2500+ OC - Ongoing Experience

[keto]

Rig in sig. It's early days, I just picked up the rig last night. Still figuring out the bios & tweaking. Still on stock cooler, as I was going to pick up an H100 but they are on recall...I'll either wait it out or pick up a different solution.

So far....I've only ramped it up to 4.0, due to lack of time/just starting out, + stock cooler. Still have the RAM at 1333, have not enabled XMP yet. Just ran 3DM11 as a basis of comparison with my other system, i7 920 (C0, not a great overclocker) @ 3.60, vid card at stock out of the box clocks. I know 3DM is more a measure of vid card performance, but I honestly thought a 2500+ at 4.0 would put a significant beating on the 920 @ 3.6, plus given I am running RAID0 on faster hardrives and SATA3, vs the 920 system which has SATA2 and 1 older HD. Not true...the 920 @ 3.6 gives P6459, the 2500+ stock gives P6281, the 2500+ @ 4.0 gives P6495. I haven't yet had time to run other benchmarks which may or may not show the 2500+ in better light.

Lots of tweaking left to explore, so not a complaint just an observation. There's lots of clock left in the 2500+ at this point, I'm still stock voltage and even with the stock cooler there's a bit of headroom as load temps are just under 70. More to follow.

[keto]

As per my post in watercooling, have an H80 on it now.

4.5 and 4.7 were easy as pie at relatively low voltages. 4.8 is proving more challenging, I can surf all day at 1.395 vcore + turbo enabled, LLC level 6, but it fails P95. Still learning about a few of the bios/power management settings.....I'm reasonably confident this is a 5.0-5.1 cpu, depending on how hot I want to run it or how much fan noise I want to put up with. Highest I've seen, with my vcore 1.425 and a full load on is a momentary spike to 68C, 66C is the spike temp @ 1.395.

My main concern is that, with LLC at 6, the idle vcore is MUCH higher than the setting...ie., if set at 1.395 it'll idle along at 1.440 with the odd momentary spike at 1.476....now, this is at 30-32C and with the cpu clocked down (C1 state?) reading 1595MHz, so I'm not panicked about it, but LLC 6 seems to be required for stability at this point. Load vcore reports 1.428 fluctuating with 1.416....but if I go to LLC5, the fluctuations go below the set vcore, and it fails P95 much sooner. I had it stable at 1.410 vcore LLC 6, but the idle spike at 1.488 motivated me to see if it would pass at a lower vcore.

Still tweaking. Haven't touched the RAM yet, it's on 'auto' at 1333 1.5v though it is rated 1866 1.65v.

[keto]

4.8 is verified stable with a 13 hour P95 blend run. 1.428vcore. Gigabyte bios is not fully implemented with respect to LLC and offset, both cannot be run at the same time. I am using LLC level 5. On the downside, this means my IDLE voltage is running at 1.476...this is not really a concern to me, as my idle temps are 30C and the chip throttles back to half of stock speed. The board is very stable, I have all power saving features enabled (C1, C3/C6, EIST, etc.).

The board is super simple to get to 4.5....I can leave everything, including vcore, on auto

My methodology to get the overclock, after much reading, was to leave the cpu multiplier at stock and put my desired 48 multiplier in each core under the turbo setting. I have manually set the RAM multi to 18.6X, again Gigabyte has a strange implementation and if you use XMP profile, it sets the cpu back to stock speed. I haven't played with my RAM timings yet, and have it set to the lowest value in bios of 3 (standard, quick, expert) and turbo (not to be confused with cpu turbo) disabled. My next step is to tighten up the settings and see how/if it affects stability.

I haven't done a ton of benchmarking yet, but here is my 3DM 11 progression (GTX580 @ stock, slightly factory overclocked speed) :
i7 920 @ 3.6 = P6459
2500K @ stock, memory @ 1333 = P6281
2500K @ 4.0, memory @ 1333 = P6495
2500K @ 4.5, memory @ 1333 = P6597
2500K @ 4.8, memory @ 1866 = P6641

As you can see, in this particular benchmark the 2500K isn't much of an upgrade from the i7 920...but there are other benchmarks where the new generation clearly outstrips the old and the hidden benefit is the power consumption, 95W on the 2500K vs 130W for the older chip - 27% improvement. Probably only a couple bucks a month on my power bill, but I'll take it.

The H80 is working out great. I leave it on the highest of the 3 settings at all times. Idle temps right around 30C, running P95 overnight it never saw 70C and averaged 65C, with my case fans on lowest setting. Ambient is constantly 21-22C.

[bradleyg5]

Why you think the 3rd core gets so much hotter? Mine does that too.

[francisw19]

Originally Posted by bradleyg5

Why you think the 3rd core gets so much hotter? Mine does that too.

Yeah mine is like that too. Maybe it's closer to another warm component (IMC, for example)? Just a guess...

[keto]

I have now disabled C3/C6.
I tried memory on 'Turbo', got a pretty instant BSOD.
I am messing around with LLC settings, trying to get a lower idle reading. If you will recall from above, I was
vcore set @ 1.435, LLC 5, bios reports vcore 1.488, CPUz 1.58,5 reports idle 1.476, load 1.428

I have changed to
vcore set @ 1.355, LLC 8, bios reports THE SAME vcore 1.488 (which I find very strange), CPUz reports idle 1.404/1.416, load 1.428. There are some strange things going on with voltage and/or voltage reporting. Some users have reported that CPUz is not reporting correct SB voltages, so I installed Gigabyte's EasyTune 6 just to look at its HW monitor. It shows the same vcore at the same time as CPUz.