Server Fault is a question and answer site for system and network administrators. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I was reading this question which references L2 cache hits as taking about 10ns: Are networks now faster than disks?

However I distinctly remember there being 10ns cache ICs for the 486 generation. A google search seems to back this up ( ). So my question is, has L2 cache just stopped getting faster? Is this a fundamental limit on memory speed that's already been reached? Or are their other issues at play?

share|improve this question

closed as not constructive by mgorven, Michael Hampton, Tom O'Connor, Dave M, mdpc Feb 15 '13 at 0:47

As it currently stands, this question is not a good fit for our Q&A format. We expect answers to be supported by facts, references, or expertise, but this question will likely solicit debate, arguments, polling, or extended discussion. If you feel that this question can be improved and possibly reopened, visit the help center for guidance.If this question can be reworded to fit the rules in the help center, please edit the question.

You're comparing very different measurements. That 10ns cache you linked has a maximum single operation rate of 10ns between operations. The 10ns to load a register from L2 cache is the complete operation. That'd be like time a old train takes to go a mile, and comparing that to the time a new train takes to go from one station to the next. – Chris S Feb 14 '13 at 6:13
up vote 3 down vote accepted

L2 cache hits (AKA L1 cache misses) on i7's are around 10 cycles, which is around 3ns. See page 9 of this document, "The load latency from the L2 CACHE is 10 cycles".

Also, this is not even close to an apples-to-apples comparison at all. The 10ns on the '486 CPU cache chip was just the time between when you had found the entry in cache to when you had the result on the cache bus. It doesn't include the time to get the address from the CPU to the cache, time to consult the cache tags to find the entry in the cache, time to ensure the cache entry is valid for the operation the CPU is going to perform, or the time to get the result from the cache bus to the CPU. At a minimum, you need to double the time to consult the cache's index to see if the entry is in the cache at all.

L1 cache misses on high-end 486 systems had about a 200ns cost. So L2 caches have gotten about 75 times faster.

share|improve this answer

Networks have been faster than discs for years - 10G and 40G Infiniband have turned that to dust. The latency of a regular disc was beaten at 100mbit. Random IO - I can not saturate 100mbit with a single disc.

I think the problem with L2 is not that it is too slow - the surrounding just became better. There is L3 cache and a faster subsystem memory wise, which was a brutal bottleneck at 486 times. L1 cache has become a more important factor, too.

Simply stated - L2 cache is where it is because that is where it makes most sense, also financially.

share|improve this answer

Not the answer you're looking for? Browse other questions tagged or ask your own question.