Battlefield 3 announced

[pjbliverpool]

And it looks like dual core CPUs are practically useless. Oh well it's past time to upgrade anyway.

Something isn't right there, there's no way a dual core should be THAT much slower than a quad. An i3 at 2.93Ghz less than 1/6th the speed of an i5 at 2.8Ghz?

If thats the way the game is supposed to behave then I'd love to know what's going on to cause it.

 

[Billy Idol]

I hope that DICE does not forget to implement destruction!! I demand that the level of destruction is increased compared to BF BC2...everything else is lame and a real downer, because imo that is what makes BF BC 2 great and unique compared to all the other one mill. shooters out there: pure destruction fest!

But I must admit, that I am a little bit nervous watching all those beta vids available: where is my destruction?! Gimme my destruction....and then my parade!

 

[tabs]

As an avid BF2 fan I'm not actually as bothered about the destruction as I perhaps should be. I really loved it in BFBC2, but if they have to sacrifice something (that is in my mind) essential, I wouldn't miss it. It would be interesting to have a server-side setting to disable destruction, for unranked servers naturally.

 

[AlNom]

Something isn't right there, there's no way a dual core should be THAT much slower than a quad. An i3 at 2.93Ghz less than 1/6th the speed of an i5 at 2.8Ghz?

If thats the way the game is supposed to behave then I'd love to know what's going on to cause it.

The lighting uses multithreading for starters.

 

[AlNom]

Something isn't right there, there's no way a dual core should be THAT much slower than a quad. An i3 at 2.93Ghz less than 1/6th the speed of an i5 at 2.8Ghz?

If thats the way the game is supposed to behave then I'd love to know what's going on to cause it.

The dynamic radiosity, for starters.
http://www.geomerics.com/downloads/radiosity_architecture.pdf
http://publications.dice.se/attachments/GDC11_LightingYouUpInBattlefield3.pdf

There's just a lot of jobs being done in parallel. Frustum culling (geometry) is done across multiple threads for instance (software z-).

http://publications.dice.se/attachments/CullingTheBattlefield.pdf

 

[pjbliverpool]

Cheers for the info, some interesting stuff there. It seems at the highest settings this game simply isn't designed to work with only 2 cores. I'd be curious to see how a tri-core fares for comparison to Xenos.

Interesting how they chose to perform a lot of that work on the CPU rather than the GPU. The reasons for this seem to be pretty console driven, i.e. DX9 level GPU's that are already overused. I wonder oif this was a PC only game if more work would have been done on the GPU.

Not that how it's done now is bad, in fact it's great to see them pushing multi-core CPU's and making use of all that power.

 

[Aeoniss]

As an avid BF2 fan I'm not actually as bothered about the destruction as I perhaps should be. I really loved it in BFBC2, but if they have to sacrifice something (that is in my mind) essential, I wouldn't miss it. It would be interesting to have a server-side setting to disable destruction, for unranked servers naturally.

DICE said that there would be far more destruction than in BFBC2.

 

[AlNom]

Interesting how they chose to perform a lot of that work on the CPU rather than the GPU. The reasons for this seem to be pretty console driven, i.e. DX9 level GPU's that are already overused. I wonder oif this was a PC only game if more work would have been done on the GPU.

Not that how it's done now is bad, in fact it's great to see them pushing multi-core CPU's and making use of all that power.

Yeah, I sort of wonder about that, especially when you look at Steam Hardware Survey, which shows only 26% have quads. At the same time, we should consider that PC gamers will probably be running significantly higher resolutions and quality of effects than consoles ever will (i.e. 1360x768, 1440x900, 1600x900 etc, higher res textures/shadows/particles/post-fx etc), which will offset the higher performance relative to the console hardware to some degree, if you see what I mean.

I'll be interested to see what sorts of options they'll expose for scaling quality/perf.

 

[Davros]

Y we should consider that PC gamers will probably be running significantly higher resolutions and quality of effects than consoles ever will (i.e. 1360x768, 1440x900, 1600x900 e

Have you been in a cave ? those arnt high pc resolutions

 

[AlNom]

I said "higher". "etc" also means the rest of the resolutions I didn't mention, which does include 2560x1600 and..... etcetera.

Anyways, the point was to consider the mid-level hardware being offset by higher PC settings and in the context of whether or not to optimize for multi-CPU.

 

[I.S.T.]

Every PC gamer will have DX10+ hardware. There may be a few who are still on XP, but I doubt it would be enough to register on the sales charts.

You can have DX10 hardware and above on XP.

 

[neliz]

You can have DX10 hardware and above on XP.

Invalid argument as BF3 won't run on XP.

 

[pjbliverpool]

Anyways, the point was to consider the mid-level hardware being offset by higher PC settings and in the context of whether or not to optimize for multi-CPU.

If this is genuinly how the final game will perform regardless of quality settings (since the test was at low anyway) then it doesn't seem like the wisest course of action from a PC market perspective.

Effectively the game won't be playable on 50-75% of users machines (dual cores).

And the decision looks even stranger when that lack of performance on the dual cores seems based on the number of threads in use rather than the total capability of the CPU. i.e. the i3, being half the i5 should be hitting around 75 fps and yet it only hits 26 fps. Surely there's a better way to re-distribute the workload so that the process of running too many threads alone (rather than the content of those threads) doesn't cripple performance.

Stranger still is that the i3 530 and Q9550 actually sport the same number of hardware threads thanks to Hyperthreading (4 each).

 

[tabs]

What sort of sound quality options are available in the Alpha though? I seem to recall a couple of settings in BFBC2 being very CPU intensive.

 

[Dresden]

I may have to take October 25th off from work.

If anyone asks, I'll tell them I was enlisted, which would be partially true.

 

[Andrew Lauritzen]

Probably based on the sales statistics?
http://gamrreview.vgchartz.com/sales/31444/battlefield-bad-company-2/ (which makes the PC version a nice 10.4% of the total sales)
PC WW sales for BC2 are around 600K units.

Not sure how much I'd trust any PC numbers these days unless they were from EA themselves considering how large a portion I imagine is Steam (and other digital distributors) who, to my knowledge, still does not give out sales numbers.

 

[tabs]

Yeah, I recall EA / DICE going on about how great the numbers were for the PC version, which apparently surprised them. No specifics though.

 

[Andrew Lauritzen]

Surely there's a better way to re-distribute the workload so that the process of running too many threads alone (rather than the content of those threads) doesn't cripple performance.

Unfortunately modern OSes are only starting to get support for cooperative task scheduling (work stealing, etc) so this is pretty typical.

People create a thread for each of their subsystems and some of the subsystems even create all their own worker threads. Then you get thread oversubscription which tanks the performance of the entire machine far below what it should be. Step 1 is using a single shared thread pool in your application but good luck getting every piece of middleware to agree on that. Thus this probably needs to be solved at the OS level (which also solves the same issue between processes), but as I mentioned, it's pretty immature so far. Tools exist (TBB, Cilk, Apple GCD, Microsoft PPL) - some better than others - but people haven't settled on anything standard yet.

Not saying that's what is definitely happening here, but it's certainly possible. Take a look at how many threads the alpha client is using (and ideally which are awake and working hard at the same time) to get an idea of if this is an issue. Also check whether it scales down proportionally on a dual core vs quad core system.

Do also note that the two processors aren't otherwise identical. The i5 has turbo and twice as much cache than the i3 among other things.

Note too that they have been saying get a quad core all along. That's hardly a budget stretch these days and I imagine a lot of people will be upgrading for BF3

 

[KKRT]

Not sure how much I'd trust any PC numbers these days unless they were from EA themselves considering how large a portion I imagine is Steam (and other digital distributors) who, to my knowledge, still does not give out sales numbers.

Vgchartz is a bullshit site and BC 2 PC version sold about the same as other versions. Here's the proof http://bfbcs.com/

 

[I.S.T.]

Invalid argument as BF3 won't run on XP.

It was more like a reminder that DX10+ Hardware is on XP than anything else. I know BF3 is Vista and up limited.