Civilization V GPU & CPU Performance In-depth

red1776 said:
Two 5870's in CF would outperform a 5970, a 5970 is a pair of 5850's

*Facepalm* Actually its based on the 5800 series, so you cannot say it is based on one card. It might be clocked to 5850, but this can be changed...
 
First of all thanks for all the great feedback guys!

where's the 5750?

We didn't include it since the performance has been so similar to the 5770 in the last few games we have tested. We cut out quite a few cards for this review as they clutter the graphs and are not necessary. Really we saw no reason to include the 5750 when the 5770 is there. Read between the lines ;)

It would be handy to know what improvement I would get running XP.

None, at least how we tested using DX10/DX11 which is not supported by XP and since we saw no performance gains when running the game in DX9 you stand to gain nothing.

*Facepalm* Actually its based on the 5800 series, so you cannot say it is based on one card. It might be clocked to 5850, but this can be changed...

That is true. However the 5970 is based on dual 5870 GPU's that have been underclocked. Saying that the 5970 is a pair of 5850 cards is very wrong as the core configuration is different, it is the same as the 5870.
 
That is true. However the 5970 is based on dual 5870 GPU's that have been underclocked. Saying that the 5970 is a pair of 5850 cards is very wrong as the core configuration is different, it is the same as the 5870.

well okay, but I think that's splitting hairs. I could have been more accurate in saying that it performs as two 5850's, but you are taking the same chip and putting "restrictor plates" on them. the 5970 is clocked as a 5850, has the same ROP's,Z-stencil,mem data rate,bandwidth, and performs very close to 2x 5850's (actually gets beat more often than not.) A card such as the Ares is actually a pair of 5870's. am I wrong in thinking that a Cypress pro is a Cypress XT with a 160 SPU cluster and 8 Texture units fused? or does the XT have other architectural differences?
 
well okay, but I think that's splitting hairs. I could have been more accurate in saying that it performs as two 5850's, but you are taking the same chip and putting "restrictor plates" on them. the 5970 is clocked as a 5850, has the same ROP's,Z-stencil,mem data rate,bandwidth, and performs very close to 2x 5850's (actually gets beat more often than not.) A card such as the Ares is actually a pair of 5870's. am I wrong in thinking that a Cypress pro is a Cypress XT with a 160 SPU cluster and 8 Texture units fused? or does the XT have other architectural differences?

Sorry didn't want to start an argument here. That said there is nothing to argue over. The Radeon HD 5970 features a pair of Radeon HD 5870 (Cypress XT) GPU's which have been down clocked to the Radeon HD 5850 operating frequency. Clock them back up and you have two Radeon HD 5870's, if you have the cooling to do so.

The core configuration of the Radoen HD 5870 and 5970 GPUs is 1600:80:32 while the Radeon HD 5850 is 1440:72:32. Anyway enough about that it is a pointless and off topic argument. Back to Civ 5 :)
 
Sorry didn't want to start an argument here. That said there is nothing to argue over. The Radeon HD 5970 features a pair of Radeon HD 5870 (Cypress XT) GPU's which have been down clocked to the Radeon HD 5850 operating frequency. Clock them back up and you have two Radeon HD 5870's, if you have the cooling to do so.

The core configuration of the Radoen HD 5870 and 5970 GPUs is 1600:80:32 while the Radeon HD 5850 is 1440:72:32. Anyway enough about that it is a pointless and off topic argument. Back to Civ 5 :)

Hey, no argument here, I actually wanted to know if there was an architectural difference between the the two GPU's :)
 
No same architecture as they are the same series. They differ in the core configuration and clock speeds.
 
The graphics in this are amazing compared to the last Civ game I've played (civ III). I was never really a fan of that game though, and I bought it thinking it was an RTS game.
 
Just a curious question. What is so special about Civ 5 that it taxes the hardware so much?
 
Well, game quality steadily declined, as did my settings. Eventually with everything at low it did ok.

I'm getting flashbacks to exactly the same feelings I had when I bought Civ 4:

Wow...erm...%^%# its not working.

Now I can't even get it to run.

But then computer has been acting weird for the last week. Time for a clean install. I just have to backup my Steam games, hope I can find a hard drive big enough.
 
Burty117, the rest of Fermi lineup is shocking to you? For example, GTX470 has been available for $250-270 on many occasions, a card that's generally a hair within HD5870 on average and costs $80-100 less.

The inferior performance of ATI cards has nothing to do with the drivers. NV's architecture is superior in DX11 games (more advanced Tessellation engine, gather4 instructions, etc.):

Besides BF:BC2 and Mafia 2, there are practically no DX11 games where 5850 beats a GTX470 or HD5870 beats a GTX480:

Lost Planet 2 - http://www.legionhardware.com/articles_pages/lost_planet_2_gpu_performance_preview,6.html
Metro 2033 - http://www.hardwarecanucks.com/foru...eforce-gtx-470-super-overclock-review-12.html
Just Cause 2 - http://www.xbitlabs.com/articles/video/display/nvidia-geforce-gtx-460_13.html#sect1
AvP - http://www.xbitlabs.com/articles/video/display/nvidia-geforce-gtx-460_11.html#sect1

Then there are BattleForge, STALKER:CoP, Dirt 2, Civ5 and even Starcraft 2 (although not DX11).

While HD5000 series clearly dominates from the performance/watt perspective, ATI no longer has the better card in any price bracket. From a price/performance perspective, GTS450 > 5750, GTX460 768 > 5770, GTX460 1GB > 5850, GTX470 > 5870. The only card with no equals is the 5970.

With HD6000, ATI will come on top once again.

I think they're shocking because they take soo much power and get really hot, had nothing to do with pricing and performance since Pricing has come down that seems reasonable, but the performance (Bar the GTX460) seems to be not alot considering the power and heat generated from the others such as the GTX470 and 480.
 
The most telling benchmark in this article is the "CPU Scaling" benchmark, showing the almost perfectly linear scaling of fps with CPU clock speed on a top-tier processor (all else being equal). As stated in the article, "Clearly, Civilization V is very CPU dependent."

For some reason, this static, turn-based (i.e., not real-time) game is coded to continuously hammer the CPU throughout gameplay, even when nothing's happening. That, to me, suggests extreme inefficiency or sloppy programming. The game is complex, nuanced, and a lot of fun, however. So here's hoping they tie up some loose ends, fix some of the more glaring bugs, and optimize Civ V in future patches. It looks like there is a *lot* of room for optimization.
 
Interesting article. Finding no issues runing game on high settings past 300 turns on a I7 950 OC to 4GHz and running a Saphire 5870 vapour chill card. CPU utillisation on the G15 keyboard monitor never goes above 25% suggesting probably only using 2 cores?
 
so i played civ5 on friends computer recently, as fun as the game is it has about a 80% crash/failure rate on startup. you can also only save 1 multiplayer game at a time, how crappy is that??
 
there's a new bundle on tigerdirect that features a Phenom X4 9750 with 4GB of DDR2. if i go with the gts 450 and upgrade some of the parts, will i be ok or would it just completely melt?
 
2 GTX 460s in SLI, updated drivers, updated SLI profile and I still get massive corruption and issues when I enable SLI. Runs fairly decent on just one card, but that's not why I have SLI... it's not new technology so I don't understand why they can't get it working correctly.
 
I'm actually playing this quite well with my Athlon X4 630 and GTS 250 @ 1080p with 2xAA. The frame rate is in the mid 30s (so long as you don't zoom all the way out), but you don't need a high frame rate in a turn based strategy game.
 
OK, having played this game now for a number of hours on my i7-930, GTX 460 1GB SLI rig at max settings (and I mean max everything - resolution, AA, etc.), the only "bumps" I run into are late into the game when it's processing all the moves of the literally dozens (sometimes over a hundred) of individual units across several continents and oceans. Graphically, it continues to look great no matter how far I am into the game - framerate isn't an issue. But it takes longer and longer in-between turns for the game to set itself up for the next round. Nothing traumatic - 10-20 seconds between rounds at the very end of a lengthy game as opposed to zero delay at the beginning. Definitely not a deal breaker.

So I find this review is spot on, especially in regards to CPU usage.

As far as the game itself, I give it a 9 out of 10 score. Beautiful graphics and animations along with off-the-charts addicting play. Don't know if I've every played a game that gives you that "one more move-itis" addiction like this one does. I never tire of watching the various units engage in their activities, especially battle skirmishes. Dropped a nuclear bomb for the first time and it was gorgeous. :) Highly recommended no matter what type of gaming genre you prefer.
 
Back