Review: Hitman 2016: PC graphics performance benchmarks

Just.... Painful....

 

Why is the Titan X performing so much worse than the 980 Ti in Ultra HD DirectX 12? 27 vs. 35 FPS sure does make a difference.

 

Thanks for the update Hilbert

They really should have sorted that delayed start up for NVidia users before releasing this though, they are just going to get it in the neck over it...

 

I can't take the pain of my 980 ti's getting there bum kicked. lol
Good show for AMD

 

Anyone have any bright ideas that could explain these huge differences in results ?

http://www.pcgameshardware.de/Hitma...Episode-2-Test-Benchmarks-DirectX-12-1193618/

Spoiler

I have no idea, it's strange. Maybe clocks ? It's unclear to me what clock these cards are effectively run at during the test

 

The other site used the latest drivers for both graphic cards types, while Guru3d used the latest official WHQL driver for nVidia cards (364.72).

The other site.
nVidia - Geforce 364.96 Hotfix
AMD - Radeon Software 16.4.2 Beta

Geforce 364.96 Hotfix driver is rumored to fix performance issues where GPU could go into incorrect p-state (performance state).

I would like to see an updated review with game tested on Geforce 364.96 Hotfix driver for nVidia cards.

 

Hilbert, can we not get reported clocks for these tests? I hate having to assume it's ~1200mhz just because that is what the reference design says

PCGH results also have another oddity btw, at the top of the DX12 graph it says "Max details ingame"

But on the actual y-axis it says (lower detail!) for both Fury and Ti

 

Thanks for the clarification, but I'm wondering for example about the 980ti reference boost clock is 1076 if I'm not mistaken, does it actually run at 1076mhz? I remember it running higher even on the reference design

 

I've personally never experienced my card going over the stock boost no matter what the temperature/tdp is, 1366 is written on the box, and 1366 is what I get.

Nonetheless I think at this point it would be reasonable to assume it was running at an average of 1200mhz in your review.

This would put the super-clocked 1350mhz (assuming it was 1350 and not higher) card at a ~13% advantage, certainly not enough to explain these results

 

Total "full" boost 2.0 clock is ~144 - 164MHz from listed base freq.

Kepler GK110B with Boost2.0 base e.g. my old 780GTX 940MHz boost 1006MHz (real 1084MHz), ~ 144Mhz


Maxwell with Boost 2.0 base e.g. my 1178MHz boost 1279MHz (real 1342MHz), ~ 164MHz.

Now when I flashed my card to base 1241Mhz it boosts to 1405MHz, still that 164MHz.

 

It's not so much that the actual boost is higher than advertised, it's that it varies.

My card will boost to 1390 if I'm not mistaken, if gigabyte 'OC mode' is enabled from their hideous guru app, but that also raises the voltage so it goes one bin (13mhz)up

Anyway, my point is *if* the reference Ti was boosting at 1200+, and not 1076 or w/e, it wouldn't account for the disparity between the PCGH results and these, does anyone that plays the game test its performance over driver sets? Could the delta be in the driver's?

 

"We have to say that we are incredible dissapointed about the quality control on DX12 from the developper. These things just shouldn't happen. Very sloppy."

Oh the irony...."dissapointed" and "developper". Very sloppy :^).

 

like Hilbert said
even a ref card that runs ie:1250 can vary in clocks depending on temps and power...the better cooled and factory bios enhaced for more power custom cards can sustain a higher boost clock for more of the time then ref.
it could mean a big difference in number

 

It's not the clocks. PCGH uses their own benchmark run ingame, not the internal benchmark. Seems like real world performance is different from benchmark performance

 

Curious why you did not include updated results using the Nvidia Geforce 364.96 Hotfix driver? In the article the results were updated on March 18 using Radeon Crimson 16.3.1 Hotfix driver, so really isn't at question of only using WHQL drivers..

 

Lets be real here for a moment and then we can start or stop pointing fingers.

Spoiler

Hilbert's Results for 1080p DX12:
Titan X = 78fps; GTX 980 Ti = 76fps.

PCGHW results for 1080p DX12:
GTX 980 Ti Superclocked = 79.9fps
- - - - - - - - - - - - - - - - - - - - -
Hilbert's Results for 1440p DX12:
Titan X = 60fps; GTX 980 Ti = 58fps.

PCGHW results for 1440p DX12:
GTX 980 Ti Superclocked = 61.9fps
- - - - - - - - - - - - - - - - - - - - -
Hilbert's Results for 4k DX12:
Titan X = [mistakes happen]; GTX 980 Ti = 35fps.

PCGHW results for 4k DX12:
GTX 980 Ti Superclocked = 37.1fps

Difference is not in GTX 980 Ti results, but in very different Fury X results. 9~15%. (They claim HQ-AF in driver. I have many more IQ improving tweaks, improving -5 TexLOD+SSAA to remove resulting shimmering => Best AF one can get and AA on top of that.)
I'll say it openly: "Those Clowns in PCGHW borked Fury X by very bad driver tweaks through registry settings."

This is my difference between 3DM Driver Standard (Left) vs. My improved IQ (right).
http://www.3dmark.com/compare/fs/8345321/fs/8345401
1.6% difference in graphics score. (should be noted, that I apparently did not enforce 3DM to run with AA, just with 1 frame pre-render limit, -5lod, HQ-AF and few minor things.)

 

You're ignoring the main point which is the relative performance of 980ti vs fury X. Guru3d didn't test episode 2 dx11

PCGH uses their own benchmark run, not the built-in benchmark. That's the main difference afaik, I think driver settings should be completely default when testing, I have no idea what you're on about in settings for amd