Overclocking questions

[DarkEnergy]

FX 8350

I've tried overclocking before but I've always stopped it because of high temps in stress tests. Well, since then I've moved the computer to the floor. I've also noticed temps are a bit lower now so I've tried overclocking again. I'm using JaysTwoCents video. I've set similar settings, CPU voltage to 1.38, CPU load line calibration to medium, CPU/NB load line calibration to regular (this is important later). I'm trying to get a 4.4Ghz clock to start with. While using OCCT for benchmarking, I noticed that after only 1 minute of stress testing it will stop with an error. At this point the temperatures are fine. I noticed in HWmonitor, under load, the CPU was reading 1.32-1.33V, not even close to the 1.38V I set it to. I went into BIOS and experimented with moving both load line calibrations one level higher. Now under stress it reads 1.344-1.356V. also I passed the 10 minute mark of testing. My first question is why would the voltage droop that low anyway? And why would the slight change fix the errors?

At 10 minutes, the CPU socket temp was 68c and core temp was ~51c. Doing some research, the max socket temp for this processor is 70c and core is 62c. So socket temp is close to the limit, but, I am running a stress test that is much harder on the CPU (OCCT) so I'm fine with that. Running an Aida64 stress test only brings socket to around 50c and core to about 43c. My other questions is, would it be safe to raise the overclock even higher? Let's say it does hit the 70c socket temp limit. I know for a fact (based on previous experience) that it will thermal throttle. However, because this is a test from software similar to Prime95 that is meant to push it way harder than other benchmarks, would it be safe anyway? Real world usage, I'm not going to be maxing out this CPU constantly and even if I was, I'm sure I wouldn't be running it with what ever instruction set pushes it hard enough to overheat. 

 

As a comparison with 1.38 voltage and 4414MHz settings, Prime95 pushes socket temp to 74c (danger) and core temp to 56, which is why I use OCCT, an ever so slightly less intense benchmark (but more intense than AIDA64). I'm using the NZXT s340 case so there is no ventilation on the back of the motherboard. The only way to cool the socket temp is to take the back panel off and blow a fan onto the back of the motherboard. 

[Terryv]

Your PSU could be the cause of Vdroop. What PSU do you have? Good quality units don't usually have problems. On my 8320 my LLC was set to high and was steady for my settings.

 

It seems that you are temperature limited at this point. You need a better cooler or you will throttle. No damage should occur as you CPU is throttling to keep temps in check.

 

Prime will stress it above and beyond any realistic real world load. I like to use handbrake to check max temps as that's the highest real load I've ever used.

 

Note: handbrake is not for stability testing.

 

Socket temp will drop along with core temp even though its always higher than core temperature. (By about 12-15 degrees in my experience) 

Edited March 19, 2017 by Terryv
Spelling

[DarkEnergy]

9 minutes ago, Terryv said:

Your PSU could be the cause of Vdroop. What PSU do you have? Good quality units don't usually have problems. On my 8320 my LLC was set to high and was steady for my settings.

 

It seems that you are temperature limited at this point. You need a better cooler or you will throttle. No damage should occur as you CPU is throttling to keep temps in check.

 

Prime will stress it above and beyond any realistic real world load. I like to use handbrake to check max temps as that's the highest real load I've ever used.

 

Note: handbrake is not for stability testing.

 

Socket temp will drop along with core temp even though its always higher than core temperature. (By about 12-15 degrees in my experience) 

PSU is a Corsair CX600M. Tier 4 of the PSU tier list on Toms Hardware. AKA, it's a bad PSU. I don't mind putting my current computer at risk as it's a bit old now. 

[Terryv]

that could very well be your problem.

 

what cooler are you using?

[DarkEnergy]

6 minutes ago, Terryv said:

that could very well be your problem.

 

what cooler are you using?

212 EVO

[Terryv]

you seem to have hit a similar voltage cap with the same cooler i was using. I stopped at 1.35 on the evo.

[SubLimation7]

49 minutes ago, DarkEnergy said:

 

Im not gonna lie, didn't read past first line or anyone else's comments. But I do have lots of overclocking experience. After typing this up I read post in entirety plus comments, this guide is perfect for ya. 

 

Note: Don't be too scared of voltage, 1.45v is acceptable under load 24/7 as long as you stay UNDER 70c

Step 1: reset everything to default

Step 2: download occt, makes it easy for this chip. 

Step 3: put LLC off to start, leave everything default and set bus speed at 240, then if that's stable keep going, I usually get to around 270 mhz on bus, these boards love bus speed, and you'll get a good chunk extra fps.

Step 4: Once youve got your bus speed, mess with your multi to get to ~4.5ghz, and set your ram speed to whatever it's stable at (usually around 1800mhz), also change your voltage to manual at 1.45v.

Step 5: go test in Occt for 5 min. If you get voltage drop on vcore raise llc until it doesn't drop. After all that, do you have headroom? If yes, continue to push multi until unstable, at that point slowly increase voltage OR just llc until it's stable without exceeding 1.45v under load or 70c.  If you didn't have more headroom, stop here and decrease voltage until you have the lowest possible voltage while still being stable. 

 

Almost forgot, under bios CPU configuration turn power saving features off

 

Or you could just buy a PC from me next time and save yourself time and money

 

Source: Tons of experience, I have an 8320 downstairs as emulator PC clocked at 4.8ghz 1.43v and won't break 53c in 22c ambient and all it has is hyper 212 cooler, and a few hidden magic tricks :).  Just threw together an 8350/gtx 690 combo I'll be selling cheap with a baby cooler on it, should be breaking national record Tuesday with it. 

[DarkEnergy]

3 minutes ago, Terryv said:

you seem to have hit a similar voltage cap with the same cooler i was using. I stopped at 1.35 on the evo.

Isn't that like stock voltage?

[Terryv]

stock for me was 1.3. I ended up settling for a 4.2 OC on 1.28v. I pushed up to 1.35 to see what the evo could do and got 4.4

[DarkEnergy]

3 minutes ago, SubLimation7 said:

Im not gonna lie, didn't read past first line or anyone else's comments. But I do have lots of overclocking experience. 

 

Note: Don't be too scared of voltage, 1.45v is acceptable under load 24/7 as long as you stay UNDER 70c

Step 1: reset everything to default

Step 2: download occt, makes it easy for this chip. 

Step 3: put LLC off to start, leave everything default and set bus speed at 240, then if that's stable keep going, I usually get to around 270 mhz on bus, these boards love bus speed, and you'll get a good chunk extra fps.

Step 4: Once youve got your bus speed, mess with your multi to get to ~4.5ghz, and set your ram speed to whatever it's stable at (usually around 1800mhz), also change your voltage to manual at 1.45v.

Step 5: go test in Occt for 5 min. If you get voltage drop on vcore raise llc until it doesn't drop. After all that, do you have headroom? If yes, continue to push multi until unstable, at that point slowly increase voltage OR just llc until it's stable without exceeding 1.45v under load or 70c.  If you didn't have more headroom, stop here and decrease voltage until you have the lowest possible voltage while still being stable. 

 

Or you could just buy a PC from me next time and save yourself time and money

 

Source: Tons of experience, I have an 8320 downstairs as emulator PC clocked at 4.8ghz 1.43v and won't break 53c in 22c ambient and all it has is hyper 212 cooler, and a few hidden magic tricks :).  Just threw together an 8350/gtx 690 combo I'll be selling cheap with a baby cooler on it, should be breaking national record Tuesday with it. 

Would have been more useful if you did read it. I already stress tested with OCCT. There would be no way I could get to 1.45v, It reaches 68c socket temp in OCCT with just 1.38. (1.34-1.35 actual voltage). If having LLC set to low decreased voltage to 1.32v causing an error in OCCT, I'm sure turning it off would just make it worse. The only way for me to NOT get an error with 1.38V is to set LLC to high. 

[SubLimation7]

7 minutes ago, DarkEnergy said:

Would have been more useful if you did read it. I already stress tested with OCCT. There would be no way I could get to 1.45v, It reaches 68c socket temp in OCCT with just 1.38. (1.34-1.35 actual voltage). If having LLC set to low decreased voltage to 1.32v causing an error in OCCT, I'm sure turning it off would just make it worse. The only way for me to NOT get an error with 1.38V is to set LLC to high. 

OK, did you try bus speed overclock? What thermal paste are you using.  Setting LLC to high is fine, but that's always the last step in overclocking.  Even then, if all you do is game nothing to really worry about, I've ran these chips at 1.6v on air for an hour plus without issue. Like you said no games will peg your cpu at 100% unfortunately.  I bet you'd be surprised what you achieved if you followed my guide, or maybe optimize your airflow and upgrade thermal paste to Kryonaut to shave 10-15c depending on current paste.  Also, only thing that matters is actual voltage under load, but your temps shouldn't be that high, I guarantee if you remount cooler with kryonaut and maybe upgrade an intake fan on case and make sure you use excellent mounting pressure with heatsink you'll shave a lot off temps. I've ocd at least 15 83xx chips with hyper 212 and never achevied less than 4.7.  But if all your after is Gaming performance you'll notice a difference if you use my oc method. 

[DarkEnergy]

2 minutes ago, SubLimation7 said:

OK, did you try bus speed overclock? What thermal paste are you using. 

Have not tried bus speed overclocking. Never even heard of it. I'm using Arctic Silver 5 and the cooler is a 212 EVO. Here's what I'm at now under load:

 

 

 

[SubLimation7]

11 minutes ago, DarkEnergy said:

Have not tried bus speed overclocking. Never even heard of it. I'm using Arctic Silver 5 and the cooler is a 212 EVO. Here's what I'm at now under load:

 

 

 

I used to use as5 exclusively, but theres better out now.  If you've had the system a few years I'm sure it could benefit from a paste swap regardless, and kryonaut has much higher thermal conductivity.  But if you do order id probably grab a  noctua fan for intake front bottom of case and one to replace stock 212 evo fan (static pressure optimized). 

 

But with all that being said it might be hard for you to justify time/money to do that just for a few hundred mhz, I'd recommend you run some gaming benchmarks and note your scores and then try my method and compare. If you do please report back with results!

 

And there's also the off chance you have a faulty chip, vrms, psu, etc. But I kind of doubt that. 

[DarkEnergy]

3 minutes ago, SubLimation7 said:

I used to use as5 exclusively, but theres better out now.  If you've had the system a few years I'm sure it could benefit from a paste swap regardless, and kryonaut has much higher thermal conductivity.  But if you do order id probably grab a  noctua fan for intake front bottom of case and one to replace stock 212 evo fan (static pressure optimized). 

 

But with all that being said it might be hard for you to justify time/money to do that just for a few hundred mhz, I'd recommend you run some gaming benchmarks and not your scores and then try my method and compare. If you do please report back! 

 

 

When I upgrade my computer (maybe in a year or two), I will definitly get better thermal paste, two fans on the CPU cooler (unless I go the watercooling route), and stronger intake fans (currently have negative case airflow pressure). For right now I'm just trying to maximize the performance of the current setup. 

[SubLimation7]

10 minutes ago, DarkEnergy said:

When I upgrade my computer (maybe in a year or two), I will definitly get better thermal paste, two fans on the CPU cooler (unless I go the watercooling route), and stronger intake fans (currently have negative case airflow pressure). For right now I'm just trying to maximize the performance of the current setup. 

I hate to sound like I'm trying to make a sale, but you should definitely check me out next time around, my pcs are guaranteed (to an extent) to perform on same level or better as custom loop liquid cooling, without all the headaches, maintenance, noise,or $1000 in extra parts, mine actually only cost a tad more than to build yourself and you get a warranty and whatnot. I love Jayztwocents, but my air cooled 1800x outperformed his custom loop 1800x in thermals, noise, frequency, and ofc price. 

 

But either way it goes good luck to you! And if you ever get some free time to try my method plz let me know results, I like to have all the data I can get. 

 

P. S.  Voltage is not right in cpuz screen for 1800x

[SpaceGhostC2C]

1 hour ago, DarkEnergy said:

FX 8350

Motherboard? That is very important in determining how stable is your voltage going to be, how much vdroop you are going to face (yes, those numbers are normal), and most important, how far you can push it safely.

 

Quote

 

As a comparison with 1.38 voltage and 4414MHz settings, Prime95 pushes socket temp to 74c (danger) and core temp to 56, which is why I use OCCT, an ever so slightly less intense benchmark (but more intense than AIDA64). I'm using the NZXT s340 case so there is no ventilation on the back of the motherboard. The only way to cool the socket temp is to take the back panel off and blow a fan onto the back of the motherboard. 

That is a good choice. Prime95 generates more heat, but OCCT fails sooner for unstable OCs.

 

Other than that, a few comments:

- 4400 May be unfeasible with the 212, depending on how much voltage your CPU needs. If you manage to get it stable without too much vcore, then you may get away with the 212.

- LLC will decrease vdroop and increase temps. At the end of the day, what you want to track is the effective vcore your CPU gets under load in OCCT. Any combination of settings that gives you the same effective vcore will probably be equally stable and generate similar heat, although there is some variation. But I'd say you need to figure out which effective load vcore you need, and then investigate the most efficient way to achieve it.

- You can let socket temps go a little bit beyond 70 as long as package temps are blow 60, and as long as it happens only in stress tests. However, large differences between package and socket temps on air cooling are likely to be caused by hot VRMs. If you have the stock heatsink around, try attaching its small fan to the VRM heatsink, it can make a drastic difference in VRM temps, and a moderate impact on socket temps.

- You'll have an easier time finding a stable overclock if you only use the multiplier first, since taking RAM out of specs at the same time as the CPU doubles the problem.

 

If you tell me your motherboard model, I may be able to give you more detailed help (I've used an ASUS and an ASRock, the latter without any LLC whatsoever).

[DarkEnergy]

2 minutes ago, SpaceGhostC2C said:

Motherboard? That is very important in determining how stable is your voltage going to be, how much vdroop you are going to face (yes, those numbers are normal), and most important, how far you can push it safely.

 

That is a good choice. Prime95 generates more heat, but OCCT fails sooner for unstable OCs.

 

Other than that, a few comments:

- 4400 May be unfeasible with the 212, depending on how much voltage your CPU needs. If you manage to get it stable without too much vcore, then you may get away with the 212.

- LLC will decrease vdroop and increase temps. At the end of the day, what you want to track is the effective vcore your CPU gets under load in OCCT. Any combination of settings that gives you the same effective vcore will probably be equally stable and generate similar heat, although there is some variation. But I'd say you need to figure out which effective load vcore you need, and then investigate the most efficient way to achieve it.

- You can let socket temps go a little bit beyond 70 as long as package temps are blow 60, and as long as it happens only in stress tests. However, large differences between package and socket temps on air cooling are likely to be caused by hot VRMs. If you have the stock heatsink around, try attaching its small fan to the VRM heatsink, it can make a drastic difference in VRM temps, and a moderate impact on socket temps.

 

If you tell me your motherboard model, I may be able to give you more detailed help (I've used an ASUS and an ASRock, the latter without any LLC whatsoever).

Everything is in my signature. It's an Asus M5A99FX Pro R2.0. For now I'm calling the 1.38 volt (1.356 actual), with LLC set to high and socket temps at 68c for 10 minutes in OCCT as stable. Not sure how I would mount a fan over the VRMs. 

[SubLimation7]

9 minutes ago, SpaceGhostC2C said:

Motherboard? That is very important in determining how stable is your voltage going to be, how much vdroop you are going to face (yes, those numbers are normal), and most important, how far you can push it safely.

 

That is a good choice. Prime95 generates more heat, but OCCT fails sooner for unstable OCs.

 

Other than that, a few comments:

- 4400 May be unfeasible with the 212, depending on how much voltage your CPU needs. If you manage to get it stable without too much vcore, then you may get away with the 212.

- LLC will decrease vdroop and increase temps. At the end of the day, what you want to track is the effective vcore your CPU gets under load in OCCT. Any combination of settings that gives you the same effective vcore will probably be equally stable and generate similar heat, although there is some variation. But I'd say you need to figure out which effective load vcore you need, and then investigate the most efficient way to achieve it.

- You can let socket temps go a little bit beyond 70 as long as package temps are blow 60, and as long as it happens only in stress tests. However, large differences between package and socket temps on air cooling are likely to be caused by hot VRMs. If you have the stock heatsink around, try attaching its small fan to the VRM heatsink, it can make a drastic difference in VRM temps, and a moderate impact on socket temps.

 

If you tell me your motherboard model, I may be able to give you more detailed help (I've used an ASUS and an ASRock, the latter without any LLC whatsoever).

Do motherboards really make a big difference in your experience?  Because they don't for me unless doing some crazy overclocking. And 4400 is more than feasible on 212, even back when 8350 had just launched I was constantly hitting 4.6-4.7ghz and I never used anything except 212 and as5 back then. Also I used to do builds in nzxt Lexa s with stock fans, and then a little later switched to nzxt phantom (phanteks now)  but didn't change much for CPU temps. What did make a huge difference was polishing and using liquid metal. Even my high oc on 1800x is on a $100 b350 chipset board with minimal vrm goodness.  Not trying to sound like a dick but I have so much experience with these chips lol and it never came down to mobo.

 

cooling your vrms won't drop core temps, which is your issue, these chips temp makes all the difference, when my 8320 was on as5 I couldn't get it past 4.6ghz, after mods and liquid metal (just reducing temps) I can run at 5ghz across all 8 cores stable. 

[SubLimation7]

2 hours ago, Terryv said:

Your PSU could be the cause of Vdroop. What PSU do you have? Good quality units don't usually have problems. On my 8320 my LLC was set to high and was steady for my settings.

 

It seems that you are temperature limited at this point. You need a better cooler or you will throttle. No damage should occur as you CPU is throttling to keep temps in check.

 

Prime will stress it above and beyond any realistic real world load. I like to use handbrake to check max temps as that's the highest real load I've ever used.

 

Note: handbrake is not for stability testing.

 

Socket temp will drop along with core temp even though its always higher than core temperature. (By about 12-15 degrees in my experience) 

Not to call everyone out but.. 8350 on sabertooth 990fx r2.0 made no difference in vdrop going from cx600 to supernova p2 1000w.  So I went from bronze tier 3? to platinum tier one and no difference. Seems that chipset and power delivery on motherboard is usual culprit. On every x99 build I've done there is no vdrop regardless of hardware, on z170 boards I always have a small amount of vdrop, on 990fx boards I always got A LOT of vdrop, however it would vary from system to system, but it was always there and power supply never changed anything. With that being said today I never use anything except tier 1 for other reasons. 

[Terryv]

4 minutes ago, SubLimation7 said:

Not to call everyone out but.. 8350 on sabertooth 990fx r2.0 made no difference in vdrop going from cx600 to supernova p2 1000w.  So I went from bronze tier 3 to platinum tier one and no difference. Seems that chipset and power delivery on motherboard is usual culprit. On every x99 build I've done there is no vdrop regardless of hardware, on z170 boards I always have a small amount of vdrop, on 990fx boards I always got A LOT of vdrop, however it would vary from system to system, but it was always there and power supply never changed anything. With that being said today I never use anything except tier 1 for other reasons. 

I've had little vdroop on my M5A99FX pro r2.0, I have a tier 2 PSU. I may have gotten lucky

[SubLimation7]

18 minutes ago, Terryv said:

I've had little vdroop on my M5A99FX pro r2.0, I have a tier 2 PSU. I may have gotten lucky

Yes I certainly wont pretend that I tested every motherboard out there, but I've never seen a 990fx board regardless of cpu not have pretty severe vdrop with LLC off.

 

Back when I was "newer" to pc building I was guessing it was power supply as well, I distinctly remember have 3 identical builds in Phantom full towers all rocking 8350s with 212 cooler (even all sabertooth 990fx r2.0 boards, we got together for a lan party and one of my friends kept crashing and he was having quite a lot more vdrop than everyone else, we fixed it with LLC, but later down the road when he got a new gpu, upgraded ram, 3 ssd's in raid 0, and some extra fans I said why not jump to 1000w psu while your at it, and I promise it changed nothing.  Now hes using one of my delidded 6700k builds and the power supply is still going strong so no regrets there, but definitely wouldn't suggest to fix your Vdrop. Funny I still have some of the boxes, and some are from more recent builds.

[SpaceGhostC2C]

46 minutes ago, SubLimation7 said:

Do motherboards really make a big difference in your experience?  Because they don't for me unless doing some crazy overclocking.

Yes, they do: to begin with, they don't even have the same options (as I said above, one of my boards doesn't even have LLC), they don't handle fast ram equally well, and they don't provide equally stable voltages. Plus they don't have the same cooling solution on their VRMs. Some motherboards can't handle 8-core CPUs at stock (before asking, OP could have any AM3+ board). Hot VRMs will throttle CPUs even if the CPU temps are fine...

 

46 minutes ago, SubLimation7 said:

And 4400 is more than feasible on 212, even back when 8350 had just launched I was constantly hitting 4.6-4.7ghz and I never used anything except 212 and as5 back then.

It all boils down to which voltages you needed for 4.4. Plus using liquid metal may be a different story. Still, I'd suggest checking old threads at overclockers.net...

46 minutes ago, SubLimation7 said:

 Not trying to sound like a dick but I have so much experience with these chips lol and it never came down to mobo.

I don't think you have experimented with enough variety of boards  

 

46 minutes ago, SubLimation7 said:

cooling your vrms won't drop core temps, which is your issue, these chips temp makes all the difference, when my 8320 was on as5 I couldn't get it past 4.6ghz, after mods and liquid metal (just reducing temps) I can run at 5ghz across all 8 cores stable. 

Core temps are your issue if your core temps are close to their limit. When socket temps are close/above their limit, on the other hand... And we all saw the SLI Krait video. It's not like it's the only thing that matters, but I wouldn't neglect it either.

 

54 minutes ago, DarkEnergy said:

Everything is in my signature. It's an Asus M5A99FX Pro R2.0. For now I'm calling the 1.38 volt (1.356 actual), with LLC set to high and socket temps at 68c for 10 minutes in OCCT as stable. Not sure how I would mount a fan over the VRMs. 

I assume it is very close to the Sabertooth in terms of settings and VRMs. If so, enough LLC will let you achieve pin-point vcore, at the expense of some additional temperature. With the Sabertooth, I got an 8370E to 4.4 on 1.28 something (1.2875, I believe), but I didn't settle for the highest LLC (I have to start it up to check exact settings), so effecitve voltage could be somewaht lower. Temps are fine, but cooler is different. On the other hand, I could never undervolt a 9370 that low (in the ASRock, though, but effective vcore is seldom below 1.34-1.35, 1.38xx in BIOS).

Regarding fan mounting, it depends on case orientation, but it takes a little improvisation and creativity - it basically comes down to what you have at hand  

[SubLimation7]

4 minutes ago, SpaceGhostC2C said:

Yes, they do: to begin with, they don't even have the same options (as I said above, one of my boards doesn't even have LLC), they don't handle fast ram equally well, and they don't provide equally stable voltages. Plus they don't have the same cooling solution on their VRMs. Some motherboards can't handle 8-core CPUs at stock (before asking, OP could have any AM3+ board). Hot VRMs will throttle CPUs even if the CPU temps are fine...

 

It all boils down to which voltages you needed for 4.4. Plus using liquid metal may be a different story. Still, I'd suggest checking old threads at overclockers.net...

I don't think you have experimented with enough variety of boards  

 

Core temps are your issue if your core temps are close to their limit. When socket temps are close/above their limit, on the other hand... And we all saw the SLI Krait video. It's not like it's the only thing that matters, but I wouldn't neglect it either.

 

I assume it is very close to the Sabertooth in terms of settings and VRMs. If so, enough LLC will let you achieve pin-point vcore, at the expense of some additional temperature. With the Sabertooth, I got an 8370E to 4.4 on 1.28 something (1.2875, I believe), but I didn't settle for the highest LLC (I have to start it up to check exact settings), so effecitve voltage could be somewaht lower. Temps are fine, but cooler is different. On the other hand, I could never undervolt a 9370 that low (in the ASRock, though, but effective vcore is seldom below 1.34-1.35, 1.38xx in BIOS).

Regarding fan mounting, it depends on case orientation, but it takes a little improvisation and creativity - it basically comes down to what you have at hand  

Look I get what your saying, but none of that applies here.  All I know is if he brought me his system I could push his clocks way higher w/o changing mobo, vrm, mosfets, psu, any of that.  Now if the question was can motherboards help stability of overclocks the answer would without a doubt be yes.  With that being said, useful info in your post for everyone else.  I just don't want the OP running out and getting new hardware just to be in the same situation, if maybe slightly slightly improved.

 

And definitely buy a mobo with overclocking features if you plan to overclock but sometimes I do leave out good info that I take for granted.

[SpaceGhostC2C]

4 minutes ago, SubLimation7 said:

  I just don't want the OP running out and getting new hardware just to be in the same situation, if maybe slightly slightly improved.

We're on the same boat, I'm also thinking in terms of how OP can get the best out of what he has. 

[DarkEnergy]

On 3/19/2017 at 7:55 PM, SpaceGhostC2C said:

-snip-

 

On 3/19/2017 at 7:49 PM, SubLimation7 said:

-snip-

 

Just an update. I decided to clean my computer today. It went from cleaning to rearranging the fans, adding a fan to the other side of the 212 EVO, 2 hours of pain in the ass and some ghetto rigging later and I'm at 64c socket after 10 minutes of OCCT. Also to note, I made some adjustments a week ago and I'm at 20.5 on the multiplier up from 20, I lowered the bus to ~216MHz, voltage set to 1.375 (1.368 actual idle and 1.344-1.356 actual under load), and LLC to High.This seems to give best performance compared to 20 on multiplier and 220MHz bus.  

 

So now the temps are pretty good as far as a 212 EVO goes. BUT, if I wanted to overclock more, I'm going to need more voltage. It's already at the lowest voltage limit, If I raised anything, I would need to increase voltage. There's no point in raising it past 1.375 since it doesn't even reach that high, which leaves setting LLC to extreme. Is that even safe with my specs? Or is it best to call it a day at 4.4GHz and 64c socket temp under load?