Sign in with
Sign up | Sign in
Your question
Closed

Dual-Core Versus Quad-Core: Part 2

Last response: in Reviews comments
Share
Anonymous
September 4, 2008 6:10:16 AM

The Intel Core 2 E8500 and the AMD Phenom X4 e9350 offer high performance within the same 65 W power envelope, and are sold at the same price point. As we’ll see AMD suffers defeat as Intel comes out on top, but this time it’s through no fault of its own.

Dual-Core Versus Quad-Core: Part 2 : Read more
September 4, 2008 6:29:51 AM

good work on getting the benchmarks out. but a better comparison would be amd quad vs amd dual. due to the fact that amd has the only real quad core. I think most if not all of us know that most apps aren't ready to scale well up to a full quad proc. keep in mind that amd has four ACTUAL cores on their procs, not two logiced out to four as intel does. Get programs that are fully optimized to run on four acutal cores, the benchmarks will change quite a bit. I honestly don't recommend a quad core to anyone for the price at this point in time unless they plan on keeping their computer upwards of at least 4 years due to the fact that software takes too long to catch up to hardware. Multi core scaling on the software side just isn't there yet. Look to amd's dual core offerings for a good price/performance ratio at this point in time.

but none the less... it's good to have some charts at this current point in time. thanks for the time put on the benchies... i'll click a sponsor or something. :-p
September 4, 2008 6:34:17 AM

Just wondering if there's a little error on the game benchmarks page... the graph for Supreme Commander and the text don't seem to agree.
Related resources
Anonymous
September 4, 2008 6:38:51 AM

little nit @ second last paragraph:

"are all examples showing that a 2.0 GHz quad core can certainly beat a sophisticated 3.16 GHz quad core"

should be 3.16 GHz dual core.
September 4, 2008 6:57:59 AM

The mainconcept analysis is wrong as well. Says the e8500 wins when it doesn't.
September 4, 2008 7:26:20 AM

nachowarrior cut the AMD true quad core BS - 4 cores total wether it be attached or not, the performance speaks for its self, plus intel was smarter not to make one huge processor etc - same as ATi's 4870 x2 - you should know that fanboy.

If you want to get technical lets compare Intel nehalem quad - no competition ;) 

Sloppy editing alright - its making AMD look good! :o  LOL
September 4, 2008 7:32:43 AM

Good idea, but the article is such a mess. Never seen anything like that on Toms. It needs editing, and needs it now. Shame.
September 4, 2008 7:36:55 AM

I still can not concieve why they are comparing a 3.0+ dual core to a 2.0ghz quad?

What is the real point of this article?
Anonymous
September 4, 2008 7:46:11 AM

one thing that the authors forget that a typical use for a computer isn't just decompressing, surfing or gaming. The typical use is decompressing AND surfing AND using a resource hog like Skype AT THE SAME TIME! Oh, did I hear BitTorrent or multiple YouTube flash videos? How about them fancy Flash Ads, about 3 of them in every one of those 20-30 open tabs in the browser? Why don't you compare a quad core and a dual core in such an environment for general performance and responsiveness?
Anonymous
September 4, 2008 7:47:08 AM

Maybe in Windows the time of the Quad core or even the Duo core hasn't arrived yet, but in Linux the multicore processors have been supported for a lot longer and I wouldn't be surprised if you find many more apps in Linux that are natively multithreaded. How about rerunning your comparison in Linux and see who's the winner there where neither processor has the advantage. Both are well supported in Linux where as many of the tests in Windows lopsidedly tainted toward Intel products. In other words, try a scenario where the processors are treated equally by the testing software.
September 4, 2008 8:08:33 AM

Thanks for the analysis catches, guys. They should have been, but weren't, caught during layout. I've adjusted the text to correctly reflect the benchmark results. Take care!
September 4, 2008 8:12:56 AM

I don't know why, but i would like to see an amd x2 @ 3.0 ghz running around those task.... It's cheaper and im really happy with it. Nice Article.
September 4, 2008 8:34:24 AM

Quote:
Intel’s fastest dual core processor.


It appears that the author is referring to the e8500 in the above statement, this would be incorrect considering the e8600 has newer stepping and a higher clock rate.

Quote:
Supreme Commander shows the same results: it runs much faster on the Intel dual core than it does on AMD’s quad core. Since the performance difference is 80%, the clock speed difference alone isn’t enough to account for the tremendous difference.


The chart shows otherwise, something maybe awry with the report.

Quote:
AMD Phenom X4 e9350


Should be AMD Phenom X4 9350e , "e" is misplaced.

Anyway, I would have liked to see what a Phenom 9950 and q6600 would have shown given the fact that their with in the same price point and would have shown the difference in efficiency and power.


September 4, 2008 11:38:14 AM

nachowarriorgood work on getting the benchmarks out. but a better comparison would be amd quad vs amd dual. due to the fact that amd has the only real quad core. I think most if not all of us know that most apps aren't ready to scale well up to a full quad proc. keep in mind that amd has four ACTUAL cores on their procs, not two logiced out to four as intel does. Get programs that are fully optimized to run on four acutal cores, the benchmarks will change quite a bit. I honestly don't recommend a quad core to anyone for the price at this point in time unless they plan on keeping their computer upwards of at least 4 years due to the fact that software takes too long to catch up to hardware. Multi core scaling on the software side just isn't there yet. Look to amd's dual core offerings for a good price/performance ratio at this point in time. but none the less... it's good to have some charts at this current point in time. thanks for the time put on the benchies... i'll click a sponsor or something. :-p


Sure it would, but the point of this article is to compare relatively similar costing processors with similar TDPs as a quad vs dual. Plus given how AMD is losing in the performance race clock for clock it emphasis on quad-core beating dual in some applications.

If it was AMD beating AMD everyone would be "so what?". Since it was (in some cases) AMD beating Intel, most go wow. The they compared the clockspeeds and wow. That really puts pressure on quad-core's performance. Great Work guys!=)
September 4, 2008 11:44:18 AM

I have done some tests on my K8 X2. Moreover, I've taken some interest in Xvid development.
- current Xvid code isn't multithreaded: it is purely single core! In fact, when I encode two videos in parallel, I get almost no speed impact from the second encoding upon the first. If you want to try a multithreaded Xvid encoder, you must compile the 1.2 CVS version.
- I bet this benchmark uses Koepi's build of Xvid 1.1.3; as far as I know, he builds it against the Pentium Pro instruction set.
- I compared Koepi's build compressing some video under Windows (32-bit) and one built directly on my K8, in Linux 64-bit + SSE2 compressing the same video: encoding speed went up by a factor of 2.5.
- ever since most Xvid developers were hired by Miro to work on Miro's MPEG4 codec, Xvid development slowed down. Many developers got interested in x264 instead.

In short, using Xvid to compare AMD and Intel processors isn't as good as it used to be. Either that, or since Xvid is one of the few very CPU-intensive benchmarks out there, you should try and build it yourselves for each platform - just to be sure. It would also be interesting to benchmark current CVS build, to see how it scales with more cores.
Anonymous
September 4, 2008 11:59:19 AM

agree with REappear. for hardcore multitaskers, like myself, the quad is clearly a winner. a couple of msn/skype windows AND 10+tabs AND running torrent AND playing Supreme Commander -its my favorite:) - on a 20x20+ map with 4-6 players does the performance hit. in this case, 2g of memory (supcom eats up between 1.3 and 1.8) and 2g of ram isn't enough anymore.

and there is a low TDP quadcore from intel, my q6600, doing 9x266@1.008 :D . its a wicked thing to see a q6600 reaching only 43C on a prime test.

AMD should put much more cache on their chips, in most of the benchs this is the reason why their CPU is so slow. oh, and efficient doesn't mean it should be this slow too.
Anonymous
September 4, 2008 12:00:53 PM

2g of memory and 2 cores... sry
September 4, 2008 12:15:01 PM

"One fact remains clear above all: our comparison has shown that the time for quad core processors just hasn’t arrived yet."

Uhh...duh.
September 4, 2008 12:26:14 PM

I think it is appropriate to run mixed benchmark with multiple tests at the same time. The outcome can be surprising. Working on quad systems fells different than on dual core,quads much more responsive if you run multiple tasks at the same time.
September 4, 2008 1:23:39 PM

Good article and outcome as expected but I must agree when conducting tests we need to run multiple apps in conjunction for a true everyday experience. I would be interested in seeing how 2xquad cores fair on some NLE video editing apps like Premier & Vegas. Can we have some test ran on Linux? .... Please? Linux is becoming more and more popular to people so this would be refreshing to see. Don't worry bout the typo's we all make mistakes, it's no big deal. No worries, Keep on ROCKn THG!!

The Silent Majority
Anonymous
September 4, 2008 1:38:23 PM

I agree with REapper, a multi-tasking benchmark will shed much more light on real-world user concerns. I also wonder what the interest of power efficiency in desktop cores is? I know everyone wants to be green, but for most desktop users that I know, heat and power aren't that big of an issue.
When it comes to multi-cpu workstations and clusters, heat IS a issue.
a b å Intel
September 4, 2008 1:58:47 PM

I have also got to agree with Reaper. When running lots of tasks, having a quad really helps. It's like looking at frame rates in games. I'd rather have an average of 60fps with dips to 45fps than an average of 100fps with dips to 20fps. A quad prevents other running apps from stuttering, even if their max speed is lower than it would be on the dual core.
September 4, 2008 2:09:23 PM

This is more like phenom vs. core 2 duo than quad core vs. dual core.
Anonymous
September 4, 2008 2:20:59 PM

this is like comparing windows xp 32-bit and 64-bit.. Applications aren't made for 64 bit so 32 bit wins. Same concept here, Compare intels Quad and amd's dual..
September 4, 2008 2:22:14 PM

Try using something other than the Overpriced (useless garbage) MSI board, use Gigabyte or Asus instead.
(you are using an excellent Asus board for the Core2, and a worthless, overpriced MSI board with the Phenom)

My question is WHY?
September 4, 2008 2:26:13 PM

For me, Tom's still has not done a proper comparison.

As others have mentioned, for the end user at home, multi-tasking performance matters. By that I mean multiple applications open and running.

I play Dark Age of Camelot and have 2 instances open, have Ventrilo up and running, and I surf the web with all these going.

Which performs better, duo or quad doing stuff like this?

That is the comparison I want to see.
Anonymous
September 4, 2008 2:31:13 PM

"One fact remains clear above all: our comparison has shown that the time for quad core processors just hasn’t arrived yet."

Actually, just about every major game that is to be released from this point on, the devs have said it will support a quad. From FarCry2, Stalker: Clear Skies, Alan Wake, Operation Flashpoint 2, ARMA2, and others.

If you are a gamer, why would you want to limit yourself now. And considering how easy and safe it is to overclock the Intel quads, it means you can be running one at the same speed as the e8500 in this test. My q6600 is at 3.01, and it is still near stock temps. I'm sure it would go faster, but I don't see the need, since like when I am running Crysis, all 4 cores are only at 75% or so.

I'm being bottlenecked by my 8800GTS 640 at the moment
September 4, 2008 3:08:15 PM

I understand where this article is trying to go, but why are you still running what are essentially single-threaded benchmarks on multi-core processors? (Yes, I realize that some of the apps are multi-threaded, but relatively few).

Try doing some "real-life" multi-task benchmarking. For instance, encode multiple videos at once and encode multiple MP3s at once. Or DVDShrink 2 or 3 discs at a time. Or combinations of all the above. It would be much easier to set up such benchmarks than to try to find instances of optimized multi-threaded applications.
September 4, 2008 3:15:37 PM

I wish the 9950BE had been included so we could have seen the low and the high of the phenoms compared to the e8500.
Anonymous
September 4, 2008 3:21:46 PM

I'm wondering why isn't a developers benchmark?.. What about a website running on Apache 2.2.9 + Postgresql 8.3 + Jmeter to generate load test? Those are all multi-threaded apps... what are you waiting for???? There are a lot of multi-threaded apps on the software developers world, and that's everyday use also.

I don't understand why you keep running single threaded apps on multi-core processors, is like using Blue Gene/L to do Powerpoint slides... doesn't make sense. Get serious about 'real life' use and start from there.

Regards
September 4, 2008 3:45:16 PM

Q6600 at Newegg $179.99, "free" OC to 3G/1333. Enough said.
September 4, 2008 4:07:44 PM

Most processes with data and instructions under 4MB don't run well on multi-threads, unless there's need to do so, such as heavy floating point calculation on separate arrays. I am now running a Phenom 9850 machine and I still wonder why I need a quad-core at home. Major game titles in the future will be heavy multi-threaders, but a fast dual-core will get people covered until late next year. Some professional software engineers that I know still know shit about multi-threaded programming, and have only vague idea about inter-thread communications and what semaphores are used for. It still takes time for them to learn and adapt the new paradigm of concurrent programming. Interestingly, objective-oriented programming seems to fit concurrent processing naturally.

Also, Intel and AMD both screwed up big time with quads, for now. Intel's current FSB model will be bottlenecking quads when all cores are demanding memory bandwidth. AMD Phenom's IMC just sucks for desktops. I'll explain why. We can split processes into two category: one that fits into 512KB L2, and the one that doesn't. For small processes, it has to be loaded into L3, then loaded to L2. For larger processes, it does the same. The only problem is that small processes don't choke IMC or FSB, but the IMC actually is bottlenecking the processing time by requiring them to be passing through L3. The problem is, L3 isn't really useful unless you have multi-threaded programs that requires shared data/instructions or really large data. That situation is usually not the case at desktop environment. There are two ways to get around this: direct L2 injection or run IMC/L3 as fast as you can to reduce latency. I haven't seen anybody doing direct L2 injection because that requires new instruction set, so both Intel and AMD are seemingly doing the second way. So, every wonder why Nehalem doesn't perform as good in games and why Intel brings back HyperTreading yet once again?

Back to the topic. Many people are still happy with their single core machines. I am still typing this reply on a Pentium M laptop. The cruel truth to Intel is that they already built the best desktop combo for the mass: Core 2 Duo. A lot of multitaskers claim the need for a quad. Huh, no quite so if they carefully monitor CPU usage. The fact is that fast dual core or fast single core with hyperthreading is all what they need. The need for more than two cores is still very specific and that doesn't justify the cost and power consumption. AMD even showed that a lackluster CPU with good chipset and perform multimedia task better than good CPU. That's hardly any news, since there was plenty MPEG-II decoding/encoding cards during 486 age.
September 4, 2008 4:08:56 PM

tracyfearsonFor me, Tom's still has not done a proper comparison.As others have mentioned, for the end user at home, multi-tasking performance matters. By that I mean multiple applications open and running.I play Dark Age of Camelot and have 2 instances open, have Ventrilo up and running, and I surf the web with all these going.Which performs better, duo or quad doing stuff like this?That is the comparison I want to see.


Tracy,

One of the challenges with a scenario like that is measuring performance accurately. If it was a seat-of-the-pants "feel"-based eval, it'd certainly get torn up. In that case, you're left with the option of running something fairly demanding (and with a consistent load) in the background as you try to run another benchmark in the foreground. It might prove to be a more accurate test, but the results aren't going to be very precise. At any rate, I'd like to hear some suggestions from the community of ways to achieve the same end while still delivering good results.

After all, when Nehalem pop up handling eight threads, this issue will only be amplified.
September 4, 2008 4:34:19 PM

scurvywombatI wish the 9950BE had been included so we could have seen the low and the high of the phenoms compared to the e8500.


Chances are 9950 will still lose to E8500 on single or dual threaded programs most of the time, and most first-person shooting games as well. The IMC and L3 structure of Phenoms aren't designed to deal with those situations, though they still perform reasonably good. For they price, you can't go wrong with both.
September 4, 2008 4:45:19 PM

megatasking vs multitasking

The E8500 will obviously run any single threaded or dual threaded application (like a game) better then the e9350 but if your into doing multiple things at the same time like: compressing, back-ups, encoding, trans-coding, rendering, surfing the web, watching video; then it is good to have a quad-core or two, and lots of ram, and a couple hard drives with NCQ.

I believe this article is geared towards showing AMD under a lime lite since it excludes an intel quadcore Q6XXX or Q9XXX, I would also would liked to have seen a comparison to the AMD X3s as well (since that is what AMD is putting up to compete with intel dual-cores)

IMO the E8500 is a better choice if you are buying a computer to play games and surf the web.

As a graphic artist with a lot of programs open, each doing different things: I would like to have a quad-core or two.

Power consumption is also a concern since you might have to get a more powerful P/S to run a quad-core (something more beefy then the e9350) and two graphics cards.
September 4, 2008 5:10:17 PM

Benchmarking needs to account for multi-core systems too. It doesnt reqire massive software changes by somebody else.
Just grab your favorite single threaded app (xvid? Lame? ..?) and run FOUR instances of it on both test systems.
who gets done first?
how much power is used?

publish results for all to see
(and no I dont know what the results will be)
September 4, 2008 5:26:58 PM

The question I see being asked in the forums all the time is E8500 versus Q9550. "Is it worth the price to get a quad now?"

Who wanted to see E8500 versus 9350e??? So, to beat a dead horse yet again, when is THG going to update the CPU charts with the latest CPUs, so we can compare all of them with various apps and games????
Anonymous
September 4, 2008 6:40:24 PM

"After all, when Nehalem pop up handling eight threads, this issue will only be amplified"

1.Nehalem won`t handle 8 threads!
HT is not about 8 threads, it`s about using the unused part of the processor WHEN POSSIBLE(cache miss&stuff).in fact it "stops" the thread waiting for a load in the cache or other time consuming operations (i didnt study that much)and runs another one until the first may become active again or the OS or the multithreaded program "awakes" a different thread.

It is obvious that the future is multithreading and just as a comment-er said, OO programming natively supports this.Games will be multithreaded. Almost all the tasks in the may, at some point, become multithreaded and those which are not, well one single core is enough to deal with them.
The main bottleneck is, as most of the people said, the OS & the software developers.

I currently study computer science in a 3d (or 2nd) world country and i know a little bit about software & hardware since i study them both. but not too much:D 
by the way: the comments were much more interesting than the article itself :) 

September 4, 2008 6:48:45 PM

Look, you've tested apples with pears.

So lets take it to the next level and make use of those 4 cores in a mature and effective way.
Bring in this: SetAffinityII (url: http://www.geocities.com/edgemeal_software/SetAffinity ).

Tune the quad core to run only system processes on core 0 and 1, run all tests/games/etc on core 2 and 3.

My guess is that it's not the hardware thats lacking performance but efficient use of it by software, and proper tuning.
September 4, 2008 7:21:58 PM

xX12amanXxI still can not concieve why they are comparing a 3.0+ dual core to a 2.0ghz quad?What is the real point of this article?


Um same thermal envelope and priceing maybe? What the hell. I swear every time AMD is compared to Intel there is always something.

Either the Intel is priced higher, or its at 45nm AMD is at 65nm, not the same thermal envelope blah freakin blah.

Why is it that people were gung ho to compare all AMD CPUs to Intel during the Prescott vs Athlon X2 era but now that AMD doesn't do so well they get mad when they compare them?

Its just like when Phenom first came out. Instead of only comparing Phenom to Athlon X2 they compared it to Core 2 Quad and Athlon X2. But people didn't like that, even though I would prefer to know which company offers the best bang for buck and not stay within just one company.

Seriously. I mean what are they supposed to do? Down clock the E8500 to 2GHz? That would be stupid really since you normally wont do that.
September 4, 2008 7:24:16 PM

mike__RO"After all, when Nehalem pop up handling eight threads, this issue will only be amplified"1.Nehalem won`t handle 8 threads!HT is not about 8 threads, it`s about using the unused part of the processor WHEN POSSIBLE(cache miss&stuff).in fact it "stops" the thread waiting for a load in the cache or other time consuming operations (i didnt study that much)and runs another one until the first may become active again or the OS or the multithreaded program "awakes" a different thread.It is obvious that the future is multithreading and just as a comment-er said, OO programming natively supports this.Games will be multithreaded. Almost all the tasks in the may, at some point, become multithreaded and those which are not, well one single core is enough to deal with them.The main bottleneck is, as most of the people said, the OS & the software developers.I currently study computer science in a 3d (or 2nd) world country and i know a little bit about software & hardware since i study them both. but not too muchby the way: the comments were much more interesting than the article itself


No not at first you are right. But Nehalem is set to go up to 8 physical cores with SMT giving it 16 threads.
Anonymous
September 4, 2008 7:54:21 PM

for sure ^ its not so much about the performance numbers... its just smoother...

like the step up from the 9800 gx2 to the gtx 280... the performance was close... but the gtx 280 just felt smoother and better

probably cause you paid more for it =P
September 4, 2008 8:08:06 PM

Why your at it could you please add an FX57 and at least half the programs only use a single core. How dumb is this going to get? We needed to add one more program we know is dual core to make the dual cores look good. Could you please do the tests over and either show core usage or any program you don't think was taking good advantage of all 4 core do them in multitasking. Run maybe a web surfing test while doing benchmarks as many would surf the web while running these programs.
September 4, 2008 8:13:22 PM

This setup they used is flawed. One cpu cost a lot more than the other. They consume the same amount of power, but that's just not important other than for laptops.

How about a review using e8400 and q6600, both at $180?
September 4, 2008 9:04:13 PM

"One of the challenges with a scenario like that is measuring performance accurately."

I agree.

That is the challenge because I do not know of a native benchmark that does this.

"If it was a seat-of-the-pants "feel"-based eval, it'd certainly get torn up."

Do it and let it get torn up. Need to start some where. Some information is better then nothing. Just be careful about making conclusions.

"In that case, you're left with the option of running something fairly demanding (and with a consistent load) in the background as you try to run another benchmark in the foreground. It might prove to be a more accurate test, but the results aren't going to be very precise."

Does it matter if they are not very precise? If the performance is close or significantly better is the question.

A lot of gamers play MMORPG. A lot of gamers play music while they game. A lot of gamers use voice programs while playing.

Set up to play songs, run a time trial in a popular game while running a voice chat with talk back and forth.

How does a quad core compare to a duo core then?

That would be a very interesting comparison.
September 4, 2008 9:08:03 PM

A $190 processor compared to one of the slowest AMD quad cores, when the FASTEST is $180...nice.

And I love how the efficiency is based purely on Crysis.
September 4, 2008 9:08:57 PM

jimmysmittyUm same thermal envelope and priceing maybe? What the hell. I swear every time AMD is compared to Intel there is always something.Either the Intel is priced higher, or its at 45nm AMD is at 65nm, not the same thermal envelope blah freakin blah.Why is it that people were gung ho to compare all AMD CPUs to Intel during the Prescott vs Athlon X2 era but now that AMD doesn't do so well they get mad when they compare them?Its just like when Phenom first came out. Instead of only comparing Phenom to Athlon X2 they compared it to Core 2 Quad and Athlon X2. But people didn't like that, even though I would prefer to know which company offers the best bang for buck and not stay within just one company.Seriously. I mean what are they supposed to do? Down clock the E8500 to 2GHz? That would be stupid really since you normally wont do that.


I honestly dont care what happened in the past but to compare a 3.1ghz dual core processor from Intel that has been proven time and time again to literally smoke anything AMD has currently to a low end 2.0ghz quad core from AMD that has been proven time again to be slower overall even being out paced by AMD's last generation x2 processors. Energy envelopes..lol what a joke everyone know's Intel has had 45nm dual's out for awhile now sounds to me like an excuse to creat this article...And pricing..well last i checked you could get a 2.6ghz quad from AMD for 190$ so how much sense does that make again?

Ask how much sense does this article make?

Next up QX9770 vs 9950 becuase they are both 140w!....Ya makes perfect sense.
September 4, 2008 9:15:29 PM

The argument about their power ratings both being 65w and that being a good reason to compare them is illogical.

They aren't measured the same way to begin with.
September 4, 2008 9:16:02 PM

To use the now outdated single-threaded XviD 1.1.3 is silly. Use much newer builds from Koepi's web site. This one in particular works much better on multi-core processors:

!