AMD FX-8150 Overclocking

[jkhoward]

AMD FX-8150 Overclocking
by Joel Howard

Quote
I have managed to overclock my FX-8150 by disabling one core on each module. I was able to over clock the processor to just under 5GHz on a Corsair H100 with the maximum temp of 61C with a load. I ran multiple system benchmarks with the CPU and received a better rating with the four core processor with every test. The reason I find this to be true is that since one core at each module are disabled you are allocating more cache to that core instead of sharing it with another core. One of the main problems with the FX-8150 CPU was memory sharing between the two cores which would slow down the clock speed. This is no longer the issue when disabling one core per module. The chip I received was a lower binned chip because it has some temp issues not allowing me to go much higher on the voltages and my motherboard will not let me go over 5GHz due to the HT/NB frequencies. I recommend doing this with the FX-8150. I have also noticed my Windows loading time has increased signifiantly and now loads before the Windows loading screen comes together. One last thing, the memory scores at the same GHz as the eight core has jumped quite a bit once disabling the cores. I am not sure why, but I think it has do with the available bandwidth from the cores to the RAM sticks. I hope this helps!

The above is a post on another forum while I was working on this overclock for the past two days. Below are some benchmarks using Passmark 7 CPU and Memory overall system scores. The idea behind this overclock is disabling one core on each module to allow the single core to have access to more cache memory increasing single threaded performance at the same clock. The way I achieved this is by overclocking the CPU with all cores enabled until it became unstable. Once that point was reached I disabled one core on each module turning my eight into a quad core CPU. I was limited to how high I could overclock do to my systems RAM and motherboard. With the right setup I feel that you can get performance close to the latest Intel processors. i5-3570K? i7-3770K? Maybe.. but realistically the speeds will closer to the i5-2500K and i7-2600K Intel processors.



Passmark 7 (x64)

EIGHT CORE SCORE @ 4.5GHz
Multi-threaded CPU: 10,567
RAM: 1,487

Single Threaded CPU: 1,320

QUAD CORE SCORE @ 4.5GHz
Multi-threaded CPU: 6,269
RAM: 1,483

Single Threaded CPU: 1,567


UPDATE:
______________________________________
EIGHT CORE SINGLE: 1,320
QUAD CORE SINGLE: 1,567

1,567 > 1,320 FOR REAL WORLD PERFORMANCE
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

To obtain the single threaded score I divided the multi-threaded score on how many cores the CPU was using. The reason the quad core has a higher speed is due to the fact that it can use all of the shared cache just for one core which was a flaw in the Bulldozer architecture.

To achieve this you must have a motherboard that supports disabling one core at each module. My current motherboard the GIGABYTE GA-990XA-UD3 does has this feature along with the GIGABYTE GA-990XA-UD5 motherboard. This test was stable at 4.5GHz and not any higher due to my CPU having a high CPU VID.. damnit! Also.. Window boots 5 seconds faster! Good luck with your overclocks!

System Specifications:
Motherboard: GIGABYTE GA-990XA-UD3
Graphics Card: GIGABYTE GV-R785OC-2GD
Power Supply: NZXT HALE82 HALE82-850-M
CPU: AMD FX-8150
SSD: SanDisk Extreme SDSSDX-120G-G25
RAM: G.SKILL Ares Series 8GB DR3 2133
Heatsink: CORSAIR H100



What do you guys think? This is my post from my personal forum: http://12tronics.com/index.php?/topic/397-amd-fx-8150-overclocking/

EDIT: NEW BENCHMARKS WITH VANTAGE NOW AT LINK ABOVE

 

[mocchan]

I've actually been very curious about this, so thank you very much

Also, would it be possible for you to test Cinebench 11.5 with your 'Quad' Core? Then compare with the FX-8150 with all modules activated (Single core test+normal CPU test). I would LOVE to see the results.

 

[jkhoward]

I could do that.. I just need some time on my hands before I can! I will try my best and get back to you ASAP. But yeah.. I was pretty impressed myself. I basically built a AMD system at work to find out it was a flop. I was determined to find a way to fix it.. so I did. But yeah.. I will try to find some time to do this. I just hate setting it back to default settings.

 

[mocchan]

I know what you mean! But thank you! :lol: Appreciate it a lot

 

[4745454b]

I must be missing something. I don't see from the passmark scores how/why the quad would be faster. Nearly everything is even, but the multi CPU score is a lot lower on the quad. Any other benchies that shows this idea better?

 

[mousseng]

I must be missing something. I don't see from the passmark scores how/why the quad would be faster. Nearly everything is even, but the multi CPU score is a lot lower on the quad. Any other benchies that shows this idea better?

Because the quad version is roughly 19% faster per core than the octo, and most games take advantage of no more than 4 cores (with BF3 being the only exception I know of, at 6 cores). Edit: This is all theoretical, of course, doing math on synthetic benchmarks.

 

[4745454b]

He got the single core score by dividing the total by the number of cores active. There was no true test of a single core. He should also run the tests a third time with 1,2,3, and 4 active to see if the scores are different as well.

 

[sarinaide]

The foundations for the test bench is correct though and many have found that disabling the psuedo core improves performance sometimes dramatically, but then again everyone knew this a long time ago, shared latency problems and deep pipelines, you remove that aspect and the FX chips are still capable despite the issues.

 

[4745454b]

Disabling half the cores won't shorten the pipeline or reduce latency.

 

[jkhoward]

I will be running a few benchmarks on my system. Today's software does not utilize eight cores in any processors unless you are doing some serious video encoding. For games and other software four cores are the most you will see you computer use. With that said, due to this reason the faster you can make those four cores the faster your performance will be will for real life applications. As I stated above.. my Windows boots much faster then it does with the eight cores enabled.

Does anyone want me to run any benchmarks besides Passmark?

 

[sarinaide]

Run them, the disabling even cores has been met with skepticism though it does work, if you can give us a fair range of synthetics just to find a median of difference made.

 

[jkhoward]

Run them, the disabling even cores has been met with skepticism though it does work, if you can give us a fair range of synthetics just to find a median of difference made.

Okay. This weekend I should be able to make some time to run my system with eight cores and then run it with the four cores at the same clock speed to get us some numbers. I will run Passmark, Cinebench and a few others just to get us some results! But.. I have seen performance gains on all of my programs alone.

Thanks guys!
http://12tronics.com/

 

[jkhoward]

Also.. I would like to try this on a motherboard that has a 'Auto Overclock' feature.. I cannot seem to get my system to reach anything higher then 4.5GHz on all eight cores. And.. the CPU VID is pretty bad on my chip being 1.365V.. DONATIONS ANYONE?!

 

[jkhoward]

I will be updating this post sometime next week with a plethora of benchmarks between the 8-core and 4-core versions of the FX-8150. Please check up at http://12tronics.com/index.php?/topic/397-amd-fx-8150-overclocking/ if you are interested!

 

[truegenius]

try cinebench, vantage, 3d mark

also try this method
1. run prime95 on all cores, then stop the prime testing on every second core of every module.
2. these tests 2 modules (4 cores) (without prime95).
3. run these tests 4 core 4 module ( by disabling 1 core per module).
4. run these tests on all cores.

and compare them

a huge list of tests

 

[jkhoward]

try cinebench, vantage, 3d mark

also try this method
1. run prime95 on all cores, then stop the prime testing on every second core of every module.
2. these tests 2 modules (4 cores) (without prime95).
3. run these tests 4 core 4 module ( by disabling 1 core per module).
4. run these tests on all cores.

and compare them

a huge list of tests

I do not see the reason to run the Prime95 test on the cores in that manner. Why would I be doing that if I will just be benchmarking?! I have added Vanage to the list of programs that I will be running. And I will be running the test on 8-cores, 4-cores (two modules,) and 4-cores (four modules.)

 

[Mousemonkey]

This topic has been moved from the section CPU & Components to section Overclocking by Mousemonkey

 

[stew198269]

hey i use the fx 8150 also and i tried everything that is going on here even got it up to just under 5ghz with only 4 cores tho needless to say tho it didnt stay stable i didnt really try much volts with it but may give it another go.
the only sweet spot i could find was at 4.5 with 8 cores on!

heres ome pics of when i got it to 4.9ghz

 

[jkhoward]

hey i use the fx 8150 also and i tried everything that is going on here even got it up to just under 5ghz with only 4 cores tho needless to say tho it didnt stay stable i didnt really try much volts with it but may give it another go.
the only sweet spot i could find was at 4.5 with 8 cores on!

heres ome pics of when i got it to 4.9ghz
http://i1261.photobucket.com/albums/ii582/stew50/DSC00416.jpg
http://i1261.photobucket.com/albums/ii582/stew50/DSC00417.jpg

To test this out you do not even need to overclock the quad core more then the eight cores. You get to a stable setup with the eight core and then disable one core at each module. In theory this will allow you to overclock a little bit higher but I was unable to due to my motherboard. I really want someone with a Asus Sabertooth to disable the four cores and run the Auto OC option to see how high the board will put the CPU.

I am stable at 4.5GHz.. I will be doing much more work trying to push it higher. I just redid my case setup and CPU thermal paste and it dropped my temperatures by 10C! I now have some more room to play around with it.

 

[stew198269]

im sure i not what u going on about but when i tried on the auto oc it only went up to 3.9ghz and then died on me it couldnt and wouldnt give it enough volts in auto but then the only way i could get over 3.9 was to do it manual in bios which the top i can get is 4.5ghz stable.
been trying today with the whole turning off 4 cores and can get 4.8 stable on 4 cores with temps as high as 61 max thats with crosshair v motherboard and h70 and all fans on max plus had to stick 2 extra in case just to be on safe side

 

[jkhoward]

Well, normally the Auto overclock will make the computer stable and give you a good test to run some benchmarks on. But I guess not! Thank you for giving that a shot. I was only able to reach 4.5GHz on my setup due to my motherboard. Can you run some benchmarks on the single core speed and post them here versus the multi-threaded score?

 

[garage1217]

Good read. Watching for more benchmarks via 3dmark and so forth.

 

[jkhoward]

Guild Wars 2 was released this weekend.. I am still here! I will be doing the benchmarks sometime this week!

 

[garage1217]

Any news on 3d benchmarks?

 

[jkhoward]

Any news on 3d benchmarks?

I have been horribly busy with school and work. This weekend I will be finalizing my setup and will be running the benchmarks. Sorry for keeping you guys waiting!