CPU bound on mass effect, E8400 @ 3.6ghz

[AzN]

So it does. Good info. The difference is there but not a huge one like going from single core to a dual core.

Phenom isn't so bad for UT3 considering a 2.6ghz is neck and neck with Q6600 @ 2.4ghz.

 

[schneiderguy]

Originally posted by: Azn

Originally posted by: schneiderguy

Originally posted by: Azn
We know that quad cores have more cache than a dual core. This could be the real reason why it performs 5-10% faster than dual cores.

The e6750 and q6700 you were comparing have the same amount of cache per core.

no. Q6700 shares the cache between cores which has 8 meg cache. While Q6750 has 4megs.

No it doesn't. The two pairs of cores each have access to 4mb of cache. The one pair of cores on the E6750 has access to the same amount.

 

[BFG10K]

Originally posted by: Azn

Not necessarily CPU bound. It could just be overhead of the game.

You're dodging the issue. The point is that you don't have to get no change in the scores to show you?re CPU limited, the performance just has to be disproportional to the workload or hardware change.

When Anandtech did their article the CPU clock speeds wasn't generating real performance gains. Only thing it mattered was architectural and cache.

So I'll ask you for the fourth time why you're asking Taltimir to underclock his CPU when according to your claim CPU clock speeds won?t make much difference?

Do you understand how your comments contradict each other?

I don't think UT3 engine is optimized for quad cores at all...

You'd be wrong then. In fact it's one of the few engines out there were a slower quad can beat a faster dual.

It's not magical? It's just the way a game was programmed. I've seen this happen in some games over the years. It doesn't matter how much gpu power you pump into the game and raw CPU mhz doesn't change a thing either.

You have absolutely no evidence to suggest this is happening in Mass Effect. The external limit (60 FPS game cap) appears to not affect Taltimar so the most obvious answer then becomes a CPU limitation.

Furthermore asking him to underclock his CPU while claiming MHz doesn?t impact the UT3 engine is laughable. It?s like you?re trying to set up a false test so you can say ?I told you so?.

 

[BFG10K]

Originally posted by: Azn

So it does. Good info. The difference is there but not a huge one like going from single core to a dual core.

Phenom isn't so bad for UT3 considering a 2.6ghz is neck and neck with Q6600 @ 2.4ghz.

See Azn, this is exactly why people get annoyed. Even when your arguments are proven wrong you respond with irrelevant rhetoric to occlude the issue and weasel out of the situation.

 

[AzN]

Originally posted by: schneiderguy

Originally posted by: Azn

Originally posted by: schneiderguy

Originally posted by: Azn
We know that quad cores have more cache than a dual core. This could be the real reason why it performs 5-10% faster than dual cores.

The e6750 and q6700 you were comparing have the same amount of cache per core.

no. Q6700 shares the cache between cores which has 8 meg cache. While Q6750 has 4megs.

No it doesn't. The two pairs of cores each have access to 4mb of cache. The one pair of cores on the E6750 has access to the same amount.

Not quite. I've seen where a quad core is actually faster than dual core even though it's not quad optimized. This only leads to cache differences.

 

[AzN]

Originally posted by: BFG10K
You're dodging the issue. The point is that you don't have to get no change in the scores to show you?re CPU limited, the performance just has to be disproportional to the workload or hardware change.

I'm not dodging anything. If anything I'm telling you there are different things than just CPU that could affect performance.

So I'll ask you for the fourth time why you're asking Taltimir to underclock his CPU when according to your claim CPU clock speeds won?t make much difference? Do you understand how your comments contradict each other?

For the second time, I told him to downclock so he can check it himself. You keep pointing the finger like it was me who said CPU clock speeds won't make the difference but it was Anandtech's article that did that I was basing my opinion on. :roll: You are quite annoying with your questions which I've already answered in previous post.

You'd be wrong then. In fact it's one of the few engines out there were a slower quad can beat a faster dual.

Kind of like when you said SP was going to make the biggest difference? :laugh: Which I corrected you. Then your flame fest started a 20 page threads? :laugh:

I was basing my opinion on original UT3 Anandtech's article. It could just be EPIC that patched the game to use more than 2 cores in the later analysis Anandtech did.

You have absolutely no evidence to suggest this is happening in Mass Effect. The external limit (60 FPS game cap) appears to not affect Taltimar so the most obvious answer then becomes a CPU limitation. Furthermore asking him to underclock his CPU while claiming MHz doesn?t impact the UT3 engine is laughable. It?s like you?re trying to set up a false test so you can say ?I told you so?.

Kind of like you have no evidence to suggest CPU is bottlenecking in 1920x1200 resolution with 4850 in mass effect. I'm just giving ideas to the equation. It's not like I thoroughly tested this game with many different CPU's with a 4850 to know. That's why I used words like it "COULD" be. I love it how you didn't read my previous post about this issue and than point fingers.

Originally posted by: Azn
Tatamir, some games do have an overhead not necessarily CPU bound like you think. Frame rates doesn't increase much when you lower or raise resolution long as you have enough GPU power to run the game. Only when the GPU runs out of steam in some resolution it is limited to does it start drop frames. You did gain when you did lower resolution. Not a huge one but it did improve non the less. Mass Effect could be one of those games.

Is this what it's all about? When you've been proven that 1 time and you've been told you so. Keeping grudges are we? :laugh:

Jesus you even did that 8800ultra analysis just to prove yourself wrong on all the previous posts about how SP makes the biggest differencee. :Q

 

[AzN]

Originally posted by: BFG10K

Originally posted by: Azn

So it does. Good info. The difference is there but not a huge one like going from single core to a dual core.

Phenom isn't so bad for UT3 considering a 2.6ghz is neck and neck with Q6600 @ 2.4ghz.

See Azn, this is exactly why people get annoyed. Even when your arguments are proven wrong you respond with irrelevant rhetoric to occlude the issue and weasel out of the situation.

Again I based my info on Anandtech's UT3 article. In that particular article UT3 didn't really increase with Quad cores. In a much later article N7 pointed UT3 does. Probably a patch.

http://www.anandtech.com/video/showdoc.aspx?i=3127&p=7

Stop pointing fingers BFG. You are probably the worst 1 here coming off with irrelevant rhetoric to weasel out of the situation when you've been proven wrong on multiple accounts.

At least I acknowledge when I'm mistaken but you? :laugh:

 

[taltamir]

Originally posted by: Azn

Originally posted by: schneiderguy

Originally posted by: Azn

Originally posted by: schneiderguy

Originally posted by: Azn
We know that quad cores have more cache than a dual core. This could be the real reason why it performs 5-10% faster than dual cores.

The e6750 and q6700 you were comparing have the same amount of cache per core.

no. Q6700 shares the cache between cores which has 8 meg cache. While Q6750 has 4megs.

No it doesn't. The two pairs of cores each have access to 4mb of cache. The one pair of cores on the E6750 has access to the same amount.

Not quite. I've seen where a quad core is actually faster than dual core even though it's not quad optimized. This only leads to cache differences.

So you have SEEN quad cores be faster, you just attribute it to them having more cache, even though they DO NOT because the cache is not shared (like in a multi GPU video card).

 

[GaiaHunter]

Thanks for the info and sorry for not returning earlier but having some ISP issues.

Thing is I need to decide between a laptop or a desktop. I can only afford the laptop + up to 2x 4850 or a new desktop system. My current desktop is an x2 6000 (yeah yeah but I got it pretty cheap on some best deal stuff some time ago).

My current mobo wont handle CF but I can get one from a friend free ( he got an Intel Quad) and add the second 4850 later if I need to invest some more cash on the laptop.

If I go some E8400 or better and need to buy decent mobo and what not my options on the laptop will dwindle.

And my 7900GT is begging for retirement.

Anyway I think I will grab my friend mobo and slap a 4850 in there or a pair of them as they are dirty cheap and buy a decent laptop.

 

[RussianSensation]

Originally posted by: chizow

If you look at the 3GHz results its very clear that 4870 CF is not scaling well compared to a single 4870 and results in the same performance as 4850 CF even at 2560. Now look at the 4GHz results and you'll see the story is entirely different, with a single 4870 outperforming the 4870 in CF @ 3GHz at 1920 and 4870CF @ 4GHz distancing itself from 4850CF.

True enough but in a game that isn't graphically intensive to begin with. I can play UT3 at 1920x1080 with every option in the menu on HIGH on a 8800GTS 320mb! So it would be similar to me showing a benchmark of Far Cry 1 and telling you it's being 'bottlenecked' at 100fps.....

Not all games will be equally gpu limited. Obviously games like HL2 are very cpu limited with today's graphics cards. But to begin with you woulnd't consider buying 4850/4870s in CF to play those types of games. So the point is rather irrelevant. You buy CF setups for Oblivion, GRID, Crysis and so on where you'll get far greater benefit from adding a 2nd card than overclocking C2Q from 3.0ghz to 4.0ghz.

 

[RussianSensation]

Originally posted by: taltamir

So you have SEEN quad cores be faster, you just attribute it to them having more cache, even though they DO NOT because the cache is not shared (like in a multi GPU video card).

Benchmarks

"Quake 4 supports only dual-core processors. True, Core 2 Duo E6850 with higher clock frequency defeats the competing quad-core solution. However, overclocking changes the situation dramatically. Although overclocked Core 2 Duo E6850 works at a little higher frequency than the overclocked Core 2 Quad Q6600, it is the quad-core CPU that wins here. The determinative factor in this case is the two L2 caches with the total capacity of 8MB." Perhaps Xbitlabs is incorrect in their assessement?

Your earlier point on minimum framerates differences with various CPU speeds is a good point.

------------------------------------------------------------------------------------------

It appears Mass Effect does not play well with ATI cards - Mass Effect benches. It could be that its engine scales better with NVidia and/or ATI's drivers aren't optimized at this point but the game is clearly not cpu bound at 1920x1200 4AA+AF. You get double the framerates by switching from 4850 to GTX 280 (and that's on a 3.0ghz C2Q not 4.0ghz!) and a 34% boost from 4850 to 4870 (in line with bandwidth and processing power benefits of 4870). Even on NVidia's side 9800GX2 scales 50% from the stock 9800GTX.

However, you can see most of the cards approach 58-61 frames at 1920x1200 without AA.

 

[chizow]

Originally posted by: RussianSensation

Originally posted by: chizow

If you look at the 3GHz results its very clear that 4870 CF is not scaling well compared to a single 4870 and results in the same performance as 4850 CF even at 2560. Now look at the 4GHz results and you'll see the story is entirely different, with a single 4870 outperforming the 4870 in CF @ 3GHz at 1920 and 4870CF @ 4GHz distancing itself from 4850CF.

True enough but in a game that isn't graphically intensive to begin with. I can play UT3 at 1920x1080 with every option in the menu on HIGH on a 8800GTS 320mb! So it would be similar to me showing a benchmark of Far Cry 1 and telling you it's being 'bottlenecked' at 100fps.....

Not all games will be equally gpu limited. Obviously games like HL2 are very cpu limited with today's graphics cards. But to begin with you woulnd't consider buying 4850/4870s in CF to play those types of games. So the point is rather irrelevant. You buy CF setups for Oblivion, GRID, Crysis and so on where you'll get far greater benefit from adding a 2nd card than overclocking C2Q from 3.0ghz to 4.0ghz.

If you were getting 100FPS then you wouldn't need to upgrade. But since you're not, the only situation you'd fully benefit from a 4870CF with a slower CPU is in heavily GPU bound situations, which only occurs at 2560 even in the games you pointed out. At that point you'd need to ask yourself if 4870CF or 4870X2 is worth it over 4850CF, or even 9800GTX SLI or 9800GX2 since you'll see they all perform about the same in CPU bottlenecked situations. You can see this in AT's benches even in the "intensive" games you mentioned. Minimal difference at 1680 and 1920 due to CPU bottlenecking.

Even some of the most intensive games out there like Crysis and Age of Conan will not scale beyond a certain FPS due to CPU limitations. Crysis is easy to see as you'll never see a benchmark with 100FPS even at very low resolutions. Age of Conan is even more obvious as 2x, 4x and 8xAA are free at 1920x1200 running ~52FPS on a 3.66GHz C2Q. So people who are complaining they can't get higher than say, 45FPS in AoC. There's a reason, its either heavily CPU limited or there's an external frame cap. Nothing will raise this frame cap except for a faster CPU (or removal of the limit if its something the Devs built into the engine).

 

[GaiaHunter]

Do the new GPU architectures really take advantage of lower resolutions?

Before AA and increasing resolution had big impacts on performance. Nowadays until you get into very very high resolutions you dont see performance impacts or very small ones.

So not getting high performance boosts by turning AA off and lowering resolutions is because of CPU bottle necking or because the new GPUs are build in a way to deal with higher resolutions?

After all you only need to guarantee a certain level of FPS in lower resolutions. I dont think AMD or Nvidia will sell more GPUs because they get 200+ FPS at 1024x768 as opposed to 100.

 

[AzN]

Originally posted by: taltamir

Originally posted by: Azn

Originally posted by: schneiderguy

Originally posted by: Azn

Originally posted by: schneiderguy

Originally posted by: Azn
We know that quad cores have more cache than a dual core. This could be the real reason why it performs 5-10% faster than dual cores.

The e6750 and q6700 you were comparing have the same amount of cache per core.

no. Q6700 shares the cache between cores which has 8 meg cache. While Q6750 has 4megs.

No it doesn't. The two pairs of cores each have access to 4mb of cache. The one pair of cores on the E6750 has access to the same amount.

Not quite. I've seen where a quad core is actually faster than dual core even though it's not quad optimized. This only leads to cache differences.

So you have SEEN quad cores be faster, you just attribute it to them having more cache, even though they DO NOT because the cache is not shared (like in a multi GPU video card).

It's not a multi gpu card first off and it doesn't work that way.

Here's an example how it might work with a quad core that is only dual core optimized.

Let say core 1 and 2 that share 4 meg cache and the cores 3 and 4 that share the other 4 meg cache. You can use core 1 to utilize all 4 meg cache and use core 3 to utilize the other 4meg cache. Which is total of 8meg cache to optimize a dual core only game.

Dual core has only 4 meg cache to share between the 2 cores. Which will always be limited to 4meg cache.

 

[AzN]

Originally posted by: chizow

Originally posted by: RussianSensation

Originally posted by: chizow

If you look at the 3GHz results its very clear that 4870 CF is not scaling well compared to a single 4870 and results in the same performance as 4850 CF even at 2560. Now look at the 4GHz results and you'll see the story is entirely different, with a single 4870 outperforming the 4870 in CF @ 3GHz at 1920 and 4870CF @ 4GHz distancing itself from 4850CF.

True enough but in a game that isn't graphically intensive to begin with. I can play UT3 at 1920x1080 with every option in the menu on HIGH on a 8800GTS 320mb! So it would be similar to me showing a benchmark of Far Cry 1 and telling you it's being 'bottlenecked' at 100fps.....

Not all games will be equally gpu limited. Obviously games like HL2 are very cpu limited with today's graphics cards. But to begin with you woulnd't consider buying 4850/4870s in CF to play those types of games. So the point is rather irrelevant. You buy CF setups for Oblivion, GRID, Crysis and so on where you'll get far greater benefit from adding a 2nd card than overclocking C2Q from 3.0ghz to 4.0ghz.

If you were getting 100FPS then you wouldn't need to upgrade. But since you're not, the only situation you'd fully benefit from a 4870CF with a slower CPU is in heavily GPU bound situations, which only occurs at 2560 even in the games you pointed out. At that point you'd need to ask yourself if 4870CF or 4870X2 is worth it over 4850CF, or even 9800GTX SLI or 9800GX2 since you'll see they all perform about the same in CPU bottlenecked situations. You can see this in AT's benches even in the "intensive" games you mentioned. Minimal difference at 1680 and 1920 due to CPU bottlenecking.

Even some of the most intensive games out there like Crysis and Age of Conan will not scale beyond a certain FPS due to CPU limitations. Crysis is easy to see as you'll never see a benchmark with 100FPS even at very low resolutions. Age of Conan is even more obvious as 2x, 4x and 8xAA are free at 1920x1200 running ~52FPS on a 3.66GHz C2Q. So people who are complaining they can't get higher than say, 45FPS in AoC. There's a reason, its either heavily CPU limited or there's an external frame cap. Nothing will raise this frame cap except for a faster CPU (or removal of the limit if its something the Devs built into the engine).

That would depend entirely on the games. Games like Age of Conan or Crysis is not one of those games that is limited to CPU because when you turn down GPU settings or go crossfire the frames rates increase tremendously. Mass effect is also much more gpu heavy than the original UT3. UT3 the game is showing it's age and was coded to scale much better with cheap GPU's at the time.

Also free AA with Age of Conan 4870x2 doesn't prove anything. R700 is that much more efficient with AA. Show some scaling with GTX280 than it might actually mean something. 280gtx is actually spitting out less FPS at the same setting which tells us this game is still GPU limited.

 

[chizow]

Originally posted by: Azn
At least I acknowledge when I'm mistaken but you? :laugh;

LMAO. I'm sure BFG will have something to add to the rest of what you wrote, but I'm not going to reply to any more of your nonsense until you do as you say, put on your Smart Guy X-Ray glasses and show us what its like to experience the "life of person who can see a bit more than others" and admit to being wrong with the following:

1) Posting some link to an AT article in this thread that you didn't and still don't understand which has nothing to do with your claim that ME isn't CPU intensive.

2) Claiming the 3870 is a faster card than the 9600GT in Age of Conan based on "Highest Playable" even when the "Apples to Apples" comparison showed the 9600GT was faster at a higher resolution and higher in-game settings.

3) Claiming the 9800GTX was much faster than the 8800GTX even when linked benchmarks showed them within a few FPS both ways. Ironically now you're claiming the 8800GTX and 4850 are at the same performance level, which by association means you're saying the 9800GTX is faster than the 4850. I'm sure there's quite a few here that would disagree with you.

I could go on but we can start with that. In the "gangsta life I live" you would've been asked to leave the conference room long ago, or at the very least muzzled if you consistently showed similar incompetence.

 

[RussianSensation]

Originally posted by: chizow

If you were getting 100FPS then you wouldn't need to upgrade. But since you're not, the only situation you'd fully benefit from a 4870CF with a slower CPU is in heavily GPU bound situations, which only occurs at 2560 even in the games you pointed out. At that point you'd need to ask yourself if 4870CF or 4870X2 is worth it over 4850CF, or even 9800GTX SLI or 9800GX2 since you'll see they all perform about the same in CPU bottlenecked situations.

But this happens every single generation! Wait 6-8 months and there will be another Far Cry, another Oblivion and another Crysis and your 4850CF setup will be utilized to its fullest and C2Q 3.0ghz will be more than sufficient to feed it. If you dont benefit from those cards now, yes you are cpu limited, but not Bottlenecked! Just look at the benches I posted above where GTX 280 doubles 4850's frames. The game is clearly GPU limited. Of course at some point you are always either gpu or cpu limited. It's always been said that CF/SLi setups are only useful for highest resolutions and AA possible. There is nothing new here. Yet 4850CF still outperforms GTX 280 in a lot of games despite this "CPU bottleneck" in resolutions lower than 2560x1600 with a "slow" 3.0ghz Quad.

Also consider the benches you linked of 3.0ghz vs. 4.0ghz Quad. We don't know at which point the limitation eases - is it 3.4ghz, 3.6, 3.8ghz or you truly need 4.0ghz? Regardless who is complaining about getting 100 vs. 120 frames in 4870CF ? If you don't need it, save your $.

People always say cpu is bottlenecking and time and time again the cpu continues to outlast 2-3 generations of graphics cards. It's funny, first people complain they are getting 15fps in Crysis at 1920x1200 on a $500 8800GTX, then they complain the game is stuck at 60FPS at 1920x1200 on a $150 graphics card...

 

[AzN]

Originally posted by: chizow

Originally posted by: Azn
At least I acknowledge when I'm mistaken but you? :laugh:

LMAO. I'm sure BFG will have something to add to the rest of what you wrote, but I'm not going to reply to any more of your nonsense until you do as you say, put on your Smart Guy X-Ray glasses and show us what its like to experience the "life of person who can see a bit more than others" and admit to being wrong with the following:

1) Posting some link to an AT article in this thread that you didn't and still don't understand which has nothing to do with your claim that ME isn't CPU intensive.

2) Claiming the 3870 is a faster card than the 9600GT in Age of Conan based on "Highest Playable" even when the "Apples to Apples" comparison showed the 9600GT was faster at a higher resolution and higher in-game settings.

3) Claiming the 9800GTX was much faster than the 8800GTX even when linked benchmarks showed them within a few FPS both ways. Ironically now you're claiming the 8800GTX and 4850 are at the same performance level, which by association means you're saying the 9800GTX is faster than the 4850. I'm sure there's quite a few here that would disagree with you.

I could go on but we can start with that. In the "gangsta life I live" you would've been asked to leave the conference room long ago, or at the very least muzzled if you consistently showed similar incompetence.

LMAO. I can't wait until BFG weasel out of that one. :laugh:

Hypocrite much? You say you aren't going to reply yet here you are replying to my post.

1. Already answered in this thread. Reading comprehension FTW!

2. If it's giving better graphic details than 9600gt and faster. Sure why not.

3. You have some major reading comprehension problems I see. Point to where I said "9800gtx is much faster than 8800gtx" I think it was you who said 9800gtx is slower than G80 because you had 8800gtx with major inferiority complex towards G92 at the time. I only mentioned 9800gtx is faster in lower resolutions without AA and G80 is faster with AA in some ridiculous higher resolutions. :brokenheart:

Your gangster life? Tough guy you. :laugh: You sell drugs and shoot guns for chump change?

Why don't you explain to me why 260gtx wasn't able to crush 4870 since you say even had a thread about ROP is the biggest factor when it comes to performance.

 

[chizow]

Originally posted by: RussianSensation
But this happens every single generation! Wait 6-8 months and there will be another Far Cry, another Oblivion and another Crysis and your 4850CF setup will be utilized to its fullest and C2Q 3.0ghz will be more than sufficient to feed it. If you dont benefit from those cards now, yes you are cpu limited, but not Bottlenecked!

That's not true at all though, especially when that 3.0GHz isn't enough to push a 4870CF's performance higher than the 4850CF set-up. Honestly what you're arguing doesn't make sense, if it comes down to semantics over "cpu limited" vs. "cpu bottlenecked" then it doesn't make sense to continue arguing.

Just look at the benches I posted above where GTX 280 doubles 4850's frames. The game is clearly GPU limited. Of course at some point you are always either gpu or cpu limited. It's always been said that CF/SLi setups are only useful for highest resolutions and AA possible. There is nothing new here. Yet 4850CF still outperforms GTX 280 in a lot of games despite this "CPU bottleneck" in resolutions lower than 2560x1600 with a "slow" 3.0ghz Quad.

Comparing the GTX 280 to a single 4850 makes no sense, especially if you're getting into GPU bound situations like 1920+4xAA. The GTX is the faster card without a doubt, how much faster depends on the game. If a 4850 is 60% of a GTX 280 and you CF'd them then of course you'd expect to see an increase over the GTX 280 with good scaling. That just means the GTX 280 isn't hitting whatever frame limits at that resolution and CPU speed. If you want to compare Apples to Apples though, you need look no further than the 4870CF vs 4850CF. When you see 1680 and 1920 no AA and there is no benefit of 4870CF over 4850CF it isn't obvious you're CPU bottlenecked?

Also consider the benches you linked of 3.0ghz vs. 4.0ghz Quad. We don't know at which point the limitation eases - is it 3.4ghz, 3.6, 3.8ghz or you truly need 4.0ghz? Regardless who is complaining about getting 100 vs. 120 frames in 4870CF ? If you don't need it, save your $.

It doesn't ease at some magical interval, it scales. I'm sure you would see an incremental increase as you increased your CPU clockspeed. We're only seeing the advantage of faster CPUs now that there are faster GPU solutions out there, although there were some convincing results before with SLI configs and Crysis (Check Derek's comments in his Crysis Tri-SLI review). I'm quite certain there is not a CPU available short of LNI2 cooled that's fully able to take advantage of things like X2 CrossFireX or GTX 280 Tri-SLI.

People always say cpu is bottlenecking and time and time again the cpu continues to outlast 2-3 generations of graphics cards. It's funny, first people complain they are getting 15fps in Crysis at 1920x1200 on a $500 8800GTX, then they complain the game is stuck at 60FPS at 1920x1200 on a $150 graphics card...

But in this case it has been 2-3 generations and we're still dealing with the same once-fast Core 2. I can guarantee you graphics cards have scaled more in the last 2 years in speed than the Core 2, if 8800GTX SLI was the fastest thing 2 years ago there's already single GPU cards that are nearly its speed that can also be CF/SLI'd. And I'm quite certain you won't get 60FPS with any single card in Crysis, probably not even at very low resolutions and settings.

 

[schneiderguy]

Originally posted by: Azn

It's not a multi gpu card first off and it doesn't work that way.

Here's an example how it might work with a quad core that is only dual core optimized.

Let say core 1 and 2 that share 4 meg cache and the cores 3 and 4 that share the other 4 meg cache. You can use core 1 to utilize all 4 meg cache and use core 3 to utilize the other 4meg cache. Which is total of 8meg cache to optimize a dual core only game.

Dual core has only 4 meg cache to share between the 2 cores. Which will always be limited to 4meg cache.

First of all, developers don't have control over what cores the threads are going to be run on, windows decides that, so there's no guarantee that the threads will each have acess to 4mb cache.

Second, even if windows decides to put the two threads on cores 1 & 3 (so each has acess to 4mb cache), what happens when the threads need to exchange data? If they were able to share cache they could just take a look in there, but instead they have to go out into the system memory to get the data that the other thread is working on, which is slower than just looking in the cache.

 

[RussianSensation]

Originally posted by: chizow

That's not true at all though, especially when that 3.0GHz isn't enough to push a 4870CF's performance higher than the 4850CF set-up. Honestly what you're arguing doesn't make sense, if it comes down to semantics over "cpu limited" vs. "cpu bottlenecked" then it doesn't make sense to continue arguing.

Just like X2 5200+ wasn't fast enough to push HL2 and 8800GTX in HL2 compared to a Q6600 at 3.0ghz..... You could have said easily back then X2 5200+ is the bottleneck. Would you tell someone you shouldn't waste your $ on getting a 4870 with a 5200+ because it'll bottleneck it over 8800GTX?

That's the exact same thing you are talking about now. You are describing being "cpu limited" (i.e. if you add a slightly faster cpu, you get slightly more performance). Mass Effect is GPU bottlenecked with a cpu limitation, it's plain and simple. You can get a greater increase in performance from upgrading to a new gpu except if you are already running something like a 4850CF setup. So in this case you can technically call it "CPU bottleneck" for that one specific game. However, there is no way I can conclude that since we do not know how well the drivers have been programmed to scale with 4870CF vs. 4850CF in this game -- yet you seem to be 100% certain we are looking at a cpu bottleneck. How about this?

GRID 2560x1600 4AA
4850 CF = 29.5
4870 CF = 30.4

If one were to just look at these 2 numbers in isolation with your logic, you'll quickly say "Oh, there is barely any performance improvement because the cpu is the bottleneck" Yet once we introduce GTX 280 SLI or 4870 X2, they completely outperform the CF setups:

GTX 280 SLI = 68.6
4870 X2 = 84.2

So please don't make statements like C2Q 3.0ghz is the "bottleneck" in a game as if it's 100% FACT.

Comparing the GTX 280 to a single 4850 makes no sense, especially if you're getting into GPU bound situations like 1920+4xAA.

Of course it makes sense. If you want to prove if a game is GPU limited, you compare GPUs not cpus..... You are just pointing out to 1 example when the game might be cpu limited after you throw a 4850CF setup on it...which only affects 5% of all gamers. Moreso, neither you nor i can say with certainty how much the 4870CF setup is being bottlenecked by a cpu (and at what speed precisely) vs. a driver issue. Look at any 4850 Cf vs. 4870 CF benches and you'll see very little improvement in some yet GTX 280 SLI outperforms them both! That's GPU limited my friend. Just because CF doesn't scale, doesn't mean the game is cpu limited.

If you want to compare Apples to Apples though, you need look no further than the 4870CF vs 4850CF. When you see 1680 and 1920 no AA and there is no benefit of 4870CF over 4850CF it isn't obvious you're CPU bottlenecked?

No, it's not obvious. Please see my links above. You cannot assume 100% causation to a cpu.

It doesn't ease at some magical interval, it scales. I'm sure you would see an incremental increase as you increased your CPU clockspeed.

There have been plenty of cases in the past where XP 2500+ was not sufficient, but once you scaled your cpu beyond XP 3000+, the benefits started to subside in a game beyond. So you cannot just assume the framerates will scale linearly from 3.0ghz to 4.0ghz to "prove" that you "need" 4.0ghz C2Q...

Look Here

E6700 @ 2.66 - 146fps
E6700 @ 2.93 - 147fps (barely any!)
E6700 @ 3.30 - 153fps
E6700 @ 3.60ghz - 151fps
E6700 @ 4.0ghz - 150fps
E6700 @ 4.2ghz - 153fps

Now imagine I linked you 2.0ghz E6700 and just 4.2ghz benches you wouldn't be able to conclude that 3.30ghz is sufficient to achieve same frames as a 4.2ghz system, would you? So your argument that 4.0ghz is required might be correct but it might not be since we do not have proper scaling graph (not to mention potential driver issues with CF).

But in this case it has been 2-3 generations and we're still dealing with the same once-fast Core 2. I can guarantee you graphics cards have scaled more in the last 2 years in speed than the Core 2, if 8800GTX SLI was the fastest thing 2 years ago there's already single GPU cards that are nearly its speed that can also be CF/SLI'd.

Yes but you games have gotten more complex as well. So far we havent said what isnt known - for CF/SLI - you want the fastest cpu you can afford. For everything else you almost always are GPU limited. Even then, you can expose the flaw in this argument since SLI GTX 280s STILL outperform 4870s in CF despite the same 3.0ghz cpu which means you are MORE gpu bottlenecked....

 

[chizow]

Originally posted by: RussianSensation

Originally posted by: chizow

That's not true at all though, especially when that 3.0GHz isn't enough to push a 4870CF's performance higher than the 4850CF set-up. Honestly what you're arguing doesn't make sense, if it comes down to semantics over "cpu limited" vs. "cpu bottlenecked" then it doesn't make sense to continue arguing.

Just like X2 5200+ wasn't fast enough to push HL2 and 8800GTX in HL2 compared to a Q6600 at 3.0ghz..... You could have said easily back then X2 5200+ is the bottleneck. Would you tell someone you shouldn't waste your $ on getting a 4870 with a 5200+ because it'll bottleneck it over 8800GTX?

Like I said the first time no one is saying don't upgrade the GPU first, just don't be surprised if you don't see as big of a difference as you expected if you have a slower CPU or if you don't see any difference compared to a "slower" solution.

I'm not talking about single cards between different generations here, there's obviously room for improvement or SLI/CF would have no purpose. I'm talking about a CF solution that should be 25% faster than another and costs nearly 2x as much that is performing IDENTICALLY in many situations, even up to 1920.

That's the exact same thing you are talking about now. You are describing being "cpu limited" (i.e. if you add a slightly faster cpu, you get slightly more performance).

No I'm not, I'm talking about what are supposed to be the fastest solutions performing the same as slower solutions with slower CPUs, but performing as they should with faster CPUs. There's obviously a huge gap between single and dual GPU solutions, the only cards that can really be compared as single GPU to multi-GPU are the GTX 280 and in some cases the 4870 and GTX 260.

Mass Effect is GPU bottlenecked with a cpu limitation, it's plain and simple. You can get a greater increase in performance from upgrading to a new gpu except if you are already running something like a 4850CF setup. So in this case you can technically call it "CPU bottleneck" for that one specific game.

And I haven't said anything to contradict this. As I said earlier Mass Effect is a bit of both, you'll certainly benefit more from a faster GPU but that doesn't mean you aren't CPU bottlenecked.

Mass Effect with 4870 CrossFire

As you can see, the 4870CF does outperform the 4850CF solution, however, when compared amongst resolutions the 4850CF is CPU bottlenecked up until 1680 and the 4870CF solutions is bottlenecked up to 1920x1200. Based on single 4870 performance and the fact its a very high resolution, 2560 being 95 FPS is probably just coincidence.

However, there is no way I can conclude that since we do not know how well the drivers have been programmed to scale with 4870CF vs. 4850CF in this game -- yet you seem to be 100% certain we are looking at a cpu bottleneck. How about this?

GRID 2560x1600 4AA
4850 CF = 29.5
4870 CF = 30.4

lol c'mon, not only is that clearly a GPU limited resolution and IQ setting, but those results are most likely due to lack of frame buffer.

If one were to just look at these 2 numbers in isolation with your logic, you'll quickly say "Oh, there is barely any performance improvement because the cpu is the bottleneck" Yet once we introduce GTX 280 SLI or 4870 X2, they completely outperform the CF setups:

GTX 280 SLI = 68.6
4870 X2 = 84.2

So please don't make statements like C2Q 3.0ghz is the "bottleneck" in a game as if it's 100% FACT.

Again, both cards have 1GB of frame buffer at a resolution that is historically GPU/frame buffer/bandwidth limited. We won't know for sure until we see tests with faster CPUs though.

Also, if you look at GTX SLI results, you'll find quite clearly that it is much more reliant on CPU cycles compared to 4870CF or X2 as it will often perform *worst* than a single GTX 280 with a slower CPU. The results are even more defined with Tri-SLI, which is why I think ATI has the better multi-GPU solution right now, as it does show better gains even at lower resolutions and with slower CPUs.

Of course it makes sense. If you want to prove if a game is GPU limited, you compare GPUs not cpus..... You are just pointing out to 1 example when the game might be cpu limited after you throw a 4850CF setup on it...which only affects 5% of all gamers.

But I'm not trying to prove a game is GPU limited....there's plenty more than 1 example, there's entire reviews showing 4850 and 4870CF performing nearly identically in many games at resolutions up to 1920.

Moreso, neither you nor i can say with certainty how much the 4870CF setup is being bottlenecked by a cpu (and at what speed precisely) vs. a driver issue. Look at any 4850 Cf vs. 4870 CF benches and you'll see very little improvement in some yet GTX 280 SLI outperforms them both! That's GPU limited my friend. Just because CF doesn't scale, doesn't mean the game is cpu limited.

Which is why I don't put too much emphasis on results from different vendors and solutions with vastly different "expected" levels of performance. This is similar to the example you keep referring to, a single card vs. multi-GPU with a slower CPU. Yes you will see more performance with the 2nd GPU, but will you see as much as with a faster CPU also? Probably not.

No, it's not obvious. Please see my links above. You cannot assume 100% causation to a cpu.

But you can when you increase the CPU clocks and that results in the expected performance gains (as linked in my original reply).

There have been plenty of cases in the past where XP 2500+ was not sufficient, but once you scaled your cpu beyond XP 3000+, the benefits started to subside in a game beyond. So you cannot just assume the framerates will scale linearly from 3.0ghz to 4.0ghz to "prove" that you "need" 4.0ghz C2Q...

Look Here

E6700 @ 2.66 - 146fps
E6700 @ 2.93 - 147fps (barely any!)
E6700 @ 3.30 - 153fps
E6700 @ 3.60ghz - 151fps
E6700 @ 4.0ghz - 150fps
E6700 @ 4.2ghz - 153fps

Now imagine I linked you 2.0ghz E6700 and just 4.2ghz benches you wouldn't be able to conclude that 3.30ghz is sufficient to achieve same frames as a 4.2ghz system, would you? So your argument that 4.0ghz is required might be correct but it might not be since we do not have proper scaling graph (not to mention potential driver issues with CF).

If you're going to link to a 4 year old article, especially one from HOCP, you're really going to have to refine your argument and point out what you're referring to. I'm not going to reset my entire CPU/GPU frame of reference circa 2004 and go at it blind.

The FEAR example is more relevant but only concludes in the obvious, that GPUs will not render infinite frames per second based on CPU speed alone. They are still bound to whatever their architecture limits them to up to their theoretical max.

Yes but you games have gotten more complex as well. So far we havent said what isnt known - for CF/SLI - you want the fastest cpu you can afford. For everything else you almost always are GPU limited. Even then, you can expose the flaw in this argument since SLI GTX 280s STILL outperform 4870s in CF despite the same 3.0ghz cpu which means you are MORE gpu bottlenecked....

Well I don't think game demands, with the exception of Crysis, are scaling as fast as GPU advancements. I think its pretty obvious when the G80/G92 made 1920x1200 a viable resolution (look at G80 reviews, yes they're still using 1280). This next generation along with the commodization of affordable CF/SLI has brought this a step further where many games are showing CPU bottlenecking even at 1920 while offering excellent performance even at 2560 (and with AA with the 4870X2). I think we're going to continue seeing this until 1) we get faster CPUs, 2) Devs start coding games to make better use of multiple cores or 3) some of the effects like physics slowing down the CPU are accelerated on the GPU.

And again with GTX 280 SLI and Tri-SLI, there's more than quite a few reviews that show performance can actually be worst than a single GTX 280 without a very fast CPU. Both Guru3D and Tweaktown have comparisons of 3 and 4GHz solutions out there that confirm this.

 

[AzN]

Originally posted by: schneiderguy

Originally posted by: Azn

It's not a multi gpu card first off and it doesn't work that way.

Here's an example how it might work with a quad core that is only dual core optimized.

Let say core 1 and 2 that share 4 meg cache and the cores 3 and 4 that share the other 4 meg cache. You can use core 1 to utilize all 4 meg cache and use core 3 to utilize the other 4meg cache. Which is total of 8meg cache to optimize a dual core only game.

Dual core has only 4 meg cache to share between the 2 cores. Which will always be limited to 4meg cache.

First of all, developers don't have control over what cores the threads are going to be run on, windows decides that, so there's no guarantee that the threads will each have acess to 4mb cache.

Second, even if windows decides to put the two threads on cores 1 & 3 (so each has acess to 4mb cache), what happens when the threads need to exchange data? If they were able to share cache they could just take a look in there, but instead they have to go out into the system memory to get the data that the other thread is working on, which is slower than just looking in the cache.

I give you that but it's only logical to use full use of CPU cache. I've never used a quad core so I don't know for sure but this is just a educated guess just looking how quad cores seem to have better performance at same clock speed as dual core. Nothing more.

 

[AzN]

Just like X2 5200+ wasn't fast enough to push HL2 and 8800GTX in HL2 compared to a Q6600 at 3.0ghz..... You could have said easily back then X2 5200+ is the bottleneck. Would you tell someone you shouldn't waste your $ on getting a 4870 with a 5200+ because it'll bottleneck it over 8800GTX? That's the exact same thing you are talking about now. You are describing being "cpu limited" (i.e. if you add a slightly faster cpu, you get slightly more performance). Mass Effect is GPU bottlenecked with a cpu limitation, it's plain and simple. You can get a greater increase in performance from upgrading to a new gpu except if you are already running something like a 4850CF setup. So in this case you can technically call it "CPU bottleneck" for that one specific game. However, there is no way I can conclude that since we do not know how well the drivers have been programmed to scale with 4870CF vs. 4850CF in this game -- yet you seem to be 100% certain we are looking at a cpu bottleneck. How about this? GRID 2560x1600 4AA 4850 CF = 29.5 4870 CF = 30.4 If one were to just look at these 2 numbers in isolation with your logic, you'll quickly say "Oh, there is barely any performance improvement because the cpu is the bottleneck" Yet once we introduce GTX 280 SLI or 4870 X2, they completely outperform the CF setups: GTX 280 SLI = 68.6 4870 X2 = 84.2 So please don't make statements like C2Q 3.0ghz is the "bottleneck" in a game as if it's 100% FACT.

Grid is just a bad example. 4850 or 4870 CF is limited to 512mb of frame buffer at that resolution while GTX280 and R700 is not.

Of course it makes sense. If you want to prove if a game is GPU limited, you compare GPUs not cpus..... You are just pointing out to 1 example when the game might be cpu limited after you throw a 4850CF setup on it...which only affects 5% of all gamers. Moreso, neither you nor i can say with certainty how much the 4870CF setup is being bottlenecked by a cpu (and at what speed precisely) vs. a driver issue. Look at any 4850 Cf vs. 4870 CF benches and you'll see very little improvement in some yet GTX 280 SLI outperforms them both! That's GPU limited my friend. Just because CF doesn't scale, doesn't mean the game is cpu limited.

Someone has the right idea. :light:

 

[AzN]

Originally posted by: Azn

Originally posted by: chizow

Originally posted by: Azn
At least I acknowledge when I'm mistaken but you? :laugh:

LMAO. I'm sure BFG will have something to add to the rest of what you wrote, but I'm not going to reply to any more of your nonsense until you do as you say, put on your Smart Guy X-Ray glasses and show us what its like to experience the "life of person who can see a bit more than others" and admit to being wrong with the following:

1) Posting some link to an AT article in this thread that you didn't and still don't understand which has nothing to do with your claim that ME isn't CPU intensive.

2) Claiming the 3870 is a faster card than the 9600GT in Age of Conan based on "Highest Playable" even when the "Apples to Apples" comparison showed the 9600GT was faster at a higher resolution and higher in-game settings.

3) Claiming the 9800GTX was much faster than the 8800GTX even when linked benchmarks showed them within a few FPS both ways. Ironically now you're claiming the 8800GTX and 4850 are at the same performance level, which by association means you're saying the 9800GTX is faster than the 4850. I'm sure there's quite a few here that would disagree with you.

I could go on but we can start with that. In the "gangsta life I live" you would've been asked to leave the conference room long ago, or at the very least muzzled if you consistently showed similar incompetence.

LMAO. I can't wait until BFG weasel out of that one. :laugh:

Hypocrite much? You say you aren't going to reply yet here you are replying to my post.

1. Already answered in this thread. Reading comprehension FTW!

2. If it's giving better graphic details than 9600gt and faster. Sure why not.

3. You have some major reading comprehension problems I see. Point to where I said "9800gtx is much faster than 8800gtx" I think it was you who said 9800gtx is slower than G80 because you had 8800gtx with major inferiority complex towards G92 at the time. I only mentioned 9800gtx is faster in lower resolutions without AA and G80 is faster with AA in some ridiculous higher resolutions. :brokenheart:

Your gangster life? Tough guy you. :laugh: You sell drugs and shoot guns for chump change?

Why don't you explain to me why 260gtx wasn't able to crush 4870 since you say even had a thread about ROP is the biggest factor when it comes to performance.

Since you can't answer your own claims I think I should give you my thought with the whole GT200 architecture which I wrote this in another forum and here when GT200 was released about a month ago.

Nvidia didn't up their texture address and filtering when they reworked their SP with GT200. The old G92 cores had 8 texture address/filter for every 16SP but GT200 has 8 texture address/filter for every 24SP. Those textures did a whole lot more for games than just higher SP clocks with G92 with modern games.

GT200 has the same 8 by 8 texturing ability just like G92 but only 10 clusters of 24 SP instead of 8 by 16SP which equals out to 80tmu. Texture fillrate was the biggest difference when comparing G92 vs G80 and why G92 was able to beat it in lower resolutions or get very close to high resolution with much lower memory bandwidth and less ROP. If they did 12 by 12 which would be the exact same number as G92 SP/texture ratio it would have 120 tmu instead of 80. GT200 is inferior far as texturing ability when you compare ratio to G92.

GeForce 9800 GTX 10.8 pixel fillrate 43.2 bilinear fillrate 21.6 FP16 fillrate 70.4 GB/s

GeForce GTX 260 16.1 pixel fillrate 41.5 bilinear fillrate 20.7 FP16 fillrate 111.9 GB/s

GeForce GTX 280 19.3 pixel fillrate 48.2 bilinear fillrate 24.1 FP16 fillrate 141.7 GB/s

Radeon HD 4850 10.0 pixel fillrate 25.0 bilinear fillrate 12.5 FP 16 Fillrate 64 GB/s

Radeon HD 4870 12.0 pixel fillrate 30.0 bilinear fillrate 15.0 FP 16 Fillrate 115.2 GB/s

Games don't need all that processing power as of yet. Most games off loads to textures and back to the memory for the most part straight from nVidia relayed to nRollo when we had the 9600gt thread. So having more fillrate makes the biggest difference when you want the performance NOW long as you aren't shader limited. Sure 280gtx has more fillrate than 9800gtx but in reality it doesn't have that much more. 260gtx has even less than 9800gtx. This is where bandwidth comes into play with GT200 where it's not so limited compared to 9800gtx which you see the performance gains from most games. Just look at any of the reviews. You will see that 260gtx isn't that far off in performance compared to 9800gtx only when AA is applied in some ridiculous high resolution does it seem like it's more faster because of bandwidth advantages. Nvidia made a future product like 2900xt tried to do. But it's still not happening.

RV770 is just that much more efficient than GT200 considering it has less filtering rates but able to surpass GT200 far as FP16 blending filter rate goes.

My speculations about GT200 about a month ago about texture to SP ratio have also been mentioned in Techreport pod cast by Scott Wasson last Saturday and how it would be able to achieve better performance with GT200. http://techreport.com/articles.x/15103

They talk about GT200 at 0:36:35 of the show.