A temporary fix for the excess PCI-E slot power draw for the reference RX 480 cards - Page 16 - Overclock.net

Forum Jump: 
Reply
 
Thread Tools
post #151 of 178 Old 07-07-2016, 02:28 PM
 
bluezone's Avatar
 
Join Date: Jan 2013
Posts: 509
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 47
PCPER assessment of new driver.
Quote:
Starting with the 16.7.1 driver, AMD will automatically reprogram the power controller on the RX 480 to better divide the power draw from the 6 available phases on the reference cards. This is a process that occurs at each and every boot, it is not a permanent change to the VBIOS. It’s possible, and likely, that future cards and partner cards may have this change integrated at a lower level, negating the need for the driver to recognize and update the controller logic. But for now, with launch RX 480s in the wild, that’s how the process works.

As I understand, what AMD is doing now is very similar to what The Stilt in the Overclock.net forums attempted earlier in the week. Power phases 1-3 that source +12V from the 6-pin connection are now given more time than phases 4-6, thus shifting the weight of power draw towards the 6-pin connector. We’ll be able to calculate that exact ratio when we show you the power consumption data from our testing, but the goal is draw less power from the PCI Express slot and more over the 6-pin connector while maintaining the exact same power and performance profiles. To be clear: this fix will not affect performance and my testing shows that to be the case.

You have to wonder why this wasn’t the direction taken by AMD engineers initially. As you will find in our results, the 6-pin connection is definitely drawing more than the 75 watts that it is rated at, but the 6-pin cabling and connectors are actually rated at 8-9A per pin, rather than 5.5A total for the PEG slot. If you are going to draw more power than rated over one of the two options, it’s clear that the over engineered and directly fed 6-pin connection is the way to go. I can only assume it was an oversight on the board team that allowed it to happen. (Also, let’s not forget that using an 8-pin connection and weighting it towards that would have prevented both issues.)

http://www.pcper.com/reviews/Graphics-Cards/AMD-Radeon-RX-480-Power-Consumption-Concerns-Fixed-1671-Driver
bluezone is offline  
Sponsored Links
Advertisement
 
post #152 of 178 Old 07-07-2016, 02:38 PM
3.5ghz
 
NomNomNom's Avatar
 
Join Date: Apr 2010
Location: Canada
Posts: 1,557
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 65
Quote:
Originally Posted by The Stilt View Post

Thanks, good to know.

Obviously, but there is really no other way they could have designed the PCB and still comply (or rather try to comply) with the PCI-E SIG specifications. If the marketing dept. demands that the card must have a 6-pin connector so that they can advertize it as a power efficient product, how else are you going to split the power draw between the PCI-E slot and the PCI-E power connector but 50/50? Unless of course you are willing to violate either of the PCI-E slot or the PCI-E power connector significantly, that is.

If they would have used a 8-pin PCI-E power connector on the card, they could have fed all of the GPU phases from it and still remain well within the PCI-E SIG specs. The other remaining power planes, which are significantly lower power could have been then fed from the PCI-E slot. That's the configuration cards like the reference Hawaii and Fiji based cards use. They draw no power for the GPU VRM from the slot, but instead feed the secondary planes (i.e VDDCI, VDDM, display interfaces) from it. That's why the PCI-E slot power draw on these cards stays at moderate levels and in nearly static level despite the varying load.

I personally believe that both the variation between the different silicon specimens, and the power consumption of the Polaris 10 ASICs ended up being higher than AMD expected. I'd expect that the power draw will be somewhat lower in the upcoming production (maybe in Q4/2016), once the yields have slightly improved.


Is there anyway of replicating this with your i2c commands?
I noticed that you said you can set the value all the way to F, how much of a change would that be?
Reason being is that I want to be able to reduce the power draw of the pci-e slots to under 20w per slot, as I have 6 gpus on a board.
"The recommended programming value is "B" (/wi06,08,1E,BB and /wi06,08,1F,B0), however you can go set it to anywhere between "B" and "F""

Thanks,

NomNomNom is offline  
post #153 of 178 Old 07-07-2016, 02:56 PM - Thread Starter
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,166
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 744
Quote:
Originally Posted by NomNomNom View Post

Is there anyway of replicating this with your i2c commands?
I noticed that you said you can set the value all the way to F, how much of a change would that be?
Reason being is that I want to be able to reduce the power draw of the pci-e slots to under 20w per slot, as I have 6 gpus on a board.
"The recommended programming value is "B" (/wi06,08,1E,BB and /wi06,08,1F,B0), however you can go set it to anywhere between "B" and "F""

Thanks,

Unfortunately such change is utterly impossible. In the very best case you'll get to 65W power draw from the PCI-E slot per card, if you don't overclock and use "F" (instead of "B") as the programming value. In order to reach < 65W levels, while keeping the performance anywhere near the default levels you need to extensively physically modify the card.

Alternatively you can modify the the motherboard itself by soldering some assisting wires to 12V pins of couple of the PCI-E slots (to the back side) and connecting the wires (power and ground) to a 12V source from a molex connector for example. Or the third (much cleaner) option would be using a product like this: EVGA Power Boost.
The Stilt is offline  
Sponsored Links
Advertisement
 
post #154 of 178 Old 07-07-2016, 02:57 PM
 
Join Date: Jan 2015
Location: Lurking over a keyboard
Posts: 5,836
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 725
@The Stilt

No worries smile.gif .

I agree and understand smile.gif , shame it just went this way, would have been a great launch if it was 8 pin and better in terms of normal and OC'ing for power sourcing/usage aspect IMO.

Yeah lets hope yields improve and AIB cards do power sourcing the better/past way smile.gif .
gupsterg is offline  
post #155 of 178 Old 07-07-2016, 03:00 PM
733mhz
 
The Mac's Avatar
 
Join Date: May 2011
Location: Boston
Posts: 2,465
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 122
it doesnt need 8 pin.

the spec has been 150 watts on the first PEG since PCIE 2.0 introduction.

The extra 2 grounds on the +2 are just for resistance reduction and sense.

unless you are using a crappy pre-pcie 2.0 power supply, then you might be in trouble.

not to mention the fact that AMD has wired up the traces on the peg connector to contain the two extra grounds.

The Mac is offline  
post #156 of 178 Old 07-07-2016, 03:01 PM
3.5ghz
 
NomNomNom's Avatar
 
Join Date: Apr 2010
Location: Canada
Posts: 1,557
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 65
Quote:
Originally Posted by The Stilt View Post

Unfortunately such change is utterly impossible. In the very best case you'll get to 65W power draw from the PCI-E slot per card, if you don't overclock and use "F" (instead of "B") as the programming value. In order to reach < 65W levels, while keeping the performance anywhere near the default levels you need to extensively physically modify the card.

Alternatively you can modify the the motherboard itself by soldering some assisting wires to 12V pins of couple of the PCI-E slots (to the back side) and connecting the wires (power and ground) to a 12V source from a molex connector for example. Or the third (much cleaner) option would be using a product like this: EVGA Power Boost.


According to this article: http://www.tomshardware.de/amd-radeon-rx-480-polaris-grafikkarte-treiber,testberichte-242146.html


16.7.1 managed to drop it to 79w, so your value of "F" should be more aggressive in shifting the power load?
I.E (/wi06,08,1E,FF and /wi06,08,1F,F0)
Also, does undervolting the core drop load from the pci-e bus? Or is the PCI-E power draw strictly reserved for the VRAM?

Thanks,

NomNomNom is offline  
post #157 of 178 Old 07-07-2016, 03:01 PM - Thread Starter
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,166
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 744
Quote:
Originally Posted by The Mac View Post

it doesnt need 8 pin.

6 pin spec has been 150 watts since PCIE 2.0 introduction.

unless you are using a crappy pre-pcie 2.0 power supply, then yo might be in trouble.

That's the reality of course, but companies such AMD need to comply (at least try to) with the specifications wink.gif
The Stilt is offline  
post #158 of 178 Old 07-07-2016, 03:06 PM
New to Overclock.net
 
JackCY's Avatar
 
Join Date: Jun 2014
Posts: 6,210
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 280
Quote:
Originally Posted by The Stilt View Post

Unfortunately such change is utterly impossible. In the very best case you'll get to 65W power draw from the PCI-E slot per card, if you don't overclock and use "F" (instead of "B") as the programming value. In order to reach < 65W levels, while keeping the performance anywhere near the default levels you need to extensively physically modify the card.

Alternatively you can modify the the motherboard itself by soldering some assisting wires to 12V pins of couple of the PCI-E slots (to the back side) and connecting the wires (power and ground) to a 12V source from a molex connector for example. Or the third (much cleaner) option would be using a product like this: EVGA Power Boost.
Or have a mobo that already has that by default without any extra.
Or a powered riser no?
JackCY is offline  
post #159 of 178 Old 07-07-2016, 03:08 PM - Thread Starter
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,166
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 744
Quote:
Originally Posted by NomNomNom View Post

According to this article: http://www.tomshardware.de/amd-radeon-rx-480-polaris-grafikkarte-treiber,testberichte-242146.html


16.7.1 managed to drop it to 79w, so your value of "F" should be more aggressive in shifting the power load?
I.E (/wi06,08,1E,FF and /wi06,08,1F,F0)
Also, does undervolting the core drop load from the pci-e bus? Or is the PCI-E power draw strictly reserved for the VRAM?

Thanks,

Increasing the value will direct more load toward the PCI-E connector, but the difference between 0xB and 0xF is extremely minor and AMD is already using 0xD in the 16.7.1 driver. Lowering the GPU power draw will definitely lower the power draw from the bus, but not by the same amount as the total consumption decreases (due VRM being split 50/50). So if you manage to decrease the GPU power draw by 10W, with "F" programming value you should see around 6W decrease in power draw from the PCI-E slot (5W at stock settings, i.e with "0" programming value).

The memory VRM is fed only from the PCI-E power connector, so the memory clocks don't affect the power draw from the PCI-E slot significantly (most likely the difference is measured in mW).
The Stilt is offline  
post #160 of 178 Old 07-07-2016, 03:09 PM - Thread Starter
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,166
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 744
Quote:
Originally Posted by JackCY View Post

Or have a mobo that already has that by default without any extra.
Or a powered riser no?

Both will definitely do the trick.
The Stilt is offline  
Reply

Quick Reply

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off