AMD Radeon Fury X 3DMark performance

[The Benjamins]

It's still bad compared to the Nvidia counter part.

They need 1200 more cores, 200GB/s more bandwidth, and watercooling to even compete that is not good at all.

This is disappointing for something that is on paper worlds ahead of Nvidia but isn't any better in real world performance.

 

 

The word on the street is there will be water cooled and air cooled versions. the best of AMD is competing with the best of Nvidia with 2 different approaches.

[Brainless906]

1) as many have pointed out Nvidia has always done better in synthetic testing. Until game benchmarks come out we wont know anything.

2) I'm going to go ahead and take out my trust salt block until the card is actually out in the wild

3) we have no idea if this was the water cooled or air cooled

4) we have no idea how it will overclock. If it overclocks anything like the TI we'll be doin great

 

 

That said if the performance does land where its looking there will be very little reason to not buy a 980 ti unless its priced very aggressively. And with AMD is with its financials they cant loose money on this.

[duriel]

The lack of VRAM made this card DOA for me to begin with... Now I'm just deciding between the 980ti or Titan X for my next build... If CaseLabs ever builds and ships my case to me...

 

why are you saying that when 4gb of hbm on the fury x is matching the titan x with 12gb of gddr5?

8gb hbm for next gen is going to absolutely blow nvidia's equivalent out of the water at that rate, especially if the vram is a limitation (which it might be, atm)

[Brainless906]

It's still bad compared to the Nvidia counter part.

They need 1200 more cores, 200GB/s more bandwidth, and watercooling to even compete that is not good at all.

This is disappointing for something that is on paper worlds ahead of Nvidia but isn't any better in real world performance.

Like do you realize how much bandwidth HBM has?

That is one of the biggest jumps in at least 7 years and the result is a joke.

 

1) Titans been in the wild forever and has multiple driver updates behind it.

2) you seem to be taking this as cold hard fact as if the card is released and been benchmarked by actual people we trust. 

3) we dont know if this was water cooled or not.

3.5) Calm down and get up off that green shaft man.

[Real_PhillBert]

Those scores are about in line with what I was expecting. I would like to see gaming performance to really compare the cards.

[marldorthegreat]

It's still bad compared to the Nvidia counter part.

They need 1200 more cores, 200GB/s more bandwidth, and watercooling to even compete that is not good at all.

This is disappointing for something that is on paper worlds ahead of Nvidia but isn't any better in real world performance.

Like do you realize how much bandwidth HBM has?

That is one of the biggest jumps in at least 7 years and the result is a joke.

Nope, nvidias back to that stage. With the tian x they are at the 250 watt tdp "limit" which nvidia never crosses, and they are going to have to increase core count.

[zMeul]

what? within 1% distance between nvidia's highest end card while taking a huge risk on hbm, and you're managing to be disappointed?

yes, that's right

AMD needed HBM to reach the Titan X in performance; Titan X that uses good ol' GDDR5

not even taking into account that Titan X has 3 times (!!!) more VRAM available to it

[zMeul]

its as fast as a titanX/980ti and is a 56% improvement over a 290X, how is that disappointing? were you expecting a bigger improvement?

yes! why did they put HBM on the darn thing if not!? HBM is not cheap, it needed a new MC - that costs money and it's reflected into final product

[Jerakl]

Just a reminder that because of GPU Boost 2.0 those Nvidia "stock clocks" are probably not stock.

Waiting for real world benchmarks. Still impressive though.

[syoumans54]

why are you saying that when 4gb of hbm on the fury x is matching the titan x with 12gb of gddr5?

8gb hbm for next gen is going to absolutely blow nvidia's equivalent out of the water at that rate, especially if the vram is a limitation (which it might be, atm)

Memory bandwidth has never been an issue. As far as I know, 12gb of gddr5 is still 12gb.... While 4gb of HBM is still 4gb. Increasing the bandwidth does not increase the storage capacity by 3. 

[The Benjamins]

yes! why did they put HBM on the darn thing if not!? HBM is not cheap, it needed a new MC - that costs money and it's reflected into final product

 

i think 56% is a big improvement.

 

they put HBm in it to invest into new tech, you have to start somewhere. sadly it does limit it to 4GB.

[Virgule]

Is the VRAM really clocked at 500 MHz ???

[Jerakl]

i think 56% is a big improvement.

 

they put HBm in it to invest into new tech, you have to start somewhere. sadly it does limit it to 4GB.

The fanboyism is strong with that one.

AMD brought something new to the industry. (Doesn't Nvidia have hbm lined up a gen or 2 down the line?)

Also GPU Boost 2, so I doubt those clocks on the Nvidia cards are stock.

[Victorious Secret]

why are you saying that when 4gb of hbm on the fury x is matching the titan x with 12gb of gddr5?

8gb hbm for next gen is going to absolutely blow nvidia's equivalent out of the water at that rate, especially if the vram is a limitation (which it might be, atm)

Uhh... Nvidia is skipping HBM1 and going to HBM2. AMD doesn't have some exclusive rights to HBM, they'll be neck and neck on that front.

People will now say "oh it's a Titan competitor for less" when it barely beats the 980 Ti, in single or dual configurations.

The hype train is crashing. It's okay. Fury is a good card and if the price point undercuts the 980 Ti it'll be great for us. But it won't, because AMD cant run a charity for a business anymore.

[Brainless906]

Well if this is where the end performance of the Fury is then Nvidia releasing the 980 Ti at such an aggressive price point (effectively killing off both the 980 and the titan x) makes sense. 

 

I "predicted" that Nvidia had to be scared of something to release a card that killed most valid reasons to buy a 980 or a titan x and got laughed at. But if the fury x came out performing like this at 700-800 bucks then the Fury x would have killed off any reason to buy a titan x or a 980...just like the 980 ti did.

 

So as great as the Fury X is (assuming these benchmarks are real end performance) Nvidia killed of two of its cards but effectively snuffed out any real reason to buy a Fury X unless it somehow manages to be cheaper. Which amd cannot afford to make it cheaper. A dick move on the part of nvidia but the financially proper move. 

 

That said all of the above is assuming this shit means anything.

 

I personally dont believe fuck all until i get real world benchmarks and pricing from people/sites i trust.

 

All this speculation bullshit is getting mad crazy, 16th couldnt come fast enough. 

[79 Potatoes]

Finally... a top end card that fits in a short case

[duriel]

 

yes, that's right
AMD needed HBM to reach the Titan X in performance; Titan X that uses good ol' GDDR5

not even taking into account that Titan X has 3 times (!!!) more VRAM available to it

 

my point is, take away 8gb of that gddr5 from the titan x and then compare

 

the fury x would completely destroy the titan x at that rate

this isn't the case though, and that kind of a comparison is what would appease to you, not me or everyone else.

 

 

Memory bandwidth has never been an issue. As far as I know, 12gb of gddr5 is still 12gb.... While 4gb of HBM is still 4gb. Increasing the bandwidth does not increase the storage capacity by 3. 

 

my point is that 4gb of hbm might be at 100% utilization while the gddr5 on the titan x is probably not completely used up, at 4k+ resolutions

with 8gb or more of hbm on the same card (won't happen this generation most likely), you might be seeing the fury x outperform the titan x since the vram wouldn't be a limitation

 

but i can't say if the fury x is limited by its 4gb of hbm or not, because i'm just seeing performance numbers. it doesn't seem to be limited though as it's performing exactly where amd wants it to pretty much (290x didn't outperform the titan until drivers were matured, at which point the 290x did exactly what it was set out to do, kill the titan for much less $)

 

 

Uhh... Nvidia is skipping HBM1 and going to HBM2. AMD doesn't have some exclusive rights to HBM, they'll be neck and neck on that front.

People will now say "oh it's a Titan competitor for less" when it barely beats the 980 Ti, in single or dual configurations.


The hype train is crashing. It's okay. Fury is a good card and if the price point undercuts the 980 Ti it'll be great for us. But it won't, because AMD cant run a charity for a business anymore.

 

i know nvidia is moving to hbm later on, which is smart of them to do, i just think it's dumb someone else in this thread is arguing amd swapped to hbm just to beat nvidia which is false, as nvidia is planning to do the same because it's just straight up better technology.

 

 

Well if this is where the end performance of the Fury is then Nvidia releasing the 980 Ti at such an aggressive price point (effectively killing off both the 980 and the titan x) makes sense. 

 

I "predicted" that Nvidia had to be scared of something to release a card that killed most valid reasons to buy a 980 or a titan x and got laughed at. But if the fury x came out performing like this at 700-800 bucks then the Fury x would have killed off any reason to buy a titan x or a 980...just like the 980 ti did.

 

So as great as the Fury X is (assuming these benchmarks are real end performance) Nvidia killed of two of its cards but effectively snuffed out any real reason to buy a Fury X unless it somehow manages to be cheaper. Which amd cannot afford to make it cheaper. A dick move on the part of nvidia but the financially proper move. 

 

That said all of the above is assuming this shit means anything.

 

I personally dont believe fuck all until i get real world benchmarks and pricing from people/sites i trust.

 

All this speculation bullshit is getting mad crazy, 16th couldnt come fast enough. 

 

i could argue it was actually a dumb move on their part, as they're already obviously much better off than amd is

if they're trying to beat amd all the time and use means like this (killing 980/titan x in favor of 980ti so there's no reason to buy an amd card at that pricerange) then they're going to be heading towards an intel/nvidia monopoly quick when amd goes out of business or ends up being bought (which will nullify their x86 license)

 

i'm not saying that will happen all because of this, but nvidia has to watch what they do or amd will go overboard and shit will hit the fan

[zMeul]

i think 56% is a big improvement.

 

they put HBm in it to invest into new tech, you have to start somewhere. sadly it does limit it to 4GB.

you should think about this: what's AMD response to Pascal? my guess? the same exact Fiji but with HBM2 - and that alone should scare the shit out of people who don't see the problem with Radeon Fury (X)

[The Benjamins]

you should think about this: what's AMD response to Pascal? my guess? the same exact Fiji but with HBM2 - and that alone should scare the shit out of people who don't see the problem with Radeon Fury (X)

 

I expect pascal to be a year away and AMD to do another architecture for the GPU and use HBM2, so I don't have any concerns because they learned how to implement HBM already and now can focus more on just adding more ram and better architecture.

[patrickjp93]

color me disappointed: all that bragging from AMD, and in the end HBM didn't do that much for it

All one had to do to confirm HBM would be a non-event is compare the 780TI bandwidth to the 290X. GPU internal bandwidth is not the bottleneck in gaming. Core strength is. For 4K the biggest bottleneck appears to be the ROPs.

[zMeul]

I expect pascal to be a year away and AMD to do another architecture for the GPU and use HBM2, so I don't have any concerns because they learned how to implement HBM already and now can focus more on just adding more ram and better architecture.

to "improve the architecture" is not done over night, as example with Pascal; and will more than 1y until we'll have desktop video cards powered by Pascal GPU

you should consider that AMD just finished Fiji and doesn't have, yet, real world performance metrics

[zMeul]

my point is, take away 8gb of that gddr5 from the titan x and then compare

already done! sort of, compare it to GTx 980Ti - it's just slightly better than it

[patrickjp93]

Of course not, but with that AIO cooler, I'm expecting results, yet I have a feeling we won't get them.

Still, GG AMD, you made a card that's on par with the absolute flagship which is an achievement.

Take a moment, look at the graph, and look at the results of the previous flagship(290X)

The performance is considerably better, and the TDP is what, 10w higher?

So how did this not do much for it?

4096 SPs against 2816 on a newer architecture. Bandwidth is not the bottleneck.

[zMeul]

All one had to do to confirm HBM would be a non-event is compare the 780TI bandwidth to the 290X. GPU internal bandwidth is not the bottleneck in gaming. Core strength is. For 4K the biggest bottleneck appears to be the ROPs.

HBM is ultimately irrelevant (!!!) yes, since nothing else changes between the system and the video card - the PCIe stays the same, RAM and CPU also

HBM only matters in GPU-VRAM transfers -- that's it!

[Brainless906]

HBM is ultimately irrelevant (!!!) yes, since nothing else changes between the system and the video card - the PCIe stays the same, RAM and CPU also

HBM only matters in GPU-VRAM transfers -- that's it!

 

 

Lol i'll be sure to save this comment for when Pascal launches...