2mb cache vs 4mb cache on CPU

So I have decided on which notebook to get and my final decision to make is the processor. I will be going with the Core 2 Duo and was wondering if a 4mb cache is going to make a world of a difference in terms of performance. It's about a $110 jump from a 1.66 ghz 2mb to a 2.0 ghz 4mb.

Thanks for your advice!

 

From my experience, the cache increase will give your more of a performance boost than the processor speed boost will.

But as it's always said on here, it depends on what you do on your computer. If you're doing a lot of processor intensive stuff, go for the better processor. If you're just doing internet/email/office apps then it's not going to be worth it.

 

Yeah, I'll be playing games. More specifically World of Warcraft and Oblivion.

 

Playing games, you probably want to go with at least the 1.83GHz processor. The processor with the best price/performance ratio right now is the T7200, the 2GHz Core 2 Duo. For only $110, I'd say it's worth it personally.

Just make sure you don't skimp on RAM, otherwise the fastest CPU in the world won't save your performance

 

Okay, thanks. I will probably just go with the 2.0 4mb Core 2 Duo then. I'll be getting 2GB ram with my system and upgrading it to 3GB in the near future.

 

The bottleneck for these games is the GPU not the CPU.
Plus, [this is a bit too geekish] the larger the cache grows the higher the latency goes.The reason that we don't see a great performance jump with 2MB->4MB cache upgrade.Of course this will be great for people who want to do certain tasks (encoding) but not much for gaming.I suggest you go for a descent GPU (like X1600) and get a mediocre CPU,as long as you don't want virtualization (VMWare and such) you can go with the 1.66 and get a better GPU.[if gaming is your most intensive thing on the laptop]

 

Yeah, I understand that the GPU would be the bottleneck but I am getting a 7950gtx card. I'd get a faster card if there was one today. I don't feel like waiting for the newer cards so I am trying to look at other things that would cause performance problems at this point.

 

If that's the case :
Get the T7200, it'll be a disgrace if you don't.

 

What do you mean latency goes up? It stays the same, at 3 cycles. Just like you don't get any more memory latency when you go from 1GB to 2GB of main memory, a larger cache is all at a low latency, period. You see benefits when your application routinely accesses more than 2MB of data that can be cached, which right now isn't terribly common, but can be depending on the usage, which is why we don't see the performance jump even in things like SuperPi, which still fits completely in 2MB of cache. Try SuperPi on higher digits of Pi, and see how quickly it will degrade performance on a 2MB cache CPU vs a 4MB cache one.

 

The 3 Cycles is for the L1 cache not the L2.
Look at it this way : Since the same kind of memory is used for making L2 cache in various Core 2 Duo laptops.There will be twice as much lines and blocks in 4MB cache as there is in 2MB of cache.So finding that particular block/line will take more time.

 

A programmer or geek I am not - but wouldn't it be quicker to find data on L2 than go out on the bus and pull it from system RAM or the hdd? Seems like it would be something like L2 - controller - wait for bus - bus time - system RAM (and the seek process again) - wait for bus - bus - controller - CPU??

 

Yes, JellyGeo, you are correct. Its much more efficient to keep everything in the cache, because it is fast, close and can be accessed much quicker. But, most applications simply don't use the cache, because the invlove very small pieces of data being requested, or there isn't enough data to fill the cache. Thats why a difference in performance between 2 MB and 4 MB cache, is only seen on some applications.

 

Maybe slightly. Even then, it's using a lookup table, so it's still very fast, and much faster than accessing the same information in main memory. And if there's only 2MB worth of cached data being accessed, it will all be at the front of the 4MB of cache in general, so there's not a speed loss there. But if you need data past the 2nd meg, you no longer have to go out to "slow" main memory, and you get your result in the cache. The speed "hit" is really a gain overall. It also doesn't need to go through every line and block every time, it only needs to go until it finds what it needs. Given the cache concurrency and other algorithms they use, in practice that latency time is not going to increase at all with 4MB of cache vs. 2MB.

 

I might be totally wrong but!

as an example.

the celeron-m has 1mb cache and the pentium-m has 2mb cache, the only other difference is the speedstep.

Now all the comparisons i have seen between the two equal speed processors have produced little or no difference.

I know they are getting to be old processors.

regards.

John.

 

Pitabred is right (not about the 3 cycles, which is indeed for L1, rather than L2), the cache latency stays the same on all chips using the same core.

Yes, in general, bigger cache means more latency, but there's enough leeway to allow up to 4MB cache without having to change the latencies. That also means that the 2MB chips don't have absolutely optimal latencies, but they're close enough, and it's a good cost-saver for Intel (rather than having to make a separate core just to save one or two cycles of latency on the L2 cache)

 

All applications certainly use the cache (apart from everything else, it's not even up to the apps. The CPU itself manages the cache and decides what to put there, and the app *can't* do anything to change it. So basically, every time the app requests data, the cpu will put it in the cache if it isn't already there.

The difference between apps is more in how cache-friendly their memory accesses are. If they do completely random memory accesses all the time, they'll almost never hit the bits of data that are in cache already, and won't benefit much from the cache however big it is.
If they tend to stick to the same areas of memory, though, almost every memory access will be a cache hit, and performance will generally go waay up. (That is one of the big things to watch out for when optimizing code)
But on average, something like 98% of all memory accesses hit the cache rather than system memory. So most apps do gain a lot from the cache.

 

I would say $110 for the upgrade is a good deal, especially for Oblivion where CPU performance is very important in towns, battlefields, and crowded AI areas.

Congrats to Pita for becoming green

 

Thanks And yeah, a faster CPU will do nothing but help you in the future. It may not be needed right now, but as games start taking advantage of multiple cores, it will be nice to have the "headroom". I have the fast CPU because I do a bit of 3D rendering and such, and I just wanted to brag about having a SuperPi 2M score of less than a minute

 

Besides cache, are FSB 533MHz and 667MHZ on CPU have a significant difference?

 

It will have good effect on the utilization of the Ram and the GPU.
Depends on the rest of the specs.

 

Think of it this way....you're dropping $2k on a new system. $100 is 2% of the total, and maybe a couple of dinners out a month for one month. Don't second guess yourself and regret it later, go for the sweet spot in the T7200.

 

$100 is 5% of the total.

 

Man, my college accounting professor wouldn't be too proud of me right now....oh well, a couple less Starbucks then.

 

haha.. anyways, do u think the extra 2mb will affect the performance on photoshop cs2?

 

Nope, not really. First, the effect of more cache isn't that noticeable in the first place, and second, Photoshop uses much larger datasets than 4MB anyway, resulting in tons of cache misses regardless of cache size. Of course, the bigger cache will help performance a bit, but I wouldn't expect to see more than 2-4% improvement.

Of course this is just a rough guess, since I don't have one of each CPU here running Photoshop.