• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

CPU Level 2 Cache Size - Is Bigger Better?

Soldato
Joined
11 Sep 2003
Posts
14,716
Location
London
Hello,

I was just wondering does anybody have some good info about CPU level 2 cache in terms of how it affects system performance?

Which tasks on the PC are noticably quicker while using a processor featuring a large slab of level 2 cache?

For example I have seen several posts where people state smaller level 2 cache sizes cripple gaming performance but have yet to see some good write-up/reviews where this has been tested?

Thanks in advance for any info! :)
 
Hey Wayne, i heard the same and have wondered this too.

In my experience it does make a little difference but the way C2D's tend to clock most users can recover any lost performance by upping clocks.

I've used the following for gaming:

e1200, e2160 (had 2 of these), e6300, e6420, e6600 (had two of these), q6600 and e8500.

In the case of the lower cached chips (e1200, e2160 and even e6300) i would say that you would need 100-200mhz extra per mb of cache lost to make up the framerate. Indeed for the e1200 you might even need up to 300mhz extra as it really is a cut down example - that said they're still awesome little chips!

For the e6420, e6600 and even e8500 the difference was far more marginal clock for clock. I tended to look at it more from the point that it simply came down to how far the chip happily clocked and then have fun! :)

The q6600 was similar to the higher cached models above but with multi threaded games offered the obvious extra performance that the additional cores added.

I'd be interested in seeing actual results though too if anyone has any. :)

gt
 
Couple of months ago I went from an X6800 (4MB L2) to Q9450 (12Mb L2).

Did not bench before and after with games but the difference was/is very noticeable to me. Everthing ingame like accessing menus or maps is now instant whereas before there was a slight delay. Now I can select a map screen or ingame menu which appears as soon as I select it. So my perception is that my gaming is much smoother. Not sure the FPS is much better as I am on a lower clock 2.66 vs 2.93. Thingsa which are Quad Core friendly like Assassins Creed, Devil May Cry4 & Lost Planet seem much smoother than before.

Same for Windows Vista now everything is instant due to the extra L2 cache. Q9450 is even quicker in Vista than my QX6850 (8MB L2 cache).
 
I'd be interested in seeing actual results though too if anyone has any. :)
Thanks for reply. I'm not sure it's the kind of review a commercial hardware website would look into as if the finding proved the Level 2 cache didn't make that much difference people may stop buying the more expensive CPU's and that wouldn't please the websites sponsors!

I'm sure there are a few areas where the extra cache makes a big difference but personally I'm not sure where? I'm guessing predictive tasks like rendering and video encoding would see a nice gain?

Couple of months ago I went from an X6800 (4MB L2) to Q9450 (12Mb L2).

Did not bench before and after with games but the difference was/is very noticeable to me.

So my perception is that my gaming is much smoother.

Oh it's a pity you wasn't able to produce a few benchies as that would have been a good read . . . although in your scenario you have doubled the cores as well as the cache so the difference would be harder to gauge i.e the extra snappiness you descibe may be more to do with the extra cores than the extra cache!

I'm looking at the cheaper Wolfdale chips with 3MB cache vs the full blown 6MB cache jobbies and trying to gauge if the extra cache is worth the large premium . . .
 
I'm looking at the cheaper Wolfdale chips with 3MB cache vs the full blown 6MB cache jobbies and trying to gauge if the extra cache is worth the large premium . . .
I would say get the 3mb.

From what i saw only the chips with less than 2mb took a big (if you could really even call it that) hit.

I'd imagine an e7200 or similar clocked up to and beyond 3.5ghz would be a mighty impressive gaming chip.

gt
 
Larger caches often have higher access latency, so when you go from a 3 meg L2 to a 12 meg L2 the latency will often increase. The is because tag checking needs to be done on more entries within the cache (as the cache is larger) but having more tag checking hardware (to ensure the same latency) will take up silicon and power so it's often not done...

Also, people should really be comparing the set associativity of their caches since, for example a 1 meg 8-way associative cache often performs just as well as a 2 meg 4-way cache. 8-way requires twice as much tag checking as 4-way. A 2 meg 8-way will often perform as well as a 4 meg 4-way, etc etc.

I can't give game benchmarks, but SPEC results ;)

(if in doubt go with the largest cache!)
 
Us Folding types have found that cache can make a huge difference. A 4mb E6600 will be almost twice as fast at crunching as a 1mb cache E2200 when both chips are running at the same speed. So that's not something that an extra 200mhz on the overclock will overcome.

But then it's been claimed that cache makes little difference to games. Someone on here did a write-up on it and came to the conclusion that it made little difference.

So things like games won't really make use of more cache, but things like video editing/converting will make use of it.
 
Back
Top Bottom