Geforce GTX 780, 770 coming in May

Gregster · 18 May 2013 at 19:10

Speed said:
Sorry but how does that prove that VRAM was at fault? Crysis 3 is one of the more demanding games out there. Any single GPU is going to struggle at that resolution and eye candy. It could easily be lack of GPU grunt. I don't see that as conclusive proof.

If you found a way to shut off VRam on a high VRam card and then test it at each amount 1GB, 2GB, 3GB etc. Then compared the results that would be far more conclusive.

Well I came from 2x2GB 680's and when testing FarCry3 with everything turned up to the full, I was getting 1fps or even less (sorry if not more specific) and basically just freezing. When I knocked down AA to X6, I was getting around 17fps.

Kaapstad · 18 May 2013 at 19:15

spoffle said:
Anyone with sense? Anyone with sense would know that it's not simply about how many pixels are being "pushed".

If that logic was correct, going from 1920x1080 to 2560x1440 would always result in a flat halving of FPS from 1920 > 2560, which isn't the case.

Seriously, this arguing about things you don't understand thing is going a bit far.

It's not pure pixel count, it's about what's being displayed and rendered on screen. These are the basics, come on.

I regularly bench @1080p and 1600p so I have a pretty good idea how the numbers stack up.

Providing you can take the CPU/system bottlenecking out of the equation you get pretty good scaling when changing resolutions. This becomes more noticeable as the resolution increases and the CPU/system influences decrease.

Kaapstad · 18 May 2013 at 19:19

spoffle said:
But anyone with sense knows that the performance drop off is directly proportional to the amount of pixels being pushed!

You have to take the CPU/system bottleneck out of the equation by going to higher resolutions then it becomes very proportional.

Gregster · 18 May 2013 at 19:28

spoffle said:
Realistically, 5760x1080 has similar performance demands to 2560x1440.

Gregster · 18 May 2013 at 19:32

Tonester0011 said:
To be fair though Greg, the performance drop off from going 1080p to 1440p is around 35%, so even though it's 60% more pixels it's not 60% more demanding.

I never said it was?

spoffle · 18 May 2013 at 19:37

I should clarify that I didn't have Titan in mind, when I have looked at benchmarks of 79XX cards in single and dual GPU configurations, the performance fall off between 2560x1440 and 5760x1080 is fairly similar.

Reaper 392 · 18 May 2013 at 19:38

I could be talking out of my bottom here, but am i right in thinking that if you compared the framerate of a game on a single screen PC compared to the same game on a triple screen PC which had the same number of pixels to render that the triple screen PC would still be slower in the majority of cases because the FOV would include more geometry?

obviously this would be a low res triple screen rig vs a high res single screen rig, but would this be the case, or are there other factors that i'm not thinking of?

Tonester0011 · 18 May 2013 at 19:39

Gregster said:
I never said it was?

I know, I'm just saying pixel count doesn't knock performance in a 1:1 case. Doubling pixels wont necessary mean you need twice the amount of GPU grunt to push the same frames. It's not always quite as black and white as it seems, other factors to account for.

Not having a dig, just saying

Wucked · 18 May 2013 at 19:40

Gregster said:
In fairness, I have seen 3.8GB used in Crysis 3 at 5760x1080, so with 4K heading our way or ultra high res gamers, 3/4 is looking a little low.

4k isn't going to be mainstream for at least a couple of years. Early adopters can have the fun dropping back down to 30hz if they like. I'm going to stay at 120hz on my 1080p for a while. And I'll meet you in any online shooter you like with your shiny new 4k screen, just don't expect to be able to kill me

Anyway, point is that by the time 4k becomes affordable enough and useful for anything other that watching movies on we'll be 2 or 3 gen's into the future. If PC gaming is still even alive at that point.

Kaapstad · 18 May 2013 at 19:40

Gregster said:
These show you are wrong when comparing the two resolutions and to tell me I am incorrect is also wrong. I would consider your statement of similar performance demands to be close but these are massively different and shows you don't really know what you are talking about.

47.8fps at 1600P (even higher resolution than you claimed has similar performance demands) against 70.8fps in Alan Wake. I don't consider a 48% swing to be backing up your "similar performance demands"

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/6.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/7.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/8.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/9.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/10.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/11.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/12.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/13.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/14.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/15.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/16.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/17.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/18.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/19.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/20.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/21.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/22.html

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan/23.html

If you compare the results between 2560 x 1600 and 5760 x 1080 there is a very close link in the results, @5760 x 1080 they score almost exactly 1/3 less than they do @1600p. This is because 1600p has almost exactly 1/3 less pixels.

QED

Gregster · 18 May 2013 at 19:58

spoffle said:
I should clarify that I didn't have Titan in mind, when I have looked at benchmarks of 79XX cards in single and dual GPU configurations, the performance fall off between 2560x1440 and 5760x1080 is fairly similar.

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_Titan_SLI/3.html

Ok, same bench (Alan Wake). The 7970 is sitting on 56.9fps at 1600P and 38.1fps at 5760x1080. Not what I consider:

Realistically, 5760x1080 has similar performance demands to 2560x1440.

When we take the 7970 CF bench, we see the same performance demands 99.3fps for 1600P and 68fps for 5760x1080. Again, not very close and shows how wrong you are. I am happy to correct you and we all learn something new every day

Tonester0011 · 18 May 2013 at 20:27

Took it upon myself to count the result Greg provided, i'm surprised with the 1600 > 5760 demands. I'll happily eat my hat. 51% more pixels and an average of 47% more demand, 1080 > 1600 though, not surprised it's basically how i said (35% more for 1440) 97% more pixels but only 44% more load.

The 7970 results were skewed because of their testing anomalies, so only shows 35% more demand for surround.

Kaapstad · 18 May 2013 at 20:32

Tonester0011 said:
Took it upon myself to count the result Greg provided, i'm surprised with the 1600 > 5760 demands. I'll happily eat my hat. 51% more pixels and an average of 47% more demand, 1080 > 1600 though, not surprised it's basically how i said (35% more for 1440) 97% more pixels but only 44% more load.

The 7970 results were skewed because of their testing anomalies, so only shows 35% more demand for surround.

The reason the results for the lower resolutions are skewed is the CPU comes into it more with bigger bottlenecks the lower the resolution gets.

Boomstick777 · 18 May 2013 at 20:49

Gregster · 18 May 2013 at 20:52

Nice mouse mat

pgi947 · 18 May 2013 at 20:52

900mhz boost, that's a bit pants isn't it? I know boost clocks are subjective but I was at least expecting a guaranteed min of 1ghz.

Edit: not for the first time I have misread something, thought it said 770. Doh.

Gregster · 18 May 2013 at 20:57

pgi947 said:
900mhz boost, that's a bit pants isn't it? I know boost clocks are subjective but I was at least expecting a guaranteed min of 1ghz.

Edit: not for the first time I have misread something, thought it said 770. Doh.

If it uses 2.0 boost, it will never hit that, it will be well into the 1Ghz and I expect touching 1100Mhz.

pgi947 · 18 May 2013 at 21:35

Not bad at all then, a little guesswork with the figures we have vs what 670/680 differences were a 780 @ say 1050mhz must be easily around stock Titan perf?

Is boost 2.0 a new Titan/700 series thing that offers better clocks?

OCPete · 18 May 2013 at 21:38

Those specs look good to me, and it all seems pretty clear now.

Its all about price.

If they can get a reference model under £500 at launch that's pricey, but I can swallow it.... just. :rolleyes:

I reckon the range will run from £499.95 upwards.

Buckster · 18 May 2013 at 21:41

I'd go to £450 but that for me would be max

can afford more- just can't justify it though