The Fury(X) Fiji Owners Thread

[JediFragger]

Was funny once, boring twice, and getting annoying after that...

Okok I'll stop being an arse

 

[Firefox35]

Was funny once, boring twice, and getting annoying after that...

Maybe just maybe if AMD didn't come out with so much BS, things like this would not happen

 

[TNA]

Okok I'll stop being an arse

If after voltage unlock it does not overclock well, then it will be well deserved. Hate all that lying crap.

 

[JediFragger]

If after voltage unlock it does not overclock well, then it will be well deserved. Hate all that lying crap.

Agreed! Although I personally think the guy just blurted it out without any real solid info (except the cooler being great) and now everyone has taken it to heart...

It's obvious that they're running it near to the ragged edge as it is!

 

[Mtom]

Agreed! Although I personally think the guy just blurted it out without any real solid info (except the cooler being great) and now everyone has taken it to heart...

It's obvious that they're running it near to the ragged edge as it is!

What makes you think that? So far it seems its just about the same in overclocking as the hawaii chips. Those usually go to around 1150 on stock volts as well. With voltage they could really go higher...i had mine to 1200 24/7, and 1280mhz while benching...and that is on air....the ones under water could run 1200-1250 easily 24/7, except if you have a really crappy chip.
I think 1250 is totally doable on the AIO cooler.

I'm more interested in this memory oc thing...someone say it doesn't do anything, others posting great improvements.

 

[flopper]

Agreed! Although I personally think the guy just blurted it out without any real solid info (except the cooler being great) and now everyone has taken it to heart...

It's obvious that they're running it near to the ragged edge as it is!

Not sure why Macri said that, normally you say its built solid with back end power to support stable power flow for those that do overclocking.
The only way to make sense of that statement he did was to say, well it overclocks better than the 980ti then. Thats a big mistake to even say and Macri should stay with memory stuff and not marketing.
Its an amatuerish thing to say.

 

[Gregster]

I don't think with voltage being able to be applied we will be seeing big overclocks. With stock voltage mine isn't very good but I do hope I am wrong.

 

[Mtom]

I don't think with voltage being able to be applied we will be seeing big overclocks. With stock voltage mine isn't very good but I do hope I am wrong.

Well this remains to be seen..my hawaii clocked well on voltage, but it didn't on stock volts. This is related to the chip leakage i think.

These are the voltages it ran.
Stock voltage 1.14-1.15
1100 +25mv (1.16-1.17)
1170 + 100mv (1.22-1.24)
1200 + 150mv (1.26-.127)
1250 + 230mv (1.3-1.32)


BTW about vram oc...with the hawaii chips, the vram oc depended on the chip (memory controller voltage)...i wonder if its the same here..if so, then voltage unlock can help vram oc as well.

when i ran my 290X at 1100 (+25mV), i could oc the vram to 1450, but is crashed instantly @1500.
@1170 + 100mV i could raise the vram clock to 1550, but crashed at 1600
@1200 +150mV it ran flawless 1650mhz.
In benches @ 1280 +250mV vram topped out at 1715MHz

This was a lightning card for the record.

 

[Dygaza]

I seriously hope we will get official ways to overclock memory. The way I see it from user benchmarks, AMD is shooting themselves in the foot if they don't allow it.

When it comes to current overclocking, can we even be sure that increasing power target actually does anything? I mean the card is throttling it's clocks in furmark, and I can't see it doing that because of temperatures. So might be powerlimit being the factor there.

Maybe the reason Fury X scales so well with memory speed, might be found from Anandtechs analyze.

http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/4

Fiji’s ROP performance is getting turbocharged for two major reasons. The first is GCN 1.2’s delta color compression, which significantly reduces the amount of memory bandwidth the ROPs consume. Since the ROPs are always memory bandwidth bottlenecked – and this was even more true on Hawaii as the ROP/bandwidth ratio fell relative to Tahiti – anything that reduces memory bandwidth needs can boost performance.

http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/23

At this point R9 Fury X’s ROPs are pushing more than 40 billion pixels per second, a better than 2x improvement over the R9 290X despite the identical ROP count, and an important reminder of the potential impact of the combination of compression and HBM’s very high memory bandwidth. AMD’s ROPs are reaching efficiency levels simply not attainable before.

 

[humbug]

I think throttling in Furmark is deliberate, and about time IMO.

Furmark stresses the GPU WAY OVER natural use, as a result it uses about 50% more power than it would under anyother user circumstance. Nvidia have been throttling Furmark for years.

Some reviewers use it to measure power consumption and then troll about how a 290X uses 350 watts when a user will never get anywhere near that.

IMO its about time AMD followed Nvidia's lead in this and added a Furmark throttle script to stop a small number of reviewers trolling.

 

[MDPlatts]

Beats whyscottys titan score for heaven 3440x1440 by 10% - the only other submitted result

1134MHz - tried 1140 but it didnt work.

3440 x 1440 - LG 34UM95P

AMD R9 FuryX @ 1134/500 - Catalyst 15.15.1004 launch drivers
Intel 4770K @ 3.5GHz Stock (+Turbo upto3.9GHz)

FPS: 35.9
Score: 905
Min FPS: 17.0
Max FPS: 74.3

 

[pmc25]

Very interesting how OCing the mem makes the diff this time, is there any word on a GPU-Z update to read it properly?

I suspect Fiji is awfully close to hitting bandwidth limits at stock clocks, in some games or certain scenarios, at 1440p and above. The Hardware.fr tests were conducted at 1440p rather than 4K ... if it had been at 4K, I expect the effect of the memory overclock (combined with GPU overclock) would be even greater.

 

[Gregster]

I do wish people would leave Furmark alone. It isn't good to use.

 

[Dygaza]

My point bringing furmark in discussion was a test where they used oc'ed (power limit 50% increase) and stock fury x. They both consumed same amount of wats. So my real point here is: Is increasing power limit from ccc actually even working atm? So that might be limiting fury's performance when you try to overclock it.

 

[Gregster]

Dirt Rally Fury X Vs Titan X (prefer max quality set in the NCP) 1080P bench.

 

[Martini1991]

I know Youtube's probably the worst thing one can use for a comparison, but there's definitely a difference in perceived performance between the Fury and Titan there.

 

[Boomstick777]

I do wish people would leave Furmark alone. It isn't good to use.

Agreed, I think now that clock throttling is baked into to new cards it means Furmark does not represent a true max TDP anyway.

So it has become a pointless potentially risky test for temps etc.

A long run of Heaven Valley can provide a good indication of temps. Would recommend that over Furmark.

 

[pmc25]

Agreed, I think now that clock throttling is baked into to new cards it means Furmark does not represent a true max TDP anyway.

So it has become a pointless potentially risky test for temps etc.

A long run of Heaven Valley can provide a good indication of temps. Would recommend that over Furmark.

I think Furmark's ok if you have a watercooled setup just to test stability of new OC clocks briefly. Testing TDPs, as you point out, is stupid. Doing anything with it on air-cooled cards (particularly today's big-die monsters) is nuts.

 

[humbug]

Dirt Rally Fury X Vs Titan X (prefer max quality set in the NCP) 1080P bench.

Most area's the TX is about 10 FPS ahead but some area's the FX is about 10 FPS ahead.

it shows a difference in architectures.

 

[humbug]

My point bringing furmark in discussion was a test where they used oc'ed (power limit 50% increase) and stock fury x. They both consumed same amount of wats. So my real point here is: Is increasing power limit from ccc actually even working atm? So that might be limiting fury's performance when you try to overclock it.

Don't need FurMark to test Power Tune.

If AMD have scripted to throttle Furmark Power Tune isn't going to work anyway

FutureMark 3DMark will push the GPU as much as any game, its quite good for testing Power Tune, Power Consumption and Thermal Properties.

I did it here a while ago, Power Tune does exactly what its supposed to do, it works very well.

http://forums.overclockers.co.uk/showthread.php?t=18620924

I really don't get why people including reviewers keep falling back to Furmark, its not representative to anything other than to its self so using it to as a universal measurement tool the way some do is idiotic, IMO.