That's part of my point. I would never buy a 780 for such a resolution. I'm just confused, I guess, it's his money and his choice - so whatever.
I still like 1680x1050 and would not upgrade if I had one. I sure as hell would not buy a $650 card for that res though. and I sure hope he has enough sense to have one heck of a cpu too.
IMO it makes more sense to buy a way cheaper card that gives you the same playable performance and then upgrade in 12-18 months. that will be cheaper in the end as the $650 780 will look like a joke once we have new 20nm cards.Well, if he plans to stick to that resolution the 780 can probably last him 3 years.
It is not swapped. It is just transferred. For example if you load next level every game asset of this level (there are tricks to read later parts on the fly) needs to be wrote into system memory. Then the part of that data (part which is for rendering purposes) is copied to video ram on graphics card using DMA which put CPU out of the loop. All that data is waiting in VRAM to be processed by GPU. For example compressed textures are stored in video ram and are decompressed by GPU for every frame they are needed. Part of that VRAM is frame buffer, where every pixel of every pre-rendered frame is stored. There may be some other thing like PhysX data, etc.
Yes, but that is buffering data before it is needed. This trick is to decrease loading times by not having to read whole world for hdd. Later parts are first buffered to RAM and then to VRAM. For example if you have a boss fight at the end of the level, there is no need to read all the boss stuff from HDD when you are entering the level. This may be loaded during your play in mid-level. But as soon as the boss needs to be rendered, all the data needs to be already in video RAM.
Graphs in review you linked show video memory usage. You need to remember that Metro LL is 360/PS3 port. Both consoles have 256MB system memory and 256 video memory. The graphs linked by me are system memory usage.
New game engines will be more efficient but new hardware possibilities will allow usage of more game assets in higher quality and less recycling which in the end will increase vram usage.
IMO it makes more sense to buy a way cheaper card that gives you the same playable performance and then upgrade in 12-18 months. that will be cheaper in the end as the $650 780 will look like a joke once we have new 20nm cards.
as long as he has a comp capable of getting the most use out of it then I guess so.I certainly would choose a less powerful card, but if he likes to max everything and also hold onto the card for years then the purchase is not entirely frivolous. Of course for a situation like 1680x1050 then an OCed 770 is even more bang for the buck than 1080+ gamers vs the 780.
I still like 1680x1050 and would not upgrade if I had one. I sure as hell would not buy a $650 card for that res though. and I sure hope he has enough sense to have one heck of a cpu too.
you do know that for MOST people that price is a factor so of course its going to be discussed.I come to this thread from time to time. What I always see is people complaining about pricing. From the top tier Titan to the now mid range 770. There were complaints about the $1000 Titan, the $650 780, and now the $400 770.
Can we give it a break? Buy something else if you don't like it. It's a personal choice.
you do know that for MOST people that price is a factor so of course its going to be discussed.
pointless back and forth is what keeps the forums alive.The problem then begins when people have their own price per performance mapped out in their minds. That's when you get the pointless back and forth.
Like I said, personal choice. There is no hard fast rule to it.
pointless back and forth is what keeps the forums alive.
I come to this thread from time to time. What I always see is people complaining about pricing. From the top tier Titan to the now mid range 770. There were complaints about the $1000 Titan, the $650 780, and now the $400 770.
Can we give it a break? Buy something else if you don't like it. It's a personal choice.
Different strokes i guess,its the principle that a card just a hairline slower is retailing for $350 less that would bug me after just spending $1,000 on another gpu.
First, let us take a look at the NVIDIA GeForce GTX 780, which was just recently released on May 23rd. We found it playable at 2560x1600 with the quality settings and tessellation set to "very high", the highest available settings, with an average framerate of 51.1 FPS. The games Analytical AA is the only form of aliasing that we could use, however it does a great job of smoothing out most rough edges in the game. We could not enable PhysX at this resolution the GTX 780 because it was too demanding, bringing the average framerate down to 41 FPS. Ideally, we want to average 45 FPS and sometimes even higher to play without lag hindering the gameplay.
What a nonargument, it's 10% slower. Overclock the GTX 780 a bit and it outperforms the Titan.Untill you go DP.
Or..as the case is with Metro 2033:
Disable features on the GTX780, because it cannot keep up with the Titan:
http://hardocp.com/article/2013/05/...deo_card_performance_iq_review/3#.Uank2diRCNI
Stop talking like GX780 and Titan are the same....they are not.
CUDA cores, texture units, RAM amount, DP....it's not the same...but yet you whine like it is?
What a nonargument, it's 10% slower. Overclock the GTX 780 a bit and it outperforms the Titan.
So 2688 stream processors vs 2304 stream processors is a "nonargument".What a nonargument, it's 10% slower. Overclock the GTX 780 a bit and it outperforms the Titan.
So 2688 stream processors vs 2304 stream processors is a "nonargument".
Or 224 texture units vs 194 texture units is a "nonargument"
Even 1/3 DP vs 1/24 DP is a "nonargument"
The 6 GB RAM vs the 3GB is also a "nonargument"
Hell, even the fact of Advanced PhysX vs Normal PhysX is a "nonargument"
Every BENEFIT the Titan has over the GTX780 is apparently a "nonargument"...go figure :awe:
Because in the end, that extra 10% of specs barely even gets you 10% more performance, which is the difference between 30FPS and 33FPS - AKA not much. And the opinion seems to be the 780 has more overclocking potential than the Titan.
Ps: 1/3DP compute is a money grab by Nvidia, they just disabled the driver setting on the 780. All of the compute hardware is there, they just don't want you to flick the switch.
Very few of those features have the slightest impact on gaming, yet you're using gaming as your primary benchmark comparison. Do you not know what you're talking about or are you purposefully being dishonest to try to justify the Titan. Which is it?So 2688 stream processors vs 2304 stream processors is a "nonargument".
Or 224 texture units vs 194 texture units is a "nonargument"
Even 1/3 DP vs 1/24 DP is a "nonargument"
The 6 GB RAM vs the 3GB is also a "nonargument"
Hell, even the fact of Advanced PhysX vs Normal PhysX is a "nonargument"
Every BENEFIT the Titan has over the GTX780 is apparently a "nonargument"...go figure :awe:
This has nothing to do about AMD, go sulk somewhere else and stop derailing the thread.Its not a non-argument in your mind when arguing about GPGPU capabilities of GCN now is it. I wonder what it would be like to have a Forum where MrK6 could keep it real. Might be fun.
It also starts with less performance.
You OC the 780 so you can enable Advanced PhysX too...then I'll start OC'ing and wave goodbye.
The end result is the same:
1/24 DP.
What was your argument?
You OC the Titan and you hit a ceiling, while someone with a 780 is more likely to be able to hit higher clocks and achieve more.
I forsee a hack to achieve 1/3DP in short time on a 780. But besides that, I doubt more than 1/10 Titan owners ever actually uses the DP ability...