BF3 Upgrade Choices

[flyboy84]

Here's my current setup:

CPU: i7 920
GPU: GTX285
PSU: 650W PC Power and Cooling

I game at 1680 x 1050, and currently run BF3 at a mix of medium and high. I'd like to make the jump to DX11 and be able to run BF3 maxed out at this resolution.

What card(s) today would do this for me, or should I wait and see how the HD 7000 series turn out?

 

[JTsyo]

Well the GTX570 is probably the best bang for the buck for BF3. Anything higher would work too. CPU should be fine.

 

[lehtv]

BF3 performance on 1680x1050, High:

GTX 285 is roughly equivalent to GTX 460: http://www.anandtech.com/bench/Product/317?vs=313

The benchmark above is without AntiAliasing, and it's probably advisable not to use AA in BF3 for optimal performance. Deferred rendering games get a big performance hit from AA.

So, comparing to GTX 460, you could increase your fps by 46% with a HD 6950, or 65% if you overclock it to 6970 speed (or beyond). The 1GB version can be had for $220, the 2GB version (unneeded on your resolution but can be useful if you upgrade the monitor) can be had for $250. A GTX 570 would increase fps by 60%, but a well cooled unit can be also be overclocked quite a bit. GTX 570 can be had for $300-350.

AMD cards take a bigger performance hit from AA in BF3, as far as I know. As long as you stick with post-processing AA only, it seems a 6950 1GB/2GB would be your best bet at the moment. Another option is to wait for the 7850/7870.

 

[flyboy84]

Wow thanks for the detailed post lehtv.

I actually got in on that deal for the HIS 6970 for $224 from Amazon last week, but the likelihood of the deal going through was slim, so I posted this thread to get serious about upgrading.

However, with the reports that people were successfully getting Amazon reps to honor that price for Sapphire and XFX boards, I gave it a try and was able to secure a Sapphire 6970. It also comes with a $30 MIR, so I'll end up getting the 6970 for 194, not too shabby!

 

[MrK6]

Wow thanks for the detailed post lehtv.

I actually got in on that deal for the HIS 6970 for $224 from Amazon last week, but the likelihood of the deal going through was slim, so I posted this thread to get serious about upgrading.

However, with the reports that people were successfully getting Amazon reps to honor that price for Sapphire and XFX boards, I gave it a try and was able to secure a Sapphire 6970. It also comes with a $30 MIR, so I'll end up getting the 6970 for 194, not too shabby!

Great grab, that's a lot of card for 1680x1050. Looks like a monitor upgrade should be next :biggrin:. Enjoy your new card!

 

[Kroze]

BF3 performance on 1680x1050, High:

GTX 285 is roughly equivalent to GTX 460: http://www.anandtech.com/bench/Product/317?vs=313

The benchmark above is without AntiAliasing, and it's probably advisable not to use AA in BF3 for optimal performance. Deferred rendering games get a big performance hit from AA.

So, comparing to GTX 460, you could increase your fps by 46% with a HD 6950, or 65% if you overclock it to 6970 speed (or beyond). The 1GB version can be had for $220, the 2GB version (unneeded on your resolution but can be useful if you upgrade the monitor) can be had for $250. A GTX 570 would increase fps by 60%, but a well cooled unit can be also be overclocked quite a bit. GTX 570 can be had for $300-350.

AMD cards take a bigger performance hit from AA in BF3, as far as I know. As long as you stick with post-processing AA only, it seems a 6950 1GB/2GB would be your best bet at the moment. Another option is to wait for the 7850/7870.

I'd like to know what is the system spec because I have a Radeon 5870 and I sure as hell do not get anywhere near what they're advertising (frame rates according to your benchmark posted)

Minimum of 54 fps? lol that's such a pipe dream. Try half of that.

 

[GoStumpy]

I'd like to know what is the system spec because I have a Radeon 5870 and I sure as hell do not get anywhere near what they're advertising (frame rates according to your benchmark posted)

Minimum of 54 fps? lol that's such a pipe dream. Try half of that.

Are you playing 1680x1050 resolution? Are you sure the GPU is the weakest link?

That's what they do in these tests...

My HD6850 is pretty much spot-on with that graph, so from what I can tell it's pretty accurate.

 

[lehtv]

I'd like to know what is the system spec because I have a Radeon 5870 and I sure as hell do not get anywhere near what they're advertising (frame rates according to your benchmark posted)

Minimum of 54 fps? lol that's such a pipe dream. Try half of that.

They test with i7 2600K. Methodology: "We used Fraps to measure frame rates during a minute of gameplay from the single-player mission Go Hunting. The performance was measured from the beginning of the mission which features a walkthrough on an aircraft carrier. This is an excellent scene as it allows for accurate testing of an outdoor environment."

http://www.techspot.com/review/458-battlefield-3-performance/page2.html

 

[Kroze]

Are you playing 1680x1050 resolution? Are you sure the GPU is the weakest link?

That's what they do in these tests...

My HD6850 is pretty much spot-on with that graph, so from what I can tell it's pretty accurate.

I play at lower resolution than that (1280x720) with everything on LOW just to get playable frame rate (35fps and up).

If I bump the graphic up to 1680x1050 on High quality, it's a joke of a slideshow.

And according to Tom's Hardware, the game isn't CPU dependent. So please what am I doing wrong? I'm on a Core i3 2100 w/8gb of ram.

They test with i7 2600K. Methodology: "We used Fraps to measure frame rates during a minute of gameplay from the single-player mission Go Hunting. The performance was measured from the beginning of the mission which features a walkthrough on an aircraft carrier. This is an excellent scene as it allows for accurate testing of an outdoor environment."

http://www.techspot.com/review/458-battlefield-3-performance/page2.html

My methodology of measuring real and actual frame rates is typing render.showfps 1 on console.

 

[MrK6]

And according to Tom's Hardware, the game isn't CPU dependent. So please what am I doing wrong? I'm on a Core i3 2100 w/8gb of ram.

My methodology of measuring real and actual frame rates is typing render.showfps 1 on console.

I've bolded the problem. If Tom's Hardware said this game isn't CPU dependent, they don't know what the hell they're talking about. Maybe the single player isn't, but who cares? Multiplayer is heavily CPU-dependent and even decent quad cores can get crushed on a loaded 64 player server on some maps. And no, hyperthreading doesn't cut it, you need a true quad core to let your 5870 breathe.

 

[Termie]

I've bolded the problem. If Tom's Hardware said this game isn't CPU dependent, they don't know what the hell they're talking about. Maybe the single player isn't, but who cares? Multiplayer is heavily CPU-dependent and even decent quad cores can get crushed on a loaded 64 player server on some maps. And no, hyperthreading doesn't cut it, you need a true quad core to let your 5870 breathe.

Yeah, pretty much, Tom's and TechSpot and Guru3d flat out fooled everyone into thinking they were providing useful benchmarks. We didn't buy this game for the single player, and telling people that a dual-core works in multiplayer is like saying built-in video works for Crysis (or BF3 for that matter).

That being said, the i3-2100 is the strongest dual-core out there. I think something else is wrong.

Kroze - I'd like to see a graph from MSI Afterburner with GPU usage and frames. That will help sort things out a bit. A 2100 should definitely be able to support at least medium if not high settings at 720p. A 5870 would produce 45 fps at 1920x1200/high if not CPU-limited, based on my testing.

As for the OP, I agree the 570 is his best choice. That being said, if he's getting by on his 285 for now, it might be worth waiting for the 7-series...then again, two or three months is a long time to wait when you're trying to enjoy BF3 now.

 

[GoStumpy]

I've bolded the problem. If Tom's Hardware said this game isn't CPU dependent, they don't know what the hell they're talking about. Maybe the single player isn't, but who cares? Multiplayer is heavily CPU-dependent and even decent quad cores can get crushed on a loaded 64 player server on some maps. And no, hyperthreading doesn't cut it, you need a true quad core to let your 5870 breathe.

I run an i3-2100 3.1ghz with an HD6850, and I get 40~ FPS with everything on HIGH or ULTRA and MSAA off. I haven't played single-player, either.
Not sure if I'm CPU or GPU limited, though! I'll try this MSI Afterburner and see if they're at 100***37; or not tonight

 

[cmdrdredd]

I've bolded the problem. If Tom's Hardware said this game isn't CPU dependent, they don't know what the hell they're talking about. Maybe the single player isn't, but who cares? Multiplayer is heavily CPU-dependent and even decent quad cores can get crushed on a loaded 64 player server on some maps. And no, hyperthreading doesn't cut it, you need a true quad core to let your 5870 breathe.

It isn't CLOCK SPEED dependent, nor does it favor Intel over AMD. a quad core CPU really helps, but I don't think more is much of a difference over a quad.

 

[GoStumpy]

Answer was, GPU is at 100% the entire time.

So, running things at 1680x1050 on an i3-2100 3.1ghz and HD6850 @ 825/1150:


Medium Settings:
Min: 48
Max: 62
Avg: 58


High Settings:
Min: 41
Max: 60
Avg: 47


Not sure how to record the CPU %, but if the GPU was 100% then I can assume the CPU wasn't 100%

 

[happy medium]

Evga gtx560ti superclocked with lifetime warranty for 200$ shipped. No rebate.

http://forums.anandtech.com/showthread.php?t=2203218

Figure in at least a 15% overclock and that will put you near gtx570/6970 performance for 100$ less

 

[lehtv]

Not sure how to record the CPU %

You can ALT+TAB to Task Manager's performance panel

 

[TheAdvocate]

That chart has been bugging me, because my MP framerates are so much higher than that SP mission. I haven't played SP yet, but I understand it is more demanding.

However, for those trying to make a MP based GPU decision, I would be a little wary of those results. Playing at those settings (1680x1050, HIGH, 0xmSAA, 16xAF) I easily average 20+ fps more on just about every map, on my sig rig (560Ti). That might be across the board for all cards, but I can only speak to mine. I routinely see triple figure FPS in less demanding situations in MP. To the point where I'm thinking about cranking up the settings some. On the low end, it probably dips as low as about 47.5 with lots of smoke, etc. 90% of my gameplay is in the 70s & 80s range.

 

[Termie]

That chart has been bugging me, because my MP framerates are so much higher than that SP mission. I haven't played SP yet, but I understand it is more demanding.

However, for those trying to make a MP based GPU decision, I would be a little wary of those results. Playing at those settings (1680x1050, HIGH, 0xmSAA, 16xAF) I easily average 20+ fps more on just about every map, on my sig rig (560Ti). That might be across the board for all cards, but I can only speak to mine. I routinely see triple figure FPS in less demanding situations in MP. To the point where I'm thinking about cranking up the settings some. On the low end, it probably dips as low as about 47.5 with lots of smoke, etc. 90% of my gameplay is in the 70s & 80s range.

Wow, that is exactly the opposite of what most of us are experiencing. Don't know what would explain that.

 

[TheAdvocate]

I always get distracted and forget to record. I'm not trying to fanboy this. Maybe I'll try to record my FPS w/ afterburner this weekend. I dont even know how to do it. Sorry, have kids, wife, etc. Hard just to get time to play, but I'll try if someone is interested and can give me quick cliffs on how to.

 

[lehtv]

Go to Afterburner settings -> monitoring tab and tick "framerate" in the Graph column

 

[ArchAngel777]

Great grab, that's a lot of card for 1680x1050. Looks like a monitor upgrade should be next :biggrin:. Enjoy your new card!

Not sure why everyone says this. I had a 460 GTX @ 822/1644/4000 (roughly the same performance of a 470 GTX) and it was severely crippled by this game on Ultra @ 1680x1050. I purchased another one, put them in SLI and is now smooth as butter. The game scales around 85-90%, which would put it above the 6970 by a substantial margin.

BTW, I used to be anti multi-GPU, my post history proves that, but I don't really notice any microstutter or input lag over my single GPU setup. I merely went SLI even expecting these problems and found that if they exist, I certainly don't see them and to be clear, I am rather picky.

Ultra Settings, Vsync, Motion Blur turned off (I hate it), 4X AA (deferred) @ 1680x1050 on a 460 GTX SLI Setup @ 822/1644/4000 is smooth as butter.

Edit BTW - I am not saying the 6970 is a bad card. That wasn't the reason for the post, the reason was that people think it is overkill for 1680x1050 which I don't agree with. It allows smooth as butter gameplay.

 

[exar333]

These big release games are coming at a terrible time for industry. If we had the new GPUs out, it would be awesome! It hurts to buy a mid-high to high-end GPU this late in the cycle.

I would say get the best card <$200.

 

[ArchAngel777]

These big release games are coming at a terrible time for industry. If we had the new GPUs out, it would be awesome! It hurts to buy a mid-high to high-end GPU this late in the cycle.

I would say get the best card <$200.

Yeah, I agree, which is why I purchased another 460 GTX for around $130 AR. I would have loved to get a 580 GTX, but the price is just too much for it be dethroned in the not too distant future. Additionally, I wanted to try multi-GPU and see for myself. It isn't that bad.

 

[TheAdvocate]

Go to Afterburner settings -> monitoring tab and tick "framerate" in the Graph column

Does that generate a log? I am not at my PC right now, but I know how to get the FPS to display while I'm playing. I've been using it with every new driver release to eyeball the fps changes, so that's what I am basing my anecdote on.

 

[redrider4life4]

Koze,

Something is def wrong with your system or the quad core processors really do make a world of difference. I'm running i7 920, 5850, and 4g of ram on a 24 inch montir at 1900 x 1200 with almost all settings on high (NO AA) and get a consistent over 30 and hovers in the mid-high 40s. I'l be upgrading to the new 78xx cards once they are released in the coming months.