Curious about a CPU and GPU Overclock

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160
I use Heaven 4.0, 3D Mark Firestrike and 11, MSi Afterburner, GPU-Z, CPU-Z, NZXT Kraken x60.
CPU: 8120
GPU: 660ti
Motherboard: Asus Sabertooth 990FX/Gen3 R2.0

My Results from Crysis 3 on ALL maxed settings except Anisotropic Filtering at 8x and MSAA at 4x.
I didn't have my tool for measuring FPS on-but it was not lagging at all. No tears, artifacts or problems, This was from a full play-through of the Second and last 2 missions:


Kraken Fan Settings:
hjj4.jpg



My Results from Catzilla
Results Screenshot(If you don't have an Account):


I scored 5021 on Firestrike.

So Crysis 3 is definitely CPU heavy.
I should've recorded the FPS.


But Anyway, my question is as follows:

When I run Prime95 Torture Test, my temperature on Extreme Fan Preset(Haven't Tested my Custom yet) is 54*C and vCore 62*C. (Yes, MOSFETs/VRM's are running hotter). If I run the regular test, it runs alot cooler with 100% CPU usage still. Around 46*C CPU, 55*C for vcore. So I can push my CPU further, but can I push it to 5gHz. Temperature wise during Torture Test it won't be stable-but as for High CPU usage it might, no? During Crysis 3, I had it set to Silent Fan mode and it reached 48*C. So I have 14*C until the Thermal Limit. I will not like to go above 58*C though personally.

Question 2, My GPU overclocking. I think my vBios or Motherboard may be limiting my overclock. I'm only able to change clock and memory frequency. It doesn't seem to raise the Voltage input or TDP when using Afterburner. (Debating flashing my vBIOS). But I was able to raise the Clock Speed a little(I mean very little) to a stable clock-but feel it has more potential:
GPU-Z Validation
Will raising PCIe voltage in my BIOS affect my Overclocking ability. BTW, PCIe Spread Spectrum is disabled.
 

mr1hm

Distinguished
Question 1: im not too sure about this one but, it's generally not a good idea to have the MOSFET/VRM's running high temperatures as this can mean that they are under a lot of stress. my old ASRock Z77 Extreme4 used to run hot MOSFET/VRMs (around 50C) and i later figured out that, those kind of temperatures can lead to random instability but, that mobo had many flaws overall for me. keeping your Core temperatures @ 62C over a elongated period of time will most likely lead to faster degradation of the chip as the recommended max temperatures are right around 62C (1.475v-1.5v for max voltages).

Question 2: there's a reason to why GPU manufacturers will limit the voltage, it's not a good idea to unlock it unless the manufacturer has allowed you to do so (as this will mean the card definitely can handle slightly higher voltages). of course, one can argue that modding their GPU's BIOS can open up possibilities to a higher overclock safely whether or not the manufacturer has locked it or not; although this is true, you're putting your GPU at risk and without the proper care/knowledge you can instantly brick your GPU. IMO, it's just not worth it to flash your GPU BIOS especially for a 660TI.

the 660TI is quite a nice video card however, i feel like it has a gimped memory interface of 192-bit which will hold you back when using demanding Anti-Aliasing settings whether you have it overclocked or not.
 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160


As far as max temp for this chip, it's stable at 62*C, and core's stop during torture test at 64*C. But for the vcore, people said to worry at 70*C+.
And the CPU's max Voltage is 1.55.

The video card runs nicely. It's a 3GB with 192 bit, so yes-it does hold back a little. But I still can get more by boosting the Core clock, and to stablize it I would have to increase the power limit. I've flashed BIOS on GPUs before, and the chances of bricking are minimal if done in a safe environment. (And not during a storm lol)

But that wasn't what I asked necessarily.
I increased the LLC to Ultra High from High due to vdroops from 1.40635 to 1.38. Now it ran at 1.416 from a 1.40635. (So temps of course increased)
The reason it hit 60*C was because I had it on silent for a little bit, and the RPM didn't actually go above 3000, (Max is 1980) Just a error when reset HWM for some reason.

Starting Torture Test:


Torture Test for 1 hour:


An Error displayed when changing my Multiplier through AS Suite 2 instead of BIOS:
3uio.jpg


A frequency only dreamed of :D Just thought I'd share it.


But what I'm asking is if it'll be ok to run at 5 or 4.8 Ghz. Obviously it won't pass the Torture Test, but it might pass the regular test from Prime95. The Torture Test tests absolute stability, pushing the CPU to limits otherwise un-achievable.
Where as the regular test runs the CPU at 100%-but it runs alot cooler. With that said, should I run at 4.8/5ghz frequency if The CPU will be within it's thermal limits of Crysis 3 and possibly Prime95 Test (not torture).
 

mr1hm

Distinguished
the limits i gave you were for safe 24/7 usage. if your running voltages at 1.5v+ then you're risking degradation whether it happens 2 weeks from now or 1 year from now.

my CPU's limits are 105C but, this does not mean i can run my CPU at any temperature above 70-72C safely. 70C-72C is considered the limit for safely operating my CPU under heavy loads for elongated periods of time, meaning, anything above this can cause degradation.

the most significant factor that contributes to CPU degradation is generally Voltage. so, if AMD has specified for that CPU that 70C is indeed the maximum (Tj Max), your chip needs to stay under this at all times but, the closer the temp gets to that limit, you may encounter CPU throttling which will make using your system almost impossible to use. if you choose that you think it'll be ok running high temperatures like that then that's your choice to make.

as far as stress tests go, nothing will stress harder than AIDA64's FPU test. in fact, P95 is believed to put unrealistic loads onto your chip and also has been known to cause degradation when running it's stress test for very long periods of time but, it works to a certain extent. i can't say if AIDA64 will test AMD CPU's realistically (as in test every single instruction set in the chip) as their newest version is designed around the Haswell architecture for Intel.

if you do a quick google search, you'll notice some user reporting that they would crash in games but be P95 stable for a extremely long time.
 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160


My computer is on about 12-14 hours a day, not 24/7. But are they using the torture test, or the standard test. Second I'll monitor my temperatures. But what I'm asking if I have 4.8 ghz for instance, at ~1.45v, and it's stable but hits thermal limits for torture but not the regular test from prime, can i run the 4.8 ghz nonstop since it will not be at 100% 100% of the time, it will be ~75% while gaming-100% video rendering and ~5-10% for browsing...etc..
As far as games crashing, I had that issue-but when I used custom torture and used all my RAM, some cores after 30min stopped-I was low on voltage. And games crashing could be from drivers, bad gpu oc's or ram...numerous things.

I understand how it degrades-I'll upgrade my CPU in a year anyway :). And AMD can't run higher do to their safety precautions. And the 8120 at 64*C -cores shutdown to prevent damage when they hit that temperature.
 

mr1hm

Distinguished
generally, i'd say yes you can run 4.8 @ 1.45v for general/light loading conditions but, at any time your CPU is under heavy load (anything above 65%-75%) and constantly being used under these conditions degradation is definitely a possibility.

although your voltages look ok in my opinion for 24/7 usage, constant temperatures above the 60C-62C mark will probably end up degrading your CPU.

i think you pretty much answered your own question. if the chip shuts down @ 64C, then i wouldn't feel comfortable with anything above 60C, if i'm feeling extremely cautious, nothing over 58C.

are you using AMD Cool n Quiet?
 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160


It's not possible to go above 62 as I've stated. The cores turn off to prevent damage :)

Tomorrow I'll try to go around 4.8-5ghz and run some tests and post my results.

Just wondering how I can create a graph as such:
http://media.bestofmicro.com/O/R/406539/original/arma-19-frot.png
 

mr1hm

Distinguished



pretty sure you just stated 64C? which one is it?

-are you running AMD Cool n Quiet? if not, for those speeds, id turn it on.

as far as creating the graph goes, i'm not too sure but, i'm sure programs like Excel and things of that nature can make one quite easily.

also, what are your temperatures looking like?
 

mr1hm

Distinguished


it actually seems that your TJ Max is variable to a certain degree although it seems as though your chip's maximum safe operating temperature is 61C: http://www.cpu-world.com/CPUs/Bulldozer/AMD-FX-Series%20FX-8120.html

i just read a random thread about keeping the chip under 55C for safe 24/7 usage.

33C seems fine for idle but, your load temps are what matters. :)

 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160


Thought you wanted my idle because my load temps are in the posts above :)
But what do you mean my TJ Max is variable, how?
 

mr1hm

Distinguished
that's an excellent question lol. i haven't seen any specification as to whether the 61C is the socket temperature, CPU temperature or Core temperature. all 3 would be somewhat different but, it seems most people keep their limits to 61C give or take a single degree.

were those load temps in the OP for 4.55 @ 1.406v? i couldn't find your voltage in the OP but, if your planning on finalizing with the 4.55ghz w/ 1.406v overclock, id say it's solid if i'm correct in assuming the 2 load temps you gave in the OP are for CPU temp and Core temp, core temp being what we want; which was 55C @ max. :)
 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160


Hmm, strange. A variable TJ Max :D

Torture Test for 1 hour:
(1.406 with ultra high LLC-1.416 under load)
 

mr1hm

Distinguished


tell me about it :heink:

looks good to me :)
 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160
Okay so here's an update. I can get the RAM at 2400 with a bus at 250, but the CPU gets hot at 4.5, 64*C almost instantly.
So I kept the bus at 233, increased the multiplier to 20.
Putting my frequency at 4676Mhz. I have the RAM at 1.7v @2133 @10-11-10-31 2T. (1T is ridiculously unpredictably unstable).
Increase CPU voltage from 1.406 to 1.45 to stabilize the higher frequency without fault.
Note that with Very High 75% LLC the Voltage boosts to ~1.464 during load.
With that said, running the Fans at 100% during Torture Test at the absolute highest settings for CPU and almost max RAM usage, the test passes for about 5 minutes before hitting 63*C and stopping one of the cores tests. But this is not a bad thing, because what will strain my CPU to this limits non stop? It's stable though.

So I went into Crysis 3 and changed my Settings to FXAA and 16x AA (For some reason my vsync is 30hz in Crysis 3-but everything else it's 60hz :??:)-But I didn't want to play the mission for the millionth time :lol:
I had Kraken Control at Extreme, not 100% during the play-through.
The Temperature of the CPU never ex ceded 52.8*C
And I will run the benchmarks and Prime95 regular test and update this with the results. Just wanted to add this for now.
Pictures, because everybody loves them. :D

Prime95 Settings:
iujn.jpg


Idle:


Thermal Limit Torture Test:


Crysis 3 Mission 2 Full play-through (Setting Images as well):
sjux.jpg




Frames, Time (ms), Min, Max, Avg
38264, 1315438, 0, 48, 29.088
(Not sure how it hit 0, it never stuttered, or lagged-I think it was just a bug)
(Working on graphing the Frames now)
jxg5.jpg


Okay so Catzilla went up to 6775, so a 3 point boost. Where as 3D mark went down! 4990 on firestrike,

Not really worth the Temperature Rise for barely any performance increase. I'll probably conitnue this when I get a couple more fans, go full push/pull(currently with 3 fans lol). And I'm going add a fan to blow from 5.25" drive bay onto VRM's, and a Sidepanel mod 140mm fan exhaust behind the CPU and VRM's. This'll further improve my temperature threshold. :bounce:
 

mr1hm

Distinguished
any kind of additional bus speed from what you already had would cause a rise in temperatures but, jumping from the 55C - 63C makes sense since the CPU's overclock is already so high; the higher you take the chip, the more steps you take into the unknown but, big jumps in temperature and voltages are generally expected the higher you go. that FPS graph just made me cringe ^^, drops like that is definitely not normal and to me, something seems to be not stable/throttling. the fact one core was stopped due to an error is probably a good indicator of an unstable overclock; sometimes when those errors happen, a WHEA error will follow with it. you can check for WHEA errors by accessing Event Viewer then clicking on "Custom View" then "Administrator" (something like that). that will show you WHEA errors as well as Kernal Power errors that took place at any time your CPU was running (although it may not show any). but, just cause it doesn't show, doesn't mean it's stable either however, if it does show, it's 99% a unstable CPU overclock caused by the RAM overclock. if it didn't show, you'll have to continue to change your settings and test your overclocks to determine what went wrong.

overclocking your RAM from the specified 1866MHz, to high speeds such as 2400MHz can hold back and limit your overclock as your IMC is being stressed. you have to check whether or not a RAM overclock is worth the jump in degrees in the CPU.
 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160


Added to a previous post. But the worker stopped because it hit 64*C, not because it's unstable. The CPU does that when it hits that temperature.
Also, I'm not running at 2400 RAM, I'm running at 2133. (Don't skim through).
As for the graph-the drops aren't too big. It's not dropping to 0, the chart is from 15-35FPS, not 0-35.
And the drops are usually when you go to cinematics, (I have a slow HDD).
And it lowered because, if you've every played the game-you know how that mission gets later on.(The drops aren't instant and rapid) That chart is playing over 20 minutes, not 60 seconds.
 

mr1hm

Distinguished
according to AMD, your CPU should shut down @ 61C not 64C. as a matter of fact, if this was even true, it would turn your system OFF not just disable a single core; however, if it does in fact disable a single core to protect itself, this would be considered CPU throttling.

- 2133MHz is still stressing your chips IMC, not to mention the good amount of voltage your pushing through it. 2400MHz will probably stress it harder but nonetheless, 2133MHz is still a good amount of stress. i think it was a good idea to clock the memory back to stock/XMP specifications.

- your load percentages are what i'm worried about (anything above 80% usage can indicate a CPU bottleneck/a limitation on CPU resources), my previous post should've mentioned "unstable/throttling/bottlenecking". looking back @ your load percentages, i'm pretty confident that your chip is a bottleneck in that game. IMO, as a competitive gamer, anything under 60FPS on a standard 60Hz refresh rate monitor is unacceptable but, for single player games, i'd generally say a dip to the mid-high 40s is generally alright. you can see otherwise but, for me, 15-35FPS @ any given moment looks like a slide show.

so, the problem is that your CPU isn't powerful enough for Crysis 3, you can try an overclock further but, i don't feel like it will fully alleviate the bottleneck.
 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160


I think your mixing up tests and results. And a bottleneck is by no way 80%+. A bottleneck is 99-100%. " bottleneck is a phenomenon by which the performance or capacity of an entire system is severely limited by a single component. Formally, a bottleneck lies on a system's critical path and provides the lowest throughput."
And my CPU "throttles" itself at 64*C (or stops a core briefly) because I have cool n quiet by CPU enabled. (SpeedStep for intel).
You also are not considering that I have a decent(But not great) Single GPU. So to further improve performance in Crysis 3, I must now OC the GPU more or get a new better one. (Not worth SLI) Simply because it is at 98-99% usage (Bottleneck). I won't be able to push more out of the game without flashing its BIOS to increase the TDP limits for a higher Clock speed. (And ending in higher benchmarkings as well)
As for memory, it is not an issue at 2133. It passes memtests and I haven't had an issue with this OC. AMD and Intel work differently, I don't know how much experience you have with AMD-but the memory controller recommends 1866. But with 990fx chipset and OC'ing the CPU's, people are able to achieve 2400+ RAM frequencies. Just because a stick of RAM is rated at a certain speed doesn't mean you can't achieve higher, otherwise the same could go for CPU's. (I'm running this 3.1Ghz CPU at 4.67Ghz)
And increasing its speed doesn't increase CPU usage by much, having too little RAM affects it more. Higher frequency increases CPU temperature (More memory controller usage-Need more voltage->CPU NB AMD, VCCIO-Intel)

Another things to note, with Crysis 3, there's an issue for SP to MP. In singleplayer, people with roughly 40fps get around 65+ in Multiplayer. Alot of Physics in SinglePlayer and other things. In BF4 without this OC i was getting an average of 50FPS.
Not sure how it'll look like a slideshow though if Blu-Ray at 1080p is 24FPS, and no one complains.
Crysis 3 had V-sync on though, so I was limited to 30FPS as well, which I mentioned. (I have a 60hz screen-which is why I was confused by this)

So I wouldn't say the CPU isn't powerful enough-I'd consider the GPU before the CPU.

Just tested it another way. Put fans on Silent mode-Played Crysis 3. At 60*C displayed by the temperature probe (100% accurate)-the game started stuttering. 61*C BOOM! it froze. I actually had to hard reset my PC :lol:. So Extreme, ~55*C. No problems.
 

mr1hm

Distinguished
well, in the end, as i've stated, it's your choice. i only stated that 80%+ may be an indicator but, your chip is @ 95%+, if the GPU usage of your 660TI is not reaching or pegged @ 99% like it should be, the CPU is having problems providing resources to the GPU.

the 660TI is a decent GPU however, the fact is, Crysis 3 depends more on your CPU as you've stated; if you notice your GPU usage not hitting 99%, OC'ing the GPU is a waste of time because we already know that your CPU's usage has hit 95%+. whether or not it stays @ 95% consistently or once in a while, its still is a good indicator that the game is too CPU intensive, not to mention the several posts that show my very old AMD Phenom II X4 960T as a comparable CPU to the FX-8120 in gaming if not better.
 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160
My GPU hits 99% constant, sometimes drops to 98, and sometimes hit 100%(Rare lol)
As for the Overclocking, should I increase the memory clock more. And keep the core clock at it's pre OC settings?
Because it's 3gb vRAM and I'm on 1080p-so the faster rendering and peed should help-I would assume so. (Because a 50 boost of core clock make crysis 3 unstable with v sync off.
Also is there a way to record average CPU usage like there is FPS?
 

Jake Wenta

Honorable
Mar 13, 2013
696
1
11,160
Here are some shocking results.
My CPU was about 60% usage through the mission. (Where ~10% was from Audio Enhancers)
I noticed my Frame drops were from loading and saving. So every checkpoint and cinematic, the game did lag for a split second. HDD issue. Also everytime I engaged the enemy and it loaded an audio for them to speak-it doesn't lag from explosions and sudden changes. Just loading-I have a 5 year old HDD with 7200RPM and 8mb or so cache, so it's garbage. lol

Highest Usage and Temperature (Silent Mode) -2170 RAM vSync off Medium Blur FXAA x16AA Mission 2 of Crysis 3:

Note that if I increase the value of any of the Clock speeds by 1, the game crashes if I have vsync disabled. At 125+ Memory, 80+ Clock with vsync, I have no issue running through the game.

CPU Usage through the Mission (Enabled once loaded the misson, disabled once Third Mission started loading-Alt-Tab Stop):

This proves it is pretty heavy for the CPU, but not bottlenecked by it by any means. Min: 25.9%, Max: 86.6%, Average:49.56%
Working on FPS Charts as it takes longer to make-And wanted this posted.
Edit:
Average FPS:34.82
Min FPS: 19
Highest FPS: 51
FPS Chart:


Not sure this is really comparable since the GPU is the Bottleneck, so comparing CPU usage to FPS doesn't really display too too much. But here it is, FPS on left, CPU Usage on the right, Seconds on the Bottom (Points used were every 5 seconds):
 

mr1hm

Distinguished
the picture you posted before were showing the CPU @ 95% which is why i told you your CPU is most definitely a bottleneck. the ones you posted just before this post show differently (it's better).

your GPU usage looks good but, there are other ways the CPU can become bottlenecked as well. it seems like your CPU is still being maxed out, whether it's just for a second or a few minutes, i don't like seeing games max out my CPU; that's for when its time for huge audio rendering sessions and even then, sometimes it won't max out.

i'm not sure how long you've had that CPU for but, why don't you just upgrade to a 3rd gen i5 Intel CPU? or even the FX-8320, which is also a very competent CPU.