I am not impressed with my Nvidia GTX 970

[kami]

This might help you tweak your Fallout 4 performance: https://www.geforce.com/whats-new/guides/fallout-4-graphics-performance-and-tweaking-guide

There are several settings that can be turned down for little to no loss in visual fidelity.

 

[SPBHM]

xcom 2 is very heavy, even people with 980 tis will get poor performance at max settings,

I think the 970 is very impressive, it's a 224bit memory bus card with 1664 cuda cores and it performs quite well on basically all games, while there are cards out there with much higher power draw and specs doing around the same (like the 390)

 

[Piroko]

I don't think a 224bit memory bus card for 310€ is impressive in 2016.

edit, to prevent flamewars, I don't consider any 28nm GPU in 2016 to be impressive, apart from the 980Ti which is so far consistently holding the performance crown - but its price isn't impressive either.

 

[Head1985]

OC gtx970 to 1500/8000 and you will have GTX980 performance level.Most reviews using GTX970 downclocked to 1100-1200Mhz.
On 1500/8000 it will be like 1-5% faster than GTX980.

 

[stahlhart]

Thread cleaned. Stay on topic.
-- stahlhart

 

[Headfoot]

You all are overestimating the performance impact of Godrays. Ultra is a lot, yes, and high is better, true.

The most performance impactful setting by a mile is Shadow Distance. The longer your shadow distance the more draw calls issued, exponentially. Ultra = 13000 distance, High = 7000, Medium = 3000 and I don't know what low is. I can comfortably run "High" e.g. 7000 on my rig below at 1080p (drop into the 50s very rarely, otherwise capped at 60). This will depend mostly on how fast your CPU and RAM is. My 290 is about on par with your 970 (2x290 still doesnt work in Fallout)

Change shadow distance to medium and godrays to high or lower and your performance problems will clean right up.

 

[Insomniator]

Turn your settings down, 1440p is no joke.

Setting everything to 'high' instead of 'ultra' in many games will completely smooth out performance while looking nearly identical outside of screenshots nerds post online to compare.

 

[tviceman]

I don't think a 224bit memory bus card for 310€ is impressive in 2016.

edit, to prevent flamewars, I don't consider any 28nm GPU in 2016 to be impressive, apart from the 980Ti which is so far consistently holding the performance crown - but its price isn't impressive either.

Citing the paper stat of a chips memory bus says very little. Case in point, Nvidia's current 128-bit GTX 960 soundly defeats the 384-bit GTX 580. Pascal and Polaris chips coming later this year will have the capability of bring equipped with a 192 bit, or perhaps even a 128 bit memory bus and running circles around current 256-bit cards.

Measure graphics cards by their real world performance and do not rely on checklist features and specs.

 

[SPBHM]

Citing the paper stat of a chips memory bus says very little. Case in point, Nvidia's current 128-bit GTX 960 soundly defeats the 384-bit GTX 580. Pascal and Polaris chips coming later this year will have the capability of bring equipped with a 192 bit, or perhaps even a 128 bit memory bus and running circles around current 256-bit cards.

Measure graphics cards by their real world performance and do not rely on checklist features and specs.

you are talking about a 40nm card with slow clocks and old architecture, when comparing newer products it should be more relevant, but in the end it's only relevant for let's say, 960 vs 970 vs 980 ti, since AMD performs so differently, but still very interesting to see the difference in raw specs vs performance.

 

[tential]

You all are overestimating the performance impact of Godrays. Ultra is a lot, yes, and high is better, true.

The most performance impactful setting by a mile is Shadow Distance. The longer your shadow distance the more draw calls issued, exponentially. Ultra = 13000 distance, High = 7000, Medium = 3000 and I don't know what low is. I can comfortably run "High" e.g. 7000 on my rig below at 1080p (drop into the 50s very rarely, otherwise capped at 60). This will depend mostly on how fast your CPU and RAM is. My 290 is about on par with your 970 (2x290 still doesnt work in Fallout)

Change shadow distance to medium and godrays to high or lower and your performance problems will clean right up.

The 290 is faster than the GTX 970 in Fallout 4, and does better as resolution increases.

Not trying to turn this into an OP should have gotten an R9 390/290, but the GTX 970 is not on par with the R9 290 in this game. Launch? Of course AMD is bad in a gameworks game. Later when no one is paying attention? Great.

Really the GTX 970 is a tier behind the R9 290 in this game (and a lot of new releases it feels like).

I think 1440p is manageable with the right settings on the R9 290.

On the GTX 970, I think you're better off going after 1080p.

I can't speak to GTX 970s with good OCs though.

Edit: The GTX 970 is literally a tier behind the R9 290 at the OPs chosen resolution of 1440p as it matches an R9 280x. So given the fact OP is playing Fallout 4 why should he be impressed with his GTX 970? It's literally slightly faster than the HD7950 I just had. It's just as fast as an HD7970, an EXTREMELY old level of performance at 1440p. It's far slower than the R9 290 at 1440p. All while being newer, and more expensive than all of those cards right now.

So yes, OP should feel unimpressed given the game he's playing and the resolution.

Edit2:

Latest game only makes it feel worse so I'd feel pretty annoyed and I really don't think it'll get better when Pascal arrives, especially with the GTX 970s memory configuration.

 

[Piroko]

Citing the paper stat of a chips memory bus says very little. Case in point, Nvidia's current 128-bit GTX 960 soundly defeats the 384-bit GTX 580. Pascal and Polaris chips coming later this year will have the capability of bring equipped with a 192 bit, or perhaps even a 128 bit memory bus and running circles around current 256-bit cards.

Measure graphics cards by their real world performance and do not rely on checklist features and specs.

Oh I'm fully aware of that. Sarcasm made me take the paper stat in place of its actual name, because its name shall not be spoken.

 

[amenx]

Something out of whack with either the 1.30 beta patch or GameGPU benches. Its not just the 970 doing miserably but the whole Nvidia lineup. The 980ti on par with Fury Nano..? MaximumPC has different results. In fact the Nvidia cards doing better with 1.30 patch then at launch.

Unless other sites can confirm 1.3 performance benches, I believe theres something amiss with gameGPU benches.

As of Sept. 15, this was the rough pecking order of cards @ 1440p:

 

[futurefields]

It's just disheartening to play a game and have it go into choppy limbo at times.

This is why you should run a performance monitor like EVGA Precision or MSI Afterburner.

THis is how they can actually give you peace of mind because a lot of the times when those games go into "choppy limbo" you would probably see your GPU is still not even struggling and that what you are really experiencing is software issues with the game itself, engine problems, etc, things that NO gpu will fix.

There are games my GTX 970 can brute force into 100's of FPS but they still never feel smooth with vsync on or off (Dead Island, Fallout New Vegas) unless you get some mod or whatever that actually fixes the game.

This is just a longstanding issue with pc gaming in general but guess what, its also starting to become pretty common over in console land. Games like Bloodborne get great reviews but when I played it at a friends house I immediatly noticed the game suffers from microstutter. It's just modern games doing so much at once, sometimes the code becomes a mess and your GPU even if you have a Intel 5ghz + GTX 980ti running at 720p the game still wouldn't feel smooth.

 

[tential]

As of Sept. 15, this was the rough pecking order of cards @ 1440p:

Or we can use results from this year. As of February 12th here are the results

 

[Arachnotronic]

Or we can use results from this year. As of February 12th here are the results

That factory overclocked 980 Ti is a monster.

 

[amenx]

Or we can use results from this year. As of February 12th here are the results

And that is funny... going into that TPU review... to the FO4 benchmarks, we find the 780ti beating both the 390x and 980.

😀 So unless we can find reviews that show some consistency... the FO4 benches at this point seem out of whack.

Aside from that, the Feb 2016 TPU review only tests 11 cards, no 290/290x results in any of the newer games tested. So how are they reaching their overall 2560x1440 summary when cards not tested in these games are lumped in the overall summary?? Again, another big question mark where we cannot extrapolate any conclusive results.

 

[Silverforce11]

Fallout 4 1.3 is vastly different to the early versions that most sites use. The renderer was overhauled & optimized. Basically poor AMD performance was fixed. Perhaps too good because some sites found AMD out performing their NV GPU counterpart.

 

[littleg]

That factory overclocked 980 Ti is a monster.

It needs to be. Apart from that AMD is winning every price point. Shows the value of halo products and taking the crown I suppose.

 

[amenx]

Fallout 4 1.3 is vastly different to the early versions that most sites use. The renderer was overhauled & optimized. Basically poor AMD performance was fixed. Perhaps too good because some sites found AMD out performing their NV GPU counterpart.

You apparently missed the inconsistencies in the results shown. Re-check the FO4 charts in the few posts above.. they are greatly at odds with gameGPU review.

 

[Silverforce11]

You apparently missed the inconsistencies in the results shown. Re-check the FO4 charts in the few posts above.. they are greatly at odds with gameGPU review.

I would be more cautious about results that show a 780Ti is faster than a 980. Especially in a GameWorks title. 😉

 

[MajinCry]

If ya ask me, average framerates are useless. It's the minimums that matter.

Would be interesting to see how a couple of the settings affect performance across all the GPU tiers; disabling NVidia's godrays implementation, and enabling async SSAO.

Also, where are these guys testing the framerates? Two good spots are at the top of Corvega and overlooking the nearby town (Arlington?), the other being Diamond City right as you enter the city and overlook it from the stairs.

 

[amenx]

I would be more cautious about results that show a 780Ti is faster than a 980. Especially in a GameWorks title. 😉

But you have no problem with benches in GW titles where AMD 280x = 970?

Again, something seriously messed up with gameGPUs benches. Look at MaximumPC benches comparing 1.3 patch results.

 

[amenx]

If ya ask me, average framerates are useless. It's the minimums that matter...

Minimum FPS on their own can be misleading. You need frame times that measure how much time spent in those minimums. If its a solitary 0.5 sec dip that does not repeat itself, it still may be called a 'minimum'. In which case the averages become more important.

 

[Silverforce11]

But you have no problem with benches in GW titles where AMD 280x = 970?

Again, something seriously messed up with gameGPUs benches. Look at MaximumPC benches comparing 1.3 patch results.

Because different vendor performance can vary wildly.

But same vendor, its actually never the case that the 780Ti is faster than 980, not in any game I've seen.

 

[Bubbleawsome]

Hey man I'll trade you. My 280x will do games on medium for 45fps at 1440p, but it's dang consistent.