Undervolting the RX5700 is back on the menu boys and grils

[exetras]

https://translate.google.com/translate?hl=en&sl=de&tl=en&u=https%3A%2F%2Fwww.hardwareluxx.de%2Findex.php%2Fartikel%2Fhardware%2Fgrafikkarten%2F50149-zweimal-rdna-als-navi-die-radeon-rx-5700-und-radeon-rx-5700-xt-im-test.html%3Fstart%3D25&sandbox=1

 

Quote

We have also devoted ourselves to undervolting, partly because AMD GPUs here often have great potential for improvement.

For undervolting we achieved the above results. For example, with the Radeon RX 5700, we were able to reduce the voltage from 0.987 to 0.937 V at almost the same clock - the power consumption drops by about 7.5%. At the same time the GPU temperature dropped by 9 ° C and the fans could turn slightly slower.

For the Radeon RX 5700 XT the undervolting even increased the clock, but we could not keep up the power consumption. However, more power with the same power consumption is certainly also to be judged as a positive result. At the same time, the GPU temperature drops, even though we let the fans spin at the same speed to keep the clock going.

 

Personal note:

With better drivers in the future and Undervolting we should be able to get some nice boost from the RX 5700 series. A bit like the Vega 56, which started low and ended high. Getting a partner RX should make this even more appealing as cooling will be exponentially better.

 

 

 

[Doobeedoo]

Very interesting, I would've expected better optimized out of the box then maybe. Oh well we'll see how it goes with new drivers over time. 

[bcredeur97]

it's so interesting how you can get more performance from undervolting just because of reduced TDP

 

easy to think "oh I need more performance let me throw MORE volts at it." but this clearly shows that isn't always the best idea

[ch3w2oy]

1 hour ago, bcredeur97 said:

it's so interesting how you can get more performance from undervolting just because of reduced TDP

 

easy to think "oh I need more performance let me throw MORE volts at it." but this clearly shows that isn't always the best idea

It's mostly due to lower voltage creating more thermal headroom.. The cooler the card, the higher it can clock.. 

[justpoet]

More voltage is usually only helpful when there's plenty of extra heat dissipation.  Undervolting a blower makes a lot of sense.

[Trixanity]

I still don't understand how AMD hasn't solved this problem one way or another. Running at massive voltages ootb is getting silly.

[The King of the Undead]

why you callin grils?

[Orcblood]

12 minutes ago, Trixanity said:

I still don't understand how AMD hasn't solved this problem one way or another. Running at massive voltages ootb is getting silly.

I think it has more to do with their chips varying alot in power requirements. They'll try to strike the best voltage/frequency balance for all cards but that'll mean there will be some cards that require less voltage that will benefit more from undervolting and other cards requiring more voltage that will benefit more from overvolting. Then you'll have those cards where the gpu power requirements are right in line with what AMD defaults it to, meaning you lost the lottery for extra self discovered performance if you get one of those 

 

The reason we don't see this with Nvidia is because they have so many different SKUs of a specific chip so undervolting and overvolting is alot less striking. 

[Dabombinable]

5 hours ago, bcredeur97 said:

it's so interesting how you can get more performance from undervolting just because of reduced TDP

 

easy to think "oh I need more performance let me throw MORE volts at it." but this clearly shows that isn't always the best idea

I've never actually needed to overvolt a graphics card, because they always hit their air-cooled clock speed limits at stock. Overvolting did nothing for them at all other than make them run hotter.

[Agosto]

These tests actually show that Navi cards aren't as factory overvolted as Vega and Polaris cards were. It also looks like AMD went for a more nvidia-like approach with total board power, with much tighter control.

[sof006]

Always found this a bit strange as you would assume undervolting would cause instability when the clocks increase. Guess not.

[GOTSpectrum]

2 minutes ago, sof006 said:

Always found this a bit strange as you would assume undervolting would cause instability when the clocks increase. Guess not.

It doesn't work on all chips, AMD pushes high voltage as standard to stabilise lower quality chips that would otherwise have to go on the reject bin. 

 

AMDs spec is much more forgiving than nvidias in that regard, they have a much wider window of acceptability.

[exetras]