i5 750@4Ghz=crossfire 6950 bottleneck?

[metilic]

Could an i5 @4 Ghz bottleneck 2x6950 2gb? Or what could be the cause of Battlefield BC2 negative scaling? The temps of the gpu's are around 65 C, so i see no temperature bottleneck. Should i overclock the CPU more? Show your creativity here my friends, I'm really stuck

p.s. I game at 1680x1050, and scaling in other games like Metro 2033, AVP, Crysis 2 is great.

 

[jasper1605]

How many fps are you getting in the game?
2 6950s with a cpu at that speed should be perfectly fine for that res. Unless you're just trying to go from 100 to 130 fps.

 

[metilic]

I get fps drops to 35-40 then up to 80 and up to 100+ and i don't know why...

 

[cadaveca]

It's not really the CPU itself, it's the low memory bandwidth. Even though 1100T is faster than FX-8150 in alot of benchmarks, FX-8150 should be better for crossfire, simply due to the extra bandwidth it's got.

Upgrading to a SandyBridge rig should give a signifigant boost to 3D performance, for example.

 

[NdMk2o1o]

What FPS do you get with a single card? when it dips down to 35-40fps what is happening on screen? have you updated to latest drivers/reverted to earlier ones? is the game fully patched? what kind of ping are you getting to the
game servers.

It's not really the CPU itself, it's the low memory bandwidth. Even though 1100T is faster than FX-8150 in alot of benchmarks, FX-8150 should be better for crossfire, simply due to the extra bandwidth it's got.

Upgrading to a SandyBridge rig should give a signifigant boost to 3D performance, for example.

Wrong thread? he has an i5 760@4ghz memory bandwidth is not an issue.

 

[metilic]

Stupid frostbite engine then...hope that bf3 wont pull a fast one on me too all games scale very good except stupid retarded BC2...

 

[metilic]

80+ with a single card..tried different drivers..ping is very good..no artifacts on screen during fps drop.

 

[cadaveca]

Wrong thread? he has an i5 760@4ghz memory bandwidth is not an issue.

Actually, in multicard, it is. Don't forget what I do here @ TPU. Benchmarks are just a couple clicks away.

Even the A8-3850 APU beats 1156 chips in memory-intensive apps:

Stupid frostbite engine then...hope that bf3 wont pull a fast one on me too all games scale very good except stupid retarded BC2...

You could just ignore the problem. I quit playing BC2 for several months due to it crashing A LOT...recent patch seems to have made things better(they fixed memory leaks).

 

[NdMk2o1o]

Actually, in multicard, it is. Don't forget what I do here @ TPU. Benchmarks are just a couple clicks away.

Not in this case it isn't and especially when you got the CPU wrong he was using so I thought you had the wrong thread by mistake, and I could care less what you do what has that got to do with it?

 

[cadaveca]

Not in this case it isn't and especially when you got the CPU wrong he was using so I thought you had the wrong thread by mistake, and I could care less what you do what has that got to do with it?

I did not get the wrong thread, you did!


What does it got to do with it, you want to know? I ran tests on Saturday for the OP. Would you like to see our conversation in PM?

You do understand i have a near duplicate of his rig here, right? You've got wrong info. You neither use ATI cards, nor understand how Crossfire works, clearly.

Like no offense dude, but clearly you're just spouting hot air here. Technically, i run benchmarks and evaluate performance for a living. That's how what I do here @ TPU relates.

 

[NdMk2o1o]

I did not get the wrong thread, you did!


What does it got to do with it, you want to know? I ran tests on Saturday for the OP. Would you like to see our conversation in PM?

So where in the hell do you get an 1100t from, that was my point and if you have something rlevant to share then share it unless the rest of us are just meant to guess what your talking about then?

I have had 3 crossfire rigs mr big reviewer so please dont tell me what I do or don't understand about, you dont know what I do for a living now do you?

By all means share your wisdom.

OP I am sorry I am out of this thread, hope you get the problem sorted

 

[cadaveca]

The 1100T is eclipsed by even an APU, even though they feature the same CPU core, with differences in CPU cache. I mention the AMD chips because although the FX-8150 is marked as lower-performing compared to a 1100T in singlethreaded apps, it does tend to shine when used in Crossfire. Please see my post above where the APU also beats 1156 system in a memory-intensive app.

It's an example of how memory is important. As I said, it's not the CPU speed that's important.

Moving over to SB provided basically double the bandwidth of other platforms. This allows it to excel in multicard configurations.


This thread is about a 6950 Crossfire bottleneck. In Crossfire, when compared with a SB system, yes the CPU is a bottleneck, but of course, it's not just the CPU causing the bottleneck, it's the differences in memory bandwidth that are the actual choke point.

 

[metilic]

So basically as a conclusion...is it my cpu or not?

 

[cadaveca]

Not 100%, no.

It depends on the app, and how memory intensive it is, and what you are comparing to. Crossfire naturally is more memory-performance sensitive as the system with two cards, not only needs to write data from memory to one card, but now to two.

As you can see by the F1-2010 graph I posted above, in that app specifically(and MOST CodeMasters games), memory bandwidth is ultimately very important. Intel chips are faster than the 1100T and AMD APUs in CPU performance, but the APU wins out of all three simple because of it's better memory performance.

So, really, it's the app and how it behaves that creates that bottleneck, clearly. If every other app in your rig works fine, it's very easy to blame the app only.

 

[NdMk2o1o]

The 1100T is eclipsed by even an APU, even though they feature the same CPU core, with differences in CPU cache. Please see my post above where the APU also beats 1156 system in a memory-intensive app.

Still why you decided to pull an 1100t out of thin air escapes me, do you not understand what I am asking, obviously not. You chose 1 game that is not the same game as the OP has an issue with where the APU beat a 1156 rig by less than 10% it proves nothing.

It's an example of how memory is important. As I said, it's not the CPU speed that's important.

again, 1 example that shows nothing, he is running at 4ghz with 1600mhz ram, was your 1156 result from that 1 bench running the same?

Moving over to SB provided basically double the bandwidth of other platforms. This allows it to excel in multicard configurations.

Again you decide to pick out the bits that support your argument the best, 1155 does NOT offer twice the bandwidth of 1156, I have owned several of both platforms and any bench will also tell you this, twice the bandwidth of 775/AM3 yes

This thread is about a 6950 Crossfire bottleneck. In Crossfire, when compared with a SB system, yes the CPU is a bottleneck, but of course, it's not just the CPU causing the bottleneck, it's the differences in memory bandwidth.

This thread is about low performance with the OP's rig playing a certain game, until your above points actually coincide with each other and we're talking about the same damn CPU and not comparing F1 with an APU and a stock p55 setup running ddr3 most likely, then I might entertain this idea a

 

[cadaveca]

This thread was opened by the OP as a direct result of our conversation over the past week or so. I do have benchmarks, and nearly the same rig as the OP, which I set up over the weekend to test for him. It might not make sense to you personally becuase you were not part of the coversation that has been taking place over the past week.

I don't really care if you personally understand my postings..it's not your thread. I'm not here to explain stuff to you, as clealy it's monday, and perhaps you'd like to argue. I have better things to do...I told the OP that his CPU was bottlenecking his cards. He opened a thread to confirm. All you've done is refute my posts, but have very little to add that proves otherwise, and are now trying to argue with me in PM. Have a good one.

I will add, that at least i know enough to not blow my chips up. You just got yours back from RMA. I do not make such mistakes, nor do I RMA parts that I know I blew up.

 

[NdMk2o1o]

Well I wasn't going to bite but tbh what the hell.

As you mentioned I PM'd you I want to make it clear I am not arguing nor have I onced tried to with you as this is the sum of my PM to you:

Got a problem with me then say it here but tbh I could care less if you were the owner of the site, I am fed up of you responding to me in the manner you do. Learn how to explain yourself before you go troll baiting someone for not having the right information, afterall you should be able to do that consider "what you do here"

Now let me get this right, is it his memory bandwidth or his CPU or the app cause you seem a little confused, heck you seem confused as to what he is actually running as you mention an 1100T out of nowhere....

I don't have anything to "prove" I never made any claims infact in a perfectly calm manner I have asked you for clarification on a number of your posts to which you have danced around and instead tried to imply that I am looking for an argument :shadedshu

Though I shouldn't nor don't expect too much more from you than that cause it is typical of what I have seen from you when someone doesn't agree with you

OP I do not think your CPU is a bottleneck or at least not to that extent, nor is it your bandwidth in my humble opinion anyway, maybe try running a profile to disable 1 card with BF2 if you get better performance with a single GPU?

 

[jasper1605]

This thread was opened by the OP as a direct result of our conversation over the past week or so. I do have benchmarks, and nearly the same rig as the OP, which I set up over the weekend to test for him. It might not make sense to you personally becuase you were not part of the coversation that has been taking place over the past week.

That indeed confused me when you talked about APUs and AMD chips to an i5-750 question lol. But that sentence clarifies that up a bit.

When you say memory bandwidth, you're referring to the speed of the RAM, or is there a deeper, larger power at work that I've neglected in my two years of nerddom?

 

[cadaveca]

When you say memory bandwidth, you're referring to the speed of the RAM, or is there a deeper, larger power at work that I've neglected in my two years of nerddom?

I refer to bandwidth efficiency, so yes, you are very right, I am NOT talking about raw MHz. AMD rigs and SKT1156 run @ about 65% efficiency, while SB is @ like 85%, and this is continuous through all memory dividers.


Not many people even look at efficiency...nor know how.

The OP made a thread a week ago asking why BC2 performance suffered. I told him I'd set up a rig and test to see if I got the same results. I did not.

When telling the OP about it, i did mention that his rig was a bottleneck to the cards, and that SB would increase 3D performance. This applies both to singlecard and multicard configs.

And sorry for the confusion, but I wasn't talking to anyone but the OP, orginally. I brought up the 1100T as an example of how this bandwidth efficiency is important. As nearly no reviewer even mentions it, I understand how this may not be something people are used to looking at.

 

[Anusha]

i do not think the CPU will bottleneck those cards. 4GHz i5 750 is like a stock 2500K (or *slightly* faster). i doubt it will bottleneck CF.

i believe negative scaling is driver related. have you measured FPS with a tool or are you talking about the *feeling*? if it is the feeling, it could well be microstuttering, which means you are getting higher FPS, but frame syncing issues are making it seem like stuttering.

have you measured what your GPU utilization is when playing the game? you can do that using Afterburner.
Go into Settings -> Monitoring tab and choose GPU load

also measure the clock speeds and make sure they are running at the 3D clocks. not underclocked BluRay or 2D clocks. something might cause these them to not run at rated 3D clocks.

are you getting this with every game?

i'm sure it is not the CPU though. that CPU, especially with the 50% overclock, is fast enough for CF. and this is not even the 6970.

 

[cadaveca]

i'm sure it is not the CPU though. that CPU, especially with the 50% overclock, is fast enough for CF. and this is not even the 6970.

If his CPU was not a bottleneck, then moving over to 1155 would have zero benefit in FPS. However, both singlecard and multicard configs get a boost in FPS from the CPU change.


I did mention to the OP that he should look at either heat or something causing throttle, or something like a virus scanner or some app eating CPU clocks, so we are on the same thread in identifying the problem with BC2.

I do not think BC2 being bottlenecked by the CPU is causing his problems in BC2, however.

 

[metilic]

Cpu usage during the game is between 70-85-90% max; GPU usage does not reach 100% on both cards, more like 90% and 40%. I know BC2 is cpu intensive but wtf it's a good quad with a 50% OC...both cards however are at 3D clocks so no problem there, temps are around 65C so no thermal throttling. The fps drops are independent of DX version used and AA/AF. This is really pissing me off, the principle of it I mean. Fps was measured with Fraps. And Anusha, this is the only game i get this crap...all new gen games are pwnd by these cards on max settings.
Driver version is 11.10 preview 3 with 11.9 cap 3 (as recommended by amd, but also used previous versions as well).

Edit for Cadaveca: the cpu usage during BC2 is exclusive for the game, no other app is using cpu; even disabled antivirus and things remain the same)

 

[cadaveca]

Now, I wonder if maybe FRAPS isn't the cause? Did you try using Afterburner's FPS display?

 

[dark2099]

If you haven't installed an AMD Catalyst Application Profile, might want to try that. I would assume that the latest drivers have the in them already, but wouldn't hurt to try.

http://sites.amd.com/us/game/downloads/Pages/crossfirex-app-profiles.aspx

 

[metilic]

I have the latest cap, also tried watching fps with afterburner, same result