9

Which of these two graphic cards would be more suitable for gaming, what are the main differences between the two, which would you recommend from the two:

1. The older model, Gainward GeForce GTX 760 Phantom 4GB DDR5 256-bit

or

2. The newer model Gainward GeForce GTX 960 Phantom 4GB DDR5 128-bit

The notable differences are the BUS memory on the two cards, and the chipset version. Basically, my question would be, that how important are 128 bits when taking into account graphic cards. Is it worth sacrificing 128 bit of bus memory, for a newer chipset? The price range is the same from where I would buy it, but the older version(1) is slowly taken out of production, I can only find two webshops which still has one on stock.

I am planning on buying a new PC for gaming, and I am interested what would be the difference between the two, which one should I choose. Or, if you could recommend another card in the same price range as this one. I am from Romania, and these cards cost 250-280$ in my country, this is the maximum I would spend on a graphic card.

  • I think you should take a look at the R9 380 as well. I've recommended it in this other question (also from Romania), and it looks pretty good. – timuzhti Dec 21 '15 at 6:24
6

To put it simply, the newer chipset you'll get with the GTX 960 won't really matter at all in comparison to the severe bottlenecking you'll experience with its memory bus size.

Explanation
On the video card itself, the memory bus connects the GPU to the VRAM. Think of it as a pipeline carrying water between two points. If you have a lot of VRAM (the water) and a small bus size to go with it (a very narrow pipe), all that water won't be able to cram into the small pipeline to get to the GPU — you need a bigger bus to account for more memory.

So, memory bus size should scale with the amount of VRAM to some extent. The exceptions to this concept come when you introduce faster clock speeds (which, in terms of this example, push the water through the pipeline faster) and overclocking (in which you manually speed up the water flow).

The cards
Now to look at the two video cards you're choosing from. The 760, while having an older chipset, has a bus size twice as large as the other card to account for the 4GB of VRAM. This bus size is still somewhat lacking, but it will do a much better job than the 128-bit bus. To get ideal processing power out of the 760, you'll need to overclock it just a bit, but to get ideal processing power out of the 960, you'll need to overclock it a ton (which ends up being more dangerous than ideal).


As for my recommendation if I have to choose between these two cards, I would definitely go with the GTX 760. But if you have enough money to upgrade even more, I'd go for anything with a large memory bus size and at least 2GB of VRAM.

  • Thanks for the prompt reply, this is what I tought also. Right now I am using an old pc, which has a GeForce 8800GTS, with 300+ bits, and I remembered that when I bought this system 6 years ago, everyone said that bits are one of the most important parts to consider when building a system for gaming. I tought that with the new technologies, they do not matter so much anymore. I am now looking into the series 9, 256 bit GeForce's, but they are rather expensive(+100-120$). What about the Radeon cards, they are considerably more cheap then the GeForces, are they so much worse than a GeForce? – Adam Baranyai Oct 15 '15 at 22:34
  • 1
    Both manufacturers put their own useful features in their cards. Nvidia cards have CUDA Cores which are extremely good with graphical programs (not necessarily restricted to games) such as Photoshop, After Effects, or 3ds Max. AMD cards are extremely good with OpenCL and OpenGL processing which is good for more general computing. At the mid range, which is about where you are, AMD gives the most bang for your buck. – Adam Oct 15 '15 at 22:46
  • Sorry for the overwhelming amount of questions, but just to be clear: after re-reading your original answer to my question, I've noticed that you say that memory bus size should scale with VRAM. Is this means, that regarding to my gaming needs, both cards would suit me well, only with the smaller memory bus size, I don't need a card with such a large VRAM? So basically, a card, with 128 bits of memory bus size, can be good for gaming, only with, let's say, only 2GB of VRAM? Or if I plan to use my system for gaming, I would need at least 256 bits, to be on the safe side? – Adam Baranyai Oct 15 '15 at 22:52
  • 2
    What I mean by that is the bus size and amount VRAM should "match" to a certain degree. So yes, a 128-bit card would be good for 2GB of VRAM or less, but nothing more. 256-bit is good for 4GB at the very most, 384-bit is good for up to about 6-8GB, and so on. Strive for at least 256 bits to be safe. – Adam Oct 15 '15 at 23:03
  • 1
    Do you have proof that a 4GB 960 is actually a worse performer than a 2GB 960? Or that a 960 never can use the extra 2GB when the card is provided with those? I agree there is a direct correlation between bit width and memory, but not in the way you describe. For example, the GTX 660 had a 192-bit bus and 2GB of video memory, but it had trouble accessing the last 512MB, as the latter 1GB was put on a single controller, and the former 1GB had two controllers. In the case of 4GB 960's simply more chips will be added for each controller. I'll say again: bandwidth(clk) matters more than bit-width. – cst1992 Dec 21 '15 at 8:12
2

I'd say you go with the 960.

The reason is that the 960 is a newer architecture, and is still going to be available if you decide you're going to add another 960 in SLI.

Bus memory is a consideration, but the way you're seeing it is not correct; there's more to it than that. For example, if you have a bus width that is more, but the memory tech is older (DDR3 instead of GDDR5), it automatically cuts the total memory bandwidth you gain in half. Depending on the memory clock, you may actually get worse performance with the DDR3 card, even though it has more bus width.

Now, I agree the bus width of the 960 is lower, and both cards are GDDR5. But the newer Maxwell architecture of the 960 utilizes a color-compression technology that increases its effective bandwidth by around 28%, bringing it closer to the 760's. It also uses a newer anti-aliasing algorithm(MFAA) that is far less resource-intensive than MSAA that you'll have to use on the 760, again requiring lesser memory bandwidth.

That said, I ask you this: what is the resolution you're using on your monitor? If it's 1920x1080(i.e. 1080p), the 960 will handle it irrespective of all these theoretical specifications. It's been designed to run modern games with that resolution in mind, and it's also more power efficient(176 W of power for the 760 vs 120 W for the 960), so the 760 runs 82C at full load and is loud, whereas the 960 never breaches 60 and is very quiet(both are very good things).

  • FXAA is post-processing. You'd really want to compare MFAA with MSAA to get similar effects. Besides which, FXAA isn't new. I have it on a old GT 220. – timuzhti Dec 21 '15 at 6:20
  • Oh, I'm sorry, you're right. It's MFAA. I've edited it. – cst1992 Dec 21 '15 at 6:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.