CIV: BE benchmarked

[CakeMonster]

Could the surprisingly 970 and 980 difference mean that we are at the point where we need the additional scaling? If they continue with this chip and just unlocking and scaling it up we should see very nice gains. I guess this proves that 970 is on the lower side, it certainly pushed me more in the direction of the 980 now...

 

[Rickyyy369]

I'd like to see a core-utilization chart, Civ5 was able to take advantage of more than 2 cores, though maybe that's only when processing turns. Looks like as far as framerate is concerned, at least in DX, it's very much limited by single-threaded throughput.

Of course, seems like any modern CPU is capable of delivering 60FPS minimums.

 

[SimianR]

It would be nice to see a review using an i5 4670K or any other middle of the road CPU that compares performance between Mantle and Nvidia cards in DX11. I'm surprised that none of these sites have done that yet..

 

[sontin]

Here: http://www.computerbase.de/2014-10/...diagramm-grafikkarten-benchmarks-in-2560-1600

Difference is around 7-10% between nVidia DX and Mantle.
50% with DX on both.

 

[Carfax83]

Here: http://www.computerbase.de/2014-10/...diagramm-grafikkarten-benchmarks-in-2560-1600

Difference is around 7-10% between nVidia DX and Mantle.
50% with DX on both.

The computerbase.de benchmarks are showing an 18% difference between the 980 and 970 which is a much different picture than Anandtech's results. But 18% or thereabouts is what I would expect, so I'm pretty certain that Anandtech's benchmark results are either an anomaly, or because of a clock speed difference between the Zotoc 970 and the reference 980.

 

[SimianR]

Here: http://www.computerbase.de/2014-10/...diagramm-grafikkarten-benchmarks-in-2560-1600

Difference is around 7-10% between nVidia DX and Mantle.
50% with DX on both.

Thanks.

The 1440P and 4K results are pretty interesting. The 290 is outperforming the 780 Ti and 970, and the 290X is faster than all of them. It's interesting because it's actually making a fairly large impact in GPU limited situations.

 

[Erenhardt]

Anyone knows if they use GPU for AI turn acceleration? I hoped that at least with mantle they will have deeper access to GPU and will be able to make use of it to speed the turn times.

Also, there are reports of bug in benchmark utility which doesn't give consistent results for mantle minimal fps.

 

[Alatar]

CPU overhead testing at computerbase, comparing AMD mantle, AMD DX11 and NV DX11 across multiple CPUs:

I think I remember reading that AMD's DX11 CPU overhead being worse was just Fud NV marketing...

 

[Spjut]

CPU overhead testing at computerbase, comparing AMD mantle, AMD DX11 and NV DX11 across multiple CPUs:


I think I remember reading that AMD's DX11 CPU overhead being worse was just Fud NV marketing...

The GTX 980 is fairly faster than the R9 290x though. Moving up from 3.5 ghz to 4.4 ghz for AMD barely made a difference, indicating a GPU bottleneck.
Mantle has the advantage at 2,5 ghz and four cores.

And really, we shouldn't expect AMD to optimize the DX11 path for games using Mantle

 

[el etro]

CPU overhead testing at computerbase, comparing AMD mantle, AMD DX11 and NV DX11 across multiple CPUs:

I think I remember reading that AMD's DX11 CPU overhead being worse was just Fud NV marketing...

where is this test?

 

[Gloomy]

If it's a CPU test, why lower the resolution but add AA? 😵

 

[KaRLiToS]

Nothing never positive about AMD Olli?

 

[frozentundra123456]

gamegpu.ru has their benchmarks up.

http://gamegpu.ru/rts-/-strategii/sid-meier-s-civilization-beyond-earth-test-gpu.html

I'd like to see what kind of improvements Mantle has when paired with an AMD cpu.

Too bad game.gpu did not test mantle with all those cpus. As an aside, I am kind of amazed at how well the 5960x does with its default low clocks. It does well in this game, and in Ryse, it is far ahead without any overclocking.

 

[Blue_Max]

For this game :- time between turns >> frames per second on screen

Yep. Even with my modern i5 machine, the time between turns is so frustrating I end up ragequitting and going back to my Civ4 collection.

SMOOTH BEATS PRETTY

 

[frozentundra123456]

An i5 4670K is going to be in a lot more gaming systems than a I7 5960X, those results are also pretty impressive. Hey guys Mantle reduces CPU dependency so lets test it on a $1000 CPU...

The test on Anand Tech was with a 4770k (certainly not a 1000 cpu) and only showed about 10 pct max, while the other one with 4670k showed 25% or so. I pretty much disregard the 768p tests. I really hate to see those coming back just to show artificially high differences (game.gpu is starting to bring them back for some games too).

I dont think anyone with even a mid range system is going to be playing at 768p, especially in a game like this where you dont need a super fast framerate.

We really need more data to see how mantle shakes out in this game. Like every other game it has been in, depending on how you manipulate the conditions, improvement can be from minimal to 40%. And if AMD is not optimizing their Dx drivers for mantle games, a true comparison becomes very difficult because you then have to compare Dx on nvidia and mantle on AMD.

 

[DiogoDX]

Nice article about Firaxis pretentions with SFR: http://techreport.com/blog/27258/civ-beyond-earth-with-mantle-aims-to-end-multi-gpu-microstuttering

Techreport about possible better latency:

All I can say is: thank goodness. Let's hope we see more of this kind of thing from AMD and major game studios in the coming months and years. Multi-GPU solutions don't have to double their FPS averages in order to achieve smoother animations or improved responsiveness. I'd much rather see a multi-GPU team producing more modest increases that the user can actually feel and experience

Meanwhile Anand:

Firaxis claims that they're working on reducing input latency rather than just improving raw frame rates, but testing for latency reductions can be difficult and in a game like Civilization it's questionable whether it will really matter much.

As for Mantle, it remains an interesting option if you have an AMD card, but so far we haven't really seen a huge benefit.

 

[ShintaiDK]

Nice article about Firaxis pretentions with SFR: http://techreport.com/blog/27258/civ-beyond-earth-with-mantle-aims-to-end-multi-gpu-microstuttering

I couldnt help noticing the 8GB 290X in the presentation. Not to mention they use an unreleased catalyst driver and an outdated nVidia driver. (I assume they mean 344.11 and not 340.11.)

And maybe its time for AMD to replace their 4960X system with the 5960X 😀

 

[3DVagabond]

And really, we shouldn't expect AMD to optimize the DX11 path for games using Mantle

So far, this has just been speculation and rationalizing. There hasn't been any sign that AMD isn't optimizing DX. Too many of their customers still use it.

 

[el etro]

If it's a CPU test, why lower the resolution but add AA? 😵

To turn the game CPU-bound. Lower the details to minimum can help too.

 

[Gloomy]

To turn the game CPU-bound. Lower the details to minimum can help too.

It would presumably be more CPU bound if it didn't have 4xMSAA...

 

[Tattare]

I wonder when the new AMD 14.9 Beta 2 Catalyst drivers will be uploaded.
4 hours till the game unlocks.

 

[Carfax83]

As an aside, I am kind of amazed at how well the 5960x does with its default low clocks. It does well in this game, and in Ryse, it is far ahead without any overclocking.

It's likely due to the cache. The 5960x has a 20mb L3 cache which is enormous compared to the 4770k's 8mb.

If it was because of multithreading support, the 3960x would be faster than the 4670k. Or the game could be using AVX2, which might explain Haswell's strong showing here.

 

[netxzero64]

So we all need to upgrade then?

 

[monstercameron]

http://www.pcper.com/reviews/Graphi...mance-Maxwell-vs-Hawaii-DX11-vs-Mantle/2560x1

pcper has some interesting results, mantle vs nvidia dx is very competitive with mantle having better frametime variance and better minimums.

 

[3DVagabond]

http://www.pcper.com/reviews/Graphi...mance-Maxwell-vs-Hawaii-DX11-vs-Mantle/2560x1

pcper has some interesting results, mantle vs nvidia dx is very competitive with mantle having better frametime variance and better minimums.

Minimums from AnandTech

using 1080 because it's likely more CPU bound. Mantle is massive in minimums.