bit-tech compares 4M & 2M L2 for C2D

[vapb400]

The question has been brought up on numerous occasions. Now it may finally be laid to rest
http://www.bit-tech.net/hardware/200...cessors/5.html

[Shpoon]

That's nice to know. Thanks for the link. I'm a little surprised that COD2 isn't much of a change when it's run at a lower resolution....

[Lestat]

i would think, and its just me,, that once you get to a certain L2 size it becomes irrelevant as the data is moving in and out so fast the larger size doesnt come into play.
motheboards, and hard drives can only push so much data.

i think its time for hard drive companies to actually start making hard drives that will push the amount of data that all this new hardware claims to be able to move.
we have had the same techniology in hard drives for years and years and years,, hell even scsi hasnt changed hardly any over the years.

raid is nice but even that doesnt push the possible amount of data needed to realize what these cpu's and ram(s) can do.

[_Eduard_]

Good. Now I just want to know how well they OC. If the E6300 clocks to 3ghz+, we have a winner here!

By the way just a question: Lets say I buy an E6300 (1.83ghz, 1067mhz FSB, 2MB) and OC it to eh 1700MHz FSB, that would be about 3GHz. Do i need mem that can handle it (DDR800+) or can I use dividers? Will it hit performance a lot?

[Danger30Q]

The Intel Bad Axe board allows for 1:1 FSB:Ram at 533mhz, 4:5 for ram at 667mhz, and 800mhz at 2:3. So at 533mhz ram speed, an FSB of 400 will have memory running at 800mhz. If you run memory at 667mhz stock, when you turn the FSB up to 400, the memory will run at 1000mhz. There really isn't much to gain for running 4:5 over 1:1. It will save you almost $200 by buying only DDR2 800 memory. That's why you see most people buying PC2 6400, DDR2 800 memory. A lot of people seem to have bought the G.Skill HZ for their Conroe ES chips: http://www.newegg.com/Product/Produc...82E16820231065

[Verisimilitude]

In the rest of the article they explain that 4MB will be more important in the future. But atm I don't really consider that a factor.

[redpriest]

Sorry guys, this is a bad comparison.

First off, they are running at low clock speed - 2.13 ghz. Why do I say this is a bad comparison? Because cache *matters* at high frequency. Memory latency is relatively constant - but core clock speed scales massively compared to memory freq - cache is designed to hide that. More is better.

Not all things benefit from cache, but some do - especially games - at high clock frequency. I don't see many games in that comparison.

Intel has 2 different designs for the 2MB L2 and 4MB L2 versions - because they have published that the die size for the 2MB L2 is smaller. Imagine the amount of resources required to build and verify 2 separate versions, and if it didn't matter so much, why bother with spending the extra money on verifying the 4MB L2 one if you "don't need it"?

[redpriest]

Also consider that Intel is trying to push these new parts as quickly as they can - since, let's face it, the old Pentiums blew. Having a 4MB L2 cache version consumes more die area, and hence lets them ship less parts. You don't do that if you don't need to.

[Lestat]

quake 4 HL 2 and CoD 2 wtf more do you want do you see the res they are running also ? maxed out ?

i have had 128, 256, 512, 1m 2m and 4m L2 cahce chips and i can tell you that there aint squat from 2 to 4 meg.

like i said once you get to a certain size its irrelevant. its like ram,, when you get to a certain point it doesnt help anything.
2gig to 4 gig of ram on a windows machine and gaming ? tell me there is a huge difference, cuz there isnt. same principal.(to a certain degree)
cache is not like memory where the more you have the better, data moves in and out of it at such a rate that you get to a point and more cache is irrelevant because the cpu is moving the data fast enough you dont need more.

and maybe the chips arent handling the larger cache properly to make a difference,, i dont know.

now that is just the way i view it from my own personal experiences.
there isnt any difference from 2 to 4meg L2. not in the current state of hardware today.
everything else outside the cpu is running like mud compared to the cpu and ram,,, so...


another analogy,,

water thru a pipe.

lets say 500 gallons of water flows at its optimal rate thru a 2 feet pipe...
so let switch it to a 4 foot pipe..
does it flow faster ? no... why ? cuz at 2ft its allow to reach maximum velocity and flow rate because the SOURCE of that flow isnt pushing enough gallons per second (let's say). increase the pump to a 1000 gallons per minute and the 2ft pipe is onyl going to restrict the flow so you have to increase the pipe to a 4ft pipe to reach maximum flow rate.

so again i fall back onto my original statement that today's current state of hard drives and other optical storage devices can't give ram, and cpu's and videocards enough data fast enough for thing to truly be realized.
tell me that a hard drive that can only do 45-60meg/sec of random data is going to give a cpu or ram, which can handles gigs of data per second, enough data to allow it to run at an optimal rate. its not.. the bottle next isnt and never was cpu's and ram or L2 ,,, ok unless you have a celeron lol..

[aintz]

good find. not 2 much improvement from 2m or 4m but noticable

[redpriest]

Sorry lestat, but you're wrong. It does make a difference. Not a 20%-30% difference, but a 6%-8% difference *is* a big difference and worth the extra cost.

I am curious which 2MB/4MB L2 cache chips you have had and tested that don't make a difference in some benchmarks, other than superpi.

3 games is not a representative sample.

quake 4 HL 2 and CoD 2 wtf more do you want do you see the res they are running also ? maxed out ?

i have had 128, 256, 512, 1m 2m and 4m L2 cahce chips and i can tell you that there aint squat from 2 to 4 meg.

like i said once you get to a certain size its irrelevant. its like ram,, when you get to a certain point it doesnt help anything.
2gig to 4 gig of ram on a windows machine and gaming ? tell me there is a huge difference, cuz there isnt. same principal.(to a certain degree)
cache is not like memory where the more you have the better, data moves in and out of it at such a rate that you get to a point and more cache is irrelevant because the cpu is moving the data fast enough you dont need more.

and maybe the chips arent handling the larger cache properly to make a difference,, i dont know.

now that is just the way i view it from my own personal experiences.
there isnt any difference from 2 to 4meg L2. not in the current state of hardware today.
everything else outside the cpu is running like mud compared to the cpu and ram,,, so...

[redpriest]

And Lestat, I do agree with you with I/O devices and such being a bottleneck, but let's look at what cache is designed to do - lower effective memory latency.

The more of it you have, you have less miss rates.

Memory is godawful slow. When you have data that is travelling around on a microprocessor in terms of *picoseconds* when you have memory arriving 40-50 ns, that is an eternity. An eternity that has to be masked by out of order execution and maybe even out of order loads. Every little bit counts. Intel has a performance analysis team, and they would not have added 4MB L2 if it didn't make a difference on some benchmarks.

Maybe not on the ones that are being tested in this comparison, but you can't make a blanket statement that states 4 MB is meaningless.

If the bottleneck in the benchmarks are used are not the CPU/RAM, you would never need to buy a new processor and RAM.

[Nathanael]

Someone should test these two in Oblivion or at least in FEAR at a high resolution.

[aintz]

should also oc them and see how it affect overlocking

[Owen1978]

I would love to see how far an E6400 can go!

[drunkenmaster]

Someone should test these two in Oblivion or at least in FEAR at a high resolution.

cod2 is basically gpu limited at most resolutions except at the lowest with lwo quality set. oblivion and fear are both also games that are very highly gpu limited so will show exactly the same results. only when the test/benchmark or application is cpu limited will cache help, and not always but mostly it will.

http://www.anandtech.com/cpuchipsets...spx?i=2795&p=4

games are not cpu limited, hence another thing showing cpu making a tiny tiny difference, other things that are really really cpu limited, divx encoding, 10% difference.

everytime you need something that hasn't been prefetched into cache you have to go to memory, it takes a certain amount of clocks to access, send and receive data, a faster cpu has more clocks per second meaning less seconds to do the same access, cache is in a way more important on a slower cpu not a faster cpu.

EDIT:- the reviews also can't really show high clock performance as only the top cpu is unlocked. you can't get a higher than 2.13Ghz 2mb cache cpu so for now theres little point for reviewers to show what a 2.93Ghz 2mb can do against a 4mb part, as there is no 2mb 2.93Ghz part. though there isn't a 2.13Ghz 4mb part it somewhat lets users know if they can save some cash and not have the cache or if they really want to buy the E6600.

[celerity]

The average performance gain with 4 MB cache is only 2-3%, in games round 5%. This means you get a few FPS more which means you wont see any difference at all. I don't think it's worth paying $100 more (E6400 vs E6600) for 4 MB cache since the performance difference is so small. I would rather spend the money on a better video card or a better mobo for OCing.

I don't know if the extra cache has a bigger performance influence on higher clock rates but it would be very interesting to see a E6400 at 3 ghz against E6600 at the same frequency.

Intel has a performance analysis team, and they would not have added 4MB L2 if it didn't make a difference on some benchmarks.

So far no tests have showed any difference between the caches, so saying Intel know what they are doing isn't really correct I think. For me, real performance matter, not in theory... I'm not excluding that the cache may have some affect on future apps and operating systems such as Vista, but there's still a long time until the releases. My guess is that apps will be more optimized for dual core CPUs in the future and the extra cache may have some affect then, but not now. If you'r planning to have that CPU for a longer time, I think it's worth paying for extra cache, otherwise, no. But as I said before I would like to see more tests!

[GoldenTiger]

behardware had a review with a lot of tests at same frequency 2MB vs. 4MB which showed between 6-15% difference depending on app/game... moreso in apps than games. It is definitely worth it.

[Prometheus]

Anand posted a 2 vs. 4 cache performance comparison that looked at the performance difference in 25 different applications. 4MB cache was 0% to 10% faster, with an average of 3.5%. You can see the results at http://www.anandtech.com/cpuchipsets...spx?i=2795&p=4