EggXpert.com

It seems that people are all looking forward to more cores for the purpose of better performance, and core race is now well under way, but when reading the article below,

betanews:

AMD: Will More CPU Cores Always Mean Better Performance?

The company that helped inaugurate the multicore era of CPUs has begun studying the question, will more cores always yield better processing? Or is there a point where the law of diminishing returns takes over? A new tool for developers to take advantage of available resources could help find the answers, and perhaps make 16 cores truly feel more powerful than eight cores.

Two years ago, at the onset of the multicore era, testers examining how simple tasks took advantage of the first CPUs with two on-board logic cores discovered less of a performance boost than they might have expected. For the earliest tests, some were shocked to discover a few tasks actually slowed down under a dual-core scheme.

..............................

I have a question for you guys-------Does adding more cores scale the performance?

That is a good question and this is IMHO only.

When dual cores came out at say 2.6GHz each core was really 1.3GHz, when quads came out well take a 2.4GHz well each core was 600Mhz each core. So really only performance for any of it is cache some CPUs have 6MB of L2 others will split them like 512 X 2 = 1MB. So MHz wise I don’t think there will be a difference in it. But when you rely on HT, Cache, and Bus Speed then there will be a difference.

That just how I see it. my two pennies worth.

I believe so, I think multiple cores are still not full supported by all software.

So far from what I know there are 5 games and one app that support multi core for now.

 

Tracer76:

That is a good question and this is IMHO only.

When dual cores came out at say 2.6GHz each core was really 1.3GHz, when quads came out well take a 2.4GHz well each core was 600Mhz each core. So really only performance for any of it is cache some CPUs have 6MB of L2 others will split them like 512 X 2 = 1MB. So MHz wise I don’t think there will be a difference in it. But when you rely on HT, Cache, and Bus Speed then there will be a difference.

That just how I see it. my two pennies worth.

Each core what?

Tracer76:

So far from what I know there are 5 games and one app that support multi core for now.

then you are sorely miss-informed?

I am also in agreement with lowboy wondering what your first post was about... even the saying that a 2.4ghz quad is actually 9.6ghz is more accurate... (though both are totally wrong... if each core ran at 600mhz you would be getting some major slowdown going from a 2.4ghz single core to a 2.4ghz quad...)

as it is, each core runs at the specified chip clock speed, it is literally like having 4 individual single core processors all on one piece of silicon 

smartalco:

Tracer76:

So far from what I know there are 5 games and one app that support multi core for now.

then you are sorely miss-informed?

I am also in agreement with lowboy wondering what your first post was about... even the saying that a 2.4ghz quad is actually 9.6ghz is more accurate... (though both are totally wrong... if each core ran at 600mhz you would be getting some major slowdown going from a 2.4ghz single core to a 2.4ghz quad...)

as it is, each core runs at the specified chip clock speed, it is literally like having 4 individual single core processors all on one piece of silicon 

 

Ok prove to me what each core really is then if its not 600 and it has to equal 2.4Ghz = 2400Mhz  Now take 2400 and divide it by 4 what do you get as the answer?  

smartalco:

  even the saying that a 2.4ghz quad is actually 9.6ghz is more accurate

Show me a 9.6GHz proc. Also you might want to read this..... http://www.insight64.com/downloads/LongLiveDualCore.pdf

 From the article even though its from 2004 it still applies. Now why would a 3.4GHz CPU deliver more performance then a 3.6GHz CPU. 

"Even this level of performance might be hard to achieve, since the latest 3.6GHz P4 actually delivers

less performance than the earlier 3.4GHz model"

Hummm from Intels website

Processor NumberΔ	Cache	Clock Speed	Front Side Bus
45 nm
Q9550	12MB L2	2.83 GHz	1333 MHz
Q9450	12MB L2	2.66 GHz	1333 MHz
Q9300	6MB L2	2.50 GHz	1333 MHz
65 nm
Q6700	8MB L2	2.66 GHz	1066 MHz
Q6600	8MB L2	2.40 GHz	1066 MHz

I don’t see any of these at 9.6Ghz CPUs like you state that there are smartalco so who makes them?

Tracer76,

I am actually quite surprised that you think that the total speed for the dual/quad core procs are each 0.5/0.25 on each core.  Here is an example that should help prove otherwise and it is one you can test yourself.  I upgraded from a single core P4 3.0GHZ to my E6750 which is 2.66GHZ on each core.  Core 2 Duo means each core runs at the listed speed, not that the total speed is the listed amount.  Back to my example though, my dual core E6750 although each core is slower than my P4 was is significantly faster then the P4 was because more steps/calculations, etc can be performed due to multiple coress.  With each core adding up to the listed amount like you are saying, this drastic of a performance increase would not happen.  In fact, I have seen many machines with multiple cores in use and every one of them acts like my example as opposed to what you state.  I work in an IT department where we have a 5 year equipment rotation and we have 2.4 and 2.8 GHz single core procs we am reaplacing with dual core systems and every single one of the 300+ systems we replaced has worked exactly as my example.

Now here is how you can test for yourself if you have the hardware.  You'll need 2 separate computer systems.  One needs to have a single core proc, the other needs to have a dual core proc, ie 2.66 GHZ, where the listed speed is slower than than the single core's speed, ie P4 3.0 GHZ.  Have the rest of the systems as near identical as you can and then run various applications on them.  The dual core will prove faster than the single core in performance.

Now to address quad cores, quad cores function in the same manner as dual cores where the listed speed is that of each core, not a total of what all 4 cores perform at.  No, you will not see 9.6 GHz by adding all the cores up but you will see more than the total 2.4 GHz you believe.  Part of the reason for this is each core requires some overhead to function and this reduces the available percentage of the processor for each core.

While I don't have time right now to look up hard specs to verify this for you, if you wish, I will do so after work tonight.

 

Edit:  Just thought of one thing, the Pentium D's may have been designed to add up to the total like you were thinking, ie each core running at 1.2 for a 2.4 GHz dual core proc.  I haven't researched the Pentium D's enough to know for sure but with the C2D line and newer multi-core procs, it does work like I was giving an example of.

On dual core and quad core CPUs @ say 2.66GHz all cores run at 2.66GHz. This is true for AMD and Intel. If the Pent D is different it is the only chip I know of but I can't say for sure about that chip. Maybe as that chip always sucked. I really don't think so because MHz is king and 2 slow cores are going to perform slowly. Designers went to multi cores because increasing MHz has just about hit max with today technologies so to increase performance they had to go to multiple cores.

Hi,

I would have to agree with Tekran...I think, lol....

It was my understanding that the clock speed was for each core.. for example, my Athlon 64x2 5200+ 2.6GHz runs each core ate 2.6GHz. In theory, my CPU is 5.2GHz total, but in real world it doesnt scale up 1:1. I think it has something to do with the L2 cache aswell.

 I think thats why AMD Athlonx64's with the 2x1MB L2 cache were named how they are when they first came out....the 5200+ = 2x 2.6GHz, the 6000+ = 2x 3.0GHz and the 6400+ = 2x 3.2GHz .  I may be wrong, but I think when I read up on the dual-core tech before I built my machine about 1 1/2 years ago, that the tech articles explained it thus. I dont know how the nameing sceame works when the L2 cache is reduced to 2x512Mb, becuase then number designations dont add up the same as the chips with the L2 @ 2x1MB. This, atleast, is how I understood the AMD CPUs......

Im not really sure how it scales up in the end.. but Im pretty possitive that each core runs at the designated speed with modern dual-core, or quad-core, CPUs. Like I said, in thoery, if both cores were maxed out, and each core were taking care of 1/2 the application processing power respectivly, then the given CPU... in thoery, is double the speed....

I confused myself now... My brain hurts, lol

 

aha, the discussion so far is interesting. but it seems all of us confused, the same with me.

i think the calculation of clock speed X4 is funny. but it is not true. theoretically, it should be so.

but if it is true for AMD processor, the intel cpu should be clock speed X2 , right?

 

No confusion as EACH core runs at Spec'd speed IE;Q9450@2.66 each core runs at 2.66GHz. Speed ratings came from different architectures used by AMD and Intel. As when AMD came out with the K8 chips they out performed Intel chips at a lower MHz because of better designed architecture of the CPU.(This caused AMD to introduce the speed ratings to help consumers compare CPUs) AMD held the performance crown. Then Intel came out with the Conroe chip with better design (clock for clock) and Intel has ruled since then. Hopefully AMD can regain ground but they have choked the last two designs.

Didn't mean to confuse anyone one, sorry, read lowboy's reply.  Sorry for the way it was written, trying to type quick before work doesn't always result in a clear post from me, I guess.  Sorry again.

Well lowboys last reply made that more simple and  so I looked up the Amdahl's law or known as the Amdahl's argument that made it even more simple to under stand after reading the Amdahl's law. So sorry If I caused any confusion  as well.

good question. I cannot explain it clearly at least based on what i know...

this is really deep

http://en.wikipedia.org/wiki/Amdahl's_law

http://www.cis.temple.edu/~shi/docs/amdahl/amdahl.html

According to Amdahl's law, the theoretical maximum speedup of using N processors would be N, namely linear speedup. However, it is not uncommon to observe more than N speedup on a machine with N processors in practice, namely super linear speedup. One possible reason is the effect of cache aggregation. In parallel computers, not only does the number of processors change, but so does the size of accumulated caches from different processors. With the larger accumulated cache size, more or even the entire data set can fit into caches, dramatically reducing memory access time and producing an additional speedup beyond that arising from pure computation.

Amdahl's law also doesn't take into account that problem sizes may be scaled with increased number of processors, which typically reduces the relative amount of non-parallelizable tasks.

For games...no.

And just to clear things up, all modern CPU's no matter the # of cores, will run a single clock frequency and thats the one that they report to you.

 So a 2.4 GHz Q6600 is really 4 cores running at 2400 MHz. If an application (such as a video encoder) is multi-threaded and can handle up to 4 threads (so one thread per core) and if you compare to a single core CPU running at 2400 Mhz, the quad core would effectively be equivalent to a 9.6 GHz single core cpu minus the performance overhead and among other losses.

Tracer67, sorry man but I don't know how to respond to what you've said. You should really pick up a few textbooks and learn a few things before giving out advice because you'll just stray people the wrong way.

As for whether or not adding more cores will increase performance, the short answer is no, and I say no because if you're asking this question, then you probably won't come across any software in the near future that will require multiple cores.

My take on it is as follows. Unlike video cards that can render floating point calculations parallel, CPUs are designed to process specific instruction sets, and typically, applications cannot divide workloads between two processors for those types of tasks. You will be able to see better performance as the number of cores increase when processing data that can be divided between multiple CPUs, such as video encoding (Adobe Photoshop, Final Cut Pro, Sony Vegas), 3D Rendering (AutoCAD, 3DS Max, MAYA), and services that require a massive queue such as SQL databases or Exchange server that can assign multiple requests to multiple CPUs. But I highly doubt you're running your very own homegrown SQL server, or will be in the near future. If you're not using applications that can distribute workload across multiple processors, adding more cores will not provide you with any performance gain. Firefox, AIM, Office applications, will all use only one processor at a time and will only need to use that processor for a very short period of time. Even if you multitask like there's no tomorrow (like I do, with at minimum 30 windows open at once), chances are you will never fully utilize two CPU cores. In environments like those, you'll find that a faster storage system will provide you much greater gains (and by faster I don't mean RAID 0 everything to hell).

For the applications you guys are typically exposed to, 99% of the time you will never fully utilize more than two cores, and if you do it will be for such a short period of time that it won't be worth the difference. There's already a thread on dual core vs. quad core ** in gaming** and the undisputed consensus so far in that thread (which you can find in the CPU section) is that games generally don't see any performance improvement going from dual core vs. quad core.

Its not a matter of what applications support multiple cores, but whether or not that application will ever have need for more than one core. There are games that support up to 8 cores, but the performance gain going from a dual core to a quad core is absolutely zero, because there's only so much your CPU can do in a game. If you think about it, what gets processed? Sound effects, network traffic, special effects, and in some games, physics processing. However, Nvidia and ATI are natively supporting physics processing on their video cards, so the need for physics processing in CPUs is declining. When you think about that, you'll realize that a Dual core Cor 2 Duo has more than enough power to process what actually gets processed in a game, as your framerate will depend on your ability to process the graphical aspect of the game, which is all handled by the video card. More cores will boost your 3DMark score, but since when was that synthetic benchmark ever a reasonable representation of gaming performance?