Overclock.net banner

1 - 20 of 60 Posts

·
SpaceStationGaming
Joined
·
9,430 Posts
Discussion Starter #1
http://www.pcgameshardware.de/Hitman-Spiel-6333/Specials/Episode-2-Test-Benchmarks-DirectX-12-1193618/






What happened to AMD's lead in this game?

Quote:
Hitman in DX12 almost always looks worse than under DX11. This is primarily the timid streaming of DX12 which, unlike DX11, doesn't always apply maximum texture resolution. We suspect that this DX12 memory management creates many headaches for developers and that the current setting is a compromise to stifle a buffer overflow, and to nip potential performance drops in the bud.
Quote:
Even before the "Sapienza" DLC, Hitman received a patch with the option to turn off these "memory protections" and, regardless of the installed video memory, set maximum quality.
In the launcher, select "disable memory protection" to "yes"; this also improves DX12 texture streaming. Contrary to expectations the symptoms did not disappear completely - they are merely weakened. DX12 without memory protection is graphically "almost" as good as DX11, usually only missing a few nuances in texture resolution. In the following comparison, we photographed a point where the differences are clearly visible:
(they show the comparison)
Although the temptation was great, because of these quality differences, we decided against testing all graphics cards under both DX11 and DX12. We will do so as soon as a (further) Patch helps. Interestingly, we at least have DX12 comparison values for the respective top models from AMD and Nvidia - click below the benchmark on "DirectX 12 vs. DirectX 11". It is striking that the GeForce GTX 980 Ti benefited only at a low resolution in DX12, while the Radeon R9 Fury X had performance left in the bank, and was able to reach the level of overclocked GM200 card - but only under DX12. Yet much more impressive and enlightening to us, in the next article section, are the effects in the CPU limitations.
 

·
Premium Member
Joined
·
10,885 Posts
Quote:
Originally Posted by dmasteR View Post

What happened to AMD's lead in this game?
They started gimping Hawaii/Fiji in preparation for the Polaris launch.

smile.gif
 

·
SpaceStationGaming
Joined
·
9,430 Posts
Discussion Starter #5
Quote:
Originally Posted by qlum View Post

Looking at these screenshots I can see where the performance boost comes from: http://www.pcgameshardware.de/commoncfm/comparison/clickSwitch.cfm?id=131725
Did you even read the rest of the article?

DX12 has issues.
Quote:
Hitman in DX12 almost always looks worse than under DX11. This is primarily the timid streaming of DX12 which, unlike DX11, doesn't always apply maximum texture resolution. We suspect that this DX12 memory management creates many headaches for developers and that the current setting is a compromise to stifle a buffer overflow, and to nip potential performance drops in the bud.
Quote:
Even before the "Sapienza" DLC, Hitman received a patch with the option to turn off these "memory protections" and, regardless of the installed video memory, set maximum quality.
In the launcher, select "disable memory protection" to "yes"; this also improves DX12 texture streaming. Contrary to expectations the symptoms did not disappear completely - they are merely weakened. DX12 without memory protection is graphically "almost" as good as DX11, usually only missing a few nuances in texture resolution. In the following comparison, we photographed a point where the differences are clearly visible:
(they show the comparison)
Although the temptation was great, because of these quality differences, we decided against testing all graphics cards under both DX11 and DX12. We will do so as soon as a (further) Patch helps. Interestingly, we at least have DX12 comparison values for the respective top models from AMD and Nvidia - click below the benchmark on "DirectX 12 vs. DirectX 11". It is striking that the GeForce GTX 980 Ti benefited only at a low resolution in DX12, while the Radeon R9 Fury X had performance left in the bank, and was able to reach the level of overclocked GM200 card - but only under DX12. Yet much more impressive and enlightening to us, in the next article section, are the effects in the CPU limitations.
Even in that picture you linked. The 980Ti looks better than the R9 Nano when comparing DX12 "Memory Protection" Form.
 

·
Banned
Joined
·
5,762 Posts
Quote:
Originally Posted by qlum View Post

Looking at these screenshots I can see where the performance boost comes from: http://www.pcgameshardware.de/commoncfm/comparison/clickSwitch.cfm?id=131725
lol comapre 980ti vs nano dx12 screens
whistle.gif
google translates it as "memory protection", prolly kicks in when 4GB starts running out.
anyway, the game really looks miles better in dx11 even than in dx12 without memory protection thingy, and fury x gets
buttkick.gif
even by 980.
 

·
Registered
Joined
·
2,001 Posts

·
SpaceStationGaming
Joined
·
9,430 Posts
Discussion Starter #8

·
Banned
Joined
·
5,762 Posts
Quote:
Originally Posted by Themisseble View Post

Let me tell you : PCGH happened.

https://www.youtube.com/watch?v=Do-Ex7VtzfA

I cant help myself..., but I hate people which believe everything on the Internet and searching benchmark that only they like. If you look at few sites you will see many contractions...
lol the graphics on fury x look downgraded like in the link above. you gotta get your eyes checked bro.

 

·
SpaceStationGaming
Joined
·
9,430 Posts
Discussion Starter #10
Quote:
Originally Posted by Klocek001 View Post

lol the graphics on fury x look downgraded like in the link above.
Good thing you pointed this out, I actually missed this. The Fury X textures are indeed more blurry compared to the 980Ti in this. Quite noticeable in fact at 1:00 mark.

https://youtu.be/Do-Ex7VtzfA?t=60

Which further proves the memory protection issue that was stated in the benchmark.
 

·
Spaghetti
Joined
·
1,130 Posts
Is it just me, or does the 980 Ti look... really really bad compared to the fury?
 

·
Premium Member
Joined
·
9,327 Posts
 

·
SpaceStationGaming
Joined
·
9,430 Posts
Discussion Starter #13
Quote:
Originally Posted by Buris View Post

Is it just me, or does the 980 Ti look... really really bad compared to the fury?
Read the article. Or even the first 10 posts....
Quote:
Hitman in DX12 almost always looks worse than under DX11. This is primarily the timid streaming of DX12 which, unlike DX11, doesn't always apply maximum texture resolution. We suspect that this DX12 memory management creates many headaches for developers and that the current setting is a compromise to stifle a buffer overflow, and to nip potential performance drops in the bud.
Quote:
Even before the "Sapienza" DLC, Hitman received a patch with the option to turn off these "memory protections" and, regardless of the installed video memory, set maximum quality.
In the launcher, select "disable memory protection" to "yes"; this also improves DX12 texture streaming. Contrary to expectations the symptoms did not disappear completely - they are merely weakened. DX12 without memory protection is graphically "almost" as good as DX11, usually only missing a few nuances in texture resolution. In the following comparison, we photographed a point where the differences are clearly visible:
(they show the comparison)
Although the temptation was great, because of these quality differences, we decided against testing all graphics cards under both DX11 and DX12. We will do so as soon as a (further) Patch helps. Interestingly, we at least have DX12 comparison values for the respective top models from AMD and Nvidia - click below the benchmark on "DirectX 12 vs. DirectX 11". It is striking that the GeForce GTX 980 Ti benefited only at a low resolution in DX12, while the Radeon R9 Fury X had performance left in the bank, and was able to reach the level of overclocked GM200 card - but only under DX12. Yet much more impressive and enlightening to us, in the next article section, are the effects in the CPU limitations.
When you compare the DX12 Memory Protection Form 980Ti vs Nano you will see the 980Ti actually looks better.

http://www.pcgameshardware.de/commoncfm/comparison/clickSwitch.cfm?id=131725
Quote:
Originally Posted by sugarhell View Post

Can't read what the yellow sign says on the Fury X.
rolleyes.gif
 

·
Registered
Joined
·
169 Posts
Quote:
Originally Posted by dmasteR View Post

Did you even read the rest of the article?

DX12 has issues.

Even in that picture you linked. The 980Ti looks better than the R9 Nano when comparing DX12 "Memory Protection" Form.
I did not read the article before as my german is kind of rusty I did now my point is still valid though just claiming that the performance is much better when the game looks much worse is kind of unfair. If anything I would just disregard any dx12 performance until the issues are fixed and you can actually compare the two fairly. Even amd vs Nvidea is kind of unfair on a game that clearly has issues at this point.
 

·
Registered
Joined
·
827 Posts
Why isnt there an overclocked Fury X included in the test?
 

·
Premium Member
Joined
·
9,327 Posts
Quote:
Originally Posted by Death Saved View Post

Why isnt there an overclocked Fury X included in the test?
Because it's an(in) overlcocker's dream
wink.gif
 

·
Registered
Joined
·
2,001 Posts
Quote:
Originally Posted by dmasteR View Post

Reference 980Ti vs PCGH using a OC'd 980Ti like they did in their previous Hitman Benchmark. Don't see your point.
And how much of an overclock you will get out? 50%?

His GTX 980Ti runs at 1200-1240MHz on benchmark
https://www.youtube.com/watch?v=AW_V78MldUs
Sometimes even higher...

Okey, lets make it clear

To get 20% boost of an OC you need to clock your core and memory clock both for 20%.
So stock GTX 980Ti boost is 1076 and memory is 1753.

Lest say you OC it to Core1500 MHz/Memory 2000MHz
39% OC for core clock
14% OC for memory clock

usually I will make it very simple. You count Mem % OC with Core% OC and just /2 = ... so 39 + 14 = 53/2 = 26.5% Thats 26.5%, if everything is perfect!
In our case (1500/2000 vs 1200/1753 (his is little higher actually) :
14% mem OC (little less actually 10%)
25% core OC

SO best case scenario of your OVERCLOCKING MIGHT
14% + 25% = 39/2 =19.5%
100 (FPS) *1.185 = 119.5 FPS
75 (FPS) * 1.185 = 89.65 FPS
50 (FPS) * 1.185 = 59.75 FPS

Of course...... nobody cared about overcking 7850 which was much better overclocker than any offer now. You could get like 45-50%... but easy normal and safe OC was like 25%.

... and stop pointing out OC on GTX 980TI you just wont get 50% OC. By overclocking your GPU to a sweet spot around 10% you will already get around 5-10% higher power consumption. But reaching for really high number like 1500MHz+ might raise your power consumption by 50% or more for getting only 20% of boost.

And no dont show me power consumption on furmark.. because fury X will run at 1000-1050MHz while GTX 980TI will run at 925-975MHz.
 

·
Banned
Joined
·
5,762 Posts

·
Premium Member
Joined
·
9,327 Posts
Quote:
Originally Posted by Klocek001 View Post

http://cdn.overclock.net/a/a7/500x1000px-LL-a7c67f87_nano.jpeg
http://www.overclock.net/content/type/61/id/2771748/width/500/height/1000/flags/LL

the wonders of HBM and async all coming together. that electric box on nano reminds me of when I played return to castle wolfenstein on my s3 savage.
Even the article is implying that this is a problem with dx12 memory management. Instead you always try to convert this into an AMD vs nvidia fanboy fight. Your bias is so obvious.

Funny thing is GP100 will gonna use HBM. Then you will worship it .

Imagine if they announce GP100 async support. I will save this comment for future reference. Thank you
smile.gif
 
  • Rep+
Reactions: Blameless
1 - 20 of 60 Posts
Top