What is the difference between CPU L2 cache 2048KB and 2 1024

Professional: Don't mislead the two upstairs, On the current multi-core CPU world, 1024 is not, several cores share 1024 L2 cache, there is no data exchange problem between caches, Intel's core series is such an architecture, 2x512, means 2 cores, each core has exclusive 512 L2 cache. AMD's U is such a design, the design of the second-level cache, there are size, speed constraints, because of Intel's U, the memory controller is outside the core, in this way, the exchange speed between the memory and the CPU can not be very fast, in order to reduce the number of low-speed exchanges between the CPU and the memory, therefore, Intel designs the second-level cache as a whole, and the capacity is relatively large, which is equivalent to, a big house is full of food, and nearby residents can quickly and easily obtain food and share resources, and AMD's U, because the memory controller is directly integrated into the CPU, his front-side bus, very fast, and, the cost of the L2 cache is relatively large in the CPU, AMD due to architectural reasons, can not design the L2 cache into a shared mode, so it can only be exclusive to each core, and then through the HT bus (AMD's proprietary front-end bus bus) to connect the 2 CPUs, this way is a bit backward, so the current AMD technology, It was abandoned by Intel for a whole generation, but the speed of the computer is not only determined by the CPU, because the AMD core integrates the memory controller, which is higher than the memory controller that Intel puts on the motherboard, the speed is higher, so, overall, the difference is not very big, but in terms of CPU monomer, Intel is still strong.

There is no difference, it's all 2m cache.

First of all, the i3 is a 2x32k L1 cache, not a 2x64k.

In addition, the whole series of i3 are dual-core CPUs, he is dual-core 4 threads, L1 cache is physically divided, and L1 is not shared whether it is data cache or instruction cache, and a single core enjoys 32K.

The i7 series is 4 cores, he is 4 cores and 8 threads, and a single core has an exclusive L1 of 32k, so he is 4x32k.

Since L1 cannot be shared, it is written as 2x32 or 4x32 for i7.

Another digression L2 is also physically divided and L2 cannot be shared, L3 is physically shared, all say shared L3, in fact, L3 is not really shared, take 2600 as an example, when you turn off hyper-threading, L3 will be logically divided into 4 parts, a single core exclusively 2M L3, when hyper-threading is turned on, L3 will be logically divided into 8 parts, when hyper-threading is turned on L1 and L2 will be logically divided, That's why there is a slight drop in single-core performance when hyper-threading is turned on.

It's based on the core number.

For example, i3 is dual-core, which means 2*256, which means 1 core with 256kb of cache.

2x256 is 256 for one core, and 2x256 for 2 cores.

Simply put, it's a core with a cache.

What is the CPU model you see?

To put it simply, 2*512KB=1MB, therefore, these two capacities are equal.

But ......If you think about it carefully, this is a big problem. A total L2 of 1MB may mean that the cache structure is a "von Neumann" structure; 2*512KB may represent 512KB instruction cache + 512KB data cache, and this separate cache structure is the "Harvard" structure. Compared with the two, the latter is more efficient.

If the CPU you see is a desktop, 1M and 2*512KB have similar meanings. It still depends on your CPU model. If you're talking about mobile phone CPU, it's likely to involve the problem of cache structure, 2*512KB is better.

2*512KB is dual-core, and each core means 512k.

But if you only use one core, you can call all caches, which is 1m