New Here - Thoughts on my initial OC?

[SouthernStyle]

Howdy, after coming to this forum for a few months now I have finally given in and signed up. After recently upgrading from an ASUS 870EVO and AMD 1055T to an ASUS Saberooth 990FX R2.0 and an FX-8350 I started getting interested into OCing. My searches led me here and I've been coming here everyday since just reading around. I've gone from knowing nothing about real OC'ing to being pretty comfortable with tweaking things just from coming to this site. Now I'm hooked, and still learning. So thanks to everyone here who has posted guides, info, etc.

I would appreciate any thoughts or advice on my current 24/7 OC. I am definitely upgrading my PSU quite soon to an AX860i before I'm very comfortable taking my OC further. Also considering seeing how much of an upgrade a Swiftech H220 would provide. Then most likely a 7950 and an SSD. This PC started as a prebuild from CyberPower PC, and since I have replaced most components that came with it. I use this PC for some gaming and lots of design work in Photoshop, Illustrator and After Effects.

Ambient temp at the time was 23.3 C.

 

[trents]

SouthernStyle, I cannot read the values in CPU-z as when you lump a bunch of images together like that and then capture the whole desktop it shrinks things down when you attach the image. The built in forum attachment tool automatically downsizes things to fit. Please break that clump up into sub units. You can initially attach up to three images with a post but you can go back and add more if necessary. Yeah, I know, it's more work.

 

[Mandrake4565]

How long have you run prime for at those clock speeds? You temps look ok, I'm actually surprised that the Noctua isn't cooling the cpu better. Also do you have a fan on the right side of the case behind the motherboard? I ask because I believe you may be the first person I've seen with an FX processor where the cpu socket temp is lower then the core/package temp.

 

[SouthernStyle]

Sorry, I'm used to Firefox. I also wanted the Prime feed in there, as I was curious if anyone had advice on what it means that Worker 6 only completes 10 tests while all the others complete 27-30. I found it odd that the module holding cores 5-6 has the fastest and the slowest cores. Core 5 did 30 tests, and 6 did only 10. While testing I found that at a 19.5 multi and an HTT of 234 my OC was 4578mhz, and during that Prime run core 3 was by far the slowest, only completing 4 tests in about the same time. And at that time core 6 was completing tests at the same rate as all other cores besides 3. I was curious as to what could cause/fix what was (seemingly to me) such drastic changes. This was at the same VCore by the way.

 

[SouthernStyle]

How long have you run prime for at those clock speeds? You temps look ok, I'm actually surprised that the Noctua isn't cooling the cpu better. Also do you have a fan on the right side of the case behind the motherboard? I ask because I believe you may be the first person I've seen with an FX processor where the cpu socket temp is lower then the core/package temp.

I've ran prime for a little over 6 hours with it once. And yes I do have a fan intaking air behind the CPU socket. It was the fan that came on the stock CPU cooler. I've been meaning to test having it as an exhaust but just haven't done it yet.

 

[Mandrake4565]

Ok 6 hours prime is a thumbs up. With the fan on the back of the motherboard I found that using it as an exhaust was preferable to intake. I believe that with it as an intake it can sometimes trick the temp sensors for the Cpu socket into believing they are cooler then they really are. either way it is beneficial but I was having an issue with stability at O/C's above 5.1. I had heard that, the temp sensors could being tricked could be an issue and found by turning it around I was no longer having stability issues. It may be an isolated case but I thought I would just give you a heads up if you push it harder. Another thing I see is, if you push harder and start finding instability you may want to raise the Bank Cycle Time(trc) 31 clocks at least to the Jedec #3 profile 37 clocks. I'm no expert with memory timings but it may cause some instability in the future. Though passing prime for 6 hours I would leave it as is for now.

 

[trents]

S. Style, do you have the green stuff still enabled?

 

[SouthernStyle]

Mandrake, I appreciate the tip. I noticed after I put that fan in that my socket temp was reporting way lower, but I did figure it was a false reading. I'll probly switch it around today while I'm thinking about it.

Trents - No, I have everything disabled.

 

[RGone]

what it means that Worker 6 only completes 10 tests while all the others complete 27-30. = Using the smaller monitor that is on my CHV + FX-8350 setup, it was very hard for me to see all the workers. I had to really adjust each worker in the woker docker to see that I had had workers fail and I could not read the failure notice. That maybe what you are seeing with a core doing 10 tests and quits and the other workers run 27 to 30 tests.
RGone...

 

[Mandrake4565]

what it means that Worker 6 only completes 10 tests while all the others complete 27-30. = Using the smaller monitor that is on my CHV + FX-8350 setup, it was very hard for me to see all the workers. I had to really adjust each worker in the woker docker to see that I had had workers fail and I could not read the failure notice. That maybe what you are seeing with a core doing 10 tests and quits and the other workers run 27 to 30 tests.
RGone...

Rgone I have had this issue myself, when passing prime 2 hours I noticed 1 or 2 of the workers did significantly less tests then the others and no there weren't any failed workers or errors. I just chalked it up to those "cores" being a little weaker. I know that my #7 guy is usually the one that fails when I'm stressing and that is the same one that did less work.

 

[RGone]

Rgone I have had this issue myself, when passing prime 2 hours I noticed 1 or 2 of the workers did significantly less tests then the others and no there weren't any failed workers or errors. I just chalked it up to those "cores" being a little weaker. I know that my #7 guy is usually the one that fails when I'm stressing and that is the same one that did less work.

Thanks Drake.

I have seen slow workers as well, but I really got tripped with the ability to see all 8 core workers easily this last go round. I usually use OCCT so I had not actually viewed the 8 cores spread out down the page. I actually had a core drop out after awhile.

Now he needs to check and see which is what with his setup. More an FYI question of him.

Yeah I hate those slow-azz workers. I have been so used to all tests completeing withing about 30secs to 1 min of each other and these FX things can be minutes apart.
RGone...

 

[trents]

Is this some subtle FX down-throttling technology we haven't caught onto yet?

 

[SouthernStyle]

Yes, I was meaning what Mandrake is referring to. Here are pics of the test runs. It's odd to me that there's such a large difference from the 2 OCs, when I went from 19.5 Multi/235 HTT = 4578mhz, to 19 Multi/240 HT = 2575mhz. Not huge changes to settings and OC speed, and both at 1.425V. I don't think I have a very great chip, and my current PSU isn't winning any prizes or anything.

While both OCs were stable, I opted for 19 Multi/240 HTT as the overall testing performance seemed more balanced.

 

[Johan45]

I've seen the same thing as Mandrake mine is 5+6 the longer it runs the further behi8nd it gets.

 

[Mandrake4565]

Caddi Daddi suggests that you use everclear on the slow workers, it helps them come back to life.

 

[trents]

I would be interested to know if you get this same delay in individual workers if you were to run Prime95 at stock speeds and voltages with all the green stuff active. Could this be an effect of the overclock?

 

[Johan45]

That's a good question Trents. It would be interesting to see if these lazy ones still fall behind.

 

[SouthernStyle]

I'm going to mess around with it tonight when I get home. I'll do a ~30min run and let you guys know what it does. The strangest thing to me is that one module goes from having both the fastest and weakest core to a completely different module doing the same thing with fairly minor changes, none of which were to voltage. I'm going to bump the voltage a bit with my current OC and see what sort of effect that has on it as well.

 

[Johan45]

Just for fun I gave it a shot on optimized defaults with only ram set to specs.



It was almost the opposite I had a couple modules that performed better the rest were about an par.

 

[Mandrake4565]

Maybe this is the reason why some of the workers are slow. Using myself as an example, when I'm trying to get X mhz oc stable ill set the Cpu V run prime, it fails 1 worker, I bump the Cpu V, it runs 2 hours and passes. Maybe it is because it is "just enough" voltage to keep the workers happy and working for 2 hours. I'm going to do an experiment when I have time and see if upping the Cpu V at a known stable oc will increase and/or have the amount of tests completed more inline with one another.