RX 480 is apparently killing pcie slots

Daniel_Chang said:
This has already been explained to you. You are being willfully ignorant at this point.
Click to expand...

So the 75 Watt limit is 75 Watt it does not state anything else , calling me names does not win you an argument. It is almost as childish as typing +1 .
 
Pieter3dnow said:
So the 75 Watt limit is 75 Watt it does not state anything else , calling me names does not win you an argument. It is almost as childish as typing +1 .
Click to expand...

Nobody is calling you names, he said you are being willfully ignorant, that's no insult. It accurately reflects your past few posts here, it genuinely seems like you are being intentionally obtuse, possibly to elicit an angry reaction.

The PCI-E specification places a 5.5A limit on the 12V from the slot. The RX480 does not respect this limit. That's it. There's literally nothing to debate here; I agree with Daniel_Chang , either you genuinely don't understand what we are talking about, or you do and you're being facetious.
 
Pieter3dnow said:
What is it is it 75 Watt yes or no ?
Click to expand...
Did you read my post?
You accept then by your logic nearly every GPU and CPU breaks the spec, but in reality they do not....
No idea why I am doing this but here goes.
By your logic the following all break PCIe spec.
AMD Radeon R9 380X Nitro (2x6pin for 150W on auxiliary) - Tom measured many peaks over 175W, multiple peaks up to 212W on those connectors combined. AMD Radeon R9 380X Nitro Power Usage - Tom's Hardware
AMD Radeon R9 Nano (1x8pin for 150W on auxiliary) - Tom measured multiple peaks over 300W on that connector. AMD Radeon R9 Nano Power Usage - Tom's Hardware
AMD Radeon R9 Fury X 4GB (2x8pin for 300W on auxiliary) - Tom measured many peaks over 350W, a fair amount over 400W. AMD Radeon R9 Fury X 4GB Power Consumption
MSI R9 390X Gaming 8G (1x8pin and 1x6pin for 225W on auxiliary) - Tom measured many peaks over 350W, a fair amount of 400W. AMD 300-Series Gaming Power Consumption
MSI R9 380X Gaming 8G (2x6pin for 150W on auxiliary) - Tom measured multiple peaks over 200W. AMD 300-Series Gaming Power Consumption
MSI R7 370 Gaming 2G (1x6pin for 75W on auxiliary) - Tom measured multiple peaks over 80W. AMD 300-Series Gaming Power Consumption

So I think that clears it up :)
Again the peaks are NOT part of the specification for the reasons I mentioned earlier, and the above are not breaking spec on the peak/max Watts.
I stuck with AMD because well I thought that might help for it to be realised that maybe it is not about peaks but average sustained Watts :)
Cheers
 
Remon said:
Seriously? Still on this? Some people are really desperate.
Click to expand...
You did see... that he asked a question which he should already know the answer to.
As a matter of course, he has been answered, again.

Perhaps you could explain it instead so he can comprehend it, then he wont ask these questions that seem to irk you.
 
GiGaBiTe said:
What spec are you reading? 11v is WAY out of the ATX spec, all positive rails must be +/-5% and all negative rails must be +/-10%. 11.6 is as low as you can go, 11.5 is out.

If you were getting 11v on a PSU, it must be the cheapest nastiest wofat IED that you can find at the bottom of the trash heap.
Click to expand...



http://www.formfactors.org/developer/specs/atx2_1.pdf

Page 20. It does have to be +-5% when for a sustained load, but Peak current is +-10%. I read all of the foot notes stating you were in spec as long as it was greater than 10.8VDC, but it does come with the caveat that is can only be that low up to 17 seconds per minute. (Which I didn't catch that part the first time I read it.) So hopefully that's enough time for it to cool back down without causing damage to anything. I'm sure as the spec was designed at 5.5A it would be fine, but at 7A+ that's putting a bit more strain on things.
 
bman212121 said:
http://www.formfactors.org/developer/specs/atx2_1.pdf

Page 20. It does have to be +-5% when for a sustained load, but Peak current is +-10%. I read all of the foot notes stating you were in spec as long as it was greater than 10.8VDC, but it does come with the caveat that is can only be that low up to 17 seconds per minute. (Which I didn't catch that part the first time I read it.) So hopefully that's enough time for it to cool back down without causing damage to anything. I'm sure as the spec was designed at 5.5A it would be fine, but at 7A+ that's putting a bit more strain on things.
Click to expand...
Where does page 20 mention peak current?
I have read the guideline in past and remember it defines peak for something like 17 seconds sustained, does not specify the instantaneous bursts.
Sorry if you were not talking about those spurs.
Cheers
 
So, I read something in the release notes for the last AMD Crimson driver about fixing a power draw issue.

Is this what they were talking about? Is this now fixed?

I have a difficult time believing this could be fixed by a driver update...
 
Zarathustra[H] said:
So, I read something in the release notes for the last AMD Crimson driver about fixing a power draw issue.

Is this what they were talking about? Is this now fixed?

I have a difficult time believing this could be fixed by a driver update...
Click to expand...

Depends on what you consider to be fixed. There are two modes. Let me illustrate it like this:

Proper Power Draw - 75|75 (75W over PCIE, and 75W over 6-pin)
Pre-driver power draw - 80|80 (or higher in some scenarios)
Post patch power draw - 75|85 (they moved some from PCIE to 6-pin to ensure that only one is out of spec, the one that can better handle it)
Compatibility mode - 75|75 (optional mode to reduce overall power consumption putting the entire card in spec, drops performance 2-3% on average)

So far, tests have shown two things. First, the performance increase of the driver largely negated the compatibility mode hit, so if you run it, performance is about the same as the prior driver. With compatibility mode off (default), the card runs slightly faster than before. Secondly, despite this, the card is still marginally out of spec in many situations, but we're talking 0.1 to 0.2A, not 2A+ like before.
 
Zarathustra[H] said:
So, I read something in the release notes for the last AMD Crimson driver about fixing a power draw issue.

Is this what they were talking about? Is this now fixed?

I have a difficult time believing this could be fixed by a driver update...
Click to expand...
less fixed more patched. But it does alleviate the issue on the slot enough to keep the chance of issue to nearly non existent. 66w/5.5amp is the max and they got it to 68w/~5.7amp.
 
Daniel_Chang said:
Depends on what you consider to be fixed. There are two modes. Let me illustrate it like this:

Proper Power Draw - 75|75 (75W over PCIE, and 75W over 6-pin)
Pre-driver power draw - 80|80 (or higher in some scenarios)
Post patch power draw - 75|85 (they moved some from PCIE to 6-pin to ensure that only one is out of spec, the one that can better handle it)
Compatibility mode - 75|75 (optional mode to reduce overall power consumption putting the entire card in spec, drops performance 2-3% on average)

So far, tests have shown two things. First, the performance increase of the driver largely negated the compatibility mode hit, so if you run it, performance is about the same as the prior driver. With compatibility mode off (default), the card runs slightly faster than before. Secondly, despite this, the card is still marginally out of spec in many situations, but we're talking 0.1 to 0.2A, not 2A+ like before.
Click to expand...


Well, the post patch mode puts them in pretty decent company. Just about every video card out there will overdraw a little on the 12v PCIe plugs, especially when overclocking. I'd feel a lot more comfortable with that, than with the initial overdraw on the motherboard.

I'm actually surprised they had this level of fine grained control at the driver level. I would have assumed this type of change would require at least a firmware upgrade, and possibly even a change to hard wired components on the board. Maybe the drivers updated the firmware on the board during install?
 
Zarathustra[H] said:
Well, the post patch mode puts them in pretty decent company. Just about every video card out there will overdraw a little on the 12v PCIe plugs, especially when overclocking. I'd feel a lot more comfortable with that, than with the initial overdraw on the motherboard.

I'm actually surprised they had this level of fine grained control at the driver level. I would have assumed this type of change would require at least a firmware upgrade, and possibly even a change to hard wired components on the board. Maybe the drivers updated the firmware on the board during install?
Click to expand...
Nah the vrm controllers supports i2c
 
CSI_PC said:
Where does page 20 mention peak current?
I have read the guideline in past and remember it defines peak for something like 17 seconds sustained, does not specify the instantaneous bursts.
Sorry if you were not talking about those spurs.
Cheers
Click to expand...

Sorry not sure why but that was the wrong link.

http://www.formfactors.org/developer/specs/Power_Supply_Design_Guide_Desktop_Platform_Rev_1_2.pdf


3.2 DC Output - REQUIRED
3.2.1 DC Voltage Regulation
The DC output voltages are required to remain within the regulation ranges shown in Table 15, when measured at the load end of the output connectors under all line, load, and environmental conditions specified in Chapter 6.
3.2.2 Output Transient Response - REQUIRED
Table 16 summarizes the expected output transient step sizes for each output. The transient load slew rate is = 1.0 A/µs.
Output voltages should remain within the regulation limits of Table 15, for instantaneous changes in load as specified in Table 16 and for the following conditions: • Simultaneous load steps on the +12 VDC, +5 VDC, and +3.3 VDC outputs (all steps occurring in the same direction) • Load-changing repetition rate of 50 Hz to 10 kHz • AC input range per Section 2.1 and Capacitive loading per Table 19.
Table 15. DC Output Voltage Regulation
Output Range Min Nom Max Unit
+12V1DC1
NOTES: 1. At +12V1DC peak loading, regulation at the +12V1DC and +12V2DC outputs can go to ±10%.
±5% +11.40 +12.00 +12.60 V
+12V2DC2
2. At +12V2DC peak loading, regulation at the +12V1DC and +12V2DC outputs can go to ±10%.
±5% +11.40 +12.00 +12.60 V +5VDC ±5% +4.75 +5.00 +5.25 V +3.3VDC3
3. Voltage tolerance is required at main connector and SATA connector (if used).
Click to expand...

Then you have to scroll all the way down to page 50 (30 pages later) to catch the notes at the bottom of the tables that state:

NOTES: 1. Peak currents may last up to 17 seconds with not more than one occurrence per minute.
Click to expand...

After reading that yet again I think I understand now what they are meaning. If you run over the rated current of the rail, you are allowed to have a tolerance of 10% at the peak load during that time. So a good example for a boxed pc would be the 300W configuration. Under normal circumstances, you would have to maintain 5% VR at 8 amps load on 12V2. But you can go up to 13A load and only have to maintain 10% regulation as long as it goes back down to 8A or less within 17 seconds. So it would actually be hard to trigger a peak load condition because the card draws a consistent amount of power. You would simply be overloading it if you were using more than the 8A allowed all of the time to trigger a peak condition, and the card would never go back to normal conditions after 17 seconds because the power draw isn't going to decrease. Back to my original post that does mean that in fact the PSU wouldn't be able to go down to as low as 11V but have to stay at 11.4V or higher to maintain spec under a static load. (In the case of the video card with a consistent 160W power draw)

So my apologies if I misinterpreted that and it does mean that hopefully it's harder to actually hit a condition where the card would be able to fry some components. The person who engineered the PSU design seems to have done a solid job trying to leave enough overhead to keep certain devices from causing problems. You could still trigger a peak condition if a HDD or cd spins up on that rail pushing the voltage down and the amperage up, but with decent wiring it should be okay.



Zarathustra[H] said:
Well, the post patch mode puts them in pretty decent company. Just about every video card out there will overdraw a little on the 12v PCIe plugs, especially when overclocking. I'd feel a lot more comfortable with that, than with the initial overdraw on the motherboard.

I'm actually surprised they had this level of fine grained control at the driver level. I would have assumed this type of change would require at least a firmware upgrade, and possibly even a change to hard wired components on the board. Maybe the drivers updated the firmware on the board during install?
Click to expand...


Ieldra said:
Nah the vrm controllers supports i2c
Click to expand...

Does any overclocking software let the end user mess with those settings? If not seems like someone might be able to reverse engineer the changes between the two drivers to figure out where the knobs are at to control this.
 
bman212121 said:
Does any overclocking software let the end user mess with those settings? If not seems like someone might be able to reverse engineer the changes between the two drivers to figure out where the knobs are at to control this.
Click to expand...

If it's actually i2c it should be fairly trivial to do...
 
Trimlock said:
Doubt it, I'm sure the initial batches are screwed but AMD should will fix any additional cards being requested.
Click to expand...

And only then they could request for re-test their cards and in those case only those "fixed" cards will be able to pass and OEM have to specifically use those "fixed" card and I hardly believe with AMD being so cheap as they are, they have to recall all their cards to fix their issues and then re-launch with PCI-SIG Logo.

At this point of the date since so long of the launched product IIRC, they are not allowed to request a re-test for PCI signature from a launched product, 60 days I think it's the grace period to recall all the cards, fix the issues at hardware level and then relaunch.. this isn't going be fixed at that level and OEM market is where the money is located.. I hardly doubt any big OEM company will be willing to use non-compliant PCI-E cards even if the issues are "treated" at software level.. honestly? and have to say again, sad. This is the result of being negligent...
 
Not being PCI-SIG certified also got implications for warranty on the other components used. So ye, good luck selling any of these to OEMs.
 
Were they every actually on that list? Most variants of the card should be compliant so not sure why they would have been removed. Even for OEMs the reference models would meet specs with the software fix. Card just needs clocked differently or set to throttle.
 
Anarchist4000 said:
Were they every actually on that list? Most variants of the card should be compliant so not sure why they would have been removed. Even for OEMs the reference models would meet specs with the software fix. Card just needs clocked differently or set to throttle.
Click to expand...

Never on the list from what I understand. Don't know if it has been submitted before.
 
It's not a big deal, they said. PCI-E allows up to 300W from the slot, they said.
 
It's been a while since I read the articles about the patch, but didn't the driver fix the motherboard connection by throwing the PSU connection further out of spec?

[Furiously Googles]

Yup, PCPER is showing 91w in Metro at 4k with 16.7.1 on the six pin connector... So yeah, I can see why PCI-SIG isn't gonna list them.
 
Araxie said:
So at the end, Good bye OEM market... sad.
Click to expand...

really? you think oem cares about 5 watts? They care about bottom dollar. OEM push crap on consumers all day everyday, I doubt they care too much about not having pci-sig.
 
Decibel said:
It's been a while since I read the articles about the patch, but didn't the driver fix the motherboard connection by throwing the PSU connection further out of spec?

[Furiously Googles]

Yup, PCPER is showing 91w in Metro at 4k with 16.7.1 on the six pin connector... So yeah, I can see why PCI-SIG isn't gonna list them.
Click to expand...

6 pin isn't the same certification body. This is 100% about the PCIE slot.
Compliance Program | PCI-SIG

Compliance Program
PCI-SIG Compliance Workshops host interoperability and compliance tests. Interoperability tests enable members to test their products against other members’ products. Compliance tests allow for product testing against PCI-SIG test modules.

Both testing types issue “pass” or “fail” results for each test area examined. To formally label products as compliant, they must score a minimum of 80 percent on interoperability tests and pass all required compliance tests.

Tested Specifications
Interoperability and compliance tests focus on the latest PCI specifications, specifically:

  • PCI Express 3.0
  • PCI Express 2.0
Test Areas
PCI Express compliance testing includes:

  • Electrical Testing - Examines platform and add-in card Transmitter and Receiver characteristics
  • Configuration Testing - Examines configuration space in PCIe devices
  • Link Protocol Testing - Examines device’s link-level protocol behavior
  • Transaction Protocol Testing - Examines devices transaction-level protocol behavior
  • Platform BIOS Testing - Examines BIOS ability to recognize and configure PCIe devices
Corresponding Test Guides by specification version can be downloaded here. These downloadable guides include test descriptions, test specifications, procedural guidelines and required tools.

 
LOL some people just can't let it go. Regardless of this pci-sig they are selling. Do you guys really believe OEMs are always looking out for consumer. They are in for cheap, as long as they get a cheap build out and make money they could care less about a sticker on the box. Drivers limited the PCI-e draw already and I am sure they are fine with extra 10 watts or so through 6 pin.
 
Drivers don't kick in when you're doing Windows updates. I'm in the Insider Edition fast ring update program and when it does its full update cycle it's about 40 or so minutes pulling power pre-drivers.

BIOS flashing would be the only non-replacement way of fixing this issue.
 
Anarchist4000 said:
Go check the integrators list. It isn't there. That's the same evidence TPU used.
Integrators List | PCI-SIG
Click to expand...

Incorrect. There's a difference between "removed" and "not yet added."

NKD said:
LOL some people just can't let it go. Regardless of this pci-sig they are selling. Do you guys really believe OEMs are always looking out for consumer. They are in for cheap, as long as they get a cheap build out and make money they could care less about a sticker on the box. Drivers limited the PCI-e draw already and I am sure they are fine with extra 10 watts or so through 6 pin.
Click to expand...

As an AMD fan, I don't like this either. As a fan of objectivity, I recognize that this is a big issue. Yes, I'm aware that Nvidia fans have blown this out of proportion previously. But you're doing the exact same thing on the other side of the spectrum. You're essentially being this guy:

07-minister.jpg
 
Algrim said:
Drivers don't kick in when you're doing Windows updates. I'm in the Insider Edition fast ring update program and when it does its full update cycle it's about 40 or so minutes pulling power pre-drivers.

BIOS flashing would be the only non-replacement way of fixing this issue.
Click to expand...

Doing much gaming while pulling updates?

If the card is running at full tilt while running Windows Update you have other issues going on.
 
Decibel said:
Doing much gaming while pulling updates?

If the card is running at full tilt while running Windows Update you have other issues going on.
Click to expand...

He's still correct. The card can technically run out of spec until the driver takes over. Is this going to be a common occurrence? No, of course not. But only a bios update can truly fix the issue.

The vast majority of users won't be affected by this. But I don't want to be that sucker in the minority either.
 
You must log in or register to reply here.
Back
Top