Overclock.net banner

1 - 10 of 10 Posts

·
Registered
Joined
·
1,643 Posts
Discussion Starter · #1 ·
Some time ago, I came across a thread where OP wondered wheter he should upgrade his brothers GPU or CPU for Battlefield 4. He currently had a QX9650 and a GTX 470, and was planning on getting a GTX 660 Ti, or a cpu-upgrade if that would be the better choice. Numerous suggestions were given, among upgrading to a R9 270x/280 or any other GCN-card, and use the Mantle API.
Thus, I decided to benchmark Battlefield 4 DX11/Mantle on a Core2Quad. More specificly, a Q6600. The benchmark-runs consists of roughly 10-15 minutes of gameplay on a 64-player server, using the built-in benchmarking-tools. All bechmarks were run at 1080p Ultra, with no MSAA. I'll use RagingCain's Frametimeanalyzer and Excel for presenting the data I recorded. Mind you, Excel still fights me, and I'll need some time to figure that out.

UPDATE 1
Test results from the 2500K has been added, and conclusion rewritten.


A quick run-down of the hardware:
Rig 1
Q6600 stock/3510mhz
4gb HyperX DDR2 800mhz
Asus P5K Premium
AMD R9 290X reference + uber
14.7 RC3

Rig 2
2500k @4700mhz
16gb Vengeance 1600mhz
Sabertooth Z77
AMD R9 290
14.7 RC3

The second rig is just for comparison, to see what's possible when the card is completely unleashed. Though, the most important thing here is DX11 vs Mantle on the C2Q. I'll update the post once I get the numbers from rig two ready.
One final thing to note, is that different maps have a different impact on the cpu (or atleast, so it seems). I therefore conducted most of the benching on a 64-player server on Zavod 311. Also, I won't be posting any detailed graphs of the frametimes because there's too much data, and as a result, looks like DX11 performs equally as Mantle. Even though it clearly *spoiler alert* does not.

Rig 1
First out is the DX11 stock run. As I had .drawgraph enabled, I noticed that both graphs did not show a thin, clear line. It was more like a thick bush, winding itself slowly up and down. Fps was hovering around 20-40, but it was stuttering a lot. While I was able to play, it was not very pleasant, and I found myself often looking at the clock to see how much I had left. As in all the runs, I took a flight over the map with a helicopter, though this proved troublesome. The cpu simply could not keep up, and failed to update the different units onscreen in a timely manner, including myself. This caused a rubber-banding effect, and made it difficult to fly. Oh well, atleast I could listen to the sweet sound of a M249 mowing down ma… No, no...That's not correct. I did listen to a sound, but it was nothing about it that was sweet. The cpu had not enough processing power to handle the audio; it frequently froze and many elements were missing completely.

1/3 is spent below 30 fps. Zavod 311

I then switched to Mantle, expecting magical things to happen. Reality slapped me hard in the face though. So hard in fact, that I ran it twice, on different maps. In the end, there was a slight difference, but I can tell you that the overall experience was at times definitaly smoother, though the stuttering was more noticable. But hey, atleast the sound actually worked this time around, and there was no rubber-banding
thumb.gif


A very slight difference if you look at time spent above X. Lancang Dam, suspect it's more cpu-intensive

It was now time to overclock. I did not want to spend too much time tweaking, and fired up a quick and dirty oc of 3510mhz. Also, I was very coarse on the voltage, and thus decided to put a 120mm Corsair-fan from a first-gen H60 to the task of cooling the motherboard and ram. I did run quite a lot of juice through both cpu, nb, sb and ram
rolleyes.gif

My first run was in Mantle, but I'll save the best for the last.

Yep, the overclock did it's job. Whereas DX11 on stock clocks spent 1/3 of the time below 30 fps, it spent more than 90% of it above 30 when ran at 3,5ghz. Jolly good job, lads. However, it was not all green fields and flowers. There was still a (very) slight rubberbanding when I flew the helicopter across the map, revealing that the cpu was pegged to the limit. It could handle BF4, but nothing more. Overall the experience was quite satisfying, but the .drawgraph still showed a thick bush.

Much, much better than @stock clocks. Definitely playable. Zavod 311

And WHAT a difference! I kid you not, I could, with a few tweaks to the settings, be playing on this pc and not been able to tell it apart from my own, except the framerate dropping every once in a while. There was still some stuttering every now and then, but that could be Mantle in itself not willing to play ball with that computer, or the fact that the OC was unstable (which it was, got a bluescreen once, but it held it together for the most part). There was no rubber-banding, the sound was perfect once again, and it was pretty darn smooth! Numbers does not completely justify Mantle. And for the first time ever, the gpu-fan ramped up! I could now actually hear the 290X.

Higher framerates across the board compared to DX11. Percentages spent above X consideribly higher. Zavod 311

Rig 2
Nothing unexpected here really. Results might have been slightly different if the server was more fast-paced, but nothing that could be noticed. Only measured, and pointless (like those times people start arguing about which gpu has the highest performance, and the delta is less than 2%
tongue.gif
).

On the gpu-side, the numbers have not changed too much.

After actually playing the game on this computer for the first time in quite some time, I decided to adjust some of my statements. At the same settings as the Q6600, this was considerably more smooth. During the 10 or so minutes of gameplay on Zavod, I only experienced one small spike on the cpu. Framerates was also higher and was kept above 60 at all times.
While the Q6600 does not pack enough of a punch to extract all the performance from a 290-series card, it is adequate even at ultra settings on the most demanding servers. Singleplayer will be a breeze, as do servers with a lower player-count. I am certain that once the OC has been dialed in properly, and the settings ingame tweaked, the gameplay will be butter smooth.

Your old 775 might just prove to be sufficent for you a few more years
thumb.gif
 

·
Registered
Joined
·
5,813 Posts
Very good read! I Think this should be stickied so that people who are looking at upgrading older rigs can see it easily.

EDIT: Also how did you get the BF4 frametimes? I'm looking at doing some benchmarking over the weekend/next week comparing some older cards.
 

·
Guru
Joined
·
1,218 Posts
Good read, but yet OCed Q6600 isn't quite comparable to OCed QX9650
tongue.gif


You would need that thing at ~4Ghz to match OP's [email protected]
 

·
Registered
Joined
·
5,813 Posts
Quote:
Originally Posted by SmOgER View Post

Good read, but yet OCed Q6600 isn't quite comparable to OCed QX9650
tongue.gif


You would need that thing at ~4Ghz to match OP's [email protected]
That's not the point, the point is to prove that mantle helps older CPU's in general. It shows that mantle is effective, it doesn't show how well mantle will help the person with the QX9650 or anything just that it will help.
 

·
New to OCN?
Joined
·
26,919 Posts
I still dont use mantle on bf4 but whenever I start my testing i will and will publish the results from c1,c2,c3,bf3 too
 

·
Registered
Joined
·
250 Posts
Awesome thread dude
 

·
Registered
Joined
·
928 Posts
Mantle usually crashes for me. What AMD driver did you use for the tests?
 

·
Premium Member
Joined
·
20,311 Posts
Quote:
Originally Posted by Coach Mcguirk View Post

Mantle usually crashes for me. What AMD driver did you use for the tests?
you've got 4GB of RAM? if so, add 4 more. it might help. i use 14.6 Beta with both my rigs (7950 and 290). Works well.

i had to move the Settings folder found in Documents>Battlefiled4 to my Desktop or else i can't log on to the game. no biggie.
 

·
Registered
Joined
·
1,643 Posts
Discussion Starter · #10 ·
Well, those 4 gigs were pretty much all the DDR2 we had in the house. I'd have to buy some more for 8gb testing, which I can't afford
tongue.gif
 
1 - 10 of 10 Posts
Top