Overclock FX-8300

[ferchkc09]

Hi All, 

 

new to overclocking but i was just experimenting with OC'ing and wanted to get opinions:

 

CPU:  FX-8300

MOBO:  ASUS M5A99FX Pro R2.0

CPU Cooler:  Cooler Master 212 Evo

 

I have used various OC guides on the interwebs and i wanted to see what temps i was getting at 4.0ghz with the following settings

 

CPU Voltage: 1.275

NB Voltage: 1.1875

 

I've disabled the power saving settings as stated in many of the guides. Here are the results of a 10 min stress test with prime 95.  I'm hovering around the 60 degree mark on the CPU and i'm not certian if those socket temps are too high.  The thing i'm struggling with is the turbo freq of 4.2ghz uses 1.375 volts.  What do you think of these temps and where do i go from here?

[SpaceGhostC2C]

I'd try to stay below 60 degrees for the package, 70 degrees for socket temps. Those temperatures seem high to me for just 1.27 vcore I don't remember the voltage I used, but I got an FX-8370E to 4.0 with lower temps on the stock cooler (which is worse than the pre-wraith cooler of the 8350 or 8370). I would expect your 212 to do much better at this point 

[ferchkc09]

thanks for the reply, that is the feeling i was getting based on some other numbers i've seen elseware.  Could it be a thermal paste/headsink application issue?  Right now the CPU is idling at 17 degrees.  

[SpaceGhostC2C]

5 minutes ago, ferchkc09 said:

thanks for the reply, that is the feeling i was getting based on some other numbers i've seen elseware.  Could it be a thermal paste/headsink application issue?  Right now the CPU is idling at 17 degrees.  

FX CPUs are known to report randomly low package temps when idle, although you can start trusting them above 30-40 C. Socket temps (the one reported under your motherboard in HW monitor) are more reliable. Still, I doubt a badly seated heatsink will cause trouble when idling. You could try re-seating the 212 (re-applying thermal paste, etc) and see if temperatures are lower.

What are your settings for LLC? I did notice in a 990FX Sabertooth that, for teh same frequency and vcore, higher LLC gave me more stability but also higher temps (I guess the effective voltage delivered to the CPU was simply higher).

[ferchkc09]

the LLC i was using was Ultra High on the CPU and High on NB

[ferchkc09]

also, i re-ran prime 95 at default settings and got a max CPU temp of 51 and package temp of 38.  Not sure if this is relevant but thought i would share

[SpaceGhostC2C]

1 minute ago, ferchkc09 said:

also, i re-ran prime 95 at default settings and got a max CPU temp of 51 and package temp of 38.  Not sure if this is relevant but thought i would share

It is. It is in fact good to run the test at stocks first to better troubleshoot your OC later 

What is the ambient temp? Those don't seem too bad for a summerish ambient temp. You could try reducing LLC in both fronts and check what happens with the temperatures and with the effective vcore. 

Also, is 1.1875 the default CPU NB voltage? If you are not touching the bus speed or OCing the ram at this point, I would leave the CPU NB voltage at stock values.

[ferchkc09]

the room temp is around 72-74 degrees.  When i loaded defaults and switched the voltages to manual, the current voltage in NB was 1.1875.  Should I change to auto?

[SpaceGhostC2C]

10 minutes ago, ferchkc09 said:

the room temp is around 72-74 degrees.  When i loaded defaults and switched the voltages to manual, the current voltage in NB was 1.1875.  Should I change to auto?

You can try, but if you never touched it, I guess it's at stock already, I just seemed to recall my CPU NB being at 1.15v before messing with it.

Also, thinking again about the temps they may be not that strange after all given your settings, although you may be overkilling for 4.0. Let me check the settings in my 8370E (currently at 4.4, also hitting 60 / 70 C in stress test) and get back to you.

[SpaceGhostC2C]

OK, I'm reading the settings from Asus AI suite so I can look at them and write to you at the same time - I hope they accurately report the bios values:

 

CPU options (Reported in "TurboV")

CPU bus frequency: 200

CPU voltage: 1.26875

CPU/NB voltage: 1.175

DRAM votlage: 1.65

 

VDD PCIE: 1.100

VDDR: 1.200

NB voltage: 1.100

VDDA voltage: 2.500

NB HT voltage: 1.200

SB voltage: 1.100

NB 1.8V voltage: 1.8

 

CPU ratio: 22

 

DIGI+ options:

 

CPU LLC: Ultra High

CPU NB LLC: High

CPU current capability: 120%

CPU NB current capability: 110%

CPU Voltage frequency: Auto, Spread Spectrum off

CPU power phase control: optimized

CPU power response control: Fast

CPU NB power response control: fast

CPU thermal control: 140 C

CPU power duty control: T. Probe Thermal

 

DRAM current capability: 110%

DRAM voltage frequency: 300KHz

DRAM power phase control: optimized

 

(CPU is running at 4400MHz, RAM is running at 2133 CL9, HT link at 2600, NB at 2400)

 

 

These settings get me to 65-70+ on socket temps depending on ambient temps on a Cooler Master V10 (which despite all its bulkiness, TEC, and marketing, is not expected to do much better than your 212, at least according to all reviews, although it may do a little better at high OCs).

It may be that this OC isn't 100% stable, but I have to change the cooler to run more/longer stress tests.

 

I hope this information is useful for your OC.

[SpaceGhostC2C]

1 hour ago, ferchkc09 said:

the room temp is around 72-74 degrees.  When i loaded defaults and switched the voltages to manual, the current voltage in NB was 1.1875.  Should I change to auto?

I run 20 minutes of OCCT now, at 24C ambient (measured on top of the PC case, probably somewhat cooler air going in), maximum 63C at socket, 55.5C package (flattened at that level after 10 minutes or so), for the setup in the previous post. Hence, although I believe you could get lower temps on the 212, it could be that all the difference is accounted by the cooler. I would check how far you can up the multiplier without increasing vcore, my hunch is that you can get a higher OC with the same voltage and temperatures you have now.

[ferchkc09]

I just started to step down the CPU Voltages and performed 15-30 minute Prime95 stress tests.  I went as low as 1.25000 volts but failed on core 5 around the 20 minute mark. 

 

I went to 1.25625 and had no failures after 30 minutes.  The results are attached.  What are your opinions?  The package temps seems much better at 55C but the socket is still around 70C.  Not sure if that is pushing it for this cooler.  The max temp shows 71 but it had been flat at 70 for the last 10 minutes of the test.

 

It also looks like the voltages are near what i have in the bios as well.  + .003 and - .008

[SpaceGhostC2C]

7 minutes ago, ferchkc09 said:

I just started to step down the CPU Voltages and performed 15-30 minute Prime95 stress tests.  I went as low as 1.25000 volts but failed on core 5 around the 20 minute mark. 

 

I went to 1.25625 and had no failures after 30 minutes.  The results are attached.  What are your opinions?  The package temps seems much better at 55C but the socket is still around 70C.  Not sure if that is pushing it for this cooler.  The max temp shows 71 but it had been flat at 70 for the last 10 minutes of the test.

 

It also looks like the voltages are near what i have in the bios as well.  + .003 and - .008

 

That's very good, which makes me think maybe you can get away with lower LLC. In any case, you got yourself 5 degrees on both temps, assuming ambient hasn't decreased much.

Sometimes high Socket temps reflect VRMs heating up. However, I don't think the package vs socket different is too high, and I also don't think your motherboard will have any problem keeping the VRMs cool at this voltages, but you may try placing a fan on top of the VRM heatsink and see if it makes any difference to socket temperatures. I have the stock cooler's fan placed there, and I don't get lower package temps in the 8370E, so maybe that can bring your socket temps a little down too.

 

Other than that (and until the purists come with the torches yelling that you need 30 days, not 30 minutes ) I'd say you're doing well and may continue to search for the max OC you can achieve with this cooler (i.e., what is the maximum frequency you can get with the original 1.27 or the current 1.25625).

[ferchkc09]

Thanks again.  

 

I'm not sure it is worth even messing with OC stuff for me, as i'm very new to this. but i did see a 22% increase in CPU Cenebench R15 scores.

 

Defaul was 518 with current setup at 625  seems like a decent increase.  I may try the turbo frequency 4.2 at this voltage and see what happens. 

 

Can you elaborate at all as what effect the LLC has?  Does it lower temps?

 

edit:  i answered my own question after reading a few articles.  Basically it helps stabilize voltages at idle and at load to what is set in the bios.  If I'm around -.004 at idle and plus .005 at load does this seem adequate or do i try to drop it to 50%?  Also, if you drop the CPU LLC to say High from Ultra High do i also drop the NB LLC?

[SpaceGhostC2C]

7 hours ago, ferchkc09 said:

edit:  i answered my own question after reading a few articles.  Basically it helps stabilize voltages at idle and at load to what is set in the bios.  If I'm around -.004 at idle and plus .005 at load does this seem adequate or do i try to drop it to 50%?  Also, if you drop the CPU LLC to say High from Ultra High do i also drop the NB LLC?

Exactly: it stabilizes voltages and reduce/eliminate the drops associated to going from idle to full load, and the spikes going from load to idle. it increases your power consupmtion as well, hence the additional heat.

The -0.004 to +0.005 range you are obtaining is practically non-existent. I have boards without LLC with 10 times that (e.g, I have to set 1.35v in bios to get 1.3-1.31 under stress). So, the question is whether you could decrease power usage and heat by reducing LLC and potentially tolerating a tiny bit of voltage fluctuations (I must say my Asus board never had vcores moving as much as my LLC-less board, regardless of the LLC setting).

 

I don't know much about what to do with teh CPU NB LLC and other settings. I didn't touch it at all in the beginning, leaving RAM at default values and working only with the CPU. Then, with teh CPU settings fixed, I set my RAM to its XMP profile and stress tests would now fail. I left the CPU settings unchanged, increased CPU NB voltage one notch, and increasedCPU NB LLC. It got stable again.

[ferchkc09]

3 hours ago, SpaceGhostC2C said:

Exactly: it stabilizes voltages and reduce/eliminate the drops associated to going from idle to full load, and the spikes going from load to idle. it increases your power consupmtion as well, hence the additional heat.

The -0.004 to +0.005 range you are obtaining is practically non-existent. I have boards without LLC with 10 times that (e.g, I have to set 1.35v in bios to get 1.3-1.31 under stress). So, the question is whether you could decrease power usage and heat by reducing LLC and potentially tolerating a tiny bit of voltage fluctuations (I must say my Asus board never had vcores moving as much as my LLC-less board, regardless of the LLC setting).

 

I don't know much about what to do with teh CPU NB LLC and other settings. I didn't touch it at all in the beginning, leaving RAM at default values and working only with the CPU. Then, with teh CPU settings fixed, I set my RAM to its XMP profile and stress tests would now fail. I left the CPU settings unchanged, increased CPU NB voltage one notch, and increasedCPU NB LLC. It got stable again.

The question that seems hard for me to find an answer is: what is a reasonable socket temp when at full load for over an hour?  I know the answer probably varies with frequencies and voltage but it would be nice to have a ball park!  I ran prime95 for 1.5 hours again last night and came in at 69C at the socket and 54C at the core.  Which is better but I honestly would rather see socket temps in the low/mid 60's if possible with air cooling to make me feel better about not fucking up my CPU ha.

 

I feel like a lot of recommendations are to put a fan on the VRM, but I honestly would like to avoid that.  If I'm seeing ~70C at full load I'm guessing that my computing style will rarely ever reach 100% utilization and if it does, not for a prolonged period time.  I only do light gaming, photo/video editing and browsing.  It just becomes, what is a reasonable temp.

 

And maybe the answer for me is, why even overclock based on what I do with the PC...

 

Edit:  Also, I heard that actually monitoring AMD Overdrive software for the "thermal Margin" numbers are a better readout for CPU temps?  My understanding is that number tells you how much temp you have remaining in the CPU at load, so as long as it is > 0 (hopefully in the teens-20's) the temps are ok.  Not sure if that is common knowledge or just a random thought I saw.

[Guest]

I have a fan aimed at the back of my motherboard. But something that helped me get low VRM and sockets temps was changing airflow from exhaust to intake. So basically reversing traditional airflow (top and rear now intake while front of case exhausts - just makes sure the CPU cooler follows suit). While this reduced CPU temps (not much on the core, honestly) it increased GPU temps. Having the intake going in onto the GPU was helpful in keeping it cool. 

 

Now my rig is all frankensteined because I started tinkering and never finished. 

[SpaceGhostC2C]

2 minutes ago, ferchkc09 said:

The question that seems hard for me to find an answer is: what is a reasonable socket temp when at full load for over an hour?  I know the answer probably varies with frequencies and voltage but it would be nice to have a ball park!  I ran prime95 for 1.5 hours again last night and came in at 69C at the socket and 54C at the core.  Which is better but I honestly would rather see socket temps in the low/mid 60's if possible with air cooling to make me feel better about not fucking up my CPU ha.

The most straight answer I can give you is, for me, socket temps below 70C, and preferably mid 60s or less.

 

Having said that, I never ran anything that makes my CPU as hot as prime95, including other benchmark/stress test programs, and computations that will use all cores to the point of making the mouse lag. So, if something is within limits in prime95, I'd say you'll be good in your everyday usage.

 

7 minutes ago, ferchkc09 said:

And maybe the answer for me is, why even overclock based on what I do with the PC...

Because you can!

I have fun doing it, and it does give some performance boost. Editing and gaming will like it, browsing couldn't care less.

 

[ferchkc09]

20 minutes ago, SpaceGhostC2C said:

The most straight answer I can give you is, for me, socket temps below 70C, and preferably mid 60s or less.

 

Having said that, I never ran anything that makes my CPU as hot as prime95, including other benchmark/stress test programs, and computations that will use all cores to the point of making the mouse lag. So, if something is within limits in prime95, I'd say you'll be good in your everyday usage.

 

Because you can!

I have fun doing it, and it does give some performance boost. Editing and gaming will like it, browsing couldn't care less.

 

that is kind of how I feel, why not because I can OC! 

 

I currently am only rocking a 120mm exhaust and 120mm intake fans, mainly to reduce noise (Corsair AF120 QE's).  I do have NZXT 120mm I could add to the intake to see if that helps bring more air over the cpu.  I may also try dropping LLC.  I had done it previously from Ultra High to High and it at ideal it seem like I was getting the same voltages, 1.248 and 1.260 so maybe that will help a tad.  I might also try lowering the multiple a point and voltage to follow.

 

I would feel much better about this if I can get the socket temp to be low/mid 60's at full load with prime95.

[ferchkc09]

i added another 120 intake and dropped 6 degrees on a 30 min test.  i might drop llc as well to see if i can mantain voltage