• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

64-Player BF3 & CPU usage

WiL11o6 said:
Update:

I switched out my 2500k for a 2600k and I made the same runs through Strike at Karkand.

On my 2500k @ 4.5ghz, my GPUs would both be 75-95% usage, it's almost always under 90%, floating around 85% and drops to 75% occasionally.

On my 2600k @ 4.5ghz, my GPUs do not go under 90% usage, very seldom does it even hit 90%, it normally sits 95-99% usage, mostly 97-99% on a 64-player server.

Flame me if you want, but it's exactly like the screenshot of the AB from the OCN forums.
Click to expand...

That's very interesting. My 2500k experience is pretty much the same as yours. I think I'll wait for Ivy Bridge though.
 
WiL11o6 said:
Update:

I switched out my 2500k for a 2600k and I made the same runs through Strike at Karkand.

On my 2500k @ 4.5ghz, my GPUs would both be 75-95% usage, it's almost always under 90%, floating around 85% and drops to 75% occasionally.

On my 2600k @ 4.5ghz, my GPUs do not go under 90% usage, very seldom does it even hit 90%, it normally sits 95-99% usage, mostly 97-99% on a 64-player server.

Flame me if you want, but it's exactly like the screenshot of the AB from the OCN forums.
Click to expand...

noticeable difference in BF3 performance?
 
Why are people flaming owners of 2600K, we can afford them, we wanted HT, we wanted the extra 2 megs of cache (something a lot of people forget, on-die cache costs money), and perhaps even a better binned chip.
 
Pitbull said:
Why are people flaming owners of 2600K, we can afford them, we wanted HT, we wanted the extra 2 megs of cache (something a lot of people forget, on-die cache costs money), and perhaps even a better binned chip.
Click to expand...
perhaps you should read back through the thread to see who started the attacking. I can afford the 2600k too but I am not paying 100 more bucks for something that will do zero for me in gaming. nearly every person here will upgrade their 2500k long before the 2600k has any advantage. even BF3 will show basically no difference at game settings that people actually use. now if you are always going to run top end cards in sli/crossfire or need the HT for other things then sure the 2600k is a good choice choice.
 
cannondale06 said:
you pay around 100 more bucks for that basically useless (for gaming) feature...
Click to expand...

Possibly, mostly true for 99% of the games.

How much did I pay for my 7970s? $1200.
How much to upgrade from 2500k (reselling 2500k for less) to 2600k? $150

Do I notice a difference between 2500k and 2600k in BF3? Probably not.
But with the 2600k, it doesn't ever go below 60fps whereas the 2500k did sometimes even hit 55fps (75% GPU usage), and that to me is unacceptable for spending $1200 in video cards. It is a peace of mind to see my GPUs running at 99% GPU usage most of the time with the 2600k, rather than seeing it fluctuate 75-95%.

$150 well spent to me.
 
WiL11o6 said:
Possibly, mostly true for 99% of the games.

How much did I pay for my 7970s? $1200.
How much to upgrade from 2500k (reselling 2500k for less) to 2600k? $150

Do I notice a difference between 2500k and 2600k in BF3? Probably not.
But with the 2600k, it doesn't ever go below 60fps whereas the 2500k did sometimes even hit 55fps (75% GPU usage), and that to me is unacceptable for spending $1200 in video cards. It is a peace of mind to see my GPUs running at 99% GPU usage most of the time with the 2600k, rather than seeing it fluctuate 75-95%.

$150 well spent to me.
Click to expand...
I spent $300 on my Q9450. Even today, it was a decision totally worth it. Besides, back then, they cost about $500+ (I went for the Microcenter $300 deal :D).

12 MB of L2 cache is awesome. I haven't tried my system on Battlefield 3, but that's alright. My Q9450 bulldozes in video encoding. :D
 
WiL11o6 said:
Update:

I switched out my 2500k for a 2600k and I made the same runs through Strike at Karkand.

On my 2500k @ 4.5ghz, my GPUs would both be 75-95% usage, it's almost always under 90%, floating around 85% and drops to 75% occasionally.

On my 2600k @ 4.5ghz, my GPUs do not go under 90% usage, very seldom does it even hit 90%, it normally sits 95-99% usage, mostly 97-99% on a 64-player server.

Flame me if you want, but it's exactly like the screenshot of the AB from the OCN forums.
Click to expand...

Why would we flame ya? If it worked for you great.. more power to you! You'll have to look long and [H]ard to find me flaming upgrades around here. On a side note.. I'll give ya a nice shiny nickel for that old POS 2500k.. my wife's PC needs an upgrade. :p
 
Because half the people here think that 2600k is a total waste for gaming, and would tell me to save my $100 because a 2500k to 2600k is totally unnecessary for gaming.
 
WiL11o6 said:
Because half the people here think that 2600k is a total waste for gaming, and would tell me to save my $100 because a 2500k to 2600k is totally unnecessary for gaming.
Click to expand...

Well they have a right to their opinion as well.. but why should you care if you are happy with your purchase? A lot of folks tell me xfiah 7970's is overkill for a single monitor. Does not mean I'm going to return one or make me enjoy them any less.
 
Silverghost said:
serious_cat_is_serious.jpg
Click to expand...

Somebody broke out the serious cat... :D
 
Skunt said:
If this is true, then damn all those peeps that kept saying get a 2500k. Every thread I went to where somebody was going to get a 2600k for gaming, he'll get loads of replies telling him to get a 2500k. Hell, I got a 2500k because of the general consensus that 2600k is of very little benefit for gaming.
Click to expand...

Yeah I got flamed for telling people to waste their money, even though I knew there would be a point where Hyperthreading would make a difference.

On the same account, those 6 core CPU's are eating this game alive...
 
Falkentyne said:
Yeah I got flamed for telling people to waste their money, even though I knew there would be a point where Hyperthreading would make a difference.

On the same account, those 6 core CPU's are eating this game alive...
Click to expand...

There is still no conclusive proof that the 2600k is better... varies a lot from user to user on this very forum. As usual.. YMMV. I'd still recommend a good OC'd 2500k to a 2600k.
 
From my observation battlefield 3 uses all 8 actual cores in my system. Don't remember the exact utilization per core but I ran BF3 windowed with task manager up and it does use all 8 of my cores.

I had to drop down from Ultra to High for the 64 player karkland maps as I would get frame drops into the single digits on ultra. High it stays in the double digits.

Frame rate is all over the place sometimes as high as eighties and other times dropping into the teens as I walk like 10 feet and a fire fight starts.

I'll have to ask my friend with an SR-2 if it uses all her 12/24 cores depending if she has hyperthreading on.

I been hearing the hypethreading off prevents stuttering too.
 
My 2500K and single 480 gtx run the game fairly well.

I see 99% gpu usage in game for the most part. Maybe dips to 98% at times but in afterburner the graph stays a the top unless I alt tab or load a map.

Interesting that the multithread works well with multi gpus.
 
When I had a single 7970, or even SLI 570s, they would be at 98-99% GPU usage most of the time on my 2500k @ 4.5ghz. Switch out SLI 570s to CF 7970s and my GPU usage drops to 75-95%. At first I thought it was the drivers, but other people here with 2500k's are having same GPU usage with CF 7970s.

If you had a single GPU setup, or two mid-range SLI/CF setup, a 2500k is enough. Even for CF 7970s a 2500k is enough I know. But I spent $1200 on video cards, I want to make sure they are working 100% worth of my money, not 75%. Especially! when I built this PC to slaughter BF3, I will spend an extra $150 to switch out the 2500k to 2600k.

Both my 7970s are working at 98-99% GPU usage all the time, on any map with the 2600k.
 
I'd say those numbers are exaggerated and not directly comparable. You need to have those two procs run on the same system at least for them to be close to comparable. Even then the MP experience in a given map can and will change drastically from game to game. If it is a given that the 2500K and 2600K are overclocked to the same speed, and you take turbo boost out of the equation, the 2600K should be marginally faster than the 2500K given it has 2MB more L3. I sincerely doubt that the difference here is going to be >= 20 FPS on a single 7970. Now when you start hitting with more than one GPU that story can change as small margins of victory may keep the GPU from starving and hence bottoming out on framerate.
That's the idea anyway. Here's some CPU charts for BF3 CPU usage on my I7 920 @ 3.78 Ghz.

The settings are:

The first set for CPU utilization is at 1920x1200 where crossfire is broken on the 7970's when launched from an Eyefinity desktop. Gametype is destroy the MCOM map is Grand Bazaar. This is a full 64 player server but this is not one of the most demanding maps.



You can see Total CPU utilization never goes above 60% during the actual game, and has an average of 52%. You can also see that the first six threads are being utilized by the game while the last two are being utilized by windows and the benchmarking software(see utilization spikes on core 4 threads every 5 seconds). The first six threads have average utilization of 47.84%, 65.02%, 45.88%, 59.22%, 43.35%, and 55.44%.This is when one 7970 is being utilized and there are 3 benchmark apps running to collect information. Afterburner, and HWiNFO64 every 5 seconds and FRAPS in standard benchmark mode. Now contrast that with the CPU usage from 52392x1050 game where Crossfire is not broken when launched from an Eyefinity desktop. Destroy the MCOM game, map Operation Metro full 64 player server. Same settings otherwise



You can see that the Total CPU utilization never goes above 80% and has an average of 59.77% this time even though the graphical demand on the GPU's has risen and you would think that CPU limitation would be less of an issue at these settings. The first six threads have average utilization of 56.32%, 72.24%, 68.21%, 53.28%, 63.55%, and 51.69. I realize this is not a 2600K but you should just see a minor difference performance advantage clock for clock for the 2600K over the 920 as the sandy bridge have had minor efficiency improvements. Both processors have the same amound of L2 and L3.

That being said I have seen these two 7970's bottom out into frame rates in the high 30s when playing on a full 64 player server, destroy the MCOM @ 1920x1200 with these exact same settings and Crossfire scaling properly(launched for extended or mirrored desktops.) on the map Strike at Karkland. So I can see that it's possible a little bit more CPU muscle might mean bottoming out in framerates that are in the high 50's instead. With multiple 7970's @1920x1200 and these settings in BF3
 
B^gD^ddy~CF~ said:
I'd say those numbers are exaggerated and not directly comparable. You need to have those two procs run on the same system at least for them to be close to comparable. Even then the MP experience in a given map can and will change drastically from game to game.
Click to expand...

not necessarily

statistically speaking, the comparison would be valid if you captured enough MP games on both processor setups and then found the mean - you could get within a 95% confidence level i bet
 
Makes me feel all warm and fuzzy having an i7-2700k vs the i5. I too got some flaming when I detailed my system (not here, but other forums) with people even saying 'shhh, don't tell him he's wasting his money on the i7, it makes the i5 cheaper for all of us.' While I have not read much on HT, I do know that it just started around when I got my last sytem back in 2002 -- I missed the boat, but my roomie who also got a new system a few months later got the HT. I figured that if HT was still being used some *10* yrs later, but was not a feature on some of the more mid-level chips that it was worth getting.

Like most of you have already noted -- the i7 can turn off the HT, can the i5 turn it on?

Plus the i7 does have more cache, faster clock and better binned -- all which point to more future proofing, and considering the overall cost difference is less than 1 bill, it's a steal.

I get like 1.53 Cinebench in single thread and 7.02 using all -- this is with stock.
 
I run 3 6970s @ 98% GPU usage OC'd to 950/1400 in BF3 on 64 player MP Conquest maps without any issue. The only time I've had less GPU usage was if I had started playing the game after using the PC for a couple days, and would get usage in the 75% range, but after a reboot, it cleared things up.

3 6970s is occupying more PCI-E bandwidth, and performs on nearly exactly the same level as 2 7970s at stock. I get exactly the same 3dmark 11 P score as 2 7970s, and the same FPS in BF3 and various other games. When you factor in OC's the 7970s may be SLIGHTLY better, maybe ~5-7% when both are OC'd to the max. Regardless, I believe my setup is just as authentic of a test for a 2500k as yours was, and my 2500k is @ 4.7ghz.

A 2500k is PLENTY for 2 7970s when OC'd to 4.7ghz+, as it was plenty for 4 6970s for myself (when I had 4 6970s), as well as my friends with quadfire setups... and 4 6970s IS faster than 2 7970s regardless of the OC on the 7970s.

/thread
 
Well, what resolution do you play at? If it's anything 1200p or higher, then even with 3x6970s it would be stressed.

I only play at 1080p and 2x7970 can demolish everything with ease MOST of the time, so need a fast CPU to boot.
 
That's_Corporate said:
not necessarily

statistically speaking, the comparison would be valid if you captured enough MP games on both processor setups and then found the mean - you could get within a 95% confidence level i bet
Click to expand...

Yes and I'd agree but the mean is not going to show you the worst case scenario. By definition it will show you the mean of the worst, best, and average scenarios. What this gentleman is looking for is will going from a 2500K to a 2600K net him a bottom out frame rate that is 20 FPS higher than he's getting with his 2500K at the same clocks. I guess we could ask him as his sig now shows him as having a 2600K.

In order to do this comparison you really need to have controls on as many of the variables as possible. Meaning getting results from people with different rigs, setups, applications running is not going to do that. It's just plain hard to do comparative tests in BF3 Multiplayer. In order to do so for this type of evaluation you'd need to test both a 2500K and a 2600K on the same rig, in the same server every time with the same number of players in the same maps. Without being able to do that the best your getting is maybe a generalized idea.

I'd be interested to hear if now that he has his 2600K he's seeing those bottom out frame rates that are 20 FPS higher than he was seeing on his 2500K in full 64 player servers in the most demanding maps.
 
^^^ I'd love to see the data as well. Controls are the biyatch on MP maps. But I'm sure you'd know pretty fast of your minimums came up 20fps pretty quickly. I'm just not seeing how. If adding HT and a few mb cache increased minimums by 20fps there would be huge noticeable gaps in the reviews numbers where there currently are not.
 
I haven't done intensive testing after getting my 2600k, however, as you guys stated earlier, my minimums didn't increase by 20. The numbers I am about the post is what I have seen using my eye looking at FRAPS fps display in game, and Afterburner, so it's not written in paper or graphs. It is only discussing the minimums I've seen, not average boost in fps.

To the best of my abilities on the 2500k and now the 2600k, I would run and stand at two places in Strike of Karkand in a 64-player conquest map with an average of 55+ players in the same server. It's the 24/7 dedicated Strike of Karkand server. The first spot is in suburban area on top of the hill overlooking the city. The second is at the gateway (or something), the one next to the warehouse.

2500k:
Through running through most of the map, I would average 75-99% GPU usage, but that was still enough to put me above 80fps. At the suburban area, I would get a min down to 55-60fps, and at the gateway, it would also drop down to 55-60fps. Always under 60fps. I have done this test here on this server about 5 times (I've done this more at other Strike of Karkand servers with same results, but to keep it more consistent).

2600k:
Through running through most of the map, I would average 75-99% just like the 2500k. However, with the 2600k, it would mostly be in the 90% GPU usage, with some drops into the 70%'s, the 2500k being mostly in the 80% GPU usage, with many drops into the 70%'s. In the suburban area, I was getting 60-65fps, and at the gateway it would be also 60-65fps. And if I remember based on memory, my GPUs weren't pegged at all, it was around 75-80% at these areas, almost same as the 2500k.

Looks more like a min of only 5-10fps, which is more than enough for me. Sure, there could be a chance that with the 2500k in all those runs, there were shit exploding everywhere so I had lower fps and the 2600k scenario didn't. After spending a week with the 2600k, I am glad I switched it out. GPU usage is up, and I don't ever see fps drop under 60fps. It may get close to dipping under, but not quite.
 
Ok good to hear...not conclusive but good to hear. I don't blame you one bit for just wanting to actually play the game and not spend hours upon hours running benchmarks and graphing the data. That's about what I expected. Your running a processessor family that's a little more efficient clock for clock than my I7 and your running it at about 700 mhz faster speed. So it stands to reason that in those worst case scenarios online where i'm getting down to say 39-40 FPS for just a second or so a little extra CPU oomph might be able to smooth that out. When those lows happen to me I see my GPU usage fall off the charts so I gotta think that's just GPU starvation that is either due to the CPU not being able to feed them quite quick enough for a very small amount of time, or that the server is not pushing responses fast enough. Those scenarios happen very rarely though. As in maybe twice in the the whole time I've had my 7970's. Both times on strike at karkland MCOM maps in a full 64 player server with me on the attacking team at the second MCOM. Both times there were so many explosions going on that you couldn't really see the opposing team. The only indication of their presence being a high volume of incoming tracer fire, grenades, noob tubes, BMP rounds as I try to crawl to the MCOM.
 
Last edited:
i got a x6 1090T at 4ghz/nb 3ghz and 2xR6970 Lighting

Ultra settings FPS 80 - 120 (Vsync ON)

GPU's usage +-95% sometimes go to 85% and 99%.
 
My GTX 470 would fluctuate between 80%-99% but when i OC'd my i7 920 from 2.66ghz to 3.66ghz, its almost always 99%. So my GPU is bottlenecking quite hard.
 
where is it reported that BF3 uses more then 4 cores EXACTLY?

i have had 1090T with 6870 crossfire setup and I KNOW it uses six cores for sure but everyone i talk to (espicially INTEL people) always say FOUR CORES ONLY

if we alt-tab to see cpu usage that is the task schedular

they say this is proved by comparing I7 3930K (six core) with HT DISABLED to I7 3770K (four core) with HT DISABLED and both cpu`s at same freq

if no frame rate increase with 3930K than BF3 uses 4 threads only right?
 
Never noticed a problem with my i5 760 personally and doing Strike @ Karkand on Ultra but than again i'm capped @ a 1680x1050 resolution as my monitor is ooooold..
 
nope this is to see if there is a reason to get a six core intel proc for BF3

does it use six cores (one way to check is benchmark with six cores enable and HT disabled)
then go into bios and disable two cores (run benchmark again)

this should be comparable to some quad core intels benchies on the net

if the frame rate dipped down then we know a six core is good to have (and bf3 uses all the cores)
if it stays about the same then we know there is no reason to get a six core.
 
The question is academic, although not unimportant, because current quad-cores from Intel are faster than BF3 demands.

It's only important because future DICE/EA games will be built on Frostbyte 2, and they may very well make use of more than four cores.

However, it should be noted that clock-speed is preferable to core count, if properly supported.
 
thats not my Q

take a six core processor and disable all HTing if it has it.

run benchmark for BF3

go into bios disable TWO MORE CORES

run benchmark

tell us your results

quick and fun test
 
I'm saying your Q isn't worth researching. It's not a bad question, but the answer it provides doesn't do anything for us.
 
sure it does

it tells you if the more cores you use the better it is

ITS EVERYHTING when it comes to this game

http://www.guru3d.com/article/geforce-gtx-690-review/12

look at that

do you see how they left out all the normal I5 and I7 CPU's?

look at bf3 cpu scaling very carefully

look how they leave 8150 at "DEFAULT" and they have a 965 OC and then 3960K

where are all the 2500, 2600, 2700, 3570 and 3770?

they left them out on purpose because with BF3 multi-core wins
the only problem is they didnt OC that 8150 otherwise it would have beat everything there except that 6 core intel
 
I really don't see how that page supports your conclusions- though I'm not terribly sure what those are.

The conclusion I draw is that if you throw enough graphics work at a system, you just have to have enough CPU- more isn't going to help. Also, AMD's chip sucks balls, and you don't actually want to overclock an 8150, it gets too damn hot.

As for leaving out 'normal' CPUs, they actually don't need them. They have Sandy Bridge in there, and Ivy isn't any faster and runs hotter when overclocked while also not overclocking as well- so they're about the same. Hyper-threading does little for games as well, so anything from a 2500k at 4.8GHz to a 3770k at 4.6GHz will trade blows- they're all the same chip.

That Sandy Bridge-E is a six-core CPU yes, but notice that it doesn't make a damn bit of difference? And notice how AMD's chip with eight half-cores is simply useless due to it's abysmal single-threaded performance?

Think long and hard about this. Right now, that means July 2012 and not some other time, anything above a $200 i5-2500k with moderate overclocking is faster than any game other than Civ-V can use. That's why we don't need to see how six core Sandy Bridge-E CPUs help us; we don't need them, and they cost over twice as much!
 
You must log in or register to reply here.
Back
Top