Help overclocking G3258 on a Asus Maximus VI Impact board

[ganons]

So what are the maximum temperatures for each of the two cores after, you stressed the cpu for a while?

I'll run a stress test with Realbench and post it later on today or did you want to know from Dolphin?

 

[Madpacket]

I don't use voltage offsets when trying to find the limits of a CPU. For your chip simply raise the voltage to 1.1 at 4Ghz and test for stability. I prefer 5-8 hours of P95 Large FFT with the sum int etc checked off. If it fails raise vcore in .025 increments until stable. I have a 50 dollar MSI mini ITX board that easily handles the G3258 at 4.4 GHz with 1.275 volts. Even highly overclocked and overvolted it runs very cool compared to an actual quad core Haswell chips. That's not to say you will reach this (or you may go higher). All depends on silicon lottery.

Good luck.

 

[ganons]

Is there a way to know which cps are good by serial number?

Edit: Did a stress test and highest temp I saw was 70 and 68 on stock. I dont have a dedicated gfx card if that makes any difference

 

[know of fence]

Is there a way to know which cps are good by serial number?

Edit: Did a stress test and highest temp I saw was 70 and 68 on stock. I dont have a dedicated gfx card if that makes any difference

So I tried RealBench stresstest and also set my CPU to 1.081V: my Pentium gets about as hot with it as with prime95, you can certainly run it for a few minutes to quickly test stability. Or once you found good settings run it for longer to make sure your system is really solid.
And yes if you use the iGPU it does contribute to the overall temperature, though it's hard to say exactly how much.

Next thing you should do is go into the BIOS set Vcore to 1.081 V (instead of auto) then raise the multiplier by 1 (to 33x), and test stability for say 10 min.
Write down: Voltage, temperatures of both cores, stress testing duration, frequency, CPU-fan RPM then move on to the next multiplier.
This way you can test how far you can go with just the default voltage.

 

[ganons]

Next thing you should do is go into the BIOS set Vcore to 1.081 V (instead of auto) then raise the multiplier by 1 (to 33x), and test stability for say 10 min.
Write down: Voltage, temperatures of both cores, stress testing duration, frequency, CPU-fan RPM then move on to the next multiplier.
This way you can test how far you can go with just the default voltage.

Where do I get CPU-fan RPM from and should I clock down my ram to 1333 which is the CPU rated speed? Does that help OC?

 

[know of fence]

Where do I get CPU-fan RPM from and should I clock down my ram to 1333 which is the CPU rated speed? Does that help OC?

Leave RAM as is, though you can find out at which speed it is running. Also leave Cache (or Uncore) at auto clock/voltage for now. Generally overclocking any kind of memory you just increase its bandwidth, which only helps in cases where the CPU or GPU is bandwidth limited. If you intend to use the iGPU then certainly faster RAM helps. But RAM is a rather error prone, I never overclocked it. Also you don't have to worry about it as long as you keep BCLK at 100 Mhz.
Bus Clock x 32 multiplier = 3200 MHz

Your board software should give you fan speed.

 

[soccerballtux]

Uhh, much, much better?

Does this mean that you might become a "G3258 troll" too, as you put it?

troll, no, I just like the grass on both sides. I thought about it, and really liked it, but realizing I wouldn't have 8 threads, lost interest

seriously, what's the IPC comparison?

 

[crashtech]

It seems like the IPC of FX is somewhere between Nehalem and Sandy, so Haswell has a pretty undisputed lead in IPC, but by how much depends on the software.

A user posted an X5550 @4.2GHz result in the Cinebench 11.5 thread and the score was nearly identical to an 8350 at the same speed. But I hear Cinebench is not AMD friendly, so grain of salt.

http://forums.anandtech.com/showpost.php?p=37637564&postcount=3

 

[crashtech]

Do you recall, what the default/auto (32x) voltage on that Pentium you got to 4.9 GHz was? Or maybe you have and would like to share a neat table with voltages and temps, just like Ian from AT. Also you've apparently tested the stock heatsink as well as the Coolermaster 212 Evo and didn't notice like a 20 degree C difference?

Test setup:

Gigabyte Z97X-UD5H
GSKILL 2x4GB F3-12800CL9D-8GBXL
be quiet! Straight Power 10 800W
Corsair H110
Samsung 850 EVO 500GB
Windows 10 Build 10525
Intel Burn Test V2
CPU-Z 1.73
Real Temp GT 3.70

VRIN was set to 1.900 for all tests, higher settings were attempted at x50 to no avail.
Above 4.6GHz, LLC was set from "Auto" to "Extreme." Vring needed to be adjusted up manually above x40 to avoid BSODs, this may be an idiosyncrasy of my mobo. Vring was not determined as precisely as Vcore due to lack of time, less voltage may or may not work at multipliers over 47. Observed voltage readings were never more than 2mV from the set value, so I have only included the set values because it looks better.


G3258 Overclocking Results:

Freq _ Vcore _ Vring __ Tcore(C)

32 ___ 1.080 _ Auto ___ 50 (Stock)
32 ___ 0.875 _ Auto ___ 44
33 ___ 0.875 _ Auto ___ 45
34 ___ 0.900 _ Auto ___ 45
35 ___ 0.925 _ Auto ___ 46
36 ___ 0.950 _ Auto ___ 48
37 ___ 0.975 _ Auto ___ 49
38 ___ 1.000 _ Auto ___ 50
39 ___ 1.025 _ Auto ___ 50
40 ___ 1.050 _ 1.050 __ 50
41 ___ 1.075 _ 1.050 __ 51
42 ___ 1.100 _ 1.050 __ 51
43 ___ 1.125 _ 1.050 __ 52
44 ___ 1.175 _ 1.050 __ 55
45 ___ 1.200 _ 1.050 __ 57
46 ___ 1.275 _ 1.075 __ 62
47 ___ 1.300 _ 1.100 __ 65
48 ___ 1.350 _ 1.125 __ 71
49 ___ 1.400 _ 1.200 __ 77
50 ___ 1.525 _ 1.250 __ BSOD

 

[richierich1212]

Thanks for the settings crash, I'll try those on my G3258 before I swap it out.