• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Witcher 3 PC Performance Thread

Genio88

Member
Those benchmarks are on Ultra/Max with tons of performance killers activated. You'll most likely be able to turn down a few settings, lose next to nothing in fidelity and reach that 60 fps.

But i want those effects, HBAO+ and Hairworks, at this point i'd rather play it at 30fps than 60fps without them
 

The Cowboy

Member
They say they're using the newest nVidia driver and "The Witcher 3 v1.02, ('semi-final' review build)." I have no idea what the Day 1 patch is going to be (v1.03?), but thats what they're using.

Edit.

Nope i'm reading it wrong, it is indeed saying 1.02 on the 1st ones and 1.02 semi-final for the 2nd set of benchmarks - weird.
 
way ahead of you I got it figured out but thanks...Nvidias strategy officially has me confused...in terms if their proprietary use and their marketing strategies....just go open standard and dnt cripple for heavens sake.

so now they are semi open to everyone but are severely crippling AMD and their older stock
It would be nice if the heavier game work FX scaled better and had more scaling options (low, high etc...). And yeah, I am for open standards as well.
Yeah but the question is ... What fps do you get at high ... I need 60^^
Probably rather fine FPS.
 

Kezen

Banned
I don't see anything wrong with those numbers. Great to see the game is very taxing even at 1080p. It will drive GPU demands and sales.
 

The Llama

Member
Those are some strange results, it's very surprising to see a 770 outperformed by a 285.

Someone did an analysis a while ago and basically find that since the Maxwell cards were released, Kepler performance relative to AMD card performance in recent games had gotten worse.
Basically, nVidia doesn't seem to be spending as much time optimizing their games for Kepler as they do for Maxwell.

Your reading it wrong, the 1st set of benchmarks is running 1.02 final build, the 2nd set of benchmarks is showing how it was with the semi-final release build.

Both say they're using v1.02. But for some reason on one set of benchmarks its listed as "The Witcher 3, Review-Version 1.02" and on the other its listed as "The Witcher 3 v1.02, ('semi-final' review build)." I couldn't tell you why they listed it that way though.
 

viveks86

Member
Sounds like a SLI issue, that will probably be addressed soon. 780 SLI vastly outperforms a single 980 and we already know that card is perfectly capable of running the game at 1080/60.

Probably the case

I'm sure i'm reading this wrong but are you suggesting that an "old preview" is better optimized than the final version with two patches and dedicated drivers?

Many features simply didn't work consistently in the old preview version, such as max draw distance, foliage visibility, HBAO+ etc, which threw any benchmarking results off. The review version had those rectified and made it more demanding. The 290X benchmark was done on the preview build. Has nothing to do with optimization.
 

Evo X

Member
They say they're using the newest nVidia driver and "The Witcher 3 v1.02, ('semi-final' review build)." I have no idea what the Day 1 patch is going to be (v1.03?), but thats what they're using.

Do they specifically mention the driver version?

Because the newest ones just came out like 30 minutes ago. Doubt that is enough time to run all the tests for the different GPU setups.

Seems misrepresentative of the final product and makes me think they are just doing it to be the first ones on the block with benchmarks.
 

Derp

Member
But i want those effects, HBAO+ and Hairworks, at this point i'd rather play it at 30fps than 60fps without them
I'm literally going to do that (except I still can't decide on HairWorks because Geralt's default hair looks better than his hair when HairWorks is on, however the HairWorks on the beasts/monsters looks incredible...). But yeah, all settings are getting cranked up and fps is staying capped at 30.
 

The Llama

Member
Do they specifically mention the driver version?

Because the newest ones just came out like 30 minutes ago. Doubt that is enough time to run all the tests for the different GPU setups.

Seems misrepresentative of the final product and makes me think they are just doing it to be the first ones on the block with benchmarks.

Yes, these are the drivers they say they are using:

Catalyst 15.4.1 beta, Geforce 352.86 WHQL

According to nVidia, 352.86 is the newest driver for The Witcher 3:

http://www.geforce.com/whats-new/articles/geforce-352-86-whql-driver-released

From what I can tell, Catalyst 15.4 was the GTA5 driver for AMD cards, and 15.4.1 just added some FreeSync improvements/bug fixes, and doesn't have any improvements or fixes for TW3.
 

The Cowboy

Member
Both say they're using v1.02. But for some reason on one set of benchmarks its listed as "The Witcher 3, Review-Version 1.02" and on the other its listed as "The Witcher 3 v1.02, ('semi-final' review build)." I couldn't tell you why they listed it that way though.

Yeah, i edited my post whilst you typed this, you are right.
 
22-25fps on a 770? Wtf happened? I guess I'll have to play it on low or some Ultra settings simply destroy the performance.

Welp. I was going to hold off purchasing a 970 till the end of next month. Looks like I might have to upgrade at the end of this month instead.
 
Many features simply didn't work consistently in the old preview version, such as max draw distance, foliage visibility, HBAO+ etc, which threw any benchmarking results off. The review version had those rectified and made it more demanding. The 290X benchmark was done on the preview build. Has nothing to do with optimization.
I'm sure i'm reading this wrong but are you suggesting that an "old preview" is better optimized than the final version with two patches and dedicated drivers?

The newer version of the game added a new graphical preset. The old Ultra was shoved down to very high, and the new ultra is even higher.

Consoles are high, very high is the previous ultra, ultra is the new ultra.
 

BigTnaples

Todd Howard's Secret GAF Account
They have 4K benchmarks up to.


A single Titan X getting 30fps with Hairworks off and SSAO not HBAO+


A single 970 hits 21FPS at the same settings.


Show me some SLI benchmarks!


These Benchmarks seem off, but we will see. Hoping I can still run 4K 30fps with Hairworks and HBAO+....
 
They have 4K benchmarks up to.


A single Titan X getting 30fps with Hairworks off and SSAO not HBAO+


A single 970 hits 21FPS at the same settings.


Show me some SLI benchmarks!


These Benchmarks seem off, but we will see. Hoping I can still run 4K 30fps with Hairworks and HBAO+....

Like I said, they did increase the ultra preset in the review version.
 

Genio88

Member
Do they specifically mention the driver version?

Because the newest ones just came out like 30 minutes ago. Doubt that is enough time to run all the tests for the different GPU setups.

Seems misrepresentative of the final product and makes me think they are just doing it to be the first ones on the block with benchmarks.

Yes they specifically say Nvidia 352.86 WHQL drivers, and they're out from more than 30 minutes actually, but i'm not so confident about those bench
 

tuxfool

Banned
so hair works calculations would in theory be offloaded onto the chosen physx medium that's what I'm getting at...is gameworks just physx framework with extensions...

No, PhysX on GPU is based on CUDA. Hairworks is based on DirectCompute. These are completely separate things.

When using physX on the gpu you can offload it to a dedicated physx card. You cannot do the same for DC.
 

The Llama

Member
Yes they specifically say Nvidia 352.86 WHQL drivers, and they're out from more than 30 minutes actually, but i'm not so confident about those bench

I wouldn't necessarily say that I'm not confident about that sites benchmarks, but I'm definitely interested in seeing more benchmarks for this game. Like I said before, I'm legitimately surprised (even as an AMD user!) that the AMD cards performed as well as they did.
 

Saintruski

Unconfirmed Member
It would be nice if the heavier game work FX scaled better and had more scaling options (low, high etc...). And yeah, I am for open standards as well.

Probably rather fine FPS.

scaling would be a plus I don't understand why it Isnt already a thing. I was kinda ok with the proprietary thing they made it they own it type thing until project cars came along and my 780 was out performed by a 960 because gameworks....that was my first offici gameworks encounter and why im do dam confused...when you start gimping your own userbase thats BS sad thing is I was forced to upgrade to a 980 Cuz AMD let them get such a stranglehold on the market gamesworks ia on like a million titles now...
 

viveks86

Member
Or, play it on the recommended High settings with a few ultra sliders.
THey mention how the latest version of the game has a high, very high, and ultra setting. So they added one more Cvar grouping. High is consoles, very high is presumably the previous ultra, and ultra is the new ultra (über?)

Wait, what? Really? Fascinating. I've backed p all the ini files. Would be interesting to see the changes
 
Are those benchmarks using the newest Nvidia drivers and Day 1 patch?

I am thinking no, so our performance tonight should be better.

352.86 are the new nVidia drivers, which they are using.
They're running on game version 1.02 which I think is with Day 1 patch, though I could be mistaken.
 

Saintruski

Unconfirmed Member
No, PhysX on GPU is based on CUDA. Hairworks is based on DirectCompute. These are completely separate things.

When using physX on the gpu you can offload it to a dedicated physx card. You cannot do the same for DC.


I got that earlier but thanks for the help :)
 

Grechy34

Member
They have 4K benchmarks up to.


A single Titan X getting 30fps with Hairworks off and SSAO not HBAO+


A single 970 hits 21FPS at the same settings.


Show me some SLI benchmarks!


These Benchmarks seem off, but we will see. Hoping I can still run 4K 30fps with Hairworks and HBAO+....

A single Titan X getting those results on 4K. Still a few years away yet.
 
Yeah ,we need high, Vhigh and ultra benches with screenshots before people flip out.

Probably a case of huge cost and little gain when you go above PS4 spec and PS4 spec gives you 60fps on hardware twice as fast.
 

Genio88

Member
I'm literally going to do that (except I still can't decide on HairWorks because Geralt's default hair looks better than his hair when HairWorks is on, however the HairWorks on the beasts/monsters looks incredible...). But yeah, all settings are getting cranked up and fps is staying capped at 30.

Indeed, i mean it's not the kind of game where 60fps makes a huge difference, like would do in a FPS or Action, instead those monsters hairworks looks pretty cool and without HBAO+ and other settings on Ultra it would look just like a console version actually
 
So there are two presets above consoles, very high and ultra ?

Wait, what? Really? Fascinating. I've backed p all the ini files. Would be interesting to see the changes

That is what they say in the article. A new very high preset showed up and the newer ultra is apparently better.
scaling would be a plus I don't understand why it Isnt already a thing. I was kinda ok with the proprietary thing they made it they own it type thing until project cars came along and my 780 was out performed by a 960 because gameworks....that was my first offici gameworks encounter and why im do dam confused...when you start gimping your own userbase thats BS sad thing is I was forced to upgrade to a 980 Cuz AMD let them get such a stranglehold on the market gamesworks ia on like a million titles now...

Yeah the march of time and GPUs does suck (for the wallet) and nvidia definitely "advances" their gameworks features based upon hardware presumptions that somewhat unfairly disadvantage older / similarly powerful cards.
A single Titan X getting those results on 4K. Still a few years away yet.

The benches partially contradict Nvidia recommendations. Perhaps that was before the newer ultra preset...
 

RE4PRR

Member
Don't really trust those benchmarks tbh, not compared to Nvidia's own guide.

Got a 5820k with 980 on 1080p so gonna max everything, want this game to push me to overclock the gpu for once.
 

Klossen

Banned
Indeed, i mean it's not the kind of game where 60fps makes a huge difference, like would do in a FPS or Action, instead those monsters hairworks looks pretty cool and without HBAO+ and other settings on Ultra it would look just like a console version actually

To each their own. Being able to run the game in 60 fps is why I prefer the PC version. Makes for a huge difference in visuals and animations having it played at double the amount of frames. Not to mention improved input and smooth camera movements. Hopefully, the benchmarks just point the Ultra preset being extremely taxing rather than the game itself being that demanding. Playing on PS4 settings really shouldn't be that demanding for 60 fps.
 

napata

Member
I don't see anything wrong with those numbers. Great to see the game is very taxing even at 1080p. It will drive GPU demands and sales.

Very taxing while being pretty meh in the graphics department. You're basically saying that developers shouldn't bother optimizing their games. Nothing in this game seems to warrant performance that bad. Games looks exactly like the ps4 version.

It looks worse than DA:I while at the same time it seems to run worse as well based on those benchmarks.

Edit: Well apparently there are now new and higher settings?
 

Kinthalis

Banned
To each their own. Being able to run the game in 60 fps is why I prefer the PC version. Makes for a huge difference in visuals and animations having it played at double the amount of frames. Not to mention improved input and smooth camera movements. Hopefully, the benchmarks just point the Ultra preset being extremely taxing rather than the game itself being that demanding. Playing on PS4 settings really shouldn't be that demanding for 60 fps.

Yep, it's evident on the 60 FPS footage we've seen. 60 FPs just brings the animationa nd world to life.
 

Bl@de

Member
Don't really trust those benchmarks tbh, not compared to Nvidia's own guide.

Got a 5820k with 980 on 1080p so gonna max everything, want this game to push me to overclock the gpu for once.

You should trust them. I've been reading the magazine for years. Their benchmarks are good. Better don't trust NVidia. Just remember: The 1.02 version has a new ultra preset.
 

Derp

Member
They have 4K benchmarks up to.


A single Titan X getting 30fps with Hairworks off and SSAO not HBAO+


A single 970 hits 21FPS at the same settings.


Show me some SLI benchmarks!


These Benchmarks seem off, but we will see. Hoping I can still run 4K 30fps with Hairworks and HBAO+....
welp.gif


Here i was thinking i could 30fps 4K on a 780 Ti if GameWorks was completely off. I could do over 30fps 4K in GTA 5 with everything maxed except msaa and advanced graphics tab.

Hopes and dreams crushed again.

Also what card do you have BigTnaples? Assuming you have a Titan X maybe overclock it a bit and you could get a small fps increase to allow room for HBAO+? Not sure about HairWorks though, since i thought that was hella demanding.
 

derExperte

Member
Yes they specifically say Nvidia 352.86 WHQL drivers, and they're out from more than 30 minutes actually, but i'm not so confident about those bench

PCGH is usually very reliable and it's a safe bet that they got those drivers before we did.
 

Easy_D

never left the stone age
What's going on?! How many different PC versions are there? :lol

Edit: If the 285 can reach a minimum of 25 I'm feeling a bit more confident in my 280X

Edit: Wait what, the 280x is getting lower frames than the 285 at 1080p? That's weird, the only thing the 285's got going for it is a higher pixel rate
 

Rodin

Member
Probably the case



Many features simply didn't work consistently in the old preview version, such as max draw distance, foliage visibility, HBAO+ etc, which threw any benchmarking results off. The review version had those rectified and made it more demanding. The 290X benchmark was done on the preview build. Has nothing to do with optimization.

Ok thanks for the clarification. So maybe the frame rate in those benchmark is butchered by foliage visibilty and max draw distance, similarly to what happened in GTA V with grass on ultra and extended fov on 100?
 

tuxfool

Banned
Supposedly AMD is also releasing a witcher3 driver, but in case they don't or their driver is delayed there is something people can try.

This is the latest Windows 10 driver but modded to work all the way down to Vista. See the thread, but the long and short of is that it has reduced overheads (one of the tests showed 10%). Note that the process is a bit involved but if people want to tinker there it is:

http://forums.guru3d.com/showthread.php?t=399099

I'll probably try this out later on W8.1 and see what happens.
 
Ok thanks for the clarification. So maybe the frame rate in those benchmark is butchered by foliage visibilty and max draw distance, similarly to what happened in GTA V with grass on ultra and extended fov on 100?
I wonder when that happened though, because DansGaming did not have such a preset. He was already running v1.02

fMXl7gw.png

Oh jeebus, I reread it and I made a mistake whilst reading so quickly. : 0

They did not add a very high option, rather the new Ultra is just that much higher.

Zwischen der Preview-Version und unserer Review-Fassung entspricht der optische Unterschied mindestens einer zusätzlichen Detailstufe, also etwa von "Sehr hoch" zu "Ultra".

"The optical differences between the preview version and our review version approximates, at the least, a new additional detail level, the equivalent of moving from "very high" to "ultra"." MY BAD.
 

tuxfool

Banned
Very taxing while being pretty meh in the graphics department. You're basically saying that developers shouldn't bother optimizing their games. Nothing in this game seems to warrant performance that bad. Games looks exactly like the ps4 version.

I was wondering when the "this game is unoptimized" post came along.
 

viveks86

Member
Ok thanks for the clarification. So maybe the frame rate in those benchmark is butchered by foliage visibilty and max draw distance, similarly to what happened in GTA V?

Pretty much, except since this is directly related to view distance, it's far more noticeable than the ultra grass in GTA V.

Oh jeebus, I reread it and I made a mistake whilst reading so quickly. : 0

They did not add a very high option, rather the new Ultra is just that much higher.


"The optical differences between the preview version and our review version approximates, at the least, a new additional detail level, the equivalent of moving from "very high" to "ultra"." MY BAD.

Ok that makes sense and is in line with my explanation as well. The actual preset is the same, it just started working the way it should in the review build.
 

Kiant

Member
So why did they add another preset?

Heres hoping there isn't a lot of dip at 1440p on the Titan X. If its around the 50fps range most of the time, Gsync should be able to pick up the slack.
 
Top Bottom