Amd Volcanic Islands details

[Ace123]

On my LGA 2011 Platform, Gtx 780 SLI scaling wasn't nearly that good, however a single card was able to overclock a higher. whether that was worth 6000 points I doubt it

14069 marks with GeForce GTX 780 at 1110/3312MHz
22363 marks with 2x GeForce GTX 780 at 1058/3305MHz

http://hwbot.org/submission/2402859_...80_14069_marks
http://hwbot.org/submission/2402854_...80_22363_marks

Lower end cards faired a little better

7770 marks with GeForce GTX 570 at 944/2202MHz
13948 marks with 2x GeForce GTX 570 at 942/2204MHz

http://hwbot.org/submission/2419059_...570_7770_marks
http://hwbot.org/submission/2419063_...70_13948_marks

7721 marks with GeForce GTX 480 at 940/1042MHz
13820 marks with 2x GeForce GTX 480 at 956/1035MHz
17666 marks with 3x GeForce GTX 480 at 891/1003MHz
21354 marks with 4x GeForce GTX 480 at 887/998MHz

as you can see here, 3 and 4 cards don't fair nearly as well......

[Olivon]

Endless Space, COH 2, Rome 2, Krater, Borderlands 2, Saints Row IV....want me to go on?

The issue is that both AMD and NVIDIA have very, very poor support when moving outside the games that reviewers benchmark.

AMD claimed a 80% scaling in the reviewer's guide, no ?

[Lanek]

Uber Mode can give negative results compared to quiet mode ?

SO AMD have done a slide where their Uber mode is slower of their quiet mode, and think use it as a selling point... this do 3 pages i heard AMD is not honest with this slides, maybe not so much ? Where is the other bars ?

[boxleitnerb]

On my LGA 2011 Platform, Gtx 780 SLI scaling wasn't nearly that good, however a single card was able to overclock a higher. whether that was worth 6000 points I doubt it

14069 marks with GeForce GTX 780 at 1110/3312MHz
22363 marks with 2x GeForce GTX 780 at 1058/3305MHz

http://hwbot.org/submission/2402859_...80_14069_marks
http://hwbot.org/submission/2402854_...80_22363_marks

Lower end cards faired a little better

7770 marks with GeForce GTX 570 at 944/2202MHz
13948 marks with 2x GeForce GTX 570 at 942/2204MHz

http://hwbot.org/submission/2419059_...570_7770_marks
http://hwbot.org/submission/2419063_...70_13948_marks

7721 marks with GeForce GTX 480 at 940/1042MHz
13820 marks with 2x GeForce GTX 480 at 956/1035MHz
17666 marks with 3x GeForce GTX 480 at 891/1003MHz
21354 marks with 4x GeForce GTX 480 at 887/998MHz

as you can see here, 3 and 4 cards don't fair nearly as well......

3DMark Performance Preset...you're kidding, right? The Performance Preset is quite CPU bottlenecked, obviously that hurts scaling.
How about some games with demanding settings?
http://www.forum-3dcenter.org/vbulle...19#post9735819

[Olivon]

http://wccftech.com/amd-radeon-r9-29...rmance-leaked/

Nothing really impressive regarding Fire Strike. 780 well-OCed (1200+) got around 5500pts+ on the GPU score.

[Ace123]

no. im not "Kidding, right". Those benchmarks are all with 3930K @ 5.1-5.2ghz, and its the only comparison I have to offer as I don't game much. If you don't agree with them, please dismiss them.

[boxleitnerb]

no. im not "Kidding, right". Those benchmarks are all with 3930K @ 5.1-5.2ghz, and its the only comparison I have to offer as I don't game much. If you don't agree with them, please dismiss them.

3DMark11 runs at 1280x720 in performance preset, I guess you know that?

It's a myth that highest oc'ed CPUs are enough for 2-4 modern highend GPUs in any setting. 5.2 GHz is only about 30% more than stock clocks, but with a second GPU you already get 100% more raw power that would require double the CPU power for perfect scaling in "tame" settings. See the discrepancy? That's why I posted the benchmarks in my link. With proper settings, the CPU becomes completely irrelevant and scaling is mostly excellent at least until the 3rd GPU (unless something like CUDA (Just Cause 2) or PhysX is involved).

[Ace123]

unfortunately, I don't have any of these GPUS available for retesting as I usually have to sell out my old hardware to be able to afford newer stuff to benchmark. A vicious cycle lol.

[Stukov]

SO AMD have done a slide where their Uber mode is slower of their quiet mode, and think use it as a selling point... this do 3 pages i heard AMD is not honest with this slides, maybe not so much ? Where is the other bars ?

I dont think you really need to pull out the spyglass to see those numbers look different (photoshopped).

[PedantOne]

no it is not photoshoped! Sometimes is difference between UBER and QUIt mode 0,1 fps and sometimes is in UBER mode slower

[Olivon]

So, what's the point with Uber Mode ?

[kromosto]

So, what's the point with Uber Mode ?

More noise???

Sent from my SM-N9000Q using Tapatalk

[FlanK3r]

from 3D Mark point not much. Games, games, games. Not 3D marks. 3D Mark could be benchmarked with AMD system FX-9000 series. And of course, there is lower score than with i7-4960x (maybe 100 points?).

[PedantOne]

So, what's the point with Uber Mode ?

if you are in Uber mode, you have Higher TDP limit for YOUR manual OC, you have more Voltage limit for your OC, you have stabile clocks without annoying Boost

PS. It seems nobody here knows what is UBER mode. In normal QUIET mode, GPU is switching clocks between 800 base and 1 GHz max clocks under max TDP envelope. Average clocks is +/- 920-960 in TDP, depends on game a tests. But if you switch button for UBER, your TDP is higher and max is still 1 GHz, but average clocks are higher then 920-960 MHz little bit.

[Olivon]

Okay, so "silent mode" is ~225W TDP limit and "uber mode" 300W ? Something like that ?

[flopper]

Okay, so "silent mode" is ~225W TDP limit and "uber mode" 300W ? Something like that ?

auto turbo within TDP= silent.
Manual no turbo = uber
seems as far I judge what it is.

[boxleitnerb]

Apparently it's the same thing you would get when you increase power and temperature target on Nvidia GPUs. Possibly with the addition of a more aggressive fan curve.

In silent mode, the card might throttle, not always maintaining the maximum boost. In uber mode, it will boost to the maximum clocks more often if not always at the expense of noise and power consumption.

[zalbard]

and sometimes is in UBER mode slower

This makes no sense. No difference does (maintained frequency is the same), though.

[boxleitnerb]

^ maybe tolerance. If it's 39.9 and 40.1 fps, it is basically the same. No two benchmark runs will give you the exact same result, there always is a (small) variance.
Additionally, if they run quiet mode first, then uber mode, the card could be warmer during the uber mode benchmarks which may influence results slightly:

idle: 32C
during first benchmark: 85C (card relatively cool in the beginning)
after first benchmark: 60C (a little cool down while changing the mode and starting the second benchmark)
during second benchmark: 94C (max temp target) -> throttling

[Stukov]

no it is not photoshoped! Sometimes is difference between UBER and QUIt mode 0,1 fps and sometimes is in UBER mode slower

I never said it WAS photoshopped, I said it looked photoshopped because the numbers look different, and they do, whether its because of manipulation or terrible compression of the image format, I can't say.

[Olivon]

Furmark 94C
3DMark 52C (100% fan)
3DMark + OC 55C (100% fan)
Metro 2033 70C (20% fan)

According to the leaker, his R9 290X sample can go up to 1130 MHz. It wasn?t stable at 1150 MHz, so instead he kept overclocking the memory, which we have to say has some potential, but not as much as we expected.

http://wccftech.com/amd-radeon-r9-29...cked-1130-mhz/

original link : http://www.chiphell.com/thread-881612-1-1.html

[boxleitnerb]

For reference:

In Metro 2033 I get 57 fps with Titan at 1006/3004 (default clocks) with CPU-PhysX in the CP selected and Adv. PhysX in the benchmark. So R9 290X is a tad faster than Titan per clock in this benchmark, likely (at least in part) due to the higher bandwidth.

[highspeed]

thats pretty sick standard firestrik score, more then standard gtx 690

[Olivon]

thats pretty sick standard firestrik score, more then standard gtx 690

Fire Strike ared OCed scores and they're not so impressive, around a GTX 780 clocked ~1275MHz.

[raghu78]

Endless Space, COH 2, Rome 2, Krater, Borderlands 2, Saints Row IV....want me to go on?

The issue is that both AMD and NVIDIA have very, very poor support when moving outside the games that reviewers benchmark.

COH 2 does not support SLI / CF. its the game's fault. similarly for total war rome 2. if the developer wants they can definitely get SLI / CF to perform well. but few developers do not bother to get multi GPU to work at all. the top AAA games like BF4, Batman Arkham Origins come out on day one with excellent SLI / CF support. so can't blame Nvidia / AMD for few developers lack of interest.