Nvidia Geforce Drivers Version WHQL 344.48 Download & Discussion

FFXIV will benefit from extra 4 threads, that's why you don't see any change even @ lower reso.

In short that new Haswell 4core/4threads is "too slow" for it, sorry to say that..

 

Hmm the game runs flawless on my GTX 980. Strange, not sure what makes the 970 so different other than it being slower.

 

Alien Isolation is freezing with these driver and that is only problem so far.

 

I have been getting +200 point performance increase in Fire Strike scores with these.

http://www.3dmark.com/3dm/4489090

Before 344.42 and 344.48 I was only getting 38xx scores

 

No, you don't understand pixel ratios.

4x resolution = 2x2=4

so 4x 1024x768 is ACTUALLY 2048x1536 because it is 4 TIMES as many pixels as 768p

Just as 4x of 2048x1536 is 4096x3072

Nvidia needs a tool tip to EXPLAIN this to users, I have mentioned this several times.

150% of 1920x1080 is actually (Not possible evenly, closest you can get is an ugly and odd 1.23x1.23=1.5129) and would be 2361.6 x 1328.4

Unless what Dice actually means by 150% is that each axis is scaled by 150% (Which WOULD be 2880x1620)

Poor wording by the developer.

Because 150% resolution then actually = 225% resolution scale

Ugh, Ok. FIrst of all> SGSSAA is NOT MSAA.

Second. There is literally no significant blurring with this bit and 4xSGSSAA. If you think so, then again, the problem is you. Because it's just as sharp as with no AA.

If you were using 4x or 8x SGSSAA with 4012C1 instead of C5, then yes. I could UNDERSTAND some sharpening. (Because there IS some blurring with C1, 4012C5 only works for 4xSGSSAA and nothing else)

But 4xSGSSAA with 4012C5 literally has no blurring of any kind


What you perceive as "Blurring" Is actually metric **** tons of high frequency information (Read: Aliasing) is being averaged and resolved in what it actually is supposed to look like (Read: Anti Aliased)
Because again it is aliased. Final Fantasy XIV has tons of texture aliasing. When you see the raw ultra sharp no AA picture, it's no wonder you think it's being blurred when it's actually being Anti Aliased. Because, it's Aliased. Even aliased shadowmaps are being Anti Aliased. Textures are being anti aliased. Because they are aliased. Foiliage, geometry are being anti aliased

http://screenshotcomparison.com/comparison/97319

If it isn't clear enough, let's blow up these already big pictures by 200% with a nearest neighbor filter which preserves all hard edges.

No AA http://i.minus.com/iPpCBJVV2QzP9.jpg
4x SGSSAA http://i.minus.com/ih2RERChbpMRY.jpg

It's easily visible, that everything with no AA is aliased. As I mentioned. Textures, foliage,shadows, geometry, information is being resolved as it properly should be with AA. Many objects are in fact resolving more detail in a more life like manner.

But this is perceived as "Blur" which is far from true at all.

Also, there's no way you will get 60FPS at 2560x1440 on max with 4xSGSSAA with a 970.

SGSSAA is FAR more demanding than MSAA.

And again, MSAA=/= SGSSAA.


BTW with 12C5 you really shouldn't be using a negative LOD bias with FFXIV. You are just creating more aliasing in game that uses Negative lod bias like crazy.


So, if you are running into problems from using too many outside tools at once.

That is not Nvidia's or the Driver's fault.

 

I confirm, on a GTX 970 SLI Strix with a screen native resolution 2556x1440 in ROG swift and DSR and available with single GPU and disparer when I go SLI.

 

Evil Within max settings

Zoom in 1080p
[/IMG]

the same zoom with DSR 2880X1620
[/IMG]

and zoom on 4K DSR


@MrBonk yes
1080*1.5
= 2880x1620

100*1.5*1.5 = 2.25x. 2.25x DSR resolution is 1.5x resolution scale on the X and Y axis at the same time. That's how bf4 handles it - 150% means it is 50% wider and 50% taller (for a total of 2.25x as many pixels) and one of the ways that it's written in DSR

 

1. Stop posting /IMGs please, you are wasting people's bandwidth. Post them as links please instead.
2. Stop using extremely compressed JPGs, you are making your comparisons hard to take seriously because so much is wrong and inaccurate with the IQ. Especially when you are using 4:2:0 Chroma Subsampling
3. It's still not 150% resolution. Again, poor wording and understanding by those who put it there. :stewpid: it's 225% the pixel count of 1920x1080 with 1920x1080 being the baseline of 100%. Therefore, it's not 150% resolution. It's 225% resolution. That's like saying 1920x1080 is 120% the resolution of 1600x900. Which it's not. It's 144%

1920 is 120% of 1600. But that doesn't =/= 120% more resolution

Proper wording for this would be "Resolution Axis Ratio/scale: 1.5x1.5" And then a second description telling you what %resolution this is and what level of SSAA it equates to

 

A problem I have with DSR, is that factors like 120% or 200% are essentially useless and pointless to me. Because they require decimal multiples that are far more complicated than they need to be. (Why should it require a decimal of 1.414375 to get 2.00x when you can get 1.96 far easier with only 1.4x1.4 )

They should just let us create custom resolutions and put a check box whether that resolution should be set for DSR rather than regular resolution.

Or specify our own ratios.

We can't go past 4k, but the biggest ratio you can get is 4x (2x2). So if you don't have a 1080p monitor, you can't use 4k like you would with GeDoSaTo

 

DSR & GSync working here mate in the few games ive got.
BF4, Tombraider,Watchdogs,Blacklist.

DSR is just an easier way of downsampling which has been around and used by many around here for quite a while, but I agree its a nice feature and now its so easy to set up is availible for the masses.

 

Native 4K works a treat

 

Thanks sherlock, like i said, SGSAA doesn't work alone for me, only works (unless it's only MSAA working) when MSAA is enabled, but then again image becomes too blurry from MSAA.

Dunno where you came up with the SGSSAA is not MSAA lol

Are you using the same flag from the post i gave the link? Maybe i'm doing something wrong.

I spoke with a guy who said he can play at 4k the game no problems with his ASUS 970, only issue seems to be the HUD messing up from downscaling which i know well too.

For me i get 30FPS at 4k or less even in areas with no people.

He says he runs a 2500k at 5Ghz, so you're saying a 2500k at 5Ghz is far better than a 4670k @ 4.5Ghz?

I'm confused

 

DSR works very nice for me, i see much sharper picture then native 2560x1080.
Not only on screenshots but also on monitor screen wen watching

 

I did notice that DSR is very wonky with games running windowed/windowed borderless. The game itself renders at the downsampled resolution, but the lack of any filtering (and probably the nature of the windowed display of content) makes it have way more aliasing than at native resolution.

While running the games in fullscreen does work properly, though.

 

it doesn't work with GSync + SLI (which PhazeDelta1 has) according to NVIDIA.

 

Depends what he finds for "play with no problem", you didnt mention what kind of fps does he get..

Also why bother with 4k? What do you get at your native, same 30fps?




No they're around equal, your 4670K has a slight edge over it..

 

Ahh yeah missed that bit.

My bad

 

One of the few threads related to bad 970 performance in FFXIV:

http://www.reddit.com/r/ffxiv/comments/2hn4gt/ffxiv_gtx970_gets_less_fps_than_hd_7850_during/

In my case i can't even put the settings on max or i stays around something and even goes to 30 something in areas with a lot of players.

I can say i disabled the setting to render objects that are not visible, helps a lot, low details for distant objects also set and also a nice difference in FPS.

Then i can say disabling animations and effects for other players and party members also helps a lot, disabling other play shadows also helps a lot.

I've been trying to stabilize this stubborn CPU that isn't stable at all at 4.5Ghz, max was 20m stress test on AIDA64 but i can play some time without it crashing the system :micro:

The problem is not temperatures is stability, and in this case i'm not sure if lower or raise voltage so i end up going back to 4.2Ghz which i don't even need to set vcore

Like i already said i barely see a difference between 4.2 and 4.5/4.6Ghz

Every other game works as it was supposed to when it comes to upgrading the GPU including with DSR.

I can say i get the same FPS at 1080p and 1440p, if i raise to 4k it goes from 50s-60FPS to 30 or less.

It must be a problem on how the drivers and this GPU renders stuff that is causing this issue.?

 

Well here is the thing, I see a difference by cpu bound games, ie going from 4.2 to 4.6ghz, even 4.6 to 4.7ghz helps a little.. For example RIFT or X Rebirth, X3 TC, DR3, GTA4, Hitman Absolution, Crysis2 mp and few more..


All those extra culling tweaks are cpu bound, idk maybe they didnt update it like in benchmark, I saw older benchmark ran at least 50-70% slower then v2.

old: 8926
new v2: no gpu oc 14735, medium gpu oc 16441




If your cpu is not stable then it can have lower fps too, undervolted and then throttles more. I've seen that myself.


^
Prefect quick test is, Cinebench15 gpu benchmark, if cpu is undervolted it can have up to 20fps difference (189, undervolted 169fps).
Vccsa - cpu system agent voltage is crucial for this difference, 2nd its cpuv. Anyway by system agent add extra 0.040 or 0.050v offset and test cinebench15 gpu benchmark again if there is any difference, 0.050v is minimal dont worry, its safe up to 0.190v.

in digi+ cpu current capability 120-130% and dram current capability 110% can help too.



Edit: Are you forcing anything special in FFXIV nvcpl profile?

 

Alright thank you, do you know if Hitman Absolution Benchmark is good to see the difference in clock speeds? I have the game on Steam but barely touched it since i played it before as part of PS3 Plus.

I could download it and check the benchmark, i think i ran it once or twice but might be confusing with another game.

The only thing i'm forcing for FFXIV is Max performance (opposed to Adaptive, i specify it for each game so it doesn't run full speed when not needed) and Single Display performance mode for all games.

I did an AIDA64 Test at 4.5Ghz 1.29v and the system BSOD at 20minutes, now i changed to 1.288v and ran to 36m, it didn't ran more cause i was bored of looking at it lol

One thing i've noticed is that RealBench v2 from ASUS immediatly crashes the system for every OC i put a Vcore on, for 4.2ghz i don't need to set a vcore, but for 4.5 i do and it crashes when gets to full load no matter the vcore i set (was between 1.2 and 1.3).

Then after the AIDA64 36m i ran IntelBurnTest x10 and didn't crash either, need to play games for a bit and see.