AMD's own testing? Fury X vs GTX 980Ti

[Techhog]

So, now that the cat is out of the bag (week late, had to sit in the corner). Who is still ready to defend AMD Fury X's awesome 95mhz overclocking capability? See, this is why I said about reading up on hardware for years and being able to correlate hidden information when presented with evidence.

 

It's a pretty nice card still, but the drivers are still pretty terrible. Can't really compete <4K.

We should wait to see what overclocking is like if/when voltage is unlocked. It has around 15C of temp headroom afterall.

[Khvarrioiren]

So, now that the cat is out of the bag (week late, had to sit in the corner). Who is still ready to defend AMD Fury X's awesome 95mhz overclocking capability? See, this is why I said about reading up on hardware for years and being able to correlate hidden information when presented with evidence.

It's a pretty nice card still, but the drivers are still pretty terrible. Can't really compete <4K.

What is there to defend, the voltages are locked. Simple as that.

[decoy11]

 

Got to 41mins in this video and look at how badly AMD tried to cheat on their benchmarks

[LAwLz]

<video>

 

Got to 41mins in this video and look at how badly AMD tried to cheat on their benchmarks

Of course AMD's own tests will be biased as fuck. I couldn't believe how stupid some people were when they said things like "AMD has no reason to lie" when I told them to not trust AMD's own benchmarks.

Of course they will make their card seem as good as possible and the competitor as bad as possible. Nvidia most likely does this too. AMD and Nvidia are not your friends. They are companies trying to take as much money from you as possible, even if it means twisting the truth to fool you.

 

Never trust benchmarks that have not been validated by an independent third party. The only people who trusted AMD's benchmarks were either naive, stupid or massive AMD fanboys. I hope this has taught them what a colossal dick AMD really is.

[LePawel]

Why would it need to be a TitanX killer when nVidia killed that one off themselves by pricing the 980ti so low?

So, 3 generations, that's how long it takes to forget that the Titan chip (Gx100) was originally the GTX X70/X80 line, and what you now call the x80 used to be an x60.

They ~doubled the money they earn from these chips, and even 980ti is still few hundred *insert currency here* above its old price.

How hard is it to see that?

They created an artificial "super card" by adding ridiculous amounts of RAM to it, and charging a K for it.

Then they create a "super card lite" by cutting the specs down slightly, but reducing the price more, and voila! 

Why are there no x50 and below desktop cards? Because they are used for higher-end cards. They bumped the numbers on all cards, invented some new ones and you get sucked in by the new "low low price".

wake the fuck up already...

[Ohlyver]

If that's the sort of results to expect then I'm impressed. AMD really seems to have done it right with fiji in my opinion, especially as this is beating the 980ti at the same price point, and as the titan x is so close in performance, with a bit of overclocking you can expect to see this comfortably beat out all the competition, especially with an AiO already attached the temps will easily be able to handle it. If I had the money I'd be racing to bag one of these right now.

I'm not a specialist ( and far from becoming one) about cards temps .. But isn't AMD running generally hotter compared to same performance in NVidia?

If so,they might have gone with the "aio" cooler so the out of the box experience would not thermal throttle ..

But I might also be dead wrong

[Khvarrioiren]

I'm not a specialist ( and far from becoming one) about cards temps .. But isn't AMD running generally hotter compared to same performance in NVidia?

If so,they might have gone with the "aio" cooler so the out of the box experience would not thermal throttle ..

But I might also be dead wrong

 

That used to be the case with reference cooled 290/x's. Usually temperature wise both vendors are somewhat on the same level.

[Briggsy]

A recent podcast by the Tech Report about Fiji pretty much nails this one to the wall, and explains everything you could possibly want to know about FuryX and Fiji in general.

 

Starting at 40:49 in the video they discuss why AMD's internal benchmarks were higher than reviewers. Honestly, the entire video is pretty educating, but this part reveals that AMD were playing to the strengths and avoiding weak points of fiji for the reviewer guide benchmarks, namely turning off AF filtering etc.

 

https://youtu.be/28CECF_Cieo?t=2449

[That Norwegian Guy]

A recent podcast by the Tech Report about Fiji pretty much nails this one to the wall, and explains everything you could possibly want to know about FuryX and Fiji in general.

 

Starting at 40:49 in the video they discuss why AMD's internal benchmarks were higher than reviewers. Honestly, the entire video is pretty educating, but this part reveals that AMD were playing to the strengths and avoiding weak points of fiji for the reviewer guide benchmarks, namely turning off AF filtering etc.

 

https://youtu.be/28CECF_Cieo?t=2449

 

Yep, choosing settings to avoid getting bottlenecked by the 64 ROPs on the Fury, which, if you are playing on 4K you should be doing anyway. As it's a card meant for 4K, only light anti-aliasing if any at all is needed, unfortunately a lot of reviewers are using MSAA and shit on 4K which really is dumb as boots.

[LAwLz]

Yep, choosing settings to avoid getting bottlenecked by the 64 ROPs on the Fury, which, if you are playing on 4K you should be doing anyway. As it's a card meant for 4K, only light anti-aliasing if any at all is needed, unfortunately a lot of reviewers are using MSAA and shit on 4K which really is dumb as boots.

AMD had anisotropic filtering set to 0...

Nobody in their right mind would do that, even on 4K. Let's not pretend like AMD was trying to tweak the settings in their benchmarks to what people actually want to use. They ignore what people would want and just set the settings to whatever would give them the biggest benefit over Nvidia. There is just no way you can spin what AMD did to anything positive. They just tried to trick people.

 

It's completely fine that they were trying to trick people by the way. AMD's own numbers are from marketing material. You shouldn't trust them for the same reason you shouldn't trust any commercials you see, because they are designed to fool you into buying their product.

[That Norwegian Guy]

fooling, tricking, blah blah

 

Their settings were always included with the original benches. You can't say they were trying to trick anyone when they were always forthright with their method.

[LAwLz]

Their settings were always included with the original benches. You can't say they were trying to trick anyone when they were always forthright with their method.

I think they were, because the benchmarks they ran made no sense and the settings were not included in the actual benchmark graphs (which is what most people will look at when comparing two different GPUs).

You can't just run benchmarks at flat out stupid settings, make graphs about it and then go "no these graphs were totally not made to trick people! See I made a note about the stupid settings in an appendix several pages later in the reviewer guide".

 

It's like the small print which is read very quickly at the end of some commercials. It's not there to remove any possibility of people getting tricked. It's there because they have to put it there, even if they don't want to.

[Notional]

A recent podcast by the Tech Report about Fiji pretty much nails this one to the wall, and explains everything you could possibly want to know about FuryX and Fiji in general.

 

Starting at 40:49 in the video they discuss why AMD's internal benchmarks were higher than reviewers. Honestly, the entire video is pretty educating, but this part reveals that AMD were playing to the strengths and avoiding weak points of fiji for the reviewer guide benchmarks, namely turning off AF filtering etc.

 

https://youtu.be/28CECF_Cieo?t=2449

 

Very interesting video indeed. Obviously setting anisotropic filtering to 0 is outright dumb. AF is so much more important than AA. 

 

But going further into the video they have very interesting debate about GameWorks.

[Khvarrioiren]

-snip-

I find it very weird to set AF to 0, since AF doesn't affect performance more than 1-2 FPS even at 4K.

[cd5josh]

in theory more watts should = more power.

 

but multiple sources are showing their benchmarks in favor of nvidia

[LAwLz]

in theory more watts should = more power.

 

but multiple sources are showing their benchmarks in favor of nvidia

Ehhh... No? Not even in theory does more power consumption = more performance.

More bhp does not equal a faster car either. There are a lot of other factors you have to take into consideration such as efficiency.

 

An example of this would be lights. An LED light can produce far more light per wattage compared to an incandescent light bulb because they are more efficient. Same thing applies to computer hardware.

[cd5josh]

Ehhh... No? Not even in theory does more power consumption = more performance.

More bhp does not equal a faster car either. There are a lot of other factors you have to take into consideration such as efficiency.

 

An example of this would be lights. An LED light can produce far more light per wattage compared to an incandescent light bulb because they are more efficient. Same thing applies to computer hardware.

 

my point being amd should not use so much more power for little to no gain

[marldorthegreat]

Looks like 4K on a single card isn't gonna be a thing yet...

It probably will be next year with the 490x. Amd has about a 50 percent increase in performance each generation.  But we may not see AMD r9 400 series for another 2 years

[Notional]

It probably will be next year with the 490x. Amd has about a 50 percent increase in performance each generation.  But we may not see AMD r9 400 series for another 2 years

 

Greenland will be released next year, on a 14nm FinFet node with HBM2. Not only will Pascal get a serious contender, but Pascal might be weaker.

[marldorthegreat]

Greenland will be released next year, on a 14nm FinFet node with HBM2. Not only will Pascal get a serious contender, but Pascal might be weaker.

We dont know that. We thought that about the 300 series when the 200 series launched. We thought this would be 20nm. 

[Texhnlyze]

Greenland will be released next year, on a 14nm FinFet node with HBM2. Not only will Pascal get a serious contender, but Pascal might be weaker.

 

Was it ever confirmed that It will be on 14nm? Link ples. I want to read.

[Notional]

We dont know that. We thought that about the 300 series when the 200 series launched. We thought this would be 20nm. 

 

Actually we do. 14nm FinFet is already up and running, and just like ZEN will use that node, so will Greenland. 20 nm at TSMC was just too leaky and poor performing, but FinFet's takes care of that. Pascal will either be 16nm FinFet+ (at TSMC) or 14nm FinFet along with Greenland on Global Foundries, though NVidia has yet to land that contract.

[Notional]

Was it ever confirmed that It will be on 14nm? Link ples. I want to read.

 

http://www.fudzilla.com/news/graphics/37712-amd-greenland-gpu-comes-in-2016-gets-finfet

[Cyrus49]

It probably will be next year with the 490x. Amd has about a 50 percent increase in performance each generation.  But we may not see AMD r9 400 series for another 2 years

 

According to some benchmarks, the 300 series is only ~10% faster than the 200 series.

[d1screet]

Now if only AMD's low-midrange cards could follow suit and see similar improvements instead of being constant rebadges...

 

That would be just great.