Fury X voltage adjustment now available

[tviceman]

GPU perf/watt is such a dull argument IMO, how much time are you really gaming? Unless it's 4hours a day...? More like 4hrs a week. So it's really a non issue.

I agree. In the context of Fiji vs. GM200 it's the least important purchasing factor.

BTW, steam says I have 18 hours played in the past two weeks.

 

[monstercameron]

If people want to pay twice as much for their graphics, doubling their heat and power output only to enable higher forms of AA then somewhere a polar bear is reading this thread, getting a haircut, and sharpening his claws.

then you enjoy fxaa man...
This poster is really complaining that superior IQ has a cost associated with it, I don't think s/he has ever heard of diminishing returns.

 

[Mr Evil]

If people want to pay twice as much for their graphics, doubling their heat and power output only to enable higher forms of AA then somewhere a polar bear is reading this thread, getting a haircut, and sharpening his claws.

While I try to keep the power consumption of my PC down, if you're worried about polar bears then VSR is waaaay down the list of things to cut out, far below driving, eating meat, having children, and holidays abroad.

 

[tential]

Downplay vsr however you want but I love the feature. Absolutely think it's amazing. Down sampling should be something supported by both vendors and it's well past overdue for it.

 

[Arachnotronic]

Downplay vsr however you want but I love the feature. Absolutely think it's amazing. Down sampling should be something supported by both vendors and it's well past overdue for it.

I agree, too...DSR/VSR are really cool features.

 

[Sabrewings]

I agree, too...DSR/VSR are really cool features.

Agreed. Absolutely love it.

 

[tviceman]

Downplay vsr however you want but I love the feature. Absolutely think it's amazing. Down sampling should be something supported by both vendors and it's well past overdue for it.

I'm not downplaying it. I'm saying it's not worth an additional $650 + all the extra power consumption and heat that goes along with running dual GPU setups.

 

[Silverforce11]

I'm not downplaying it. I'm saying it's not worth an additional $650 + all the extra power consumption and heat that goes along with running dual GPU setups.

I personally think people who run quad GPU setups are a bit crazy (especially prior gen where CF/SLI scaling beyond 2 GPUs were abysmal).. but its their passion. Whatever rocks their boat.

 

[soccerballtux]

I personally think people who run quad GPU setups are a bit crazy (especially prior gen where CF/SLI scaling beyond 2 GPUs were abysmal).. but its their passion. Whatever rocks their boat.

doesn't AFR scale perfectly, CPU aside? it works great when you're targeting 60fps and each card can do 20 on its own

 

[alcoholbob]

doesn't AFR scale perfectly, CPU aside? it works great when you're targeting 60fps and each card can do 20 on its own

There's barely any scaling past a 2nd card.

Typically at 4k you get scaling like this:

1 card 40 fps
2 cards 60 fps
3 cards 68 fps
4 cards 73 fps

 

[3DVagabond]

There's barely any scaling past a 2nd card.

Typically at 4k you get scaling like this:

1 card 40 fps
2 cards 60 fps
3 cards 68 fps
4 cards 73 fps

It's way better than that.

Firestrike ultra:
1 card = 4025
2 card = 7557
3 card = 10801
4 card = 13605

http://www.eteknix.com/amd-fury-x-quadfire-results/

It’s all well and good buying and flaunting these graphics cards, but what we all want to see is how it fares in a benchmark. Thankfully, that has already been done in the form of 3DMark Firestrike Ultra (4K), along with single increments of each card. With the addition of each card, we see almost perfect scaling, losing only around 12% performance per card per addition

 

[Silverforce11]

Synthetics.

 

[3DVagabond]

Synthetics.

This shows the capability. Typically what reduces scaling is the CPU, not the GPU's (assuming working crossfire profile). With DX11 we are starting to become pretty seriously CPU bound in many games. Can't get DX12 soon enough. I don't know why MSFT waited so long.

 

[XiandreX]

every chips power goes up with voltage as you know, however note that this is a massive and dense die ~600mm^2. applying voltage to the massive shader array is expensive and ~100+mV turns out to be alot of power. Maybe you should check out oc 290xs or 780tis.

The performance gained versus wattage increase via overvolting/overclocking
is terrible imho for the Fury X.
I guess I worded it incorrectly.

 

[XiandreX]

Its because the GPU is wide. Every core needs a portion of power at certain level of frequency. In theory, the wider GPU you have the slower it should be clocked. AMD pushes their GPUs to the limits of reasonable power consumption and performance ratio.

The other side is where you downclock and undervolt the core. Wider GPU with slower cores, will be more powerful in the same power envelope than narrow, but high clocked one.

And thats exactly what differs Maxwell GPUs from AMD ones. AMD gets more power efficient with every step down on the scale of voltage, without loosing significant amounts of performance. I've already brought to this forum link from other sites/threads about full Hawaii being able to be squeezed to 145W of TDP while having 85-90% of nominal performance.

That exact reason is why Green500 list says GCN is the most power efficient architecture that's ever been on this planet.
http://www.green500.org/news/green500-list-november-2014

I know its not exactly relevant to this thread. Just for information .

Thank you very much for your helpful post.

 

[tential]

I'm not downplaying it. I'm saying it's not worth an additional $650 + all the extra power consumption and heat that goes along with running dual GPU setups.

I mean the powerconsumption as you already said... who cares?

An additional $650? Well ya, if you want to max out triple A games on a lower res screen and use VSR sure.

IMO, VSR/DSR provide great upgrade paths. You can 4K VSR/DSR then upgrade to a 4K display when you're ready. I'd get a card now, then get a Vizio P Series refresh when it comes out later this year.

I can't decide what to do personally as no choice seems like a good choice but ya, VSR, especially from double your resolution? Why not....

 

[ocre]

Thank you very much for your helpful post.

You can down clock and undervolt any card. This isnt some special feature for GCN.

The power savings seems impressive for Hawaii because AMD pushed the clocks aggressively. But, no matter what GPU you have...undervolting and undreclocking will use less power.

 

[soccerballtux]

I agree. In the context of Fiji vs. GM200 it's the least important purchasing factor.

BTW, steam says I have 18 hours played in the past two weeks.

OTOH, if I used Crossfire, I would definitely be happy about Fury Nano, both the power used with 2 cards, and because of the heat.

 

[Cookie Monster]

Its because the GPU is wide. Every core needs a portion of power at certain level of frequency. In theory, the wider GPU you have the slower it should be clocked. AMD pushes their GPUs to the limits of reasonable power consumption and performance ratio.

The other side is where you downclock and undervolt the core. Wider GPU with slower cores, will be more powerful in the same power envelope than narrow, but high clocked one.

And thats exactly what differs Maxwell GPUs from AMD ones. AMD gets more power efficient with every step down on the scale of voltage, without loosing significant amounts of performance. I've already brought to this forum link from other sites/threads about full Hawaii being able to be squeezed to 145W of TDP while having 85-90% of nominal performance.

That exact reason is why Green500 list says GCN is the most power efficient architecture that's ever been on this planet.
http://www.green500.org/news/green500-list-november-2014

I know its not exactly relevant to this thread. Just for information .

Let me ask you some questions. Why is Fiji/Hawaii considered "wide" and Kepler/Maxwell "narrow" (if that is what you were trying to imply)? Because ones clocked higher and the other?

And where are you getting 145W of TDP while having 85-90% performance figures? For perspective, a R290X has a TDP of 250W. GTX970 has a TDP of 145W. Both according to TPU perf summary is pretty damn close in performance@stock.

Id argue that GCN isn't the most power efficient architecture especially in 3D applications. But thats for another time..

 

[tviceman]

OTOH, if I used Crossfire, I would definitely be happy about Fury Nano, both the power used with 2 cards, and because of the heat.

I get what you're saying and you are right about lower power cards; that is the beauty of being power efficient.

But lets wait for Nano reviews. It could be less efficient than AMD had led on. They did, after all, lead us to believe that Fury X was the fastest most efficient GPU ever made when in fact it's clearly neither.

 

[showb1z]

And where are you getting 145W of TDP while having 85-90% performance figures? For perspective, a R290X has a TDP of 250W. GTX970 has a TDP of 145W. Both according to TPU perf summary is pretty damn close in performance@stock.

Id argue that GCN isn't the most power efficient architecture especially in 3D applications. But thats for another time..

TDP isn't the same as power consumption. No doubt Maxwell is more efficient than GCN, but the difference isn't as big as those numbers make it look like:

100W becomes 50W in the real world. Properly cooled GCN is quite a bit more efficient.

 

[Glo.]

Let me ask you some questions. Why is Fiji/Hawaii considered "wide" and Kepler/Maxwell "narrow" (if that is what you were trying to imply)? Because ones clocked higher and the other?

And where are you getting 145W of TDP while having 85-90% performance figures? For perspective, a R290X has a TDP of 250W. GTX970 has a TDP of 145W. Both according to TPU perf summary is pretty damn close in performance@stock.

Id argue that GCN isn't the most power efficient architecture especially in 3D applications. But thats for another time..

http://forums.macrumors.com/threads/nmp-2015-where-is-it.1849662/page-8#post-20880421

Power efficiency is an oft-used negative against the large-die Hawaii chips, but I've been playing with powertune settings and Furmark recently as an experiment to fit a "hot and noisy" AMD card into an SFF with limited cooling.

Actually, I stand by an earlier post I made that says I think AMD pushed Hawaii silicon too far.
With both GPU-Z and Furmark able to report power consumptions, I can see a 100W reduction in power consumption on 290X cards for as little as 5% performance loss.

If you have a Hawaii card, I urge you to crank power limits down in the overdrive tab of CCC and see what the resulting clockspeed is under full load. Even in a worst-case scenario, I'm seeing a typical clockspeed of 850MHz with the slider all the way to the left at -50%

That means that Hawaii (the two samples I personally own, at least) can run at 850+MHz on only 145W (half the 290W TDP). As mentioned, that's a worst-case scenario using a power-virus like Furmark. Under real gaming situations (I was messing around with Alien Isolation on 1440p ultra settings) the clocks averaged about 925MHz yet my PC was inaudible; Fans that normally hum along at 55% were barely spinning at 30% during my gameplay.

As Nvidia has proved, you can make a 28nm chip run efficiently. I think the design of Hawaii holds up very well under vastly reduced power constraints - AMD just pushed it outside its comfort zone in order to get the most out of it.

In saying that, the "underpowered" 290X is around the same performance as my GTX970 and also the same cost - significantly higher than a GTX960 4GB. I don't know if die-harvested 290 cards deal with power limit caps like the cherry-picked 290X cards.

People also undervolted and lowered the TDP in programs of Fury X to 225W. And it was able to run on 1035 MHz at that TDP. 225W and 1035 MHz. 1000 MHz doable for full Fiji chip at 175W? From 225W you get 8.5 TFLOPS of compute power. Nothing in the entire world comes close to this power efficiency. 37 TFLOPS from 1W.
http://linustechtips.com/main/topic/405259-fury-x-and-the-r9-nano-experiment/

I think that AMD blocked Fury X a bit because of excessive temperature on VRMs. Also to note that the 225W of TDP may be wrong, because he thinks its 375 counted from dual 8 pin connectors. The real TDP may be 275W and in reality that 40% reduction can account for 175W - exactly the same as Fury Nano. But that is only my theory.

 

[Silverforce11]

Crypto-coin miners knew full well the potential efficiency gains with an undervolt/downclock on GCN. Its not new. I got my R290 down to ~160W @ ~880mhz from its stock of 947mhz. That's from ~230-240W gaming load. 10% performance drop for ~80W less usage, or an improvement in perf/w to close to 970 levels.

The TPU article even showed Fury X with an undervolt + OC combo using less power than stock Fury X.

For Nano, all they have to do is bin for chips with lower vcore, full dies running lower clocks, bang, massive perf/w gains. The interesting result will be whether user undervolting on top of that, can improve perf/w even further or its at its limits. Would make for a great mITX GPU if price is right.

 

[Glo.]

The problem with that TPU test is that they tested it with only one game, which completely don't react to AMD hardware. I would like to see what OC does in Far Cry 4, for example.

 

[tential]

The problem with that TPU test is that they tested it with only one game, which completely don't react to AMD hardware. I would like to see what OC does in Far Cry 4, for example.

I want more extensive testing done but I'm not holding my breath about the results.