Intel Core i5 2500K overclockers

[Mud]

I'd like to get an overview of the speeds people run this CPU at.

 

I'm running mine at 4.0GHz at 1.180 core voltage. I've tested at 4.5GHz as well, but it gave me 0 extra frames on the ramp. Quite obviously making my GPU the bottleneck.

 

Anyway, I'll be happy to hear what others get out of the 2500K

 

Edit:

Perhaps a handy reference:

I get 66fps in the Smerch Hunt EN mission (loading the mission and then not pressing any other button apart from Pause and CRTL+Pause)

I normally run with Vsync on and maxfps set to 60, but for this test I disabled them.

Edited July 19, 2011 by Mud

[Pilotasso]

4.5Ghz with 1.24V in manual mode. I think I can do with less but havent tested extensively yet. Besides this is lower than the rated voltage so Im not worried. :)

 

[Bolt-1]

Right there with you Pilotasso. Although voltage is slightly higher but temps are still great. :worthy:

 

[Rhinox]

the i7 2500k/2600K pack with P67 chipsets provide -imo- amazing potential, especially in RAM bandwith.

Could you explain this? AFAIK, 2500k/2600k are almost insensitive to memory speed (and therefore bandwidth). I've tested my rig with memory running at 1066/1333/1600MHz. It has impact only on results of some synthetic tests, but nearly none on LO/DCS. Moreover, RAM bandwidth is influenced most of all by memory controller, which in case of i5/i7 is part of cpu. So P67 chipset has very little to do with it...

[Ghanja]

Running Prime for 5 minutes does not say much about stability at all - run it for a couple of hours and see what it does ... :music_whistling:

[EtherealN]

Most preferably, run it for 12 hours starting in the morning, and make sure you do this on a hot day. You need the burn test to really be the heaviest test your machine is ever going to see, otherwise it's fairly pointless to run the test as a stability confirmation anyhow.

 

A couple minutes is fine if you just want to make a quick check on your cooling and ensure that you don't outright BSOD or any such funky business, but once you are established at the OC you'll be using "in production", you need a really heavy burn test.

[Bolt-1]

@ Ghanja

A couple minutes is fine if you just want to make a quick check on your cooling .

 

And thats exactly what I was doing for the post, as I already know its stable, with a run of prime 95 for 11 hours while I was @ work. The temps for that run were only 3 or 4 deg. higher over all. With no errors at all.

Edited July 20, 2011 by Bolt-1

[Rhinox]

Most preferably, run it for 12 hours starting in the morning, and make sure you do this on a hot day

IMHO, any stress-test longer than 3-4 hours does not make sense. Or better said, it does not necessary show stability as a function of overclocking. There are many more factors, which contribute to the overall stability, i.e. if you do not use ECC-memory modules (which desktop does?) you get up to ~1 memory error per day. I see it on my server log, which shows 3-6 "soft-errors" per week (16GB ECC in 4 modules). Or power-source: if you do not use quite expensive "on-line" UPS, there is a high chance electric voltage fluctuation might bring your computer down after a few hours, despite of filtering done in PSU.

 

So if your system survives 12hour stress-test, ok. If not, it may be caused by many other things, not only because of aggressive overclocking...

[EtherealN]

IMHO, any stress-test longer than 3-4 hours does not make sense.

 

As I indicated in my post, the point is mainly to ensure that your run includes all variables. For example, testing it on a cool morning might show it doing just fine, but a couple days later you're struck by a heat spell and your computer finds itself overheating and faulting on the afternoon.

 

As for your list of errors... Seriously, if your machine does not survive being fully tasked for 12 hours, something is wrong with the equipment. A properly set up computer should never be "brought down" by anything less than a power outage. (Or the user manually turning it off, obviously.) My machines run 24/7 and when I'm not using them they run Folding on CPU and GPU, as long as there's no power outage I can go out on a contract for a week and come back to find it still chugging along fine. (I have actually done exactly this a couple times.) This is out in the hicks, no UPS, with old wires and a somewhat flaky grid that someties pop lightbulbs. The machine is just fine though.

 

An involuntary shutdown is never acceptable and should never be considered part of normal operation.

 

So if your system survives 12hour stress-test, ok. If not, it may be caused by many other things, not only because of aggressive overclocking...

 

What's the point of overclocking a broken system though? ;)

[Rhinox]

...As for your list of errors... Seriously, if your machine does not survive being fully tasked for 12 hours, something is wrong with the equipment.

That is your opinion, and it is incorrect opinion. As I said previously, memory errors are quite common and natural for DDR/SDR chips, even if very rare (I'd say a moderate server with ~dozen of RAM-modules, running 24/7, hits at least a few of them per week). Even your PC has them, you just do not have any way to check it. In 12 hours, I'd say you have ~30% chance to hit one memory error (which does not necessary mean instability, as this might happen in unused or non-critical portion).

 

But because these errors are rare and because of price, cheap "pc-world" sticks mostly with non-ecc. That level of "stability" is enough for PC, so small probability of non-recoverable memory errors is natural. It does not mean there is something wrong with hardware, or the system is broken. It just means you got what you paid for... :-)