AMD FX-6300 significantly better performance if only 4 cores enabled

Toggle sidebar Toggle sidebar
P

Phantomlvr

Commendable
Jun 19, 2016
8
0
1,510
AMD FX-6300
Gigabyte GA-990FXA-UD3 (Not sure rev, I think 3 or 4)
Gigabyte HD7770 1GB
16GB Corsair Vengeance RAM


So I was tweaking my overclock settings (CPU and Graphics) and using Unigine Valley to benchmark. Last night I managed to get small increases which i was happy with. Today I noticed in Windows Resource Monitor that only 4 of the 6 cores were displaying and of those 2 were parked. I followed the steps noted in other posts (msconfig>boot>advanced options, unchecked both processor and memory options, restarted. Note that the pulldowns showed only 4 processors. I think memory was only seeing half my RAM as well). When the PC came back up, resource monitor showed all 6 cores now and 16 GB of RAM. I ran Unigine to bask in the joyful improvements only to get stats that were 50-60% of the results with 4 cores. In the Unigine stats that I saved for each run, you can see that it only identified 2 cpu modules (4 cores) yesterday and 3 (6 cores) today (the "x2" and "x3" in the CPU model line). Both of these runs are at the same GPU, Mem and GPU Voltage settings.

Screenshots of tests
4-cores
6-cores

If I go back into the boot settings and force only 4-cores the stats improve again. Any reason why 6-cores is significantly worse than 4 and anyway to get 6-cores working efficiently?

Thanks
 
Sort by date Sort by votes
P

Phantomlvr

Commendable
Jun 19, 2016
8
0
1,510


Thanks for the feedback. Unengine Valley clearly shows the symptoms of the issue. I have changed the settings, rebooted and re-tested numerous times and every time there is significant performance degradation with 6-cores enabled. In all of the test cases that I am using for my argument the tests were run shortly after a fresh reboot, after the start-up processes had settled and with no other apps running except my AV and the hardware monitoring tool, Afterburner or HWMonitor.
 
Upvote 0 Downvote
P

Phantomlvr

Commendable
Jun 19, 2016
8
0
1,510


Thanks for the response. After reading your message I enabled logging in MSI Afterburner and re-ran the tests

MSI Afterburner log - 6-cores

MSI Afterburner log - 4-cores

While the CPU usage looks about the same the interesting difference is the behaviour of the GPU. With 4-cores it pretty much runs maxed for the whole test. With 6-cores though it looks to average about 60-75%. :??:

I can't spend more time this morning on it, but I will have another look this evening.
I'm thinking of trying the following:
- setting the CPU back to default and see if that changes anything
- disabling parking on the CPUs, just out of curiousity

I'm not at all familiar with pc hw architecture so I'm clutching at straws. Is it possible that the overhead of the extra cores is consuming resources that the system needs to run the GPU?

I have a Silverstone 600W power supply btw in case the power question comes up.
 
Upvote 0 Downvote
P

Phantomlvr

Commendable
Jun 19, 2016
8
0
1,510


Sorry for no response yesterday, I was updating my drivers and ran into trouble with Win7 updates. I am still trying to fix that. I have downloaded Furmark and will give that a run and post once I'm done with the Win7 stuff. Overall the system feel less responsive with the extra cores enabled

Cheers
 
Upvote 0 Downvote
gamerk316

gamerk316

Glorious
Jul 8, 2008
13,233
130
44,840
Odd; Unigene is heavily GPU performance biased, and I really can't see a case where disabling cores hurts performance in the way you're seeing.

First thing to check is to ensure with all the cores enabled, clock speed is correct. If I'm reading those log's from Afterburner right, the CPU is running at 1.2 GHz, which is WAY underclocked.
 
Upvote 0 Downvote
G

GhislainG

Titan
Jul 16, 2008
23,962
5
74,965
You read this thread incorrectly; disabling cores improves performance. Unigine Valley benchmark score is higher with 4 cores than it is with 6 cores; this is what's odd.
 
Upvote 0 Downvote
gamerk316

gamerk316

Glorious
Jul 8, 2008
13,233
130
44,840


I know, but from a technical perspective, that doesn't make any sense.
 
Upvote 0 Downvote
P

Phantomlvr

Commendable
Jun 19, 2016
8
0
1,510


Thanks mate. Standard clock for the GPU is 1050Mhz so until I work this out I won't push the GPU much more. I'm even thinking of dropping it back to factory just for comparison between the 2 settings. To make sure heat/power/something else aren't the cause. (EDIT Corrected CPU to GPU)
 
Upvote 0 Downvote
P

Phantomlvr

Commendable
Jun 19, 2016
8
0
1,510


So after updating my drivers (and dealing with Win7 updates) I ran Furmark. With Furmark I don't see any difference between the number of cores. I only get ~19fps (yay :( ). Looking at the MSI Afterburner graphs Furmark barely touches the CPU and therefore the GPU runs at 100% for the whole test. Re-testing with Valley after the driver update shows the same results as before. Valley is clearly more CPU intensive than Furmark from the Afterburner graphs. One thing I just thought of that I haven't done is to compare the MSI Afterburner graphs if I run the AMD Overdrive benchmark, which is CPU only. This might show whether the problem is CPU or CPU+GPU related. I'll do that test later and post.

Thanks for sticking with me.
 
Upvote 0 Downvote
P

Phantomlvr

Commendable
Jun 19, 2016
8
0
1,510
Thanks. Because it was altering the CPU settings that caused the change, I was leaning away from it being a GPU issue.

Isn't it the case though that, regardless of Valley's usefulness as a benchmark tool for my system, it is still concerning that the results are worse with the extra cores? I would have thought that any benchmark tool would have shown better results with the full 6-cores enabled. Maybe the tool might not accurately represent framerates, etc but the results would have been better when I enabled the cores?

 
Upvote 0 Downvote
P

Phantomlvr

Commendable
Jun 19, 2016
8
0
1,510
Drumroll please... Problem is sorted. Thanks so much for your help on this. I would not have resolved this as quickly without your assistance.

TL;DR - Changing my power settings from Powersave to Balanced or Performance. (Plus installing KB2645594 & KB2646060 but I don't know if they actually did anything other than help explain the symptoms.)

Reason (IMHO): CPU usage using Unigine Valley was at the threshold of parking(or whatever), which meant the OS was inefficiently yo-yoing the cores, causing slowness (But that's a guess from the symptoms and the info in KB2645594)

Details: I believe the reason that 4-cores was better than 6-cores initially was because the CPU usage for each core was higher on average and therefore they weren't trying to powerdown. Under heavy load everything worked nicely but reduce the load a little and things started yo-yoing. One of the patches specifically mentioned moderate CPU load as a symptom. Then reading another article on disabling CPU parking which talked about using powercfg commands and then I clicked and thought "what are my power settings"

So thanks a ton! Time to go test some of the games that had been jittery in the past.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom