l2 cache and stuttering in games

[madh83]

Looking over the benchmarks it doesn't seem like the cache makes much of a difference in gaming. I was curious though, if the difference from frame rates is that the processor with less cache is just a little slower the whole time, or if they're the same speed at some points but the one with less cache tends to stutter more? I know that in game stuttering usually has more to do with video and system ram, but just curious because I've never had any experience with core 2 duos before.

 

[madh83]

Anyone?

 

[Cheex]

Quick example...

You will see ~ a 7%-10% increase in performance of some games moving up from 2MB to 4MB of L2 Cache.

 

[tigersty1e]

Originally posted by: Cheex
Quick example...

You will see ~ a 7%-10% increase in performance of some games moving up from 2MB to 4MB of L2 Cache.

I think he knows this, but was asking if this was a constant 7-10% decrease or a 20-30% decrease at certain points in the benchmark bringing the average down to 7-10%.


I think it's a constant decrease. No stuttering in the benchmark.

 

[lyssword]

nah, stuttering has more to do with system memory/vid card, like you mentioned. Mostly video card though. However if ther is a lot of physics in the background it's possible that cpu is culprit. I'm pretty sure the slowdown is linear though..

 

[tallman45]

The fastest components are your Processor/Memory/Cache/GPU/HDD

If there is not the data the App needs in Core memory or in the processor Cache then it must go retrive it from the HDD

A HDD subsystem must be tuned to your specific needs or it will ALWYAS be your bottleneck Aka chokepoint

Are you running a single drive ? Is it operating properly, does it have errors/bad sectors, fragmented ? That is the place to look 1st

 

[Borealis7]

Originally posted by: tallman45
The fastest components are your Processor/Memory/Cache/GPU/HDD

Cache is faster than RAM, otherwise it wouldnt be used.

the tradeoff is price vs speed vs size

 

[Cheex]

Originally posted by: Borealis7

Originally posted by: tallman45
The fastest components are your Processor/Memory/Cache/GPU/HDD

Cache is faster than RAM, otherwise it wouldnt be used.

the tradeoff is price vs speed vs size

Ya...it is...

CPU>L2>RAM>GPU>HDD

 

[jdkick]

Can you post the specs of your system?

Without the specs, i'd guess that either your GPU is insufficient (low frame rate) or you don't have enough system memory (frequent hard disk activity during game play).

 

[Tempered81]

his stuttering probably comes from a fragmented harddrive caching into system ram midgame. Nothing to do with the cpu. Like these guys said, its just a small performance decrease in FPS going to a smaller cpu l2 cache.

 

[madh83]

Originally posted by: tigersty1e

Originally posted by: Cheex
Quick example...

You will see ~ a 7%-10% increase in performance of some games moving up from 2MB to 4MB of L2 Cache.

I think he knows this, but was asking if this was a constant 7-10% decrease or a 20-30% decrease at certain points in the benchmark bringing the average down to 7-10%.


I think it's a constant decrease. No stuttering in the benchmark.

Exactly what I was asking. I'm just wondering about that very specific topic, I'm not asking how to fix stuttering on my own system.

 

[The-Noid]

CPU-L1-L2-L3-GPU-PPU-GPURAM-SYSTEMRAM-HARDDRIVE CACHE-HARD DRIVE-USB

GPU and GPU-RAM are actually quite fast for things that are left on the card. Much faster than system memory.

 

[justly]

In a purely theoretical sense it would seem that performance dips (stuttering might be to hash of a description) caused by a smaller cache would happen more often.
If you take two identical processor cores with the exception of cache, both running at the same frequency, both should have equal performance when retrieving data from cache. However, when a cache miss occurs that program will essentially stop to wait for the data to be retrieved from main memory or the hard drive.
So, I would have to say that any application/program/benchmark that shows degradation in performance between two processors having the same architecture and different levels of cache is technically stuttering.

Determining how this degradation effects user experience is a much more difficult question, one I wish review sites would address.
I suppose its possible that these theoretical dips in performance (if short enough) could go completely unnoticed, something similar to how a tube tv flickers on and off without being noticed.

I wish I could give you a definitive answer but without having these processors to test for myself all I can provide you with is theory.

Considering the topic is about stuttering you might find this interesting even though it is slightly off topic (its regarding the AMD Phenom).
This quote was taken from Aceshardware, and was written by the same person that did the LostCircuits review where a Phenom 9600 achieved a higher average frame rate than a Intel QX9770 in a UT3 benchmark with frame smoothing enabled (effectively capping max frame rate).

"This appeared to be another oddball at first but I am running UT3 now, using the neural impulse actuator (a headband to read out neuromuscular signals and translate them in to game commands) and all I can say is that on the Phenom platform, it runs 100% smooth even with the extra workload of doing all the FFTs on several channels simultaneously for the different signal components, whereas there are mouse lags and other chops even on the C2Q X9650. With any Core2Duos or Athlon 64, UT3 it becomes very tedious to play under heavy load in this configuration.

Keep in mind that this depends on the number of bots. If the bots are disabled / reduced and it is just a walk through the level, then it really doesn't matter what CPU you are running because it is just processing the scenery. But have a number of bots shooting at you and at each other, that's a totally different situation and, mind me saying, more relevant for the gamer."

EDIT to provide links

 

[madh83]

Interesting...hmm..yea it's hard to find information on the subject in most reviews. Thanks for the links justly, that discussion thread sounds a little too technical for a non-programmer though.

 

[GundamF91]

Check out the Xbit lab review here, it's very clear that L2 cache is important in games and other applications. Higher CPU speed does not compensate for relatively small cache amount. See the results

http://www.xbitlabs.com/articl...int/celeron-e1200.html

 

[lopri]

I have to say that the visible stuttering has little, if any, to do with the size of L2. While playing Oblivion, I would see the FPS difference between E6600 and E2140 (heck add Opteron 165 to the mix) in the City area. (like Market District) There was ~10FPS difference but the E2140 (and Opteron 165) was just as fluid as E6600. On the other hand, where E2160 hiccuped, E6600 hiccuped, too. Now, this is just one game and I'd think every game will handle cache-memory differently. But under ideal situation, L2 is just too fast to cause visible stuttering. As many have mentioned, there are many other things to consider (such as video card drivers, system memory, hard disks, etc.) before L2 when it comes to stuttering.