Fable Legends: AMD and Nvidia go head-to-head in latest DirectX 12 benchmark

[Dygaza]

In that techreport bench they did used factory overclocked version if 980ti though. Aftermarket cards are cool if used in tests over references, but they should label them in sheets aswell.

 

[tommybhoy]

He knows that.

He also knows that anything that isn't GM200 generally gets beaten by AMD equivalent price point...

 

[Gregster]

In that techreport bench they did used factory overclocked version if 980ti though. Aftermarket cards are cool if used in tests over references, but they should label them in sheets aswell.

The problem is - It is hard to find a non overclocked card to try and make the Fury X look better. I know what you are saying but when you look at the 980Tis at OcUK

https://www.overclockers.co.uk/productlist.php?groupid=701&catid=1914&subid=1402

The cheapest card is the Zotac AMP edition and that boosts to 1355Mhz without touching anything, so it would be disingenuous to hunt down a bog standard Nvidia reference card. When you look at the Fury X, it has its clocks and that is it, no OC cards are available to buy.

Plus when you factor in overclocking, the Zotac AMD edition (which is far cheaper than the cheapest Fury X) which was tested had an OC of 1481Mhz to 1550Mhz, so again, having owned a fury X and seeing it's stable overclock being a poultry 1100Mhz (50 mhz over stock), I think the Fury X is a poor choice for DX12 over the 980Ti.

http://www.guru3d.com/articles_pages/zotac_geforce_gtx_980_ti_amp_extreme_review,35.html

 

[tommybhoy]

Rinse/repeat, while ignoring the volume sales.

 

[mururoa]

It's not realistic to expect them to hunt down a 980 Ti running at the specced reference clocks, since such cards hardly exist; my bog standard reference card boosts to close to 1300 out of the box.

 

[mmj_uk]

It's not realistic to expect them to hunt down a 980 Ti running at the specced reference clocks, since such cards hardly exist; my bog standard reference card boosts to close to 1300 out of the box.

It's impossible I would say, the 'reference' clocks are a minimum guarantee more than anything. I am yet to see an NVidia GPU that doesn't boost a few hundred megahertz above advertised speeds.

As for the Humbug crying about a top end clocking 980Ti 'skewing' results, you can get them for the price of Fury X so how is that skewing anything? AMD's system of reducing the tessellation factor down to an optimum (that is low) level for their GPU's for maximum performance, rather than rendering the application requested setting is skewing benchmark results more than anything else.

 

[fs123]

How have they used Async more in AOTS?

And have you actually looked at the AOTS bench thread? lol

Since the Nvidia cards aren't seeing any real difference between DX11 and DX12 I can only assume that AMD has simply improved their DX12 drivers a lot over the DX11 drivers rather than any significant boost given by DX12 itself..

If Async compute was being used significantly then I would expect some real difference between the Async codepath and the Nvidia codepath of the game since the Nvidia version of the game does not use Async at all.

 

[fs123]

It's impossible I would say, the 'reference' clocks are a minimum guarantee more than anything. I am yet to see an NVidia GPU that doesn't boost a few hundred megahertz above advertised speeds.

As for the Humbug crying about a top end clocking 980Ti 'skewing' results, you can get them for the price of Fury X so how is that skewing anything? AMD's system of reducing the tessellation factor down to an optimum (that is low) level for their GPU's for maximum performance, rather than rendering the application requested setting is skewing benchmark results more than anything else.

AMD does not reduce the tessellation by default. I have tested it and in Witcher 3 the tessellation can be reduced 'manually' via CCC but otherwise it is the same as Nvidia's by default and is refelcted in the performance. Prove it otherwise or stop repeating the lie.

 

[Gregster]

Since the Nvidia cards aren't seeing any real difference between DX11 and DX12 I can only assume that AMD has simply improved their DX12 drivers a lot over the DX11 drivers rather than any significant boost given by DX12 itself..

If Async compute was being used significantly then I would expect some real difference between the Async codepath and the Nvidia codepath of the game since the Nvidia version of the game does not use Async at all.

What I find head scratching is how in AOTS - DX11 gets more frames than DX12 on Nvidia. I know Mantle was great for AMD and gave some great performance boosts but still didn't overtake Nvidia (DX11), so it just shows how well Nvidia had DX11 coded or how badly DX12 is on Nvidia at the mo (pick one).

I think it is great to see a couple of early DX12 demo's and bench's but truthfully, it is far to early to call and I expect some massive improvements once devs and drivers are matured for DX12.

 

[shankly1985]

What I find head scratching is how in AOTS - DX11 gets more frames than DX12 on Nvidia. I know Mantle was great for AMD and gave some great performance boosts but still didn't overtake Nvidia (DX11), so it just shows how well Nvidia had DX11 coded or how badly DX12 is on Nvidia at the mo (pick one).

I think it is great to see a couple of early DX12 demo's and bench's but truthfully, it is far to early to call and I expect some massive improvements once devs and drivers are matured for DX12.

It take over take Nvidia DX11 stop kidding yourself. Benchmarks on here speak for themselves and yes I am ignoring runs that needs some heavy OC. For the Avg joe Mantle was far better performance to anything else.

 

[CAT-THE-FIFTH]

What i want to see is more mid range builds with cards like my 670 in benchmarks not 980s and furys with the lowest being a 960, and the latest cpus with i7s mainly, all the damn time in benchmarks. It's like they think PC users ONLY have the latest things. When most have a little bit older but still really good hardware like the i5 3570k and 670 for examples.

I really want to see some Kepler benchmarks and on older GCN based cards like the HD7970. These are all advertised as being able to run DX12 so it would be nice to see actual performance too.

Plenty of these older cards were current only a year ago or even less,so hardly that old for plenty of people.

I still remember people recommending the GTX770 and R9 280X cards over the GTX960 in the earlier parts of this year too.

 

[Rroff]

Really hard to find Kepler numbers its almost like they are actively avoiding benching the titan/780/780ti.

There are some 680 and 7970 numbers out there which do quite well though the 7970 takes a bit of a lead over the 680 - at a guess it looks like the memory bandwidth is holding some of the older cards back as those with bigger buses/more bandwidth like the 7970 pull ahead a bit compared to cards in a similar ballpark that have more cut back VRAM.

 

[tommybhoy]

Anandtech has 7970/680 results on various cpu's.

For the Avg joe Mantle was far better performance to anything else.

Put in a lot of hours on Mantle, while Nvidia DX11 is great, what you got on screen wasn't on the same level as Mantle, Mantle was miles ahead and gave us insight to what's coming on DX12, can't wait for titles to land.

 

[Gregster]

It take over take Nvidia DX11 stop kidding yourself. Benchmarks on here speak for themselves and yes I am ignoring runs that needs some heavy OC. For the Avg joe Mantle was far better performance to anything else.

Just looked at all of our Mantle Vs DX11 bench results and Nvidia are winning all of them, so stop being blind. And discount overclocking on a site called Overclockers as much as you like, the numbers are there and they don't lie.

 

[shankly1985]

What I find strange and I have said this before on this forum. Since when did Overclocking become real word numbers?

Its one reason I dont follow overclocking bench thread great for competition but that where it ends.

Its not what you get out the box. One reason you see all my results are from stock it tell people a much bigger picture!

I also pointed out we should have stock bench runs but I was shot down because this is a Overclockers forum lol Yeah everyone who buys from here Overclocks there hardware.
Not a chance we are small fish when it comes to overclocking.

Just looked at all of our Mantle Vs DX11 bench results and Nvidia are winning all of them, so stop being blind. And discount overclocking on a site called Overclockers as much as you like, the numbers are there and they don't lie.

You correct number don't lie, but you also need to understand these are not real world numbers..

 

[flopper]

Anandtech has 7970/680 results on various cpu's.



Put in a lot of hours on Mantle, while Nvidia DX11 is great, what you got on screen wasn't on the same level as Mantle, Mantle was miles ahead and gave us insight to what's coming on DX12, can't wait for titles to land.

Latency was the main thing with Mantle.
Nvidia had nothing agaisnt that.
Dx12 will offer the same smoothness for AMD.

 

[Gregster]

Having used both AMD and Nvidia on DX11 and Mantle, there was nothing between them. Both ran great and smooth in the games I played, albeit performance was shocking with Mantle on the FX in BF4.

 

[humbug]

Have you been talking with Scone on Steam again Greg ^^^^^

In that techreport bench they did used factory overclocked version if 980ti though. Aftermarket cards are cool if used in tests over references, but they should label them in sheets aswell.

Its an example of what i was talking about, they are using overclocked Nvidia cards and then not telling you what clocks its actually running at.
These people are just an advertising arm of Nvidia and really should be treated as such.

 

[llamadal]

Zzż

 

[CAT-THE-FIFTH]

Zzż

LOL,still asleep at 1PM!!