• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Dual vs Quad

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

rainless

Old Member
Joined
Jul 20, 2006
At what point does the speed advantage of a dual core CPU erase the core advantage of a quad?

Would, for example, an E8400 @ 4.5ghz be able to stomp a 9550 @ 3.65ghz in Adobe Premiere/multimedia rendering?

I'm looking mainly for links, but if anyone has experience I'll take that into consideration as well.
 
When I saw the title I thought "You gotta be kidding me, not another dual vs quad thread"....alas your question is very valid and seems easily answered by running a few rendering sequences to compare performance/base speeds.

If someone can suggest such rendering runs I'd be happy to run those for you to see what the data says :)

Edit...here is one done by Graysky : x264 Benchmark, a HD MPEG2 rendering : http://www.techarp.com/showarticle.aspx?artno=442
 
If you are talking about multi-threaded apps, (rendering, etc), and you are talking about two chips using the same architecture (45nm vs. 45nm), you can pretty much multiply the frequency by the number of cores to see which is more powerful. This doesn't work for single threaded applications (gaming) because they can't make use of the extra cores. So the higher clock speed wins.
 
And then there are programs that multi-thread but do not scale 100%. For example, they may only get 150% the performance of a single core on a dual core.

For programs that scale perfectly (or almost perfectly), just multiply frequency by number of cores. I BELIEVE (but could be wrong, since it's not my area of expertise) encoding / image processing falls in this category (if the program is multi-threaded).
 
If im reading the data right, a Q6600 at 3.6 hammered an E8400 at 4.2

That's pretty damned alarming...

Even if quads don't run well on my motherboard it's STILL alarming.

...almost alarming enough to make me pick up an i7.

And then there are programs that multi-thread but do not scale 100%. For example, they may only get 150% the performance of a single core on a dual core.

For programs that scale perfectly (or almost perfectly), just multiply frequency by number of cores. I BELIEVE (but could be wrong, since it's not my area of expertise) encoding / image processing falls in this category (if the program is multi-threaded).

Yes. Adobe Premiere is a multithread app. (Hell... even Photoshop is multi-thread.)
 
Last edited:
For instance I got a quad at work that is 2.0Ghz (booo damn thing is too slow sometimes for single threaded apps). Anyways comparing it to my home PC that was at 4.05Ghz dual core, yes it could come close to a 1:1 in 3D Renderings where Dual could pull ahead depending on the task at hand.

General rule I apply is...
1 Core = 100% Each additional core is roughly 80% effective.
So a Dual @ 4.0Ghz = 7.2Ghz
A Quad @ 2.0Ghz = 6.8Ghz
Which would be nearly right on for 80% of the tests I did with rendering times on certain scenes I've made.
 
For instance I got a quad at work that is 2.0Ghz (booo damn thing is too slow sometimes for single threaded apps). Anyways comparing it to my home PC that was at 4.05Ghz dual core, yes it could come close to a 1:1 in 3D Renderings where Dual could pull ahead depending on the task at hand.

General rule I apply is...
1 Core = 100% Each additional core is roughly 80% effective.
So a Dual @ 4.0Ghz = 7.2Ghz
A Quad @ 2.0Ghz = 6.8Ghz
Which would be nearly right on for 80% of the tests I did with rendering times on certain scenes I've made.

Interesting way of looking at it...

So architecture has little to do with it? 45nm etc...

I myself am looking at quad vs. dual for a workstation with rendering etc...
 
For instance I got a quad at work that is 2.0Ghz (booo damn thing is too slow sometimes for single threaded apps). Anyways comparing it to my home PC that was at 4.05Ghz dual core, yes it could come close to a 1:1 in 3D Renderings where Dual could pull ahead depending on the task at hand.

General rule I apply is...
1 Core = 100% Each additional core is roughly 80% effective.
So a Dual @ 4.0Ghz = 7.2Ghz
A Quad @ 2.0Ghz = 6.8Ghz
Which would be nearly right on for 80% of the tests I did with rendering times on certain scenes I've made.

Thanks deathman! Good to see you! I haven't seen you in a while...

Interesting way of looking at it...

So architecture has little to do with it? 45nm etc...

I myself am looking at quad vs. dual for a workstation with rendering etc...

Another thing you want to look at is cache:

http://www.ocforums.com/showpost.php?p=5982536&postcount=1

My E8400 has 6megs of cache... but a q8200 only has two megs of cache. I learned how important cache was in the 65nm C2D days. I have a very strong overclocking B2 E6400... I could get just as high as most E6600s could get, but I would get stomped by their cache everytime.

So a q8200 running at 3ghz should net 10,200 using deathman's theory... but because of the cache... it wouldn't be much faster than my 4ghz E8400 @ 7,200.

Now a q9550 would positively STOMP my *** at almost any speed. So if you're going to go quad... you've got to go BIG. Q9550... Q9650... i7...

...otherwise you might as well get a dual core.
 
12mb cache is overkill, which is why I guess core i7 only has 8mb inclusive as well. Core i7 in some benchmarks is more than 100% scaling from each core due to the fact it has hyperthreading, which shows in cinebench.
 
Thanks deathman! Good to see you! I haven't seen you in a while...



Another thing you want to look at is cache:

http://www.ocforums.com/showpost.php?p=5982536&postcount=1

My E8400 has 6megs of cache... but a q8200 only has two megs of cache. I learned how important cache was in the 65nm C2D days. I have a very strong overclocking B2 E6400... I could get just as high as most E6600s could get, but I would get stomped by their cache everytime.

So a q8200 running at 3ghz should net 10,200 using deathman's theory... but because of the cache... it wouldn't be much faster than my 4ghz E8400 @ 7,200.

Now a q9550 would positively STOMP my *** at almost any speed. So if you're going to go quad... you've got to go BIG. Q9550... Q9650... i7...

...otherwise you might as well get a dual core.

Reminds me of the old celeron vs. pentium debates with respect to clockspeed or cache benefits - history repeats itself:D

Thanks for the info :beer:
 
Interesting way of looking at it...

So architecture has little to do with it? 45nm etc...

I myself am looking at quad vs. dual for a workstation with rendering etc...

Well the architecture between mine are the same if not mistaken. Both 45nm yadda yadda yadda, just the Quad I know has more cache but its still 3Meg per core if you want to divide it up if not mistaken.

Thanks deathman! Good to see you! I haven't seen you in a while...

Haven't seen me? :( Man Im all over the place just jumping from section to section :) Just slacking lately in the game area.



rainless said:
Another thing you want to look at is cache:

http://www.ocforums.com/showpost.php?p=5982536&postcount=1

My E8400 has 6megs of cache... but a q8200 only has two megs of cache. I learned how important cache was in the 65nm C2D days. I have a very strong overclocking B2 E6400... I could get just as high as most E6600s could get, but I would get stomped by their cache everytime.

So a q8200 running at 3ghz should net 10,200 using deathman's theory... but because of the cache... it wouldn't be much faster than my 4ghz E8400 @ 7,200.

Now a q9550 would positively STOMP my *** at almost any speed. So if you're going to go quad... you've got to go BIG. Q9550... Q9650... i7...

...otherwise you might as well get a dual core.

Cache is important in some things, and in others it isn't. Its a toss up it seems. Certain games you can see a difference in while others not so much. Take your pick really but 2-3Megs of cache per core IMO seems to be a nice idea right now for intel's, AMD's thats a different story since they use L3 cache.

12mb cache is overkill, which is why I guess core i7 only has 8mb inclusive as well. Core i7 in some benchmarks is more than 100% scaling from each core due to the fact it has hyperthreading, which shows in cinebench.

12mb isn't overkill, at least in the C2D/C2Q erra. Without the onboard controllers the more it could get the better since it took longer to get the data in/out of the CPU itself to the memory. Now its not nearly as necessary but still its important.
 
Large cache is only important when a program accesses many different parts of memory unpredictably, and repeatedly. Extremely few programs exhibit that kind of memory behaviour, so I would guess large caches (>3-4MB) is not all that important. Most memory-intensive programs access memory sequentially, and don't really benefit from bigger cache (since by the time it needs the same data again for the next iteration, it's going to be a cache-miss anyways, unless you have a few hundred MBs of cache).
 
Large cache is only important when a program accesses many different parts of memory unpredictably, and repeatedly. Extremely few programs exhibit that kind of memory behaviour, so I would guess large caches (>3-4MB) is not all that important. Most memory-intensive programs access memory sequentially, and don't really benefit from bigger cache (since by the time it needs the same data again for the next iteration, it's going to be a cache-miss anyways, unless you have a few hundred MBs of cache).

depends on the applications and what your doing really.

Which is why I created this thread specifically about Adobe Premiere Pro and other multimedia rendering programs. ;)

I could care less how games "perform"... I mean I can PLAY all the games I've tried to play. So the underlying performance doesn't concern me.

What I'm concerned about is how long I have to sit in front of the computer and wait for things to render. (For some odd reason it always seems like Flash takes longer than Premiere Pro... maybe that will be different when I switch to Vista 64...)
 
quad core for you definetly, Adobe products have been multicore / processor optimized almost since their invention.
 
Back