Gamer's choice: How much does cache matter? 2mb vs 4mb benchmarks

[Phosphate]

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2802

I'm mainly into gaming, not other benchmarks but check out this. It seems for 3D performance and games only clock speed matters. Actually the 4 mb cache seems to make no difference at all for this application.

Looks like I'll be getting a e6400 for sure now. Anyone with a conflicting article? Remember I'm just talking games and 3D.

[FunkyRider]

the extra 2MB cache is a performance premium, if you don't wanna pay the extra $$$, then it doesn't matter

[Phosphate]

the extra 2MB cache is a performance premium, if you don't wanna pay the extra $$$, then it doesn't matter

Yeah the extra cache has no effect on games (which is all I care about) , but it does on other non-game benchmarks.

So it seems that 2 mb cache is more than enough for these applications.

[BottomsUp]

i went with the E6300 partly because of that article instead of the E6600 i originally wanted. I used the extra $$$ to buy a better video card. It think/hope i made a good decision.

[burningrave101]

Yeah the extra cache has no effect on games (which is all I care about) , but it does on other non-game benchmarks.

So it seems that 2 mb cache is more than enough for these applications.

Gaming is one of the areas that extra cache does make a difference but you'd have to be testing a CPU intensive game to really notice it. With gaming you need to be more worried about buying high-end graphics cards and less worried about getting the fastest CPU since all the heavy load is on the GPU in most games.

[Phosphate]

Gaming is one of the areas that extra cache does make a difference but you'd have to be testing a CPU intensive game to really notice it. With gaming you need to be more worried about buying high-end graphics cards and less worried about getting the fastest CPU since all the heavy load is on the GPU in most games.

I know some games are more CPU intensive, but they tested a lot of games here and no difference. Just look through the test results. When e6400 or e6300 is overclocked to approximately the same Ghz as e6800 there is no difference in frame rates.

Maybe in some kind of really heavy real time strategy there could be a difference but not in stuff like BF2, oblivion, and F.E.A.R. (maybe later on the 4mb cache will matter but for now I can't find evidence that I should be considering this for the stuff I do other than the extra multiplier the e6600 has)

[775911]

In some apps it has a 5% or so effect, not worth worrying about really.

Quite funny that the 'slower' chip gave better performance sometimes due to high fsb benefits

[DeToNaToR.cl]

i went with the E6300 partly because of that article instead of the E6600 i originally wanted. I used the extra $$$ to buy a better video card. It think/hope i made a good decision.


That was a clever decision...

If you dont plan to overclock extremely seems that E6300 is the best option...with a nice board like DS3.

They money you save can be used to get more ram or a better video card, and that will trully improve game performance :)

[thunderstruck!]

I want as much cache as possible for gaming, so I'm getting an E6600. I also think that the extra cache will give it extra life when it comes time to upgrade in 2008-9.

[theteamaqua]

below r "maximum" for 24/7 on tower 120:

E6300 : 7x500 3.5GHz
E6400: 8x480 3.8Ghz

all on DS3, assuming dual pi 32m stable for 24/7

i think they r good choices best clock / dollar ratio

im getting either 6400 or 6600, still deciding as i will get kentsfield when it comes out

[Gun_Strife]

I want as much cache as possible for gaming, so I'm getting an E6600. I also think that the extra cache will give it extra life when it comes time to upgrade in 2008-9.


wow i cant have the same rig for hat long
but its a good point if ur keeping it for over a year id go with the e6600
if ur not gonna keep it that long id go with the 63 or 64
as the dif today isnt great

[Phosphate]

I want as much cache as possible for gaming, so I'm getting an E6600. I also think that the extra cache will give it extra life when it comes time to upgrade in 2008-9.

I see what you are saying but I just hate spending money on gear when I don't need it. Its amazing how upgrading things slightly just jacks up the total cost. The future? Maybe tommorows games will need more L2, but by that time there will be other CPUs anyway. Besides I still need to buy a DX10 video card.

I guess what I found out was that Intel is not short changing us an L2 cache with the allendales as far as gaming applications go:cool: (unlike my celeron which is L2 cache deficient :p: )


Now the question is how much does the extra cache matter at speeds above what they tested?

[boomheadshot45]

Good choice, i'm go with the E6300 too, i'm probaly going to get to 3.0 ghz easily on air, damn, i love conroe!

[realsmasher]

hasn't that test a serious issue ?

they are driving e6400 and 6300 on nice fsb, so ram also does.

but how can you compare that ?

and think about : 2,6ghz e6300 with fast ram sometimes even looses agains 6600stock.

anyway : to make a good test, use 6600 at multi 8 and test it agains 6400 with same memory speeds. those tests here show nothing in my opinion.

[Phosphate]

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2802

anyway : to make a good test, use 6600 at multi 8 and test it agains 6400 with same memory speeds. those tests here show nothing in my opinion.

You are right that would be a better best and I think they actually did that in some other article and there was a 3-10% difference but that was in non gaming stuff.


What I took away from this is that at these speeds tested *for games* the 2 mb cache is plenty and adding more is not improving performance (unlike celeron that is L2 deficient)

I think the take away point is that intel didn't shortchange us on L2 cache with e6300/e6400.

[Phosphate]

?

and think about : 2,6ghz e6300 with fast ram sometimes even looses agains 6600stock.




You are right on F.E.A.R. that happened. That seems to be the exception. You know I didn't look that carefully after seeing this http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2802&p=7 so I missed that one.

[realsmasher]

the problem is, that you don't know if it comes form cpu clocks or from faster memory.

anyway, i'll buy my 6600 mainly because of multi. if it's also faster because more cache, i wont be angry about that ;)

[mikael999]

because they are not overclokig the e6600 in this article is because it is a test of how far they can go with BUDGET E6300 E6400 and imo the E6600 isnt a budget cpu ;)

[burningrave101]

I don't know, the E6600 doesn't cost a whole lot more then what people were paying for Opteron 165's and you see how many people flocked to get one of them for a lower cost overclocking CPU.

[realsmasher]

where i live, the 6600 is just 50€ more expensive then 6400 and 6400 is 50€ more then 6300€.

i don't think that you can say : "oh 6400 is 250€, so this is a budget cpu, but 6600 for 300€, oh no that expensive, you cannot buy this"

with that logic, 6400 wouldn't be bugdet, too.

but apart from 6400 only offers 1 multi more then 6300, 6600 has 1 multi AND doubled cache at the same higher costs.

so in my opinion, the step from 6300 to 6400 is bigger then from 6400 to 6600.

[JWMc]

I wish they had tested World of Warcraft. With all sorts of scripted addons in the mix the CPU load that WoW creates is in a league of its own. I currently have an AthlonXP 3200+ Barton 400fsb with dual channel Corsair DDR400 and it's nowhere near enough to put the bottleneck on the video card in WoW.

I've been waiting a long time for an upgrade so I wanted to nail this irritating problem of occasional game freezes in WoW with a sledgehammer so I went for the E6600. Yes I'm sure it's not server lag, my mouse cursor freezes and nobody else complains of the 'lag spike'.

-JWM

[mdzcpa]

No matter how you slice it, the extra 2MB of cache is always helpful for gaming. More so in RTS games, but offering more performance accross the board.

The issue with some of the testing with 2 vs 4 MB of cache in gaming is just like it is with any CPU testing in games....its always going to be GPU limited on higher end 3D FPS games.

That said, when new generations of GPUs come out which can handle the current high end gaming titles better, you'll see not only the CPU, but the cache as well, make a difference.

[uOpt]

The same applied to 1024 KB Cache AMD64 versus 512 KB cache AMD64s at the same clockspeed. Pretty much doesn't matter if you look at real-world applications.

It is very rare to find an application that actually makes heavy use of more than 512 KB cache.

mdzcpa, you can easily test your theory by raising clockspeed a little. If you see an improvement from the slightly higher clockspeed but you don't see the same improvement from the bigger cache then the cache is useless and you were not GPU limited enough to just hide it although it was there.

Personally I have some applications that are very very fast on a 4 MB Conroe, up to 50% and in one case 100% faster at the same clockspeed than AMD64. I attribute that to Conroe's new prefetching mechanism in combination with the 4 MB cache. I have not been able to verify that theory as I don't have performance counters running and I don't have a 2 MB Conroe CPU to compare to either.

However, for these applications I am sure to get a 4 MB Conroes no matter what.

Besides, the 2 MB models have too low multipliers to reach their natural overclocking sweet spot (if same as with r5 ESes) with current boards, so you need a E6600 or better anyway.

[Dominus]

If you want to try a CPU intensive game try 'PlanetSide (http://planetside.station.sony.com/reserves/)' - The only problem is you won't be able to reproduce a situation/instance exactly and so tests wouldn't be so accurate. Mind you, with a weeks worth a benching on it you may get some pretty decent results.

[Phosphate]

I see what you guys are saying about GPU limited, but how GPU limited was the testbed when it was running two X1900XT in crossfire? To me that is a pretty powerful card set-up.

[ea6gka]

i am going for teh 6400 , the 2 extra MB the 6600 has does not justify its higher price in terms of the performance gains except ofcourse the multiplier...5% loss is really negligible

[Zebo]

people buy the 4mb chips for the multiplier not the cache.

[Phosphate]

people buy the 4mb chips for the multiplier not the cache.

Yeah I might get the e6600 for that reason alone until I get enough overclocking skills to squeeze max performance out the the e6400. Also it may be that the 9x multiplier allows better overclocks than 8x multiplier. Not sure if 10x multiplier is improving things though.

[Zebo]

Boards ive been seeing, and looking at Anandtech (http://www.anandtech.com/mb/showdoc.aspx?i=2799) and wes' work, can only hit 400Mhz.. Meaning 3600Mhz with a 6600 or 3200 Mhz with a 6400 or 2800Mhz with a 6300... Screw cache that's some BIG Mhz differences.

Sure there are some extraordinary clockers here that get lucky with a great board going well over 400Mhz...but I would'nt bet on it. Wes at AT has loads of experiance with OCing and can't do it so I have no reason to believe I can either. So 6600 for me, at least... maybe 6700 just to be sure I can max chip out.

[AndrewZorn]

/also buying E66 for the 9x/clocks rather than cache
though extra cache is always nice

[GoldenTiger]

I decided to switch to the E6600 instead of the E6400, mainly due to multiplier but somewhat due to cache... I need the CPU to last as long as it can, so a 400+mhz difference and extra cache combined should help things out longevity-wise :). The multiplier was the big thing.

[Brahmzy]

I decided to switch to the E6600 instead of the E6400, mainly due to multiplier but somewhat due to cache... I need the CPU to last as long as it can, so a 400+mhz difference and extra cache combined should help things out longevity-wise :). The multiplier was the big thing.

My reasons exactly for the E6600. :) Plus, When I pre-ordered, I had a bunch of extry cash sittin' in PayPal with no home.

[Nullvoid]

The extra cache does seem to matter, *especially* for gaming. Check out this part of the conroe review over at matbe: http://www.matbe.com/articles/lire/306/merom-et-conroe-test-des-core-2-duo/page13.php

It shows a performance boost of between 8.7% and 16.2% in various different games when moving from 2mb to 4mb cache.

Whether that justifies the extra cost will be up to everyone to decide themselves.

[ewitte]

Overclocking is always an option. You can't make a 2MB cache CPU work like there is 4MB though.