Why isn't the Fury X more powerful

[MoonlightSylv]

I'm not saying that I'm not satisfied with my card, I am, and I'm also not fanboying. Now that that's out of the way.

 

Why isn't the Fury X more powerful? It has 4096 Stream Processors and 8.6 TFLOPs. Wouldn't these specs translate to a much more powerful card or do specs not matter at all on this thing? I'm genuinely curious as to why a card that on paper is stronger than a 980Ti isn't faster.

[Oryzaki]

Because Stream processors are not the same as a cuda cores and 8.6 TFLOPS is what console fanboys use and look where that got them.

[Evolution.]

specs dont mean everything it's also about how the games utilize the resources 

[PCGuy_5960]

You can't compare cores and TFLOPs between different architrectures. AMD usually needs more TFLOPs to get the same performance as Nvidia

[Vegetable]

Ehhh stream processors and TFLOPs dont directly relate to performance in games really at all. A cuda core (nvidia cards have this) is not directly comparable to a stream proc in any way

[Oryzaki]

1 minute ago, PCGuy_5960 said:

You can't compare cores and TFLOPs between different architrectures. AMD usually needs more TFLOPs to get the same performance as Nvidia

Essentially TFLOPS mean jack shit to IRL performance 

[Oryzaki]

1 minute ago, Vegetable said:

Ehhh stream processors and TFLOPs dont directly relate to performance in games really at all. A cuda core (nvidia cards have this) is not directly comparable to a stream proc in any way

They have the same processes but more Stream processors are required to get the same amount of performance as a single cuda core

[Darkseth]

Fury X has problems in really "using" all those Cores at the best efficiency.

and yea, 4gb Vram is pretty low. Even it it were HBM8, it wouldn't mean much. Works most of the times, in some games, 4gb are bottlenecking even in 1080p.

Also, stuff like Driver overhead in DX11, Driver optimizations for games, Game optimizations for the Architecture itself.. 

[Oryzaki]

https://superuser.com/questions/649019/differentiate-cuda-coresnvidia-and-stream-processorati-amd

[themctipers]

because it competes with a 980ti iirc, and a heavily overclocked 980ti is just a middle of the road 1070 in gaming performance.

[PCGuy_5960]

2 minutes ago, themctipers said:

because it competes with a 980ti iirc, and a heavily overclocked 980ti is just a middle of the road 1070 in gaming performance.

A heavily overclocked 980 Ti is better than a 1070

[Darkseth]

Untill you heavily overclock the 1070 too

[themctipers]

Just now, PCGuy_5960 said:

A heavily overclocked 980 Ti is better than a 1070

minus power efficiency and whatever novideo shoved into that card

 

im talking like 185-250mhz range, i really have no idea how a 980ti overclocks but im pretty sure thats all you can get without 95c hellfire in your room 

[themctipers]

Just now, Darkseth said:

Untill you heavily overclock the 1070 too

haha nice meme

 

with a locked bios that is encrypted, and a voltage lock of 1.04v on my card..........

[loculus]

18 minutes ago, Oryzaki said:

https://superuser.com/questions/649019/differentiate-cuda-coresnvidia-and-stream-processorati-amd

Be warned that this answer is old and is prior to GCN architecture even (and recall with GCN, which was significantly more programmable, AMD started pushing more on compute). In HD 6970 AMD was still using VLIW4 (TeraScale 3).

 

The capabilities and arrangements of the shaders / processing cores / whatever you want to call them have changed over time for both companies (okay and Intel too).

 

 

If you look at some non-gaming benchmarks, you do see some scenarios where you got high usage of all the resources in Fury X and sometimes a significant lead over GTX 980 Ti. Also some games as well. Depends on the exact workload and optimization of course, and obviously on compute if you're looking at OpenCL, Nvidia doesn't care as much about that because they're all about their CUDA ecosystem.

 

Also there are just some scenarios where HBM is less suited (the bandwidth was great because the design is so wide, but at higher latency IIRC), or the 4 GB of VRAM is an issue.

[Darkseth]

Just now, themctipers said:

haha nice meme

 

with a locked bios that is encrypted, and a voltage lock of 1.04v on my card..........

Afterburner --> Voltage to +100, and Voltage should cap at 1.093 Volt^^" On EVERY pascal so far. Haven't heared 1 single unit, that didn't go that far up.

 

still, 2.1 Ghz is quite common for most 1070 Unity at higher voltage. The worser ones shoulds till do 2050+

[The Sloth]

23 minutes ago, Oryzaki said:

what console fanboys use and look where that got them.

not very comparable, a console can do much more with 1.8 teraflops than a similar spec PC can, since DEV has direct access to hardware. 

[themctipers]

Just now, Darkseth said:

Afterburner --> Voltage to +100, and Voltage should cap at 1.093 Volt^^" On EVERY pascal so far. Haven't heared 1 single unit, that didn't go that far up.

 

still, 2.1 Ghz is quite common for most 1070 Unity at higher voltage. The worser ones shoulds till do 2050+

not a thing  tried it but it did nothing

mine did 1950 at first, then with a reinstall of windows and sipping some of amd's FineWine^TM technology, it now is doing 2050mhz at 160w.. yeah, im power limited too.

 

 

[Oryzaki]

Just now, nerdslayer1 said:

not very comparable, a console can do much more with 1.8 teraflops than a similar spec PC can, since DEV has direct access to hardware. 

 I was talking about how they said because the were getting 8 or what ever amount of terraflops on the xbox one x it would be as good as a 1080. Also optimization is optimizeing the drivers of the hardware.

[Darkseth]

Just now, themctipers said:

not a thing  tried it but it did nothing

mine did 1950 at first, then with a reinstall of windows and sipping some of amd's FineWine^TM technology, it now is doing 2050mhz at 160w.. yeah, im power limited too.

 

 

What..?! Which model do you have?

 

No 1070 is powerlimited at 160 Watt ^^ Stock (Founders) can go up to 170 Watt (150 with 113% PT). Custom models usually can go up to 190-200+

 

But still... not worth cranking it up so high anyway^^" +50% Powerdraw for +1-2 fps.. useless

 

I have my GTX 1080 at 0.8 Volt and 1821 Mhz (but Memory to +500), and i feel zero difference thanks to G-Sync ^^" And my GTX 1080 draws only 120-135~ Watt. And is still faster, than EVERY hard Overclocked 1070 / 980 ti

[themctipers]

Just now, Darkseth said:

What..?! Which model do you have?

 

No 1070 is powerlimited at 160 Watt ^^ Stock (Founders) can go up to 170 Watt (150 with 113% PT). Custom models usually can go up to 190-200+

 

But still... not worth cranking it up so high anyway^^" +50% Powerdraw for +1-2 fps.. useless

 

I have my GTX 1080 at 0.8 Volt and 1821 Mhz (but Memory to +500), and i feel zero difference thanks to G-Sync ^^" And my GTX 1080 draws only 120-135~ Watt. And is still faster, than EVERY hard Overclocked 1070 / 980 ti

gtx 1070 zotac (not amp/minin/whatever, just zotac)

mine can only +13% on power limit. sorta worth, +4-5fps 

[The Sloth]

Just now, Oryzaki said:

 I was talking about how they said because the were getting 8 or what ever amount of terraflops on the xbox one x it would be as good as a 1080. Also optimization is optimizeing the drivers of the hardware.

it would be better than a gtx 1060, not as powerful as a gtx 1070, but close 

 

[Oryzaki]

3 minutes ago, nerdslayer1 said:

it would be better than a gtx 1060, not as powerful as a gtx 1070, but close 

 

My GTX 1060 can get artificial 4k at 60fps soooo... some games weren't even going to be fake 4k 

[Oryzaki]

Nope

[The Sloth]

2 minutes ago, Oryzaki said:

My GTX 1060 can get artificial 4k at 60fps soooo...

so? who cares, consoles can do a lot more with less powerful hardware. i was pointing out the facts Tflops matter but not always, Xbox throw around numbers does mean something. 

for 600$, 4k 30FPS ( most AAA will have to checkerboard) gaming, 4k HDR BlueRay and much more advantages, bringing console into PC gaming is kinda dumb.