Cache vs OF

flyrice

When I was looking into buy a processor, I narrowed it down between an AMD Athlon 64 3400+ Clawhammer with a 2.2 Operating Frequency (OF) and a 1MB L2 Cache and an AMD Athlon 64 3400+ Newcastle with a 2.4 OF and a 512K L2 Cache. The price difference was 6 bucks (aka negligble) so I went with the bigger cache.

I guess the question I have is, what's more important for overall speed and stability? ie which is better (higher OF or more cache memory?)

mmonnin

Well if you want speed, the OF is the speed. It IS the frequency. Stability is the same for both.

shwaip

flyrice wrote:

When I was looking into buy a processor, I narrowed it down between an AMD Athlon 64 3400+ Clawhammer with a 2.2 Operating Frequency (OF) and a 1MB L2 Cache and an AMD Athlon 64 3400+ Newcastle with a 2.4 OF and a 512K L2 Cache. The price difference was 6 bucks (aka negligble) so I went with the bigger cache.

I guess the question I have is, what's more important for overall speed and stability? ie which is better (higher OF or more cache memory?)

some apps will perform better with the 512k more cache, some with the .2GHz higher frequency. They are rated the same, so their performace will be similar.

Straight_Man

Well, at real high CPU speeds, cache can be used both for input and output. Let's say RAM is slower that CPU rate as far as genning output that maybe goes to RAM. Then bigger cache will let the CPU stumble less, as it has more room for things pended in or out-- cache can be used for both I (input precaching) and O(output post-caching). That is one reason why on real fast processors, especially the multicored ones, caches will be huge. Rest of system main functions are running slower, so things need to be pended to allow CPU to run so much faster. Smaller cache CPU with EXTREMELY fast RAM might work as well as a larger cache CPu with RAM of same speed or even a BIT slower. BUT, give both the CPUs slower RAM, not the fastest video card, and some RAM intense tasks, the larger cached one should handle more tasks or threads at once. Even when CPUs were much slower, and L2 and L3 cache, if they both existed, were increased in size to larger sizes, the CPUs were more stable than with small amounts of those caches.

The faster the CPU, the faster it is likely to fill cache up if rest of system is not clocked the same as CPU and CPU is under a large load compared to real capacity. And rest of system has never been clocked to result in same throughput of data as CPU and cache can pass through. So, there always has been a proven benefit to bigger cache (IE more stability under higher load), and the faster the CPU the more benefit you get as to stability out of larger cache. A CPU with no cache to work with stumbles, and the faster it is, the faster and worse it stumbles if it needs and has no cache space free for a certain number of time slices per CPU design(CPU can only pend work in its pipe for so long, before it clears pipe, and if there is no place to put results they get LOST as they get wiped or "bit bucketed") for a certain number of time slices per CPU design(CPU can only pend work in its pipe for so long, before it clears pipe, and if there is no place to put results they get LOST as they get wiped or "bit bucketed"). CPUs are not intended to pause thier core calc pipes. They are intended to send results to cache, whence it goes to RAM or whereever, and get work from cache. They historically have had tiny L1 cache, typically bigger L2 and L3 cache once the progress in CPU design allowed for both L2 and L3 (as chip arrays on motherboard to start with, not even on die). RAM was one heck of a lot slower responding than L1 or even L2, typically.