Okay, so by now most people know that the all the modern AMD chips run on an architecture where instead of individual cores it uses "Modules" where in each module there are 2 cores that share L2 and L3 cache. This is often the attributing factor to the lower IPC that these chips experience. BUT, there is a bios setting in most motherboards (Gigabyte, Asrock, and MSI for sure) that have an option of enabling One Core per CU (computational unit). This setting essentially turns the FX 8350 into a Quad core but giving each core full access to the L2 and L3 cache instead of simply turning off two modules. I'm working on the benchmarks and results to see just how powerful these cores are when they aren't restricted by sharing resources.
Stay tuned, the numbers are getting interesting ;)
*Edit* The purpose of this isn't to compare 4 module/ 8 cores to 4 module/ 4 cores, it's really to compare 2 modules/ 4 cores vs 4 modules/ 4 cores.
I will be including 8 core results as a baseline, and I do NOT expect 4m/4c to outperform 4m/8c.
it's simply to see IF there is a delta between the module vs core structure and measure that change.
It's basically to see what impact sharing resources between cores and having fully independent cores makes with these cpus. More of an experiment than anything.
For example I've seen substantial gains in both Planetside 2 and Guild Wars 2. But I'm working on all the metrics, so stick around :) we'll see if it does anything useful in the end.
From my understanding the penalty for loading both threads on a module has decreased significantly on more recent releases such as steamroller and presumably excavator as well.
It will certainly be interesting to see the results though.
I ran some tests like this a long time ago now. In most things you get anywhere between 5-10% extra performance. But in the grand scheme of things it isn't really beneficial.
This is such a turbo vs displacement discussion lolz... in the end, it's the efficiency for the specific trajectory that counts, you can test all you want, you'll never get results that mean anything to anyone else...
For starters, do you run tests in Windows? Great, that doesn't know how to handle multiple cores if you throw a stick at it lolz...
Hardware is so boring these days... everything just runs everything, and discussions about hardware performance are such spitting competitions. The fact of the matter is, that an APU or a Pentium Anniversary runs most things just fine, and if it weren't for the software downgrade of Windows, it would also run 4k games at 250 fps...
Run controlled tests that mimic your normal workloads. My hypothesis is that your tests will see performance increases in some workloads and decreases in others.
What I'm saying is, you might see an increase in synthetic tests, but decreases in real world use (most likely unnoticeable without some sort of software monitoring telling you there is a difference) . Or vice versa. So just run tests on what you normally do.
if you have a motherboard that has the capabillity of doing this, (Asus boards dont have this capabillity), Then you will increase the per core performance of the cpu, because you will get a 256bit fpu availeble for 1 core.
However, like i said, Asus boards can only disable complete modules, not just individual cores. So in the case of an Asus board, you will decrease the performance by disabling modules.
So you need to be sure, that your mobo, is capable to disable specific cores 2/4/6/8
The OP specifically did not mention ASUS leading us to believe that indeed ASUS boards do not support this. Otherwise I am sure he would have mentioned ASUS originally, considering that ASUS boards are well esteemed.
Stackz0r likely felt like you were originally saying this because no mention of ASUS was directly made in the original post that you were in a way chiding the OP by sharing information about this from your experience. However I find this accusation to be wrong on his part, and your defense is valid.
However, you seemed to outlash a bit harshly towards Stackz0r considering that he was just concerned about OP's feelings, there isn't a need to threaten him.
It seemed to me as if she assumed that the OP did not know that Asus boards do not support this, and that he may possibly have an Asus motherboard, and possibly inferring that he would not be able to perform said test.
What she did not see however was that he had already begun his testing, therefore he obviously does not have an Asus board. Her wording in her post was directed at the OP (I get that the info is still useful, and that English is not her first language, but she still should have worded it differently) which could be perceived as an insult to the OP's knowledge. A way of saying "been there done that, that's not going to work" even though he already had it working.
If I was the OP, and Mistery were a personal friend of mine, my response would have been something along the lines of "I know what the fuck I'm doing." But she's not, so I tried to approach the situation quasi-politely. Some people just word things like Mistery did subconsciously, and many of us are guilty of it at some point (cognitive dissonance is a human instinct), so I was pointing it out so that she can avoid that for future reference.