AMD in Trouble? RX 480 Powergate

I am about to grab some popcorn too. Let this play out. Cuz its just too much fun watching it from the sidelines. My parts cant get here soon enough so I can enjoy my oboard graphics and grab me a 1070 or something. Only if nvidia get off their fucking founders edition crack pike and give me a damn reference card.
 
Ieldra said:
It's all coming from the 24pin ATX




yes but on the board the power is running through share traces that are then split to each pci-e connector



this is a very simple tactic, you're essentially going to keep asking me prove and present evidence of things until I either get bored or spend half my life walking you through things, it's simply a proposition you know I will never agree with, and you will try to paint that as evidence of some kind of flaw or bias. Meh. You could at least be original.
Click to expand...
Proving and presenting evidence of your findings is the correct way of going about what you are trying to do. It's the scientific method. Otherwise why should anyone listen to you and not heavily debate your claim?
 
so you think the reviewers are not using a scientific method to figure out what they are seeing?

This is an easy problem to fix for AMD, its nothing big.
 
Ieldra said:
No the limit is per slot, the spec places a 75w limit per slot. In practice for multi-gpu what you have is most of the power drawn from external connectors
Click to expand...
But your made the argument that violating this in a single slot isn't a big deal, it's when you run crossfire that it becomes a big deal. If it's per slot and violating it on one slot isn't a big deal, why would it be on another
 
  • Like
Reactions: NKD
like this
Ieldra said:
IT IS NOT COMPLIANT WITH PCI-E OR ATX SPECIFICATIONS. PERIOD. There is no debate about this whatsoever, so my claiming it is non-compliant is 100% correct and corroborated by SEVEN independent measurements of which 6 are review websites, one is a user, and 2 are retail samples.
Click to expand...

Congrats! now only if nvidia could make a mother fucking reference gtx 1070 available so I could buy one. Before the AIB rx 480 cards show me otherwise if they are doing 1500. My money is waiting damnit. who is gonna take it first.
 
Sheppard said:
But your made the argument that violating this in a single slot isn't a big deal, it's when you run crossfire that it becomes a big deal. If it's per slot and violating it on one slot isn't a big deal, why would it be on another
Click to expand...


The power traces for the slots come from the same place, so if there is more being pulled that speced, and those traces aren't speced for that kind of power draw, you are going to run across problems.
 
razor1 said:
so you think the reviewers are not using a scientific method to figure out what they are seeing?

This is an easy problem to fix for AMD, its nothing big.
Click to expand...
Questioning their method is required if their findings go against the findings of the certifying body.
 
  • Like
Reactions: NKD
like this
I bet AMD couldn't come up with the performance number to equalize 970 so they had to add some juice to get higher core clocks and hence in the process exceeded 150W. They have pretty much done this time to time. Anyone remember the 290 / 290X fiasco. They were overvolted to the hilt, running a mockery at 94C.
 
razor1 said:
so you think the reviewers are not using a scientific method to figure out what they are seeing?

This is an easy problem to fix for AMD, its nothing big.
Click to expand...

Yea thats what I have been saying But Ieldra is on a roll, lol. They can just set a hard limit to 150w on reference cards. bump up the speed a little on the fan to retain boost clocks. LOL problem solved!
 
razor1 said:
The power traces for the slots come from the same place, so if there is more being pulled that speced, and those traces aren't speced for that kind of power draw, you are going to run across problems.
Click to expand...
If it's coming from the same place, running crossfire at all would violate it
 
NKD said:
Yea thats what I have been saying But Ieldra is on a roll, lol. They can just set a hard limit to 150w on reference cards. bump up the speed a little on the fan to retain boost clocks. LOL problem solved!
Click to expand...

lol i've been saying a bios update or driver change should fix this, you're the one being dramatic :p
 
Sheppard said:
Questioning their method is required if their findings go against the findings of the certifying body.
Click to expand...


yet you have 6 reivew sites that have found this so far, and this isn't alarming to you?

You have Kyle that posted an article about AMD and Polaris, everything is up in arms warfare, but when you have data from independent sources, you don't seem to be bothered.

Those specs are made for ALL computer parts to be COMPATIBLE and COMPLIANT to each other and the rest of the system.
 
Sheppard said:
If it's coming from the same place, running crossfire at all would violate it
Click to expand...

Not it wouldn't. Imagine you're driving a tiny car down the load line. You exit the 24pin ATX and find yourself on a highway, then you must take a turn to get your PCI-E slot. The highway can hadnle 300W, the road to your slot only handles 75. It's not rocket science
 
razor1 said:
Its out of spec, it has to be fixed otherwise they can't get the pci-e certification, pretty simple, then OEM's will have to make non PCI-e complaint systems, why would they want to do that? They won't. Either AMD fixes this problem or they loss OEM sales.
Click to expand...
As far as I can tell, no reputable OEM is selling systems with RX 480s in them. This is not the case for NVidia's new cards.

That the big OEMs doen't seem interested in AMD's new wonder-card speaks volumes.
 
Sheppard said:
Questioning their method is required if their findings go against the findings of the certifying body.
Click to expand...
If you have ever worked with a certifying body then you would know they do not test every product manufactured just the ones referenced for the testing. If several reviewers are getting different results that is not that unlikely changes can be made after certification or the boards used in certification where altered to pass or different enough. (for the record which can not be verified I have worked with UL certification as well as Factory Mutual)
 
pcgeekesq said:
As far as I can tell, no reputable OEM is selling systems with RX 480s in them. This is not the case for NVidia's new cards.

That the big OEMs doen't seem interested in AMD's new wonder-card speaks volumes.
Click to expand...


Its possible they aren't but usually you have to wait for a month or two before we see new cards coming into their systems, just have to get rid of old cards first.
 
NKD said:
So the certification department fucked up? if its certified amd has every right to sell it without question, they did their job so they may have been okay with 10 watt over if they saw that happening in the lab. Not like AMD just slapped it lying to them.
Click to expand...

Sorry, that's not how trademarks work. When the PCI Express spec says "5.5 amps max draw from the 12V pins" and a card bearing the PCI Express mark draws more than that, it puts the validity of the mark at risk. That's not something the PCI_SIG, which owns the mark, would want.

Trademarks are considered a consumer protection mechanism. Fraudulent use of a trademark is considered counterfeiting.
 
pcgeekesq said:
Sorry, that's not how trademarks work. When the PCI Express spec says "5.5 amps max draw from the 12V pins" and a card bearing the PCI Express mark draws more than that, it puts the validity of the mark at risk. That's not something the PCI_SIG, which owns the mark, would want.

Trademarks are considered a consumer protection mechanism. Fraudulent use of a trademark is considered counterfeiting.
Click to expand...
Think you missed the verification go ahead part. It was certified therefore no slapping of trademark used.
 
JustReason said:
Think you missed the verification go ahead part. It was certified therefore no slapping of trademark used.
Click to expand...

I think you missed the 170W power usage. It is certified PCI-E compliant, and in testing it is not. Whether computers spontaneously combust or not is actually irrelevant.
 
So let me get this straight. A single pcie port being violated by 5-10w is no big deal but when you use crossfire it becomes a problem, but the specification is per port, but all the ports get their power from common traces but somehow running crossfire doesn't violate this even though they are sharing the same power delivery traces?

I believe that you are incorrect.
 
Ieldra said:
I think you missed the 170W power usage. It is certified PCI-E compliant, and in testing it is not.
Click to expand...
So being certified says nothing of what it used necessarily. Yes there is a problem but fraudulent use of certification is not one of them. Makes me wonder how many here screaming foul did the same with the 970 specs.
 
Sheppard said:
So let me get this straight. A single pcie port being violated by 5-10w is no big deal but when you use crossfire it becomes a problem, but the specification is per port, but all the ports get their power from common traces but somehow running crossfire doesn't violate this even though they are sharing the same power delivery traces?

I believe that you are incorrect.
Click to expand...

You have been believing exactly what you want to from the outset

JustReason said:
So being certified says nothing of what it used necessarily. Yes there is a problem but fraudulent use of certification is not one of them. Makes me wonder how many here screaming foul did the same with the 970 specs.
Click to expand...
Let me get this straight, being PCI-E certified, which stipulates a 75 MAXIMUM on the PCI-E slot, means nothing when it comes to actual power draw ?
 
Sheppard said:
So let me get this straight. A single pcie port being violated by 5-10w is no big deal but when you use crossfire it becomes a problem, but the specification is per port, but all the ports get their power from common traces but somehow running crossfire doesn't violate this even though they are sharing the same power delivery traces?

I believe that you are incorrect.
Click to expand...


It doesn't matter, if you use a cheaper motherboard, they might not have the power delivery systems that can even withstand a small increase like that from one card.

You have to understand having spikes is one thing, having a sustained load over a certain limit is quite different.

If the motherboard fails because of it, then you got another problem the warranty of the motherboard is compromised by the screw up.
 
AMD should be held accountable for this. If any of the companies get away breaking rules and regulations, they must be penalized to the heaviest to prevent this from happening.
 
Presbytier said:
If you have ever worked with a certifying body then you would know they do not test every product manufactured just the ones referenced for the testing. If several reviewers are getting different results that is not that unlikely changes can be made after certification or the boards used in certification where altered to pass or different enough. (for the record which can not be verified I have worked with UL certification as well as Factory Mutual)
Click to expand...
I work in aviation and I hold a repairmens certificate. I'm very much familiar with what a "first article" is.
 
Its certified at 75 watts RMS, I know that certain motherboard makers (Asus) puts in components that could probably allow greater power over the board. But running at 80watts won't instantly cause your board to blow up in flames, but it most likely will lessen the life span of it.

Also, if you have a cheap board, the last thing you want is more power to be pulled over those traces. Especially if the board is a thin one, that may cause a severe heating problem.
 
Sheppard said:
I work in aviation and I hold a repairmens certificate. I'm very much familiar with what a "first article" is.
Click to expand...

Then you should understand this, lets say you use a part in an airplane that causes another part to fail, would you be held accountable for what you did?
 
razor1 said:
They just need to fix it, they don't need to be penalized.
Click to expand...

This level of control can also be done in the drivers. Although if they place it in the drivers it can be altered very easily and I don't think AMD would want to get out.
 
Sad state when one has to go else where to get real info. Seems that some found about 79W being pulled on the 12V rail. Spec is 66W but that is before the tolerances allowed. Seems based on the 1.1 revision with +9% on 12V rail that figure is then ~72W which then the 79W is still above spec but not too far off.

Also AMD will have 3 months to fix the issue before any fines or other punishments ensue.
 
Trimlock said:
Its certified at 75 watts RMS, I know that certain motherboard makers (Asus) puts in components that could probably allow greater power over the board. But running at 80watts won't instantly cause your board to blow up in flames, but it most likely will lessen the life span of it.

Also, if you have a cheap board, the last thing you want is more power to be pulled over those traces. Especially if the board is a thin one, that may cause a severe heating problem.
Click to expand...

RMS and DC don't mix...
 
Sheppard said:
RMS is also measured differently.
Click to expand...

It largely depends on the maker, but the certifying company has a pretty hard line limit on the spec. At the same time a few pieces of equipment that falls out of spec aren't going to cause rampant destruction either. Much like running other items out of spec, you run the risk of damaging all components involved ... over time.
 
Sheppard said:
Questioning their method is required if their findings go against the findings of the certifying body.
Click to expand...
The certifying body, PCI-SIG, allows the vendor to select the card to be tested. It's really an honor system, PCI-SIG doesn't pull random samples off the production line and test them, and doesn't use game benchmarks to test for compliance with the spec. Instead the add-in card is put in a "compliance mode" for testing.

Details of the PCI-SIG PCI-Express compliance testing are here. You can also read about it here.
Interestingly, power consumption of add-in cards doesn't seem to be part of the test that the PCI-SIG performs.
 
I couldn't easily find the 2.0 revision without being a paid member of PCI-SIG. Best I could find was the 1.1 revision. So not sure how the baselines and tolerances changed.
 
Here is the issue Leldra.
You noticed a little thing, and you saw an opportunity to blow something up without quite understanding what you were talking about. You've made the argument that pulling a little extra power from a single PCI-E port isn't a big deal and that it becomes an even bigger deal when running crossfire. You then made the argument that the specification is for each slot. Ok if the specification is for each slot and it isn't compounding then it isn't a huge issue on two slots. You then made the argument that all slots power the power from common traces. ok so its now VERY evident that those traces are capable of providing much more than the value of a single slot because they would have to, to support crossfire correct? Some motherboards like mine have 4 slots and even more. Therefore there is no issue with the traces being able to provide enough power, even for crossfire, and the specification is for each slot and if the extra power on a single slot isn't a big deal then the extra power on another slot isn't a big deal because those slots have their own specification.
That being said a certifying body has passed the card and you are trying to argue against what that body has said therefore you must provide and present evidence of that argument you are presenting. It is not the job of the presentee to verify your argument for you by providing data/

razor1 said:
Then you should understand this, lets say you use a part in an airplane that causes another part to fail, would you be held accountable for what you did?
Click to expand...
IF the part used in the airplane caused a failure. A failure we have yet to verify occurring in any way in this case.
 
Sheppard said:
IF the part used in the airplane caused a failure. A failure we have yet to verify occurring in any way in this case.
Click to expand...

So you would do something like that and wait for a plane to come down? Tell me what planes ya work on cause I'm not flying on them lol, J/K I am 100% sure you would never do that you are arguing for arguing sake, jokes aside, this is what an actuary is for........
 
You must log in or register to reply here.
Back
Top