So i think i broke my motherboard! replacement?

[carling220]

I spent more than you suggested, but also much less than a new cpu and motherboard. I bought 2 new fans, the 212 evo, and an asus board for about £200 all in. I'd need the fans come what may, so take that away and its £180. For a new cpu and board of any note it'd cost a fair bit more!

I also decided i'm not a serious gamer really anyway, everything ran fine before, except GTA. So I will overclock my 2500k, which is good learning experience.

Ive looked at bench marks now online I found, the OC 2500k isn't that far behind the newer OC's I5's at all, not many FPS behind from what I saw. I'll use the money I saved on my new GPU, and then sell my 560Ti and 770 windforce to make some of that back!

Thanks for the advice, I think this was a better route! Learnt a lot also looking in to all this. The only challenge now is to fit everything together again without breaking it, lol!

 

[BonzaiDuck]

I spent more than you suggested, but also much less than a new cpu and motherboard. I bought 2 new fans, the 212 evo, and an asus board for about £200 all in. I'd need the fans come what may, so take that away and its £180. For a new cpu and board of any note it'd cost a fair bit more!

I also decided i'm not a serious gamer really anyway, everything ran fine before, except GTA. So I will overclock my 2500k, which is good learning experience.

Ive looked at bench marks now online I found, the OC 2500k isn't that far behind the newer OC's I5's at all, not many FPS behind from what I saw. I'll use the money I saved on my new GPU, and then sell my 560Ti and 770 windforce to make some of that back!

Thanks for the advice, I think this was a better route! Learnt a lot also looking in to all this. The only challenge now is to fit everything together again without breaking it, lol!

Wow, Man! I Like to see optimism! You seem happier-than-a-pig-in-s***!!

That's great! If you want some tips, send me a PM with questions. First step: Set it all up at stock with "auto" voltages; get the memory configured and thoroughly tested -- HCI Memtest 64 is good for that.

Then you want to make a note of the default voltages showing in the BIOS column of monitored values. [I'd keep a notebook with dates and a format for writing down settings.] I'm sure all those ASUS boards present it that way, next to the settings column. Configure it for "Turbo" mode, find the multiplier that defines the maximum turbo multiplier. Then edge up with multiplier a notch at a time until the system fails -- to find the maximum stable speed for "stock voltage" settings. You could stop there, but read a couple or three online OC'ing guides, and don't hesitate to PM and ask.

 

[carling220]

Wow, Man! I Like to see optimism! You seem happier-than-a-pig-in-s***!!

That's great! If you want some tips, send me a PM with questions. First step: Set it all up at stock with "auto" voltages; get the memory configured and thoroughly tested -- HCI Memtest 64 is good for that.

Then you want to make a note of the default voltages showing in the BIOS column of monitored values. [I'd keep a notebook with dates and a format for writing down settings.] I'm sure all those ASUS boards present it that way, next to the settings column. Configure it for "Turbo" mode, find the multiplier that defines the maximum turbo multiplier. Then edge up with multiplier a notch at a time until the system fails -- to find the maximum stable speed for "stock voltage" settings. You could stop there, but read a couple or three online OC'ing guides, and don't hesitate to PM and ask.

thanks for the info, just waiting for my board to be delivered, fingers crossed today.

I do have one important question actually since you offered. Before I overclock do I need a new power supply? Mine is a 650w supply. I think it should probably be OK, been fine for years. I guess it depends on how you overclock with regard to voltages? Have to keep an eye on this for a new GPU also.

 

[BonzaiDuck]

thanks for the info, just waiting for my board to be delivered, fingers crossed today.

I do have one important question actually since you offered. Before I overclock do I need a new power supply? Mine is a 650w supply. I think it should probably be OK, been fine for years. I guess it depends on how you overclock with regard to voltages? Have to keep an eye on this for a new GPU also.

"fine for years . . " How many years?

Who actually manufactured the PSU (the OEM)?

They slowly deteriorate over time. Certain practices are bad for even good PSUs: for instance, if your computer goes to sleep and maintains that sleep state for days at a time without use, it will really tax some PSU components because it isn't providing any cooling for them.

Usually, a good PSU (such as one made by Seasonic -- even re-branded) should have a 5-year warranty.

Do you have -- or can you get -- one of those little PSU testers?

If the unit behaves properly now, you could test it and give it a try.

For the wattage, here's my recent experience. I used a 650W Seasonic semi-modular unit to build my 2700K system with the refurbished Z68/Gen3 motherboard. It was, essentially, brand-new -- used to just test a couple motherboards, or run for no more than a few days.

This is a big discussion that pops up in the Power Supplies forum: Whether the PSU online "calculators" (like Extreme Outervision) are accurate; how much the rated wattage should be to use with this or that graphics cards -- more succinctly -- 2x graphics cards in SLI/Crossfire.

For two GTX 970's, with the processor OC'd for turbo 4.7Ghz or 137W under burn-test/LinX stress-test, and the 970s OC'd to maybe 1450/3750 core/memory speeds, the 650W seemed to be enough. That is, the 650W would cover the extreme and unlikely power-draw (more than mere gaming) with about 50W to spare. The system has about 4 fans, one HDD and two SSDs . . . optical drive. Some folks would feel more comfortable with the SLI running from a 750W unit.

I can't tell you what to do with the old PSU. My Seasonics all have 5-year warranties and I needlessly avoid Bronze-rated for Gold or Platinum units. If they've run non-stop longer than the 5-year window, I'm inclined to replace them -- but only for major parts replacement -- like motherboard, RAM and processor -- with overclocking intentions. Often, you will be able to trace other hardware failures to a dying PSU. Or they will begin to have trouble going through sleep and hibernate cycles.

 

[carling220]

Thanks, OCZ PSU it is.

Quick question about mobo now i've swapped it.

New mobo is installed now. Everything seems to run fine for the minute. I've read people tend to re-install windows 7? is this definitely required? The board is a 77 chipset and old one was a 68, so potentially need to install new chipset drivers?

Thanks!

 

[BonzaiDuck]

Thanks, OCZ PSU it is.

Quick question about mobo now i've swapped it.

New mobo is installed now. Everything seems to run fine for the minute. I've read people tend to re-install windows 7? is this definitely required? The board is a 77 chipset and old one was a 68, so potentially need to install new chipset drivers?

Thanks!

You may be lucky I got back to you so soon if you haven't moved forward with this or decided to do a complete fresh Windows installation anyway.

If you actually got your Z77 system to boot from the Z68 windows boot-drive, it is as I might have predicted.

Before you do anything else, install the Z77 chipset drivers. After that, I'd reinstall (clean install for graphics) any additional hardware drivers -- starting with the SATA drivers (AHCI or RAID depending on your original setup -- don't get mixed up about the appropriate BIOS mode -- and the BIOS mode of the Z77 board should match that of the Windows boot disk you likely tried raising to desktop with the new board).

Then, check your event logs. I'm actually hoping you had the chance to keep these "blue and tidy" before your problem with the original motherboard. If not, there are benign errors, and "more sinister" errors which can be corrected. I'm assuming you've also run SFC /SCANNOW or otherwise didn't have any trouble with corrupted OS files up to this point. I'd also schedule some CHKDSK operations for next boot-up through Explorer->[HDD]: properties. On reboot, the usual recognizable summary of the CHKDSK operations should appear in an early Event Log information item soon after successful reboot.

You actually could be "good to go" on this. Make sure everything remains stable -- that all the basic operations including wake, sleep, hibernate etc. seem to work smoothly. Fact of the matter, you'll want to run some basic preliminary stress-tests on the stock configuration before you get started on that over-clocking business. You'll want to install your testing software if everything seems kosher -- OCCT, Prime95, LinX, IBT. Monitoring software: ASUS AI Suite, HWMonitor, CPU-Z, GPU-Z. As in the Boy Scouts: "Be Prepared." Or "the 6P principle:" "Proper Planning Prevents Piss-Poor Performance."

Probably should've mentioned this earlier in this post: You'll have to "re-activate" for the new hardware -- likely. If it hasn't alerted you to a reactivation requirement, it should and more likely soon. If you have to, go through the phone-activation process and be sure to answer the questions correctly -- "this windows was only installed on one computer."

I may have mentioned -- I bought this EBay bundle back in September, because I wanted the Sabertooth Z77 board. The owner included the original Windows white-box, since it was mated to the board. When I damaged the board and couldn't used it, I ordered a replacement Z68-V Pro/Gen3 -- still processor-compatible with both SB and IB. I installed the white-box Windows 7 Ultimate, went through the slightly-more-complicated activation, and -- "Voila!"

 

[BonzaiDuck]

PS

If Windows had been originally configured for AHCI BIOS storage mode, and IF you used the Windows native MSAHCI driver, you shouldn't have to fiddle with any SATA AHCI driver reinstall. If instead you had used the intel drivers, then install the ones for that particular motherboard. However, as I recollect from recent conversations here, some driver revisions covered BOTH Z68 and Z77 chipsets, so you may not even need to reinstall those. Or rather, you may only have to upgrade them.

See the recent discussion under "Intel R S T" or the equivalent phrasing -- probably in "Memory and Storage" forum.

PPS or whatever. . . . Keep an eyeball on your Device-Manager->"Ide ATA/ATAPI controller" or "Storage controller" nodes . . items . . . devices. "Yellow-'!'-bang" indicators should be addressed for all devices generally, only including storage.

 

[carling220]

Hi, thanks. Everything seems to work without even upgraded drivers.

Well I did the online reactivation, took 2 minutes. Everything works fine. CPU temps don't seem to go over 55 degrees with the new 212 cooler and new corsair case fan.

Maybe I do need to install the new chipset drivers though! I ran firestrike the 3d Mark test, and the score was much lower now, from 7200 to 6200! The graphics score was the same, but the physics score went from 78 to 3900! I think the physics score is about the CPU? Could this be caused by not having the correct drivers for the board?

edit: so i reset the PC, and the score went to 6783, with physics score going up to 5287. So physics score used to be 7200, then 3900, now 5287...strange

 

[carling220]

Problem solved. So, I did two things. First was installing the chipset drivers like you said, thanks. That brought the firestrike score back up to 6900. Then I went in to the bios, and out of three modes, efficiency, normal, and performance, I changed the mode to performance. It offers these modes on the UI on the main screen. Now it's back up 7200 for fire strike.

Will be interesting to see if OC's the CPU has any meaningful contribution towards the physics test on fire strike.

edit: So after pressing performance mode in the bios, I did the fire strike as I said. It said the CPU was at a constant 4330 MHz during the test, and the temperatures increased too. GTA V now works much better at a steady 50-60fps.

Has this setting in the ASUS bios done some sort of automatic overclock?

 

[BonzaiDuck]

Problem solved. So, I did two things. First was installing the chipset drivers like you said, thanks. That brought the firestrike score back up to 6900. Then I went in to the bios, and out of three modes, efficiency, normal, and performance, I changed the mode to performance. It offers these modes on the UI on the main screen. Now it's back up 7200 for fire strike.

Will be interesting to see if OC's the CPU has any meaningful contribution towards the physics test on fire strike.

edit: So after pressing performance mode in the bios, I did the fire strike as I said. It said the CPU was at a constant 4330 MHz during the test, and the temperatures increased too. GTA V now works much better at a steady 50-60fps.

Has this setting in the ASUS bios done some sort of automatic overclock?

It seemed to me that an AI Over-clocking feature was in the BIOS as well as AI Suite software for all those boards beginning at least with Z68.

And it would have done auto-overclocking, to get a 2500K up to 4.33Ghz.

We all discovered that those motherboard features always over-volt the processor needlessly, but you'll have to make your own assessment as to how that shakes out with the Z77 chipset. I'm pretty sure it would still be a shortcoming.

I used auto-overclocking on my boards just to get a baseline of what the processors would do -- a variation on the theme of finding highest clock at "auto" voltage, which I mentioned.

After that, I did all my tweaks manually in the BIOS, a little at a time, step by step.

When I used the BIOS AI-Tweaker overclock feature, the voltage was going as high as 1.44V just for 4.3 or 4.4 Ghz. That was, of course, the Z68 boards. My very highest voltages now show in the monitoring software at no more than 1.37V maximum -- and drooped voltage of 1.35V under severe load @ 4.7 Ghz. And this result follows enough tweaks and adjustments to be rock-stable running LinX for 50 iterations, or 20+ iterations of Intel Burn Test. I try to avoid more than one or two long-running tests, and avoid running Prime95 more than 5 hours or so, even if I may have run it for 9 hours once or twice. You could run it for 18 hours with good cooling and voltages under 1.37V maximum (some will say 1.4, and a lot of hot-dawgs felt comfortable with 1.44, but that's needless for 24/7 use).

With OCCT:CPU, you can trap errors in under 3.5 hours that might take longer with the other tests, and it doesn't heat-stress the processor as much as the latter. But I use all those tests, including Prime95 sFFT and LFFT -- methodically. It has to pass 50 iterations of "affinitized" LinX to get my Good Housekeeping OC award, and then I still do some fine tuning.

1.37V was the "upper safe boundary" when it was last published for 32nm socket-1366 chips, and the Sandy-Bridgers are 32nm processors. Also, it is the drooped voltage that is more significant for slow deterioration; if tests show the processor unloading itself and pegging 1.37+ -- no problem. It's not a hard and fast rule. They had published that spec in consideration of their anticipated RMA/warranty costs. Nobody I've ever heard of has had one of these processor buy the farm -- so to speak -- they're all still running at the user's original favorite overclocks.

In that business context, it's all about statistics and levels of risk.

Certainly, you will reduce temperatures a bit if you fine-tune the voltage and clocks for stability, and higher temperatures are also the bane of stable overclocks.

 

[Candymancan21]

I dunno if this was mentioned but arctic silver 5 is non conductive.. If you get paste on the cpu pins it doesn't cause shorts... This was discussed like 15 years ago... Maybe some of you are too young to remember or weren't in PC's back then... But people used to get AS5 under graphics card gpus and the cpus and it didn't cause any issues.

Also if you bend pins.. just bend them back with a small mini set of pliers or something.. It isn't hard.. I've been bending pins for years... Although I can understand that these pinless cpu's probably make it a bit more difficult because all you had to do back then was bend the pin to where the cpu fit in the socket lol

 

[BonzaiDuck]

I dunno if this was mentioned but arctic silver 5 is non conductive.. If you get paste on the cpu pins it doesn't cause shorts... This was discussed like 15 years ago... Maybe some of you are too young to remember or weren't in PC's back then... But people used to get AS5 under graphics card gpus and the cpus and it didn't cause any issues.

Also if you bend pins.. just bend them back with a small mini set of pliers or something.. It isn't hard.. I've been bending pins for years... Although I can understand that these pinless cpu's probably make it a bit more difficult because all you had to do back then was bend the pin to where the cpu fit in the socket lol

AS5 may not be "conductive" but as I understood it (and from "like 15 years ago") it has some capacitance. Personally, I can understand the OP's actions that led to his problem -- now solved. I wouldn't want any sort of contaminant on the CPU socket pins. Not so much as a human hair.

I just stopped using AS5 some 7 years ago -- mostly for minor differences in performance and not for the electrical properties.

 

[carling220]

Interesting, thanks for the information again.

Opening CPUID monitor it says the CPU VCore voltage, which I believe is what I am after is saying around 1.3 V. Min 1V, max 1.344V. From a quick google, 1.3V I think is considered fairly tame? Whether 1.3V at 4.3 GHz is considered good, I am not sure. Temperatures are fine under load, get to about 72 at max in GTA V, which is very CPU intensive. Sits about 45 degrees C average when browsing etc, tad high but not dangerous.

It seems then you are saying that with proper optimisation, I can probably get more bang for the buck, probably get a higher clock speed, and keep roughly the same temperature profile, if I messed around a bit with everything.


edit: So I went in to the BIOS. The CPU voltage in there says 1.1V, not 1.3V as CPUID monitor says. In advanced menu it says something like 'all cores attempt to boost to 4333mhz'. So maybe in this mode, it just extends the boost up to 4.3, where as stock wasn't it like 3.7? So potentially my CPU might only go up to 4.3 under heavy load. And sitting here now, maybe it is only 3.3?

 

[PhIlLy ChEeSe]

If your letting the Bios change the over clock it will change the V core, so when you SEE 1.3 in CPU-Z then in the bios 1.1v its cause the bios isn't over clocked yet. Also the V core will change if you have energy savings mode on, fluctuate as the CPU changes speed often.
That CPU you have is one of the best over clocking CPU'S ever, I think I had mine up to 5600 with water cooling and a few cores shut down. You'll keep busy for a while also, as the only limit you will have is the cooling. Should you want higher over clocks you'll need a water cooler.

 

[BonzaiDuck]

Interesting, thanks for the information again.

Opening CPUID monitor it says the CPU VCore voltage, which I believe is what I am after is saying around 1.3 V. Min 1V, max 1.344V. From a quick google, 1.3V I think is considered fairly tame? Whether 1.3V at 4.3 GHz is considered good, I am not sure. Temperatures are fine under load, get to about 72 at max in GTA V, which is very CPU intensive. Sits about 45 degrees C average when browsing etc, tad high but not dangerous.

It seems then you are saying that with proper optimisation, I can probably get more bang for the buck, probably get a higher clock speed, and keep roughly the same temperature profile, if I messed around a bit with everything.


edit: So I went in to the BIOS. The CPU voltage in there says 1.1V, not 1.3V as CPUID monitor says. In advanced menu it says something like 'all cores attempt to boost to 4333mhz'. So maybe in this mode, it just extends the boost up to 4.3, where as stock wasn't it like 3.7? So potentially my CPU might only go up to 4.3 under heavy load. And sitting here now, maybe it is only 3.3?

The BIOS monitor will show the voltage at the base multiplier for the processor, which is either 33 or 34. You might want to look at it with EIST enabled and then disabled (with the required reboots to BIOS), to see if EIST has any affect on the reported value. My vague recollection is that it doesn't. I could look at my notes from four years ago -- because I still have them.

I'm speculating somewhat, but I'd think the 1.3V reflects the auto-overclocking's excessive voltage with EIST turned off in BIOS. And I'd begin your sojourn of overclocking with both EIST and C1E turned on -- all the default energy-saving features. With EIST enabled, Windows monitoring and CPUID CPU_Z would report an idle value between 0.9 and 1.1V.

So, yes, I'll reply to say that you should be able to have the processor "turbo" on all cores to a nominal 4.5 Ghz to show a peak voltage close to what you already see.

You should read some overclocking guides from online resources, and not just one of them, for OC'ing both Sandy Bridge and Ivy-Bridge processors on the Z77 chipset -- preferably using the ASUS P8Z68 and P8Z77 models and their BIOS'es as reference.

As I said, take inventory and keep a notebook. I don't have the Z77 BIOS at my fingertips, and you should look through the menus to find an item resembling "Extra voltage for turbo mode," in addition to the Offset and other voltage settings in the tuning or "tweaker" menu. If you do find this second voltage setting, it can be used with the Offset voltage for settings that keep the actual VCORE from exceeding VID. If you don't have it and can't find it, it would mean that the Offset voltage would be your sole means of voltage adjustment.

Load-line Calibration -- LLC -- will decrease the default "droop" built in to the circuitry which keeps the same thing from happening under heavy load. You may eventually want to tune LLC so it shows a droop of ~2mV under load conditions. The default droop is somewhere in the neighborhood of 6mV to 8mV.

So for instance, you barely stress the processor so all cores still show maximum turbo speed, the peak turbo voltage will register in a monitoring program. When you severely stress the processor, the voltage will be so many mV below that value -- the "vDroop."

But read the over-clocking guides enough so that you can turn off any auto-overclocking, set everything back to default, and begin raising the turbo-multiplier above its default ~36 or 37 a notch at a time, to find the point where "auto" default voltage is becoming insufficient to maintain stability. That would be your first step.

You also have the prerogative to simply use the auto-overclocking for the stable 4.3Ghz and leave it alone, but it WILL be unnecessarily volted, and you CAN achieve higher clocks at similar voltages while keeping temperatures manageable.

 

[carling220]

Thanks for the great reply. I will definitely at some stage manually overclock. I am away for a few days starting at the weekend etc so will probably wait, read up, and become very familiar with what I am doing. I had a look in the BIOS and am not 100% sure on everything. I took screenshots, and some things I noticed:

At the top in yellow it says all cores target CPU turbo mode speed 4236MHz, and target DRAM speed of 1372MHz. AI overclock tuner is actually on manual, but it says even on manual, the CPU ratio and BCLK frequency are still optimized automatically. The EPU power saving mode is disabled.

CPU voltage it says is set to 1.152V, in 'offset mode', as opposed to manual mode. The CPU offset voltage is set to auto. Also on auto, are voltages for DRAM, VCCSA, PCH, and CPU PLL. Only thing I saw about turbo, was turbo ratio, which is set to 'by all cores'.


So yeah will stick with this auto for now. However, I noticed something a bit weird. At idle, or just sitting here now typing, the cores in CPUID monitor range from 37-52 degrees C. Why would it hit 50 degrees without any real stress? When I run firestrike however, at no point during the entire firestrike test does CPU temperature exceed 63 degrees C. Even in GTA V at 100% load for a while, the most I ever see is 72 degrees C. So under load the temperature is very well controlled. Could it be due to fan speed changing with load? Or is 30-50 at idle reasonable?

 

[BonzaiDuck]

Thanks for the great reply. I will definitely at some stage manually overclock. I am away for a few days starting at the weekend etc so will probably wait, read up, and become very familiar with what I am doing. I had a look in the BIOS and am not 100% sure on everything. I took screenshots, and some things I noticed:

At the top in yellow it says all cores target CPU turbo mode speed 4236MHz, and target DRAM speed of 1372MHz. AI overclock tuner is actually on manual, but it says even on manual, the CPU ratio and BCLK frequency are still optimized automatically. The EPU power saving mode is disabled.

CPU voltage it says is set to 1.152V, in 'offset mode', as opposed to manual mode. The CPU offset voltage is set to auto. Also on auto, are voltages for DRAM, VCCSA, PCH, and CPU PLL. Only thing I saw about turbo, was turbo ratio, which is set to 'by all cores'.


So yeah will stick with this auto for now. However, I noticed something a bit weird. At idle, or just sitting here now typing, the cores in CPUID monitor range from 37-52 degrees C. Why would it hit 50 degrees without any real stress? When I run firestrike however, at no point during the entire firestrike test does CPU temperature exceed 63 degrees C. Even in GTA V at 100% load for a while, the most I ever see is 72 degrees C. So under load the temperature is very well controlled. Could it be due to fan speed changing with load? Or is 30-50 at idle reasonable?

There is a tolerance range or expected calibration error in the core temperature sensors. It had always seemed to me that you would find the four cores varying around a core-average by +/- 5 or 6C. But this would not be evident in the "current values" column of HWMonitor. Running at idle for some few minutes, my 2700K system, HWMonitor shows the core-maximums trapped in the "Max" column as ranging from 36C to 46C.

So I'm wondering what monitor software you are running for your figures, or whether you're looking at the "Maximum" column or the "Current value." For instance, the column of current values for my system shows -- at any given instant, something similar to 35, 35, 34, 29, and all the cores are fluctuating in that range. The Maximum column will trap some instant of temperature for any core that doesn't represent an equilibrium state over a longer period of time.

Somehow I only suspect this explains the numbers you present. Idle temperatures don't much show the expected +/- tolerance range; it's most likely to show under high load. Or -- a wider range would be evident in maximum values over some period of time, since at idle -- the cores vary considerably in usage.

You MIGHT be able to reduce the range of variation under load conditions by lapping the CPU IHS flat (and get rid of the nickel plate on it), with a similar treatment to the heatsink base. If you use a micronized diamond paste or something like Cool-laboratory's Liquid Ultra thermal compound, you might see an overall improvement in cooling across the cores in LOAD conditions by some few C degrees. Between the lapping and a better TIM, I had always concluded that these little increments of improvement were additive.

Those core sensors are less reliable in any sense, though, at IDLE conditions. There was even confirmation for previous-gen Intel chips like my E8600 that one or more core sensors would be "stuck" at an abnormally high level. So at idle, my E8600 might show 51C, and while beginning to load the cores to stress them, I would see no change until the actual temperature had passed 51C.