Radeon R9 285 3DMark FireStrike Performance Revealed

[btarunr]

AMD put out its own 3DMark FireStrike performance numbers for its soon-to-launch Radeon R9 285 performance-segment graphics card. Running on a test-bed powered by a Core i7-4960X six-core processor and 16 GB of DDR3-1866 memory, the R9 285 scored P7066 (performance preset) and X3513 (extreme preset); compared to an NVIDIA GeForce GTX 760 (337.88 WHQL driver) scoring P5650 and X2774, on the same test-bed. For a card that's in the same price-range (the R9 285 will start at $249) as the GTX 760 (currently going for as low as $239), that's an impressive performance uplift. The R9 285 will start selling on the 2nd of September, 2014.



View at TechPowerUp Main Site

 

[FrustratedGarrett]

What are the FireStrike scores of the R9 280X?

 

[Sempron Guy]

^around 9-10k range I believe

 

[SecretaryKennai]

^around 9-10k range I believe

Around 3500 too. X score is for extreme, 9-10k is standard firestrike. So it might be a slight performance uplift from the tahiti GPU on a 256 bit bus, and consumes power between a 7870 and 7950 which comprises a roughly 40W range between 100 and 140.

 

[nunomoreira10]

not that bad, aroud the same perfomance as an overclocked r9 280

 

[Big_Vulture]

How is it going against 750ti Maxwell? Any more efficient?

 

[Steevo]

How is it going against 750ti Maxwell? Any more efficient?

While we are at it we will also calculate how well a smart car does against a Tesla. Then a prius against a ferrari, after which we will perform a test of a city bus against a tank.

People who are concerned with that level of efficiency aren't concerned with 20 more watts, but it is supposed to use significantly less power yes.

 

[Patriot]

How is it going against 750ti Maxwell? Any more efficient?

Last I checked the 265 was just as efficient as a 750ti ... it just is a lot more powerful.

At 190w for 3.3Tflops the 285 is on par with the 750ti at its 1.3Tflops.
Scaling efficency up to that big of a die is nice. Curious what the DP performance is.
I want this for Distributed computing if it has good DP.

 

[ensabrenoir]

.....this is cool and all but theyre aiming at a gpu thats going to be replaced shortly. Sorta like chevy coming out with a comercial saying their new Maliabu is better than a 2010 toyota camary. Momentary victory only.... Coming out with a card that matches or beats price and efficiency of the 750ti and faster than a 770 or something would be awesome. But still nice none the less.

 

[arbiter]

.....this is cool and all but theyre aiming at a gpu thats going to be replaced shortly. Sorta like chevy coming out with a comercial saying their new Maliabu is better than a 2010 toyota camary. Momentary victory only.... Coming out with a card that matches or beats price and efficiency of the 750ti and faster than a 770 or something would be awesome. But still nice none the less.

Yea Given how AMD tends to do benchmarks and rls them i put little faith in their word or graphs they put out. On top of that its synthetic benchmark and i put very little faith in those as they don't really mean much in real world use. I think if they wanted to do right comparison it would be vs 770 as new 800 series are due in a month or so the competing card should be around there or so.

 

[natr0n]

AMD using competitions processor to test unreleased gpu.

Seems odd

 

[Fourstaff]

AMD using competitions processor to test unreleased gpu.

Seems odd

That was my first thought too. However, you wouldn't want to hold back your product because you refuse to use your competitor's system.

 

[lastcalaveras]

looking forward to the 4gb edition they mentioned also it'd be nice if it has no crossfire bridge connectors.

 

[BiggieShady]

Last I checked the 265 was just as efficient as a 750ti ... it just is a lot more powerful.

At 190w for 3.3Tflops the 285 is on par with the 750ti at its 1.3Tflops.
Scaling efficency up to that big of a die is nice. Curious what the DP performance is.
I want this for Distributed computing if it has good DP.

Please, please understand that efficiency is the same thing as performance per watt metric. As you can see on this graph, Maxwell's efficiency is in its own league.

 

[arbiter]

Last I checked the 265 was just as efficient as a 750ti ... it just is a lot more powerful.

Please, please understand that efficiency is the same thing as performance per watt metric. As you can see on this graph, Maxwell's efficiency is in its own league.

Um how is r7 265 that has a listed TDP of 150watts just as efficient as 750ti is so lower power you can get away without a 6pin PCI-e connector? r7 265 is faster but 750ti isn't that much slower and as the graph proves much better in efficiency wise.

 

[|cpm]

Anonymous (smell like NDA) sources told Hardware.fr that powertune limit the Tonga GPU power envelope to 150W

Given than we can expect the 285 to be 1.6 faster than a TI 750 @ 1080p
Extrapolating the gaming power consumption of the 285 from the 270X (quoted 180W but ingame use peaking around 122) I would expect the 285 to peak around 130W

That would translate as "85% as efficient" (/ TI 750)

Now how does GPU efficiency scale ? While we expect MAXWELL GPU to be power efficient does the power efficiency scale in a linear manner ? In other word should we expect the bigger Maxwell be as efficient as the smaller parts? It was my understanding that with bigger chips (of the same kind) efficiency decrease.

 

[the54thvoid]

I can't wait to see this discussion come full circle. AMD triumphed in efficiency years back with 5 series. They lost the plot with the Hawaii core in going for performance at any cost, even then using power tune to keep it real.
Nvidia had Fermi and after that slowly improved. In the Fermi days AMD slagged NV off for it's power and heat signature. With R9 series it's odd how power efficiency becomes an out of fashion metric. Now Maxwell looks to bring it right back on track, AMD have released Tonga to provide GTX 670 (sorry, GTX 760) beating performance at better than Hawaii efficiencies.
And now AMD energy efficiency is important again. Colour me unimpressed. Not by AMD but by reactions to the 285.
I want to see the real power of Tonga. Not a (potentially) gimped down sales niche product. AMD are toying with their product stack methinks (fair enough I say) but what lurks in the Tonga architecture, beyond that which they are releasing now?

 

[prazola]

I do better with R9 280X + Q9550 ( http://www.3dmark.com/3dm/3860049 ). Maybe this refresh is only a bus cut and DP units reduction.

 

[Sony Xperia S]

I do better with R9 280X + Q9550 ( http://www.3dmark.com/3dm/3860049 ).

Hmm, nice. This several years old baby is still making solid gaming configurations. How is it in games?

 

[esrever]

I do better with R9 280X + Q9550 ( http://www.3dmark.com/3dm/3860049 ). Maybe this refresh is only a bus cut and DP units reduction.

Well the 280x is a tier higher than the 285 which will simply replace the 280. You also seem to have a decent OC on that. The refresh simply adds true audio, more geometry engines, more asynchronous compute cues and the freesync support while making it cheaper for AMD to make the chips.

 

[Steevo]

I can't wait to see this discussion come full circle. AMD triumphed in efficiency years back with 5 series. They lost the plot with the Hawaii core in going for performance at any cost, even then using power tune to keep it real.
Nvidia had Fermi and after that slowly improved. In the Fermi days AMD slagged NV off for it's power and heat signature. With R9 series it's odd how power efficiency becomes an out of fashion metric. Now Maxwell looks to bring it right back on track, AMD have released Tonga to provide GTX 670 (sorry, GTX 760) beating performance at better than Hawaii efficiencies.
And now AMD energy efficiency is important again. Colour me unimpressed. Not by AMD but by reactions to the 285.
I want to see the real power of Tonga. Not a (potentially) gimped down sales niche product. AMD are toying with their product stack methinks (fair enough I say) but what lurks in the Tonga architecture, beyond that which they are releasing now?

performance per Sq cm? Or performance per watt? Heat issues were the cooler and die area for the wattage.

Hawaii was a new die, and both companies are yet unable to move to new processes like they both want, so a overbuilt chip like the 750Ti, that underperforms just shows that anyone can make a chip more power efficient by clocking it down. how about the performance per dollar, and relative performance it can't really do 1080, but for the same price you can get 26% more performance and 34% more power draw both in average (almost linear in performance to power.....strange....) from a HD7870/270/265. Plus the overclocking potential for a 7870 is magnitudes better, so it will last you longer with the same supported DX features.

 

[prazola]

Hmm, nice. This several years old baby is still making solid gaming configurations. How is it in games?

Not bad, but 790i ultra sli has a bad AHCI support that cause some driver issues and a very delicate memory controller, very hard to optimize timings.
BF4 runs well, DX11 or Mantle handle Ultra without drops, BF3 low fps on 64 servers.

This is my primary PC ( http://www.3dmark.com/fs/2257954 ), it runs well too and has margin to OC more.

So, all good...exept for mkv compression, because I can't find a decent OpenCL utility.