Civilization V

The other new game in our benchmark suite is Civilization 5, the latest incarnation in Firaxis Games’ series of turn-based strategy games. Civ 5 gives us an interesting look at things that not even RTSes can match, with a much weaker focus on shading in the game world, and a much greater focus on creating the geometry needed to bring such a world to life. In doing so it uses a slew of DirectX 11 technologies, including tessellation for said geometry and compute shaders for on-the-fly texture decompression.

We have always considered Civ 5 an interesting game both for its near-complete use of the DX11 feature set, and because of its interesting performance characteristics. 2 weeks ago we called it CPU limited based on the fact that once we had sufficiently powerful cards, AMD and NVIDIA results tended to clump together despite any difference in their respective cards’ speed. With the Forceware Release 265 drivers, NVIDIA has blown this assumption apart, with NVIDIA’s more powerful cards launching ahead at 1920 and lower. We appear to be quite GPU limited on the NVIDIA side all of a sudden, which is about as drastic change as we could expect. Furthermore NVIDIA is holding their cards close to their chest on this – they’ve obviously found a wonder optimization, but they aren’t ready to say what it is.

In any case while AMD has always trailed NVIDIA in single card performance in Civ 5, with these driver changes it’s quite monumental. The GTX 560 Ti is 44% faster than the 6950 at 1920, 80% faster at 1680, and even the GTX 460 768MB can edge out the 6970 at 1920. Whatever NVIDIA has done, it has made Civilization V quite a lot faster and smoother at 1920 and 1680, particularly when a large number of units are on screen.

Among NVIDIA’s own cards the lead has actually shrunk some compared to our earlier games. The GTX 470 has an edge on the GTX 560, and the 560 in turn is down to a 25%-30% lead over the GTX 460 1GB. We don’t seem to be ROP or memory bandwidth limited, so perhaps this is a case of the GF104/GF114 architecture’s shaders underperforming?

HAWX Battlefield: Bad Company 2
Comments Locked

87 Comments

View All Comments

  • SolMiester - Wednesday, January 26, 2011 - link

    Hi can we please have some benchies on this feature please...While surround is not an option for the single nVidia card, 3D is, and it is hard to judge performance with this enabled. Readers need to know if 3D is an viable option with this card at perhaps 16x10 or 19x10

    Ta
  • DarknRahl - Wednesday, January 26, 2011 - link

    I agree with some of the other comments and find the conclusion rather odd. I didn't really get why all the comparison's were done with the 460 either, yes this is the replacement card for the 460 but isn't really to relevant as far a purchasing decision goes at this moment in time. I found HardOCP's review far more informative, especially as they get down to brass tacs; the price to performance. In both instances the 560 doesn't make too much sense.

    I still enjoy reading your reviews of other products, particularly power supplies and CPUs.
  • kallogan - Thursday, January 27, 2011 - link

    Put those damn power connectors at the top of the graphic card, think about mini-itx users !!! Though this one is 9 inches, should fit in my sugo sg05 anyway.
  • neo_moco - Thursday, January 27, 2011 - link

    I really don`t understand the conclusion :

    The radeon 6950 wins hands down in 1920x and 2560x in almost all important games: crysis , metro , bad company2, mass effect 2 and wolfenstein

    The geforce wins only the games who not a lot of people play : civ 5 , battleforge, hawx , dirt 2

    Add to that others tests : 6950 wins in the popular call of duty , vantage .
    In 3dmark 11 the geforce is 15 % weaker(guru3d) so the conclusion as i see it : radeon 6950 1 gb is aprox. 5-10 % better and not the other way .
  • neo_moco - Thursday, January 27, 2011 - link

    after 15 months of the radeon 5850 launch we get a card 10 % better for the same price; i dont get the enthusiasm of some people over this cards ; its kind of lame
  • HangFire - Thursday, January 27, 2011 - link

    "By using a center-mounted fan, NVIDIA went with a design that recirculates some air in the case of the computer instead of a blower that fully exhausts all hot air."

    I don't think I've ever seen a middle or high end NVIDIA card that fully exhausted all hot air. Maybe it exists, I certainly haven't owned them all, perhaps in the customized AIB vendor pantheon there have been a few.

    This is not just a nitpick. When swapping out an 8800GT to an 8800 Ultra a few years back, I thought I was taking a step forwards in case temperatures, because the single-slot GT was internally vented and the Ultra had that second slot vent. I didn't notice the little gap, or the little slots in the reference cooler shroud.

    That swap began a comical few months of debugging and massive case ventilation upgrades. Not just the Ultra got hot, everything in that case got hot. Adding a second 120mm input fan and another exhaust, an across-the-PCI-slots 92mm fan finally got things under control (no side panel vent). Dropping it into a real gamer case later was even better. (No, I didn't pay $800 for the Ultra, either).

    I'm not a fan of, um, graphics card fans that blow in two directions at once, I call them hard drive cookers, and they can introduce some real challenges in managing case flow. But I no longer run under the assumption that a rear-venting double slot cooler is necessarily better.

    I'd like to see some case-cooker and noise testing with the new GTX 560 Ti reference, some AIB variants, and similar wattage rear venting cards. In particular, I'd like to see what temps a hard drive forward of the card endures, above, along, and below.
  • WRH - Saturday, January 29, 2011 - link

    I know that comparing an on CPU chip graphics with GPU cards like the ones on this list would be like comparing a Prius to a Ferrari but I would like to see the test results just the same. Sandy Bridge on CPU GPUs come in two models (2000 and 3000). It would be interesting to see just how far below the Radeon HD4870 they are and see how the 2000 and 3000 compare.

Log in

Don't have an account? Sign up now