Playing with an i5 3570k, thoughts?

[wand3r3r]

I am just testing out a 3570k/Gigabye Z77 UD3H board.

Simply dropping 45 as the multiplier while not touching anything else resulted in 4.5GHz and the voltage seems to vary between 1.2159 and 1.2209. The temperatures peak around the 80ish mark max 78 -> 84 in one core. (prime95)

Edit. These voltages were waay off. I noticed the BIOS was way higher so I am now checking volts from CPUz.

I guess these temps are acceptable but I'm sure I should manually start dropping/setting the voltage (also to ensure it doesn't rise above 1.22 as the temps are fairly high). It appears the other alternative is to lock the voltage and increase the multiplier until it fails.

Anandtech got this on the board/3770k.

At 1.100 volts, 4.5 GHz. 70ºC at load 1.116 volts.

At 1.150 volts, 4.6 GHz. 75ºC 1.164 volts.

At 1.200 volts, 4.7 GHz. 82ºC with PovRay, 86ºC with OCCT, and a load voltage of 1.212 volts.

At 1.250 volts, the board successfully booted at 4.8 GHz, with 1.272 volts under load and 89ºC with PovRay - but this was not stable due to the memory errors in PovRay, suggesting more voltage is required. Given the current load temperature, I was unwilling to push the voltage further.

Any thoughts on this?

 

[WhoBeDaPlaya]

Didn't they get those results with a 3770K?

 

[wand3r3r]

Didn't they get those results with a 3770K?

Ah good catch, you're right.
I just went straight into the page where this board was.

 

[WhoBeDaPlaya]

Heh, no worries. I actually built a 3570K / UD3H setup just a few weeks back.
Got 4.5GHz with a fixed 1.2Vcore. Max temp (CoreTemp) during IBT is 79C using IC Diamond and an Antec Kuhler 620 w/push-pull fans.

 

[jkauff]

I've been running my 3570K at stock voltage using a 45 multiplier for months now with no problems. I don't do stress testing, but running two-hour Handbrake encodes with all four cores at 100% I rarely get higher than 70c on any core. Using Thermaltake air cooling with Arctic Silver TIM, no discrete GPU.

 

[lehtv]

I've been running my 3570K at stock voltage using a 45 multiplier for months now with no problems. I don't do stress testing, but running two-hour Handbrake encodes with all four cores at 100% I rarely get higher than 70c on any core. Using Thermaltake air cooling with Arctic Silver TIM, no discrete GPU.

You probably mean stock VID. VID is the hard coded voltage that the CPU tells the motherboard's voltage regulator to supply at a given clock speed. The higher the clock speed, the higher the VID in order to remain stable. Running at 4.5GHz at truly stock voltage would mean applying a high enough negative offset voltage to achieve the same VCore as when running at stock clocks.

 

[TJCS]

I've been running my 3570K at stock voltage using a 45 multiplier for months now with no problems. I don't do stress testing, but running two-hour Handbrake encodes with all four cores at 100% I rarely get higher than 70c on any core. Using Thermaltake air cooling with Arctic Silver TIM, no discrete GPU.

Not trying to jinx you or anything but during my testing I have had HandBrake CLI crash just right before the end of a 8hr high-bitrate 2-Pass encode, while the lower-quality encodes are all error free.

Sometimes, it could be just the opposite. The HQ encodes go through fine but the lower-quality encodes causes WHEA errors.

 

[jkauff]

Not trying to jinx you or anything but during my testing I have had HandBrake CLI crash just right before the end of a 8hr high-bitrate 2-Pass encode, while the lower-quality encodes are all error free.

Sometimes, it could be just the opposite. The HQ encodes go through fine but the lower-quality encodes causes WHEA errors.

Are you using last HB release or nightlies? I've found the recent nightlies to be more stable than the stable release.

 

[TJCS]

Are you using last HB release or nightlies? I've found the recent nightlies to be more stable than the stable release.

I do not solely use HandBrake to confirm my stability results, however when there is abnormal behavior it is usually a good indication that something is not right.

I am using the current official 0.9.8, but I haven't had any stability issues with the program itself since the last couple of official releases.

The only time I have problems with HandBrake stability was due to improper overclock/bios settings. I can validate this by either WHEA errors in the event log, or using other methods of stability testing to confirm this.

Ever since I obtained my stable bios settings months ago, I have had zero Handbrake crashes.

 

[MrSquished]

I've been trying to find my OC with my 3570 - yesterday with my vcore holding at 1.2 I was at 4.5 ran prime for a couple hours no problem, i have alerts set for any WHEA errors, temps with a Hyper Evo were peaking from 68-76 on prime, gaming in the mid-high 50's. However Linx was a beast on my temps at 1.2v, they were going into the 90's. I didn't like that so I dialed back down to a 4.4ghz OC, vcore holding at 1.7 and everything runs fine. Linx still gets into the 80's but since nothing I do in real life comes close to that kind of stress I don't worry.

Oh I also downloaded a trial of Aida64 after going back down to 4.4 - stress testing with this chip at 4.4 w 1.17v my peak temps are from like 66-74. again playing BF3 my temps go up to the lower 50's. Ambient temp is 25C, Idle temps are basically 30-32C.

If I had a better cooler I think I could run this chip at 4.5-4.6 with 1.2v no sweat but while it's tempting, I am not buying another cooler for a couple hundred megahertz I don't need with money I don't have.

Board is an Asrock extreme 4

 

[TJCS]

@PixelSquish

I think you meant to say 1.17v @ 4.4? :sneaky:

What level is your LLC at?

 

[MrSquished]

@PixelSquish

I think you meant to say 1.17v @ 4.4? :sneaky:

What level is your LLC at?

correct! my bad. Will edit my post to reflect that.

LLC is at Level 3.

 

[TJCS]

correct! my bad. Will edit my post to reflect that.

LLC is at Level 3.

Not sure how extensively you have tested your settings, but I think you have some room for lowering temps at your current oc. (I have the same motherboard).

Have you tried LLC Level 4?

 

[2is]

You probably mean stock VID. VID is the hard coded voltage that the CPU tells the motherboard's voltage regulator to supply at a given clock speed. The higher the clock speed, the higher the VID in order to remain stable. Running at 4.5GHz at truly stock voltage would mean applying a high enough negative offset voltage to achieve the same VCore as when running at stock clocks.

I'm thinking the same. A lot of people think leaving the voltage setting alone means its running at the same voltage at their overclocked speeds as it was at the default speeds, when in-fact, they could very well be running higher voltage than those of us who fiddled around with our offsets.

 

[tweakboy]

Nice rig, nice OC, now go use it game on! ... good job grabbin 16GB ram, have fun

 

[MrSquished]

Not sure how extensively you have tested your settings, but I think you have some room for lowering temps at your current oc. (I have the same motherboard).

Have you tried LLC Level 4?

you are very correct again.

i really need to get an SSD because re-booting into windows is so slow - because you know it could be taking 10 seconds to when you can start ripping off proggies. actually, it's not so much booting into windows but how long it takes for the desktop to become usable. so it's made me loath to tinker too much.

but i did go back and muck about - i did lower to LLC level 4 and lo and behold i found taht i had forgotten to lower my offset voltage setting from + .050 that i was using on my 4.5 OC testing. i lowered it to +.020 and my temps went down a nice bit. ran prime for an hour and then 5X Linx, no WHEA errors since then either - will run prime all night tonight but it looks good. prime on large fft took no core past 70 degrees this time, Linx did get the two middle cores to hit 85 but more often than not they were hovering around 80 while the two end cores never broke 80. my wager is playing bf3 will be right around 50.

in CPU-Z under 100% load in Linx Core Voltage shows up as a pretty steady 1.12 going up to 1.128 or down to 1.116 at the most. this is of course at 4.4 where we are staying until i can afford to splurge on a higher end cooler. which will be after an SSD.

also for some reason Core Temp will always lock up while running Linx. does fine in Prime though. need to use realtemp to monitor when running linx.

 

[MrSquished]

What is more accurate for voltage - CPU-Z or Core Temp? Big difference in my screen shot below when running prime - was also the same difference when running 5x Linx run earlier:

 

[tracerbullet]

If I remember right, Core Temp shows VID which is sort of the "requested" voltage by the CPU, and CPU-Z is more like the actual voltage being given to it by the board. I don't know for certain so let someone else verify. I'll regret venturing that response if it's totally wrong.

 

[MrSquished]

well 9 hours in prime ran just fine with CPU-Z showing the core voltage to be 1.112 at 4.4ghz

 

[coffeejunkee]

Looks like a decent chip, my 3570K can only dream about that.

But how about actually playing a game, no crashes/whea errors? And do you have c3/c6 states enabled or disabled?

 

[MrSquished]

Looks like a decent chip, my 3570K can only dream about that.

But how about actually playing a game, no crashes/whea errors? And do you have c3/c6 states enabled or disabled?

i've been wondering about the c3/c6 states what exactly are the plusses and minuses, but they are enabled.

i have an alert to go off for any WHEA errors as i was getting them while tweaking my oc and i haven't had one since at these voltages and these clocks for the last 3 days. my offset voltage going down by a smidgen will get me WHEA errors.

i've played 4-5 hour bf3 sessions since these settings without a hiccup. no crashes doing anything.

 

[coffeejunkee]

Ok, sounds good. I'm asking because i've noticed prime/linx testing means basically nothing for my cpu. It can happily complete 20 runs Linx at a vcore that will make BFP4F crash in minutes.

C3/C6 are powersaving states which completely shut off parts of the cpu. They are also required for turbo boost at stock. If you can leave them enabled that's good, but I had to disable them at higher freqs as they caused some crashes/whea errors.

 

[Idontcare]

i have an alert to go off for any WHEA errors as i was getting them while tweaking my oc...

How do you set that up? Sounds handy.

 

[MrSquished]

How do you set that up? Sounds handy.

i had no idea what a WHEA was until last week because of reading about it on AT. good to pay it forward.

go to your Event Viewer under Adminastrative Options, open up Warnings, then go to WHEA logger and double click it. on the right hand side pane a bunch of options appear under Actions - under actions go to "Attach Task to this Custom View"

Name your task, hit next. Then 'When an Event is Logged' is next, click next - then under Action select Display a Message. then finish.

voila. you get a little pop up box when a WHEA happens.

 

[Idontcare]

i had no idea what a WHEA was until last week because of reading about it on AT. good to pay it forward.

go to your Event Viewer under Adminastrative Options, open up Warnings, then go to WHEA logger and double click it. on the right hand side pane a bunch of options appear under Actions - under actions go to "Attach Task to this Custom View"

Name your task, hit next. Then 'When an Event is Logged' is next, click next - then under Action select Display a Message. then finish.

voila. you get a little pop up box when a WHEA happens.

:thumbsup: thanks!