Overclock.net banner
Cancel

6 MB of L3 cache or 12 MB of L2 cache?

5002 Views 30 Replies 16 Participants Last post by  Blameless, 
echohunter
which one has more advantage?
Reply
Save
Like
1 - 20 of 31 Posts
bucdan
l2 id say
Reply
Save
Like
M
Depends on the chips you are comparing.
Reply
Save
Like
gerikoh
8mb of L3
See less See more
Reply
Save
Like
L
At that level, there is no point. The chip depends more than the cache.
Reply
Save
Like
X
Well it seems your comparing an amd Phenom II 940 and an intel Q9550 and clock per clock there pretty even. and each chip uses the cache that they have in differant ways so comparing them is pointless.
Reply
Save
Like
j0z3
Quote:

Originally Posted by low strife View Post
At that level, there is no point. The chip depends more than the cache.
Right you are.

Maybe hes looking at Phenom 2 and q9*50
See less See more
Reply
Save
Like
whe3ls
they are about equal
Reply
Save
Like
M
If you're looking at a phenom 2, the L3 is actually pretty slow on those chips. Faster than L2 but it's not the best.
Reply
Save
Like
ZealotKi11er
L1>L2>L3 but when i comes do PII and C2Q they are different. AMD Cpus have intergrated memory controller which makes them less need of L2 but C2Q need more because they miss it. You see i7 has L3 too and not much L2 because of it intergrated memory controller.
Reply
Save
Like
G
In real-world applications? It doesn't matter. In benchmarks, assuming there is no other difference between the two processors, the one with the L3 cache will benchmark better (like superpi) since the L3 cache is more quickly accessed by the cores vs. L2 cache.

I'd go with the one with more cache in general. Keep in mind, cpu's with L3 cache also have L2 cache.

EDIT--looks like I have my info incorrect--L3 cache is actually slower than L2 cache? L3 cache feeds into L2 cache which feeds into L1 cache, which then goes to the core, yes? If so, I don't understand what the point of having L3 cache even is--just have more L2 cache...?
Reply
Save
Like
echohunter
I am comparing Phenom II 920 and Q9550. I know that they are very close in performance. I wanted to know if the L3 in Ph 2 make any difference. If the Ph 2 also had 12 MB of L2 instead of L3 would it make a huge difference?
Reply
Save
Like
L
Quote:


Originally Posted by echohunter
View Post

I am comparing Phenom II 920 and Q9550. I know that they are very close in performance. I wanted to know if the L3 in PII make any difference.

Not really... The only reason you would get one or the other is how much your budget is. AMD rigs, without exception, are cheaper.
See less See more
Reply
Save
Like
M
You're not going to find any usefulness from buying one or the other just for the cache
Reply
Save
Like
T
Quote:


Originally Posted by guyladouche
View Post

EDIT--looks like I have my info incorrect--L3 cache is actually slower than L2 cache? L3 cache feeds into L2 cache which feeds into L1 cache, which then goes to the core, yes? If so, I don't understand what the point of having L3 cache even is--just have more L2 cache...?

Difference is in hit rates. The L3 has a much greater hit rate while being slower. Hit rate refers to the likeliness of the cache throwing an erroneous datum in or out. When this happens the processor has to wait to fetch/send it again from/to the RAM which takes extra time. The progressive is thusly L1->L2->L3, where L2 and L3 are similarly copies of RAM data. The processor cascades down the caches to find the correct value, with the highest likeliness of correctness being in the L3. In this way it avoids having to access the RAM which is much more time consuming.

It's almost like gears on a car; the analogy is that the one with only L2 cache is a four-speed manual transmission while the CPU with L3 cache is like a manual transmission with five speeds.
See less See more
Reply
Save
Like
A
Quote:
Larger caches have better hit rates but longer latency. To address this tradeoff, many computers use multiple levels of cache, with small fast caches backed up by larger slower caches.

Multi-level caches generally operate by checking the smallest Level 1 (L1) cache first; if it hits, the processor proceeds at high speed. If the smaller cache misses, the next larger cache (L2) is checked, and so on, before external memory is checked.

wiki
Can you see that the application might decide which is better. If it finds what it needs in the AMD L2 it would be faster in theory than the Intel because AMD has less to search. But if AMD does not find and must go to L3 while Intel with larger L2 may find so would be faster.

Why are people saying L3 is faster? I just don't get that?

Any way The advantage goes to smaller cache as long as what is being hunted is found. That is why it is a design strategy and balance by the engineers as to what is the optimal balance/trade off.

Also the AMD and Intel's have different storage rules exclusive versus inclusive. But I don't think you kids want to hear.

There are no absolutes, no one size fits all. Notice w/I7 Intel is down sizing the L2 and increasing the L3 same as the 920? Smaller L2 and adding L3 is better in most cases but not always. So bye bye C2D.
See less See more
Reply
Save
Like
T
>>Why are people saying L3 is faster? I just don't get that?

No, L3 is faster. You misunderstood the wiki article. If it fails the L2 cache it has to retrieve the data from the RAM which is much slower than the L3 cache.
Reply
Save
Like
A
Quote:

Originally Posted by test tube View Post
>>Why are people saying L3 is faster? I just don't get that?

No, L3 is faster. You misunderstood the wiki article. If it fails the L2 cache it has to retrieve the data from the RAM which is much slower than the L3 cache.
No you don't understand it goes looking to L1 then L2 then L3 then RAM.

If not then you diagram L1, L2, L3?
See less See more
Reply
Save
Like
T
Okay. The tradeoff is between latency and accurary.
2 levels of cache:

Processors
|^
v|
Level 1: Fastest, least accurate
|^
v|
Level 2: Moderate speed, moderate accuracy
|^
v|
RAM: Very slow compared to cache, extremely accurate

3 levels of cache:
Processors
|^
v|
Level 1: Fastest, least accurate
|^
v|
Level 2: Moderate speed, moderate accuracy
|^
v|
Level 3: Slow speed, high accuracy
|^
v|
RAM: Very slow compared to cache, extremely accurate

By going to the level 3 cache, which is very likely to have the accurate data, it avoids going to the RAM completely, whereas with only level 2 cache it would have to go straight to the RAM which is slower.
See less See more
Reply
Save
Like
T
There's a good article here comparing a processor with an L3 cache to roughly the same processor without one: http://www.tomshardware.com/reviews/...n-duo,664.html

This picture too shows how as the size (which the speed of data is based on, smaller is faster) of the cache increases the accuracy (likeliness to return erroneous data) of the cache decreases.

http://upload.wikimedia.org/wikipedi...e,missrate.png
Reply
Save
Like
1 - 20 of 31 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Overclock.net
A forum community dedicated to overclocking enthusiasts and testing the limits of computing. Come join the discussion about computing, builds, collections, displays, models, styles, scales, specifications, reviews, accessories, classifieds, and more!
Full Forum Listing
Explore Our Forums
Hardware News Intel CPUs PC Gaming NVIDIA Water Cooling
Top