[mariushm]

If what Tom's Hardware says about three power phases being powered by motherboard slot and three by the pci express connector, a simple fix would be disconnecting the 12v input from the pci express slot and making a connection to the pci-e connector to draw power from there.

The card has an inductor meant to filter the 12v input, I guess from the motherboard, because there's another inductor right by the pci-e 6pin connector to filter that input.

 

So I guess someone could desolder that inductor and solder on one side of the inductor a couple of cables and then solder the ends of the cables to the pci-e 6 pin connector.

See

 

 Desolder the inductor that says R47 on it, solder a couple of AWG18 wires on side close to the mosfets and solder the other end at the pci-e connector.

It's just a guess, I don't have the card and I don't even know if what I say is workable, maybe there's some other protections or changes required to make this happen

[zMeul]

2 minutes ago, Kobz360 said:

All I am saying is, is that the newer the board, the better it's going to be built. A cheap $50 AM2 motherboard from 2006 will be nowhere near as good quality wise as a cheap $50 motherboard from 2016

I bet you quite the contrary, as I said to @Sintezza, CPU power requirements went down in recent years

[zMeul]

4 minutes ago, Sintezza said:

TDP doesnt realy tell a whole lot.

TDP stands for thermal design power, and is a number measured in watts of which your cooling solution should be capable to dissipate from the cpu die.

Its not TPD or total power draw.

TDP for the ASIC is identical to power draw - ASIC being the CPU or GPU

when it comes to the PCB as an assembly, that's something else

 

just to give you an idea how they used to build mobos:

GA-X38-DS5

GA-H170-D3H

 

do you see the density difference in electronic components 

[Sintezza]

2 minutes ago, zMeul said:

TDP for the ASIC is identical to power draw - ASIC being the CPU or GPU

when it comes to the PCB as an assembly, that's something else

like i said TDP is thermal design power, and thats something diffrent,

then total powerdraw.

[maulemall]

It sux that there isn't anyone on this board who can do anything more than cut and paste.

Wouldn't it be great if someone on this board had actual experience and the ability to make this test instead of cutting and pasting someone else's "opinions".

Well it's a blessing that we have people who can scream how "AMD LIED TO US AGAIN" with no real proof.

 

[Kobz360]

11 minutes ago, zMeul said:

PCie gen2 and gen3 are electrically compatible, leaving to the side the fact it can only provide half the bandwidth oh a gen3

until couple of years ago, even Intel's PCH had PCie gen2 lanes

True in a way yes, though you still have CPUs like the 8350 with relatively high TDPs.

 

Regardless, just because CPU power requirements have (generally) gone down, it doesn't mean board manufacturers will halt any other improvements in their boards. The CPU is not the only thing board partners think about when making a board. Are we really debating if motherboard quality has remained stagnant over the past decade? 

Once again though, I've said it before and I'll say it again.. we will see when it's tested with the cheap modern boards. What we do know for certain is that cheap decade old boards are definitely at risk, as seen in that video. Even if we find that cheap boards today are fine with it, it still doesn't change the fact that AMD should try and fix the issue (if it turns out to be highly problematic). 

[Prysin]

29 minutes ago, zMeul said:

read my post above yours

 

and ps: AMD hasn't manufactured anything new (chipsets) for couple of years - they are still on PCIe gen2

Gen 2 for AM3+

Gen3 x16 for Steamroller (Kaveri)

Gen3 x8 for Excavator (Carrizo/Bristol Ridge)

i have no idea why they cut the amount of PCIe 3 lanes in Excavator. Although, performance wise, you prob wont see much of a difference. As proven by these two articles:

 

https://www.techpowerup.com/reviews/AMD/R9_Fury_X_PCI-Express_Scaling/

https://www.pugetsystems.com/labs/articles/Impact-of-PCI-E-Speed-on-Gaming-Performance-518/

[kokobash]

Just as others stated before. Wait for AMD's official response or action before adding more fuel to the fire. It has only been like a day or two since this was discovered.

 

Other question though, wouldn't the cpu maxed out first before the gpu on older platforms? i.e. AM2, AM3, 775 boards? if that is the case, we wouldnt be seeing spikes on the gpu since it will be under utilized

[Briggsy]

I want to see someone test quadfire 480's on this motherboard for no reason

 

[Kobz360]

So before this becomes another news topic, Imma post this here. So I found (on WCCF) a link to this where all these guys talk about how 480's are killing their boards. I'm slightly suspicious though due to the fact that the majority of them created their accounts on the same day (July 1st). They also have no physical proof of anything happened. They could be trolls although.. that's a lot of effort to go to just to make a point. Time will tell boys..

 

https://community.amd.com/thread/202410

[gilang01]

https://community.amd.com/thread/202410

https://community.amd.com/thread/202429

 

O dear.

OOO dear.

I have an rx480 on the way and I hope my ECS H77H2-M3 can handle it...

or I'm screwed for good.

 

[79wjd]

9 minutes ago, Kobz360 said:

So before this becomes another news topic, Imma post this here. So I found (on WCCF) a link to this where all these guys talk about how 480's are killing their boards. I'm slightly suspicious though due to the fact that the majority of them created their accounts on the same day (July 1st). They also have no proof of anything happening with no physical proof. They could be trolls although.. that's a lot of effort to go to just to make a point. Time will tell boys..

 

https://community.amd.com/thread/202410

I don't think the join date really means anything -- a lot of people join forums when they have a problem that they need help working through. I don't even remember some of the forums that I've made accounts on in order to troubleshoot some issue. Whether or not the posts or trolls or not is still hard to tell, but the main thread in that post is at the very least believable -- a 970 board with an overclocked 480 (if at stock it's pulling a bit over spec, overclocked is probably ridiculously over spec). 

[wolfsbane3083]

Don't know if this has been posted directly here yet:

These guys drew 125w from the PCI-E slot and also had over 700mv of ripple!

[zMeul]

2 minutes ago, wolfsbane3083 said:

Don't know if this has been posted directly here yet:

These guys drew 125w from the PCI-E slot and also had over 700mv of ripple!

tell them to do the same test on a cheap mobo