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

4 cores 3.92GHz or 5 cores 3GHz?

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

xtcvv2

New Member
Joined
Oct 2, 2012
So just a quick question.

Is it better to have 4 cores at 3.92GHz or to have 5 cores at 3GHz? :eh?:
 
Open ended question... depends on the application.. but honestly, who knows since the clockspeed difference is so big. Id almost imagine 4Ghz 4 cores would be better... Test it and report back...
 
Test how? With some stress testing/benchmarking program?
 
I thought you had an application in mind this was to be used for? What type of applications are you using? If its just gaming, the 4 core. If whatever you are doing is multi threaded, then, well its a crapshoot you need to test.
 
Well I use my pc for gaming, video editing and photo editing.

So if I would be able to OC the 5 cores to something higher then would that be preferred?
 
Assuming a perfect multithreading scenario, (meaning, it is completely down to the processors themselves and the program is made in a way that scales perfectly with the number of cores and processor speed) four cores at 3.92 GHz is going to be faster by about 5%. In single threaded applications, the higher core speed will be 30% faster. Given the two choices in the first post, I would go with four cores, no question.

In the real world, the answer may vary. For example, if the program doesn't scale well beyond four threads, you would get a huge performance penalty for the extra core and slower clock speed. The program could also hit other bottlenecks, such as memory bandwidth or hard disk throughput. However, this case is pretty clear cut, so I would be surprised if I was incorrect.
 
Okay, so when I have my cpu back to 4 cores at 3.92GHz what should I have my NB freq and DRAM at? I usually have everything except FSB, CPU Multiplier and CPU Voltage on auto. I'm not very good at this kind of stuff.
 
I just put my cpu back to 4 cores at 3.92GHz and I started stress testing and I got an error on core #2...I didn't get that before and I have the exact same settings I did before.

Does this mean I have broken my cpu?
________________________________

Scratch that, it just passed a stress test...don't know why it would first report an error...
 
Last edited:
So just a quick question.

Is it better to have 4 cores at 3.92GHz or to have 5 cores at 3GHz? :eh?:

Open ended question... depends on the application.. but honestly, who knows since the clockspeed difference is so big. Id almost imagine 4Ghz 4 cores would be better... Test it and report back...

Phenom II core scaling is ~3.85x per 4 cores...multi-threading efficiency decreases for every core/thread you add, but lets assume ~4.8 for 5 cores.

3.92 * 3.85x = 15.09 "eBay GHz" :D
3 * 4.8x = 14.4 "eBay GHz"

No benching is needed, 3.9 GHz will win in all multi-threaded scenarios (and win up to 30% in single thread) provided similar CPU-NB and DRAM clocks.
 
Last edited:
I just put my cpu back to 4 cores at 3.92GHz and I started stress testing and I got an error on core #2...I didn't get that before and I have the exact same settings I did before.

Does this mean I have broken my cpu?
________________________________

Scratch that, it just passed a stress test...don't know why it would first report an error...
Your processor is fine. It is incredibly difficult to kill a modern processor. Or you have to be incredibly stupid, like me.

What program are you using to stress test? If you aren't using LinX, you can get it off my website here (their download servers have been offline for months). To run it, click the "All" button after launching the program, change the number of iterations to something large (>50), and hit Start. Keep an eye on temperatures and shut the program down if they get too high. You will want to run this for a few hours.

What temperatures are you seeing under the stress test? If you aren't using CoreTemp, you can get it here. This program is written by our very own The Coolest. It is the best out there.
 
thideras, Prime95 is optimal on AMD platforms. I wholly agree about intel platforms though.

CoreTemp is written by The Coolest! That's really cool. I had been recommending it to people so they could read Core VID but other members here discourage it because they think it is inaccurate.
 
I've found that LinX works just as well on AMD systems. It always finds instability faster and producer higher temperatures. :shrug:
 
I've found that LinX works just as well on AMD systems. It always finds instability faster and producer higher temperatures. :shrug:
I've found it possible in the past to pass 100 pass LinX but fail prime rather quickly :confused:

Also on the new BD CPUs LinX is not even close as far as heat output as Prime95, nor does it catch errors as quickly, I can run almost 5 GHz in LinX no problem but fail Prime95 at 200-300 MHz lower for same voltage. :eh?:
 
I will admit that I haven't run it on Bulldozer. My server is running Magny Cours. That is probably why we are seeing different results. I keep forgetting there are large differences between releases on AMD chips.
 
Back