6 cores vs 4 questions and thoughts

illfindu · Jul 27, 2011

I personally own a AMD Phenom II X6 1090T. Now I have for a long time seen many many benchmarks showing is short comings and I both under stand them and feel there findings make sense but I posse a question to the community because to be honest I'm not sure I know the answer. First off let me put forth a situation you have a generic W7 PC with the before mentioned CPU a high end GPU and 8GB of ram. Now lets say that system had several back ground programming running on a regular basis virus software, voip servers, skype ,aim things of this nature small but numerous enough to take up lets says 10-20% of the processor at any given time if you where to boot up a heavy usage processes lets say adobe photo shop of flash catalyst or a game that can use all four cores. Would the system see a performance boost over a Quad core AMD of the same speed due to the fact that 4/6 cores could be 100% taken up bye the program/game and the other 2 cores could handle the back ground programs. On paper I would think the answer would be that the x6 would outperform a x4 in this situation but my under standing of how the OS threads commands from programs to up to 6 cores isn't that solid. Secondly if this is the case how come we never see multi tasking benchmarks I feel they would be much more real life it might just be me but I commonly run multiple programs at once some of which take alot of CPU power for instance does a x6 beat a x4 in 3dmark if your also encodeing a video? I think the idea of some one having some thing running like a video encoder or 7-zip running while they might wanna play a game or do some photo shop work is not out of the question. Comments ideas Note: I am simply trying to get a under standing not start a flame war about Intel's current superiority to AMD when it comes to there top end offerings.

jsc · Jul 27, 2011

The quad core, 8 thread i5-2600K seems to have the edge even with multiple tasks/threads running.

henydiah · Jul 27, 2011

From my experince, I use 6core phenom for editing video, it's my job, and I don't use it for game, I feel my system fast when rendering, because I overclock it to 3.8ghz for dailly using till Now. And photoshop alway's used

brandondiep · Jul 27, 2011

Intel cores are just more efficient than amd period. They are faster clock for clock, use less power and overclock like hell. At the 200 price range the 2500k is the winner compared to amd's 6 core line-up even at multitasking. Just be happy with your cpu and not beat yourself up because you didn't buy the Intel.

illfindu · Jul 27, 2011

the point of this was never to debate the speed factor im more then aware that intel is alot faster per cycle and more efficient my point was why is the six core AMD not shocked against the our core amd in a multitasking benchmark and for that matter why are there no multi tasking benchmarks or at lest none iv seen.

brandondiep · Jul 27, 2011

illfindu :

That made no sense at all, you should proof read before you post. Intel still wins in multi-tasking like the above poster. /end thread

illfindu · Jul 27, 2011

Okay ya sorry was making a fast post before I had to go. First off let me make it clear one last time I am not bring Intel in to this at all really my question is more a general one about why we don't see multitasking benchmarks and if in a situation with several programs running a x6 would beat out a x4.

k1114 · Jul 27, 2011

The debate on number of cores vs performance is a very complicated matter as number of cores is only one part of the equation. http://www.tomshardware.com/reviews/sandy-bridge-core-i7-2600k-core-i5-2500k,2833-15.html As you can see from this, even if a program is multithreaded it handles the threads differently. The i5-2500k vs the i7-875k or i7 950 the i5 wins in most but loses in others. These are just separately run and not multitasking but then it just gets that much more complicated. But if say a phenom x4 vs an x6 with same clocks then of course the x6 will win, same goes for a 2600k vs a 2500k with same clocks.

The deciding factor is going to be price/performance where the x6 rivals the i5. In most cases the i5 wins but there are certain situations where the x6 wins so is dependent on each person's needs.

A little off topic but if the system as a hdd then it will be a big bottleneck in multitasking, much more than anything. Long renders are best done overnight with nothing else running and there are probably better things to do than to play a game while rendering if it's just short and during the day.

illfindu · Jul 27, 2011

K114 that was alot more help I think we are getting to the heart of my question. Is a OS able to lets say dedicate 4 of 6 cores to a game that can use all four cores and then use the remaining 4 cores to run back ground and system operations?

henydiah · Jul 27, 2011

brandondiep :

I know SB fast clock and the beast for game (benchmark) but this is real for live, and don't think so about multitasking .., why Workstation PC in the world do not use SandyBridge ... ? For multitasking is the best is xeon and opteron ( more core and thread ), and I build pc follow from videocopilot, videohive, creative cow etc and use firepro he he

. And this one : www.videoguys.com/Guide/E/Videoguys+System+recommendations+for+Video+Editing/0x4aebb06ba071d2b6a2cd784ce243a6c6.aspx.

pepe2907 · Jul 27, 2011

Because it's actually difficult to make decent multytasking tests. First question - mix what with what? There are so many variants. And the result may be influenced by so many factors. And as more tasks you add to the test, things get more complicated and possible variants rise exponentially. A multitasking test is never "pure". But I saw some level of multitasking tests not once - like video processing while zipping etc.

jonnyrb · Jul 27, 2011

illfindu :

I saw a few similar answers and thoughts posted to my knowledge as well. AMD is less efficient per core to it's Intel counter part, so they sort of.. / tried to make up for it by slamming cores on there. They don't get utilized optimally by the OS or any older games, or modern for that matter.

As for your programming "running" in the background.. that is SOOO nominal that it's almost irrelevant. More than anything they just sit there bogging down your RAM. run a heavy virus search while talking to someone on skype and watching youtube. You CPU's Won't spike worth a damn other than initially opening the APP

Video encoding is one of the few places you WILL see a nice change with cores. As I noticed this was an example of yours.

king smp · Jul 27, 2011

"They don't get utilized optimally by the OS or any older games, or modern for that matter" - johnyrb

I have to respectfully disagree

Modern games are using more cores all the time

and Linux and Win7 will utilize many cores very well

Win7 ultimate 64bit will handle dual sockets with up to 32 cores if I got it right

OSes are multhreaded software which easily run at least 500 threads at a time

so the more cores to use for the threads the better

Win7 Thread Ordering service has come a long way from the XP days

for overall general use (not just gaming-there is other things to do on a computer besides game!) a 6 core is a good thing

if your building a computer for the sole purpose of gaming then it doesnt make sense
but if do something silly like work on a computer (VMWare or Virtual Box VMs, photoshop,Maya,Autocad,MS Office,video editing with Vegas Pro or Adobe etc)
then it is better to have more cores especially if you need to have alot of apps at one time
with my dual core I am always wanting more cores
I could easily use four-six cores
and am planning on a new build just for that reason

ghnader hsmithot · Jul 27, 2011

jsc :

Multitasking benchmarks are few mate.

However there is a multitasking benchie called The Advanced Multi Process Test
http://www.passmark.com/products/pt_advmulti.htm

And its hard to make a multitask benchmark.
I mean can you accurate first switch on rendering after that 30seconds later add another proceess and another.It wont be accurate.

cobra5000 · Jul 27, 2011

brandondiep :

Kinda funny how you tell him to proofread when he CLEARLY stated " I am simply trying to get a under standing not start a flame war about Intel's current superiority to AMD when it comes to there top end offerings." Yet you post REPEATEDLY about intel's "superiority".
Just sayin...

jonnyrb · Jul 27, 2011

king smp :

I love the tone in which you type that in if that makes sense fyi lol You sir are very correct, as far as the OS using up to 32 cores.. I never stated it wouldn't use them. however you will never see a load as heavy on the higher up cores. Even on my quad core, cores 2,3 (3,4 since they count 0 :O) the usage is often a solid 15% lower. I do perhaps have a skewed reading on this as I am often playing games..

king smp · Jul 27, 2011

oh i like games also
I just get a little tired of everybody on Toms being so concerned about gaming
i remember toms being more hardcore modding and work applications
years ago
more cores are better for application usage
that is why workstations will have opterons and xeons
for serious work there is no denying that more cores are better

BTW Long live Bioshock
cant wait for Bioshock Infinite

now I got to get back to a powerpoint presentation LOL

gokanis · Jul 27, 2011

cobra5000 :

Thank you, was waiting for someone to point that out.

I figured he was either in a hurry or it was translated from another language.

illfindu · Jul 28, 2011

I never meant for this to start in to a competition even as a AMD fan boy I commonly point out to people I'm helping in other threads that the i5-2500k is one of the best deals out there and slaughters the AMD ii x6. I guess I should have expressed more clearly that I just wanted peoples options on seeing more multitasking benchmarks. And many of you did comment to that effect and I really appreciate that. I would like to thank the community on all the good feed back.

pat · Jul 28, 2011

Well. there is benchmark and there is real life. Benchmark are most of the time done on clean and optimized computer while real life computer are usually junky computer running plenty of unnecessary stuff, that we keep because, really, the computer can run them without really be bothered.

Then there is hyperthreading. Hyperthreading make use of "holes" in a thread to run another one in between. When a thread is using most of the core, then no hole is available so hyperthreading is not really helping there, sometime slowing the main thread. Just like cars merging in the highway. When few cars are going on the highway, merging is easy as there there is enough hole for other cars. But add cars and congestion happen, everything is slowed down. Then, adding another lane helps. Just like adding another core.

So, what is better? 4 cores, 6 cores or hyperthreading? 4 cores will handle almost everything fast for normal use. Hyperthreading will improve thing on with easy apps that do not taxes the core to the point that no holes are available for the other thread to keep going. Heavy stuff like rendering, editing compressing, everything that will make the cores using more than 75% of its time loaded will definitively benefit for another real core. I have the 6 cores and I can tell you that is faster and smoother than my previous 4 cores when doing some video stuff. I'm not doing video stuff all my day thou.. not even every week. That's means that my computer is not loaded most of the time. So, probably that hyperthreading would do good at waiting after me. Because, when I just surf the web, or doing normal stuff, my cpu run at 800 MHz. Yeah. Cool and quiet is much more usefull than hyperthreading.

But which one is better? Simple. It depend of what you are doing with your computer. I would say that 4 cores is now a good run for a general computer, not because it is needed, but now, they are inexpensive, so why not? And in my opinion, 6 cores is smoother and sometime faster than 4. But definitively smoother. Adding a RAID 0 stripe or a SSD HDD will help to. Because the data that your app uses is most of the time stored somewhere, likely the HDD, which is slow compared to memory at moving bits. So, a fast cpu will be processing data fast while waiiiiiiiiiiiiiiiiiiiiting for other to come by.

So, to keep my CPU busy when doing video stuff, a RAID 0 stripe do great. As well As fast memory in good quantity. Beacuse when you have a fast CPU, you don't want it to use the HDD as cache because you don't have enough RAM. No, you want the data moving at fast pace. A F1 car won't go faster than a minivan in traffic jam.

I don't think I've properly answered you question yet. Real life multitasking benchmark. My real life is probably different than yours, so a benchmark that will make a system doing good may have my cry for more speed while make you happy. My own benchmark is as follow. Can my system finish the small job while I'll go to make a coffee. Answer has to be yes. Next, can my system finish a medium job in the AM or the PM. Answer has to be yes. And can my system ever finish a big job? Yes, because a job that big won't make you waiting in front of the computer anyway, so when you'll come back, the job will be done (most likely) without you knowing exactly when it just finished.

In real life, speed difference between cpu are often normalized by "life" Because we all have a life. While having a fast computer to brag at your friend may be nice, it won't make you type faster nor surf the web faster. Human are the slowest link in the computer field and the speed at which you type may have an application on a fast system finish slower than one on a slower system. Because the 2 second speed advantage you may have will likely be kill by you reading the dialog box. looking at the key to press, moving your finger and actually press it 4 seconds after the task completed...

king smp · Jul 28, 2011

It is rough to make a duplicatable multitasking benchmark utilizing more than one program
there is benchmarking suites like Sisoft Sandra and Passmark V7 that tests
all different components of a system
also using toms benchmarks and anandtech you can look at variety of bench marks
including multithreaded ones which take advantage of many cores and this
way can compare two cpus together in a realistic way
obviously if you are doing alot of encoding then you would concentrate on multithreaded
for example
look at this anandtech bench
http://www.anandtech.com/bench/Product/146?vs=102

that compares a 1090t to a 965BE
as you can see the 1090t in the encoding and other multithreaded tests
dominates the 965BE
also in the gaming the difference is MINIMAL
the difference between a 1090t and a 965BE in gaming
doesnt even affect real life gaming
if you do other things beside game and for overall computer usage
you are much better off with the 1090t at stock speeds
when you bring in the OC factor the BE will win

but me personally would choose the 1090t everytime
with budget not a factor

gostrath · Mar 16, 2012

This seems to have become a flame war. I am interested in this topic for a simple reason, We am CPU hungry. Our company uses rackmount servers for two types of tasks: virtual machine hosting and data storage/access. All of these machines are basically headless servers, so we waste no power on graphics, USB, firewire, audio, etc.

Our virtual machines are processor hungry in that we have scaled up CPU utilization to try to have nearly 100% CPU utilization. Our virtual machines can be load balanced from hypervisor to hypervisor on the fly to try and produce roughly equal CPU utilization on all virtualization boxes. In this scenario, we CAN maintain equal CPU utilization across all cores. If we move too many VMs onto a server, we can maintain 100% core utilization across the CPU.

Data storage access doesn't sound like much, but it can be just as CPU intensive. For example, our storage bricks have large RAID arrays to store and retrieve data. Add to that the fact that each storage brick continuously accesses metadata on other bricks in order to satisfy data queries. These two factors alone do not really stress the CPU. The real issue, though, is our memory based SQL queries on slave servers. Due to the fact that we load many of the data tables in RAM, disk latency becomes a non-issue, but CPU speed and memory latency become issues. Run massively concurrent, multi-table join queries on a large memory based, read-only databases and you will see what I mean. High speed data-mining baby.

These 2 server based examples clearly show that high CPU utilization in today's virtualization and distributed storage environments can be an issue. This brings us back to the original question. Which is better, a faster CPU or a CPU with more cores. The answer is; it depends. If the software in question is multi-core aware but CAN'T always organize its workloads in a non-blocking manner, then faster CPUs with less cores will most likely yield better performance as several cores will end up idling at various times. This generally occurs in video games, for example. However, if your software CAN organize it's workload in a non-blocking manner, then you have a real question. The reason for this is that, in a non-blocking workload, all cores CAN be maintained at 100% utilization. Since we have this second situation, I am interested in the answer.

As far as whether you can run a program and play a game at the same time; gaming uses, for the most part, blocking threads which makes uniform core utilization hard to maintain. This will skew any attempted real world analysis and will vary on a game-to-game basis, IMHO. Therefore, I would say that if you are running CPU intensive blocking threads; get a low core count, high speed CPU. In the event of full core loading with non-blocking threads, I believe higher core count will yield better results. The chart at AnandTech (http://www.anandtech.com/bench/Product/147?vs=362) seems to confirm this. It indicates that in many of the multi-threaded renderings and encodings, more cores wins. You will also note that they do not ALWAYS win. This, I think, is due in part to the blocking vs non-blocking threads used by the software in the benchmark.

NOTE: I used the Phenom II X6 1055T and the Phenom II X4 980BE as they have the closest passmark synthetic scores. This means that, in terms of 100% utilization of processing power, they are equivalent. I chose not to find equivalent Intel CPUs because determining cores and equivalent performance is not as easy as with AMD (X2=2 core, X4 = 4 core, X6 = 6 core).

6 cores vs 4 questions and thoughts

Distinguished

Champion

Splendid

Distinguished

Distinguished

Distinguished

Distinguished

Titan

Distinguished

Splendid

Distinguished

Distinguished

Splendid

Distinguished

Distinguished

Distinguished

Splendid

Distinguished

Distinguished

Expert

Splendid

Honorable

Share this page