Another Question about CPU bottlenecks

[cobhc]

Ok guys, I have my Q6600 running at 3.6 ghz and my 5870 at 950/1300 and the Dirt 2 Bench with everything maxed out (including 8xAA) at 1920x1080 I get an average of 65fps, which is pretty good IMO. However I see plenty of benchmarks online hitting 78 with a similarly clocked 5870, and a similarly clocked i7.

Now I was under the impression that at 1920x1080 and higher (especially with high levels of AA as well) you're a lot more GPU bound, and that I shouldn't be seeing as much difference in frame rates between my CPU and an i7.

I don't have any more benchmarks at the moment to compare, but I'm definitely not seeing similar performance to a lot of benchmarks out there on high image quality settings.

Should I be seeing this amount of difference in performance between my CPU and an i7 setup when running 1920x1080 and 4/8x AA?

 

[cainer]

some vids for ya..

dunno what the rest of his setup was so i dont know if these results would mean much to ya but it shows the i7 beating the q6600 in every test so the difference your getting in the dirt2 bench is probly about right

 

[raralis]

Will you notice any difference between 65fps and 78fps?

Is your min fps dropping below 30?

If your answer is "no" to both, keep your current cpu.

 

[Ste]

Hi,

I think that a lot of the minimum fps scenes are still caused by CPU bottlenecks and that has an effect on the average. I have a 5870 at 900/1300 and just upgraded from a Q6600 @ 3.6 to an i7 920 @ 4.0.

I went from a 41fps minimum (1080p, maxed, 4 AA) to a 59.9 minimum (as before but now 8x AA) by going from the Q6600 to the i7. The 59.9 was with vsync on, too, it's probably higher.

I also wonder whether the old P35 platform represents a bottleneck compared to the X58 (tri channel DDR3 for example).

Either way, I think we're at the point now where 1080p can be considered a low enough resolution to get CPU bottlenecked on the big graphics cards. I have a similar situation in Crysis, too - massive improvement.

 

[Ste]

btw - 60hz is a massive deal in racing games. My laptimes have gone down and I swear it's because of the consistently fluid experience...

 

[coupe69]

See here for an interesting article on CPU bottlenecks. Q6600 @ 3.6GHz is more than enough, you'll not notice any benefit from clocking it any higher.

 

[suj999]

I don't think you will see any bottlenecks now.. but I think in some future games you will, so its about time you start to think about an upgrade, and do the upgrade around the end of the year or Q4

 

[Ste]

They don't compare the older achitecture, though.

On a single 5870 (900/1300) the 3.6 Q6600 -> 4.0 i7 transition has a significant impact on gaming performance at 19x12. At least, in the three games I have tried it does - Dirt 2 (DX11), Crysis, GTA IV.

Note that this is my definition of 'significant' and that others may vary...

 

[cobhc]

Hmm, maybe I should wait till the new Intel chips come out next year. I might look at a better motherboard for my chip though. The Q6600 isn't stable at 3.6ghz really, it's ok for benches, but not much more, and I'm sure it's the board as I can do 3.2ghz on 1.28v with max prime temps of 50c, I think the PWM temps are the problem as they hit 80+ under load.

 

[DK06CTR]

Well I have just gone from a E6750 clocked at 3.6ghz system to a core i7 at 3.9ghz with both systems have had the 5870 with it clocked at 1000 core 1300 memory and with dirt 2 i had around 56fps average and around 40fps minimum. now with the i7 got 81 fps average and 70fps minimum thats with 1920x1080 res ultra settings 8xmsaa so i was having a rather large bottleneck that i wouldnt of expected.

 

[cobhc]

Thanks for the feedback DK.

I really think I'm going to leave it, if I'm honest I can't really afford the £500 or so needed for the upgrade at the moment and I might as well leave it till next year and see what comes up.

 

[Bz786]

yeah cool