AMD To Issue Software Fix To Address RX 480 Power Consumption Problems

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.


No idea how they test it. IIRC AMD mentioned the testing in the RX480 AMA.
 
So the computer world was afire with the 970 3.5Gb issue, and then AMD sends out a card that draws so much power through the Mobo, it breaks the mobo. I wish the card makers would stop with this, and...I hope fate makes this as big of an issue, because I bought a 970 an my performance never suffered. Internet? you gonna string this one up too?
 
If there are so many brilliant people with all the answers, why don't you take a job with AMD, save the company, become famous, CEO and make millions while you laugh at the poor stupid people over at Nvidia and Intel?
 
Problem is for some older MB and not related to new products, also AMD is quick to fix it and i doubt the performance will be changed since they just take less power from the slot and more from the 6 pin connector.

Some blame AMD but as you know the PC industry is vast, it`s simply impossible to test new products with everything out there. That`s why even with MB manufacturers you will see that in their support documents you can see only a select number of products that are shown as compatible, in theory everything goes but on some occasions there might be issues.

Example, Canon 1DX Mk2 having issues with pictures saved on certain Sandisk memory cards ... And if this issues is solved easy by a driver update or BIOS there shouldn`t be a big fuss about it.
 


This is not true.

If you actually read the spec, it states that in section 6.9...

An adapter must never consume more power than what was specified in the most recently
received Set_Slot_Power_Limit Message or the minimum value specified in the corresponding
form factor specification, whichever is higher.

This Set_Slot_Power_Limit is controlled by the motherboard, not the graphics card.

As the graphics card can't tell the motherboard what type of PCIe power limit it has!

And indeed, nearly all mainstream motherboards use the default slot power limit of 75.

The graphics card is exceeding spec, as the graphics card doesn't choose the Set_Slot_Power_Limit.

The PCIE spec DOES NOT allow for ANYTHING REMOTELY CLOSE TO 300W.

The PCI-E spec allows for a MAXIMUM of 12V @ 5.5A = 66W. Considering an 8% voltage tolerance that goes up to 71.3W.

https://www.reddit.com/r/Amd/comments/4qfwd4/rx480_fails_pcie_specification/d4vekzc

That comment explains it perfectly. The '300W' spec is including a 6 pin and an 8 pin with the slot. That's 75+75+150.
 

I'm sure AMD already has many brilliant people whose advice gets readily ignored by senior staff. You don't become CEO of a company overnight, people with over a decade of seniority who are dead-set in their ways usually get picked for that, not genius upstarts who want to reinvent everything and shoot for the moon.
 
Why are guys talking performance costs? AMD can simply do the same as Nvidia with last years 960 issue. Move the extra power draw from the PCI slot to the PCI-E power plug. The GTX960 in the test after all was only 1 watt less the RX480
 

Assuming the board is physically capable of doing so and that the AUX MOSFETs are capable of handling their TDP going up by ~20%.

With the FETs already running at 96C, even a 10% increase on the AUX FETs would push the AUX FETs in the neighborhood of 105C and that's the board temperature, junction temperatures would be much higher still. Power semiconductors may be rugged but you still don't want to run them at their 150-175C maximum junction temperature.
 
https://pcisig.com/specifications

According to the standard the PCI E 225 rev. 1 was established back in March, 2008 @ 300 watts for the slot. They also verified that going over 75 Watts was not a problem.

Think of it this way. You can build a motor to push out 600 HP and put that motor to run at 200 HP. Just because someone makes that motor go to 210 HP isn't going to make it explode instantly.
 
nVidia and AMD really cannot afford for this to damage AMD's image. While I remain a nVidia only user, I understand the need to have a viable competitor to keep other companies honest, and without AMD, nVidia would start to rot from the inside out. It happened with Intel back a decade ago or so.
 

With 8Gb chips costing more than twice as much as 4Gbits one and GDDR5 costing twice as much as regular DRAM does, that's ~$10 of the board manufacturers' margins getting pissed away right there.
 
Samsung discontinued their previous generation of 4Gbits chips and the new generation is at the "production samples" stage. I guess availability is the main reason for the 8GT/s chips on 4GB boards.

Unless AMD specified Samsung as mandatory for the reference boards, AIBs could have used Micron or other 7GT/s GDDR5.
 


I wish people would stop trying to compare the two "issues" as equal. Nvidia missed a paper spec detail (on purpose or on accident is debatable). The GPU still technically *has* 4GB VRAM, it's just that .5GB of it is not directly accessible. The humorous thing is that the majority of 970 owners didn't even know there was an issue until they read about it...then they started complaining about being cheated. I know...I was one of them. But then I got over it pretty quick and gamed on.

And has been reported countless times by tech websites including Tom's, it takes a LOT of effort to make the "issue" even show up. Even at 1440p there are only a handful of games that use more than 3GB VRAM anyway. Nobody's motherboards got fried.
 
Good luck with that. 3 out of the 6 VCore VRMs are tied directly to the PCI-E 12 Volt pins. That is why the slot supplies half of the power to the card. Since the biggest draw of power at stock speeds is 27% over the spec, AMD is going to need to under volt by 30% at least to stay within the slot specs:
50 minute mark and beyond
https://www.twitch.tv/buildzoid/v/75850933
 

First off, power is not linearly related to voltage; it's proportional to voltage squared. So decreasing voltage a little can have a large effect on power. And undervolting would make the card run cooler, and cooler cards draw less power everything else equal.
2nd, they likely have the option to simply make the power delivery asymmetric, with the 6 pin connector VRMs delivering more power per phase than the other VRMs. I don't know if this is a good option, or something they'd do, but it's a possibility.
 
The link to the AMD site where people were complaining about their motherboards getting fried was very interesting. It seemed all the motherboards that were fried were ASRock motherboards or did I read that wrong? Either way it seems best not to overclock the 480 until more testing is done.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom