• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

The Fury Pro Fiji Owners Thread

Lokken86 said:
Ooh.
Wonder how fast it could go without the nitro cooling, like standard water or something.

Just noticed the 1TB memory bandwidth!
Click to expand...

That overclock got a score of 10033 marks in 3DMark Firestrike Extreme.
How does that compare to an overclocked TitanX/980Ti?
 
AthlonXP1800 said:
What a massive waste of bandwidth with no improvements on 1TB matched GTX 980 Ti performance on air at 3 times less memory bandwidth. :o
Click to expand...

fs123 said:
Which means AMD wasted their time concentrating on higher bandwidth when they really should have increased core performance.
Click to expand...

oh, dear :( can we get more trolling? do you guys even understand wtf you are talking about?
Let me put this in a way that even you guys can understand (though I still have doubts):
HBM was developed and used by AMD in fiji because:
-gets rid of hot and power hungry GDDR5
-simplifies PCB design, look at the cards size
-even at 500Mhz it gives a lot more bandwidth than superclocked GDDR5. next gen gpus from nvidia and amd are gonna be huge leap in performance. Look even nvidia is doing HBM next year, so probably they do need that bandwidth. So if nvidia expects their next GPU to need hbm type bandwidth then AMD is not far behind as well, or far ahead (whichever way you look at it)
-having hbm in fiji is pipecleaning process: old arch, old process tech, brand new memory controller and memory interface. Come next year, AMD will have plenty of in the field experience with HBM, and when releasing next gen GPU, they only will need to worry about process and their GPU arch changes, while nVidia will be trying to sort out new GPU arch, new memory controller, new memory interface, new process node at once. Companies much more resourceful and experienced than nvidia rarely pull of such release without issues, most of the time they don't even try such things in order to minimize the risks of something out of all new things going wrong.

and if some crazy enthusiast OCs HBM to 1Ghz, how is this AMD fault, kids? How did AMD waste time while some OCer was playing around with AMD card in his basement?

and mister smart arse fs123, please enlighten us, how would you increase core performance on 28nm node with GCN arch? Please, the floor is yours, if you want I can bring paper and pencil for you, but several hundred million pounds/dollars will be in your expense, while you are coming up with this revolutionary way of designing chips.
 
Last edited:
muziqaz said:
oh, dear :( can we get more trolling? do you guys even understand wtf you are talking about?
Let me put this in a way that even you guys can understand (though I still have doubts):
HBM was developed and used by AMD in fiji because:
-gets rid of hot and power hungry GDDR5
-simplifies PCB design, look at the cards size
-even at 500Mhz it gives a lot more bandwidth than superclocked GDDR5. next gen gpus from nvidia and amd are gonna be huge leap in performance. Look even nvidia is doing HBM next year, so probably they do need that bandwidth. So if nvidia expects their next GPU to need hbm type bandwidth then AMD is not far behind as well, or far ahead (whichever way you look at it)
-having hbm in fiji is pipecleaning process: old arch, old process tech, brand new memory controller and memory interface. Come next year, AMD will have plenty of in the field experience with HBM, and when releasing next gen GPU, they only will need to worry about process and their GPU arch changes, while nVidia will be trying to sort out new GPU arch, new memory controller, new memory interface, new process node at once. Companies much more resourceful and experienced than nvidia rarely pull of such release without issues, most of the time they don't even try such things in order to minimize the risks of something out of all new things going wrong.

and if some crazy enthusiast OCs HBM to 1Ghz, how is this AMD fault, kids? How did AMD waste time while some OCer was playing around with AMD card in his basement?

and mister smart arse fs123, please enlighten us, how would you increase core performance on 28nm node with GCN arch? Please, the floor is yours, if you want I can bring paper and pencil for you, but several hundred million pounds/dollars will be in your expense, while you are coming up with this revolutionary way of designing chips.
Click to expand...

The Fiji core has nearly twice as many stream processors yet the performance is not double of a 290X. Maybe they should have tried to make the stream processors perform better instead of packing more in a 28nm process and presumably increasing costs.
Nvidia has far less CUDA cores but the performance is better. All I'm saying is maybe optimizing the existing architecture would have been better than just hoping more is better. ;

Theoretically with the huge amount of bandwidth and large number of stream processors Fiji should have destroyed everything on the market but something is not quite right in practice. Maybe DX12 will unleash the power...
 
Last edited:
fs123 said:
The Fiji core has nearly twice as many stream processors yet the performance is not double of a 290X. Maybe they should have tried to make the stream processors perform better instead of packing more in a 28nm process and presumably increasing costs.
Nvidia has far less CUDA cores but the performance is better. All I'm saying is maybe optimizing the existing architecture would have been better than just hoping more is better. ;

Theoretically with the huge amount of bandwidth and large number of stream processors Fiji should have destroyed everything on the market but something is not quite right in practice. Maybe DX12 will unleash the power...
Click to expand...

4096/2816=1.45
So no not nearly double. And yes Fiji in majority of decently written games and programs is that much faster. The rest of the games just need some driver tweaking, as any other new GPU always needs.
And, again, instead of talking out of your arse, please learn what actually 'make the stream processors perform better' mean. What is the point on investing in dead end architecture which will be replaced by new arch year after? Is AMD ****ing money? Learn the economics of chips design and manufacturing, then come here and make these idiotic statements.
So you are saying that nvidia cuda cores are exactly the same as AMD shaders?
So AMD FX CPU cores are exactly the same as Intel Haswell, or Skylake?
Remember that nvidia cores clock faster, which is due to their specific design, and lack of DP capabilities (no wonder nvidia released their quadros silently just few days ago, since they don't want to look stupid when they brand new cards get slaughtered by competition and their own old arch cards in compute).

Theoretically you shouldn't be talking about chips design costs, but in practice you are here, but hopefully you realize you have no clue what you are on about and you just stop posting here.
Same with AMD, theoretically their card should do much better, in practice we hope they will sort out their performance driver to unleash the beast, and there is this elephant in the room called dx12. The card has been out just a month or so, and everyone from green side already think that fiji cannot improve in performance.
How easily everyone forgets that 290x jumped in performance a lot. Omega driver done amazing things last year, and when 300 series came out, so called rebranded 290x was now in same league as 980gtx, when just a month before release it was trading blows with 970gtx. And 290x is how old?
Now assume that AMD driver team brings out just a small % of performance compared to what they done to 290x, we will be looking at undisputed king of the hill.
Just a bit over a month old, people.
 
fs123 said:
The Fiji core has nearly twice as many stream processors yet the performance is not double of a 290X. Maybe they should have tried to make the stream processors perform better instead of packing more in a 28nm process and presumably increasing costs.
Nvidia has far less CUDA cores but the performance is better. All I'm saying is maybe optimizing the existing architecture would have been better than just hoping more is better. ;

Theoretically with the huge amount of bandwidth and large number of stream processors Fiji should have destroyed everything on the market but something is not quite right in practice. Maybe DX12 will unleash the power...
Click to expand...

its different designs of technology.
AMD has a more flexible approach that ages well and are done for dx12 games and compute performance.
the 390 for example is simply a gamers card as its best in class for 90% of gamers. The rest buy a fury x or a Titan x.

Issue is 28nm as that is holding back performance more than anything else.
previous years AMD/Nvidia just had a die shrink added double performance due to the shrink itself which will also happen next year to 16nm which will add double performance with existing technology.

at 28nm you dont get good performance above a 390 for your money.
 
Last edited:
funny how everyone looks at the 10k score

and no one looks at the graphics score which is higher then a titanx at 1500-1600mhz could ever do on air but okaaaaaaaaaay

we have the experts here again at work...:rolleyes:


he used a 4790k with the standard cooler and not overclocked at all

also if you read his posts he said he just had enough cooling stuff left for 1 fast run so he just put 2 volts on the chip and made 1400

he doesnt even know until now if he could have done way more or the same with less volts on the chip so we still dont know what really can be done with the chip maybe 1600?

we will know in the future i hope
 
HardStyler3 said:
funny how everyone looks at the 10k score

and no one looks at the graphics score which is higher then a titanx at 1500-1600mhz could ever do on air but okaaaaaaaaaay

we have the experts here again at work...:rolleyes:


he used a 4790k with the standard cooler and not overclocked at all

also if you read his posts he said he just had enough cooling stuff left for 1 fast run so he just put 2 volts on the chip and made 1400

he doesnt even know until now if he could have done way more or the same with less volts on the chip so we still dont know what really can be done with the chip maybe 1600?

we will know in the future i hope
Click to expand...

I tried to look at the graphics score but I was on my phone and it was too small to read. I also assumed his cpu would be better than that.

Edit; Looks like it would at least equal 8packs best score with a 1700mhz Titan X with the same cpu. Might need a little more on core but doable.
 
Last edited:
HardStyler3 said:
funny how everyone looks at the 10k score

and no one looks at the graphics score which is higher then a titanx at 1500-1600mhz could ever do on air but okaaaaaaaaaay

we have the experts here again at work...:rolleyes:


he used a 4790k with the standard cooler and not overclocked at all

also if you read his posts he said he just had enough cooling stuff left for 1 fast run so he just put 2 volts on the chip and made 1400

he doesnt even know until now if he could have done way more or the same with less volts on the chip so we still dont know what really can be done with the chip maybe 1600?

we will know in the future i hope
Click to expand...

+1

When guys like this and 8pack get more knowledge and time with the chip who knows how far they can take it. That Graphics score of 11511 is only 647 points behind 8packs Ln2 run on a 980ti.

11735, GFX Score 12158, 980 Ti LN2 Cooled, Physics Score 27544, CPU 5960X @5.76 LN2 Cooled - 8 Pack - Link HOF

12160, GFX Score 12759, TitanX LN2 Cooled, Physics Score 27276, CPU 5960X @5.626 LN2 Cooled - 8 Pack - Link

His Titan X is 1248 faster so not at all bad for a first run with a 4790k that was not heavily overclocked.
 
Last edited:
Who really gives a flying frack about LN2 results, It has no real world usability, I can't play a game with it on ln2 and that's why I buy my hardware, For gaming. Over 99% of owners of all brands of cards will never run ln2 so why care?
 
YB2ndBEST said:
I'm just here for boobies
Click to expand...

A cracking pair :eek:

tumblr_lrhhn3k5fs1r0dt8ho1_500.jpg


:D
 
nashathedog said:
Who really gives a flying frack about LN2 results, It has no real world usability, I can't play a game with it on ln2 and that's why I buy my hardware, For gaming. Over 99% of owners of all brands of cards will never run ln2 so why care?
Click to expand...

yeah but it can be an indicator what could be possible with watercooling and unlocked voltages sure we will never get there because you cant cool it normally but like when a 1400mhz /1000mhz hbm can do that maybe we can get 1250/600-700 with good watercooling which would be very nice
 
You must log in or register to reply here.
Back
Top Bottom