Yet another G3258 Overclock Thread

[staykov]

Hello guys,
I heard about G3258 some few months ago and decided to get one so I can refresh and improve my overclocking skills. Haven't been as excited about a CPU's overclocking capabilities in the last 4-5 years to be honest.

So I bought the CPU and got the whole rig. Unfortunately I wasn't able to move past 4.5GHz @ 1.31V. I know about the silicon lottery and all, but since so many people were able to clock it successfully at 4.8GHz, I still wanted to share my experience in case I missed something. I still understand that this particular G3258 sample might not be a good overclocker.

The rig:
CPU: Intel Pentium G3258
CPU Cooler: Corsair H100i
MB: Gigabyte Z87X-UD3H (Last bios v.10b)
RAM: Corsair Vengeance Pro 2400, 2x8 Kit
GPU: XFX Radeon R9 290 Double Dissipation
PSU: Fractal Design Newton R3 1000W Platinum
SSD: Samsung 840 Evo

(I know this rig is an overkill for this CPU, but it was built about the overclock, not to pursue "the best bang for the buck")

Reaching 4.5GHz was as simple as it gets:
*Set vcore at 1.31
*Set LLC at Medium
*Set Multiplier at 45
*XMP Disabled
*F10

The result was good. Rock-solid stability after 9 hours of OCCT Large Data Set torture testing. No errors whatsoever. Did a 9 hour Prime95 blended test as well just for safety - still good. Load temp 65C, Idle temp 23C.
This worked flawlessly with XMP-enabled as well.

However when I tried running a 4.6 everything went nuts. So I decided to read a little more on the webs, regarding the proper ways of overclocking a Haswell-based cpu. I am currently testing the CPU on prime95 with the following settings:
*LLC = Turbo
*vcore = 1.395
*vrin = 1.900
*vRing = Normal, offset +0.100
* I tried playing with Uncore, but frankly I wasn't able to do better than the Auto setting so I returned it to Auto

I disabled all PowerSaving functions:
*TurboBoost = Disabled
*CPU Enhance Halt (C1E) = Disabled
*C3/C6 State Support = Disabled
*CPU Thermal Monitor = Disabled
*CPU EIST Support = Disabled

*Changed the PLL to South Bridge - some clockers claim this would improve stability.

So I'm currently running Prime95 Large FFTs with these settings and it appears to be stable, but it's still early to say as it failed on vcore 1.385, about 30 mins in. I am doing this with XMP disabled.

Initially I thought that this PSU might not have been the best in terms of ripple control and what not, so I switched to Antec HCP-1000 Platinum, but it did not improve the situation so I ruled out PSU performance issues.

Basically that's it. I'd really appreciate any input. I am not a skilled overclocker so I could have easily missed something. For example I didn't change any of the chipset voltage settings. Maybe that's important?
My goal is to reach 4.8. Do you think it's possible?

Thanks in advance.

P.S.
vcore =1.395
vrin=1.950
mutliplier=46
I am 2 hours in Prime95 Large FFTs seems to be stable so far. Will have to pass the 24h test before it's official.
I'd like to go a bit faster though. 4.7 or 4.8 would be ideal.

[Suffokation]

Man, i'd really love to pick up a Pentium bench just to hammer the shit out of them.

Anyways, nice post.

[Sauron]

It's very likely just an unlucky chip. 1.395 is a very high voltage, I wouldn't run it that high even if it does turn out to be stable.

[FellTheSky]

1.395 is just way too high for haswell, the voltage jump from 4.5 to 4.8 also shows the chip is hitting the wall. Also, at 4.5 uses 1.31, and at 4.6 uses 1.39?

 

Have you tried to increase the blck a bit? maybe multiplier x45 and 102/103 blck?

[flibberdipper]

An then there's me. XMP? Yes. Voltage? 1.1. Multiplier? 41. I hit an easy 4.2GHz on mine on the stock cooler, but it doesn't seem like it wants to go any higher at that voltage, and I don't like where the temps go either.

[staykov]

It may be high for a Haswell although opinions seem to differ on the web, regarding this matter.

Even if we agree that it is high, this cpu still cost me only $60 so I wouldn't care if it melted away tomorrow.

However, for the sake of overclocking mastery I'd love to tweak settings in a way that will allow me a lower vcore.

IMO, simply upping the vcore in order to achieve stability is brutish and inelegant. I've seen a 4770k run at 4.9Ghz at 1.25v, 

so I really don't feel a sense of accomplishment, running a G3258 at 4.6 on 1.395v

 

Update:

It wasn't stable at 4.6 anyway. Crashed after 3 hours of Prime95.

Back to 4.5 @ 1.31

I don't think this chip will perform any better.

[runit3]

It may be high for a Haswell although opinions seem to differ on the web, regarding this matter.

Even if we agree that it is high, this cpu still cost me only $60 so I wouldn't care if it melted away tomorrow.

However, for the sake of overclocking mastery I'd love to tweak settings in a way that will allow me a lower vcore.

IMO, simply upping the vcore in order to achieve stability is brutish and inelegant. I've seen a 4770k run at 4.9Ghz at 1.25v, 

so I really don't feel a sense of accomplishment, running a G3258 at 4.6 on 1.395v

 

Update:

It wasn't stable at 4.6 anyway. Crashed after 3 hours of Prime95.

Back to 4.5 @ 1.31

I don't think this chip will perform any better.

1) are you running any offsets?

2) where is cache mult and voltage?

3) enable XMP

4) stop using Prime95. Use XTU or AIDA. I've had systems 12+hr stable in XTU and they've never seen a BSOD despite screen-freezing on Prime95 after 45 seconds. It's more trouble than it's worth for Haswell and doesn't reveal anything meaningful besides max theoretical temperatures under synthetic load (which are meaningless -much like IBT).

5) I'll go pull some BIOS screens of the G3258 OC I'm about to ship out in an hour or two and post them up here if you would like.

[staykov]

Hello,

1) I'm only running the vRing with an offset of +0.100 as noted, everything else is on static or auto

2) If by cache mult and volt you mean uncore, I have only frequency settings for the uncore. I don't know where the voltages are, haven't seen any in the settings.

3) XMP is enabled

4) With all due respect, I don't believe this is a good advice. Using fault-tolerant software to benchmark stability is like trying to nail Jello to a tree. Prime95 can be used to achieve max temperatures when Small FFTs are selected, LargeFFTs however are mostly used for error checking and that's what I am after. I'm not using only Prime95, I'm also using OCCT and Cinebench.

5) Sure. That'd be interesting, although my idea of stability does not match yours, so I don't think I'll be interested in applying the same settings.

[lovingGamer]

G3258 and r9 290 really? (no bottleneck?)

[staykov]

It's a CPU test bench, obviously the GPU doesn't matter too much.

[ProKoN]

Given your results I dont think your gonna hit 4.7-4.8 under 1.5V

 

show the pentium NO MERCY.

 

I have two  of these chips in my possesion right now. One does 4.8GHz @ 1.48V the other only does 4.6GHz@1.45V

 

Luck of the draw. always keep ring ratio at stock when trying to figure out maximum core frequency.

 

dont be gentle on this chip