Spikes are not as bad as sustain. You are also comparing a custom card vs a reference one. In the case of the 960 we would have to blame asus not nvidia.
AMD: Speculation, Rumors, and Discussion (Archive)
- Thread starter iMacmatician
- Start date
- Status
If you would just read the whole test of the ASUS 960 you would see that no other 960 tested did this. Apart from that the ASUS spikes but the average sustained power draw is below the limit imposed by the connectors installed. Hardware.fr, HT4U and Toms have tested the RX480 with the Powerlimit in "Wattman" moved to the max and have measured over 200W average. Imho that is not acceptable for a reference card.
Most cards sustain less than 50W, even the 1080 does. And yes it could be better but you have to add the OC and added voltage(IDK if it has) anyways AMD reference board is very good and maybe thats why we dont see that spikes of 300W on the card.
Besides that im quite surprise there is no stock at all(Amazon doesn't even have this listed yet...) and in newegg its all sold out. This page track the stock and it have been the same picture since launch day: https://www.nowinstock.net/computers/videocards/amd/rx480/
There are also misleading info about the memory chips, in newegg the 4GB cards says it has 8Gb memory which is wrong(it has 7Gb). Very surprise how bad AMD prepare this lunch overall...Raja says that AMD biggest problem right now is its image(everyone think is doomed and no1 wants to work there)(yes that was generalizing and based of Rajas statements) So AMD needs to invest in clearing hes image so talent people would look to get a job there(although almost all the talent is going to the mobile industry which makes things very hard for AMD).
Hope the launch of 470 would be better and more smooth.
So your saying the disabled the software blocks put in place for the card to run properly and it doesn't run properly ? Big shock there.
The average consumer buying a $200 video card is hardly going to be one that disable the power limit in wattman
I don't think the excess power draw matters from a practical point of view, but try to get the facts straight: as has been widely reported, only the Asus Strix GTX 960 does it, not other boards (incl. the Nvidia reference board.) And I believe even there the average power remains below the limit.
I think that's the most plausible scenario. It came out fine during measurements, and then they cranked up the clocks to make it more competitive after the high clocks of Pascal became known. And they didn't remeasure that part. A RX 470 with the promised, much better perf/W, would point in the same direction.
First, there was the PCIe specification, which for whatever reason, limited 12v on PEG to 5,5A for the regularly found "dumb" PEG (if memory serves).
Then there was violation of the spec. And when you violate a spec, proof of burden is on you that you can do so without any negative consquences.
I agree, if there wasn't a spec with a defined ampere limit, and everyone just had agreed on a 75-ish watt limit for no hard reasons, burden of proof would indeed lay with the people claiming imminent damage.
As Wattman is part of the official driver package and as it seems enough to just set the Powerlimit to 150%, I can not follow your assumption.
I wonder what the
This is not the first graphics card to violate specs though. The only difference is that somehow violating them via the PCIe connector is universally accepted and ok while violating them over PCIe slot is not ok. I wonder who decides which violation is ok and which one isn't.
As for proof, there is no problem until proven otherwise.
This is not the first graphics card to violate specs though. The only difference is that somehow violating them via the PCIe connector is universally accepted and ok while violating them over PCIe slot is not ok. I wonder who decides which violation is ok and which one isn't.
As for proof, there is no problem until proven otherwise.
Ext3h
Regular
Seriously guys. What are you even talking about.
If the slot or any trace on the motherboard is underrated, it's resulting a higher resistance and consequently in an increased power draw via the PCIe connector instead.
Going above 75W on the PCIe slot would only be a real issue if there wasn't a secondary power source to balance. But as it stands, and if you don't forget to plug in the PCIe connector, you simply won't be able measure this type of >75W load on he PCIe slot if the board can't provide it.
Do you all think it's a coincidence that this power draw was only measured when using boards with 2, 3 or even 4 PCIe 16x slots? If the power supply on the board offers less resistance, e.g. simply because there's a total of at least (yes, some of these boards test with have even extra PCIe connectors next to the slots) 6x 12V pins supplying the board , but only 2x 12V on the PCIe connector, it's no surprise that - without active balancing - the majority is drawn via the board.
Measure this on a board which is actually not rated for this current, and you won't be able to reproduce these results.
This only could be a possible problem if the card didn't have an extra connector at all, as the draw couldn't balance any longer. But as long as AMD isn't going to push the PCIe slot-only RX 470 above 75W, I'm not seeing any reason to worry about.
If the slot or any trace on the motherboard is underrated, it's resulting a higher resistance and consequently in an increased power draw via the PCIe connector instead.
Going above 75W on the PCIe slot would only be a real issue if there wasn't a secondary power source to balance. But as it stands, and if you don't forget to plug in the PCIe connector, you simply won't be able measure this type of >75W load on he PCIe slot if the board can't provide it.
Do you all think it's a coincidence that this power draw was only measured when using boards with 2, 3 or even 4 PCIe 16x slots? If the power supply on the board offers less resistance, e.g. simply because there's a total of at least (yes, some of these boards test with have even extra PCIe connectors next to the slots) 6x 12V pins supplying the board , but only 2x 12V on the PCIe connector, it's no surprise that - without active balancing - the majority is drawn via the board.
Measure this on a board which is actually not rated for this current, and you won't be able to reproduce these results.
This only could be a possible problem if the card didn't have an extra connector at all, as the draw couldn't balance any longer. But as long as AMD isn't going to push the PCIe slot-only RX 470 above 75W, I'm not seeing any reason to worry about.
edit: Might be late to the party...
Looking at the table 4.1 of PCI Express Card Electromechanical Specification Revision 3.0, it is quite clear and non-negotiable, that it's 5.5A (+/- 8%) for the 12 volt supply via the slot.
Partner card? Or did a reference card this as well? And no, I would also caution people to buy any other card, when I measure violation of the PCIe specs.
I wonder too. Of course, it is not ok for other cards to violate the specs.
@Tridam says that boot time power supply calibration and anti aging compensation are disabled on the RX 480. I wonder why that is, and how much power-efficiency is really being left on the table.
http://www.hardware.fr/articles/951-2/polaris-10-5-7-milliards-transistors-14-nm.html (last paragraph)
http://www.hardware.fr/articles/951-2/polaris-10-5-7-milliards-transistors-14-nm.html (last paragraph)
D
Deleted member 2197
Guest
D
Deleted member 13524
Guest
Yet everyone seems to be ignoring actual technical knowledge and instead trying to learn something about circuits by looking and at and pasting here that PCIe electrical spec sheet.
The epitome of ridiculousness was probably when I saw a comment saying "It's increasing current output so it's even worse". Because apparently a better solution would be to increase the 12V feed to 14V and screw up all the 12V circuitry?
I'm a bit disappointed at this thread, really.
D
Deleted member 2197
Guest
Except, it wasn't. Our Asus Maximus VII Hero has exactly one PEG×16 slot and we were seeing up to 6.4 Ampere in gaming loads non-overlocked.
Last edited:
You need to look at average demand and distribution rather than trying to judge this from the spur peaks as that is instantaneous bursts that are not usually an issue (although it can cause noise).
It is sustained draw that you are interested in, that would be around the average to maybe a little bit higher, that said the 960 averages under 60W and the 480 averages 80W in those charts.
If you want to analyze the spikes-spurs, you really would need the data and zoom the window so those 1-2ms spikes can be seen in better detail over time, even then due to their behaviour it would be ok.
Ask yourself why Tom's Hardware was OK with the 155W peaks from the 480.
BTW your last chart is mixing two different distributions; you are showing the mainboard ATX/PCI express riser slot for 480 and the auxiliary PEG for the 960.
The auxiliary PEG can take much more punishment as it is fully rated to 8A per 12V contact, the problem for the PCI express slot is that it can be anywhere from 5.5A to the slot riser being tighter to spec at 5.5A IMO, where the ATX12V and PEG connectors are rated higher although quality-price of the PSU and motherboard will be a factor (more so for the ATX12V 24-pin connector) and supplied cables ideally 18AWG and 16AWG (for the 24-pin connector) for the 12V wires.
AMD would had been better going further out of spec over the auxiliary PEG and reduced the demand over the PCI express slot IMO.
This post is excluding mGPU and OC considerations.
Cheers
Last edited:
D
Deleted member 2197
Guest
This won't help in this case. The fact is that the RX 480 pass the PCIe compliance test. My wife checked it, because she is now testing the card for a future OEM PC build. The motherboard makers will also ask the PCI-SIG when they make these kind of decisions, and they don't care about the forums or the reviews.
Also a lot of factory overclocked cards are not pass the PCIe compliance test, and this could be a problem for the whole industry, because for example ASUS don't want to say that the new super ROG card is a nice product, but you might void the motherboard warranty if buy it. These companies make too much money on the factory overclocked product. PCI-SIG might able to do something, but my wife said they just care about the reference products.
- Status
