2600K @ 5GHZ, How common is it really???

[Diogenes2]

Hardware monitor shows 1.46 - 1.48 ..

 

[s1njin]

You can call a 4.4GHz 2500K @ 1.25V at "stock voltage" because thats what the chip requests at 3.4GHz turbo, but you almost definitely have to use fixed volts or offsets which is clearly not the factory default mode.

I think the confusion arises because some people OCed their SBs at auto voltages and assume that their OCed chips use the same voltage at load as a non-OCed one.

Well, if the chip requests 1.25v at stock voltage when in turbo, than setting mine to run 1.275 full time is only 2% higher than what it would pull running turbo.

Is it safe to assume that the chip is primarily in turbo mode when playing games?

Since that's pretty much all my PC does (its either gaming or off), running my chip w/ 2% more volts to get to a 4.5 OC is a no-brainer to me.

 

[BTRY B 529th FA BN]

Intel knows. The max VID for the SKU is not specified willy-nilly.

There is a physics-based reason, bounded by an economic one (minimizing infield fails that are covered by warranty), which is data-driven and empirically derived for the product.

The only voodoo involved is that there is a whole heap of people out there who know jack squat about device physics and lifetime reliability analysis, and the void of ignorance gets filled in with a whole lot of guessing compounded by arrogance.

The middle-ground comes in when people want to knowingly reduce their CPU's lifetime but only to the extent that the odds of it dieing within 2yrs is increased only a little tiny bit. People will say "I don't care if my CPU lasts 10yrs, I only need it to last me 2yrs...so how much more cowbell Vcc boost can I give it?".

That's when it helps to have some education and experience (industry experience, not enthusiast zuber-overclocker garage experience)

LOL

 

[podspi]

Hardware monitor shows 1.46 - 1.48 ..

That's pretty awesome, congrats

 

[edplayer]

I've hung onto my e8600 @ 4.5 Ghz. for years, and was toying w/ the idea of grabbing a 2600K setup,
but I would only do so if it can go 5GHz + on air.
I've read maybe only 2-3 out of 100 chips will do this.

and who said that? How many cpus did they test?

Kyle posted that ~2 out of 100 can do 5GHz at stock voltage (as he was told by one from within Intel). Don't recall if it was 2600K, 2500K or both.

 

[Rvenger]

Thats about the volts I was pulling @5ghz on my 2500k, the temps were no more then 80c at max load for over an hour which is fine by my own standards but thats just too much voltage to run daily all year then I'm comfortable with. I would run this at 4.8 if I wanted a max OC daily but I'm using this thing as a server so 4.5ghz is fine for me.

Surutcra - Are you throttling your Vcore at the speed you are at now? When you are running idle does your clock speeds stay the same along with the Vcore?

 

[pantsaregood]

If a SB CPU hit 5.0GHz on stock voltage, that would be very impressive. That's ~50% OC on a 2500K/2600K. i7 920s were known for their common 50% OCs, but they generally required a pretty decent voltage bump and ran fairly hot.

I don't see it happening. Even at 1/50, a 50% boost in clock speed seems absurd for stock voltage.

 

[john3850]

Read a few posts here and there about how fast i7-990x degrade when benched at 5.5ghz
even with high end water do to high vcore.
How come the 2600ks seem to hold up so much better at 1.5+ vcore.
I know there two different chips types but there both made on the 32nm.
I wouldn't run 1.5v on a 32nm daily but I would run a few 3dmarks at 1.5v on water.

 

[PlasmaBomb]

Intel knows. The max VID for the SKU is not specified willy-nilly.

There is a physics-based reason, bounded by an economic one (minimizing infield fails that are covered by warranty), which is data-driven and empirically derived for the product.

The only voodoo involved is that there is a whole heap of people out there who know jack squat about device physics and lifetime reliability analysis, and the void of ignorance gets filled in with a whole lot of guessing compounded by arrogance.

The middle-ground comes in when people want to knowingly reduce their CPU's lifetime but only to the extent that the odds of it dieing within 2yrs is increased only a little tiny bit. People will say "I don't care if my CPU lasts 10yrs, I only need it to last me 2yrs...so how much more cowbell Vcc boost can I give it?".

That's when it helps to have some education and experience (industry experience, not enthusiast zuber-overclocker garage experience)

Just a quick question(s) IDC - presumably Intel know the "safe 24/7" voltages based on stock clocks - and increasing the clocks without adjusting Vcc will still increase "wear".

Also with so many power saving features, do Intel predict using "realistic usage" or "24/7 flat out usage"?

Or are CPUs tested using an accelerated wear methodology, and then the limits extrapolated?

cowbell :thumbsup:

Hopefully this makes sense... it's late...

LOL

I presume that the LOL is in reference to cowbell...

 

[tweakboy]

e8600 @ 4.5 Ghz. Man stick with that until 2012 at least I dont think you will see a HUGE difference. Wait for Ivybridge.

Nice, ya you know whats up. thx for message.

Or we wait for 11nm in 2013 then pull the trigger.. hehe

 

[at80eighty]

I presume that the LOL is in reference to cowbell...

think that whole post was gold. truth and funny as hell

 

[T-man]

So anyone else @ 5Ghz on air, 24/7?
what volts, motherboard, etc....

 

[richierich1212]

Not 24/7 because I don't feel the extra speed is worth the jump in voltage. Only using a Corsair A70 with Artic Silver MX-3. 5.1GHz with 1.48vcore (4.5GHz only needs 1.272vcore):

 

[ThankMeLater]

So anyone else @ 5Ghz on air, 24/7?
what volts, motherboard, etc....

Hi there. I work for a financial services organisation, and part of my job is making things go fast. As such, I've built ~30 SB overclocked servers running at around 5Ghz on the following platform: -

1. SB 2600k
2. Asus Maximus Extreme IV
3. Crazy custom air cooling.

What I've noticed is this based on this platform:-

1. All the chips will run to at least 4.0-4.1Ghz with stock volts.
2. All of them have made 4.8Ghz, the only variation is the volts used to get there. I have consistently used around 1.350vcore to get to 4.8 - with the exception of a few very good chips, I've found that most of them hit a wall at this speed and it requires larger jumps in voltages (and associated cooling) to get them to even 4.9Ghz. The fastest I've gotten so far has been 5.0Ghz at 1.355vcore.

Let me qualify what I'm doing though.

1. We don't use SpeedStep.
2. We don't use HyperThreading.
3. Temps at 4.8Ghz are around high 50's, low 60's - at 5.0Ghz they're around mid 70's.
4. The CPU's run at 100% utilisation on 3 of the 4 cores for around 10 hours a day, Monday to Friday, and this has been the case for ~2 months. We haven't had a machine crash even once.

I'm intrigued to see what people's settings are - I'm of the opinion that the voltages I've been using to bump past 4.8Ghz just aren't worth it - I'll happily run anywhere up to absolute maximum 1.38vcore, but am happier around 1.35vcore - on the mobo I'm consistently able to get 4.8Ghz. The stories I'm seeing of people pushing 1.42vcore or whatever to get to 5.0Ghz seem to me to be ridiculously high - or has anyone tested 1.40 consistently 24/7 with no problems?

 

[Morg.]

Every single 2500k or 2600k chip can do 5Ghz, very few of them need 'dangerous' voltage to do it, and very few of them can do this on air while remaining under 70°C.

Last 2500k stable OC I had in my hands was 4.9 @ 1.3911 vCore or something like that .. temperature wall more than anything (air mugen2).

As TML said, past a certain point it's more about records than actual useful performance.

Either way, if you want 5Ghz and can't bin CPU's, you'll have to resort to WC if you want that 5 Ghz.

The safe voltage for an i5-2500k is and will remain unknown .. but by looking at the fact that most conservative folks think it's 1.38v, I would guess 1.43v has to be more or less it

 

[ThankMeLater]

Last 2500k stable OC I had in my hands was 4.9 @ 1.3911 vCore or something like that

...still seems like a lot of volts, but maybe it's because my own experience has exclusively been on the Asus Maximus IV Extreme - I average about 1.355vCore for 4.9Ghz... this is with BIOS rev. 1303.

 

[KIAman]

I also went from E8500 @ 4.5ghz to 2600k and the difference was huge from a user experience perspective. I mostly game and develop software and do occasional graphics and in those tasks, the 2600k even at stock is significantly faster.

My 2600k can do 5Ghz but I don't need all that power so I scaled it back a little with a significant drop in voltage required (and temperature).

 

[yottabit]

Hi there. I work for a financial services organisation, and part of my job is making things go fast. As such, I've built ~30 SB overclocked servers running at around 5Ghz on the following platform: -

1. SB 2600k
2. Asus Maximus Extreme IV
3. Crazy custom air cooling.

What I've noticed is this based on this platform:-

1. All the chips will run to at least 4.0-4.1Ghz with stock volts.
2. All of them have made 4.8Ghz, the only variation is the volts used to get there. I have consistently used around 1.350vcore to get to 4.8 - with the exception of a few very good chips, I've found that most of them hit a wall at this speed and it requires larger jumps in voltages (and associated cooling) to get them to even 4.9Ghz. The fastest I've gotten so far has been 5.0Ghz at 1.355vcore.

Let me qualify what I'm doing though.

1. We don't use SpeedStep.
2. We don't use HyperThreading.
3. Temps at 4.8Ghz are around high 50's, low 60's - at 5.0Ghz they're around mid 70's.
4. The CPU's run at 100% utilisation on 3 of the 4 cores for around 10 hours a day, Monday to Friday, and this has been the case for ~2 months. We haven't had a machine crash even once.

I'm intrigued to see what people's settings are - I'm of the opinion that the voltages I've been using to bump past 4.8Ghz just aren't worth it - I'll happily run anywhere up to absolute maximum 1.38vcore, but am happier around 1.35vcore - on the mobo I'm consistently able to get 4.8Ghz. The stories I'm seeing of people pushing 1.42vcore or whatever to get to 5.0Ghz seem to me to be ridiculously high - or has anyone tested 1.40 consistently 24/7 with no problems?

They really let you overclock the servers? Are you the head IT there? I guess I don't understand what kind of application it is... I'm awfully surprised you're even using non-ECC memory, let alone overclocking. Why not use Xeon parts? And do they not care about power consumption? How do you test stability?

I don't mean to sound like a jerk, I'm just very surprised to see this being used in a server environment and I hope you understand why I'd be skeptical. I feel like there's a lot to be learned here

 

[ThankMeLater]

They really let you overclock the servers? Are you the head IT there? I guess I don't understand what kind of application it is... I'm awfully surprised you're even using non-ECC memory, let alone overclocking. Why not use Xeon parts? And do they not care about power consumption? How do you test stability?

I don't mean to sound like a jerk, I'm just very surprised to see this being used in a server environment and I hope you understand why I'd be skeptical. I feel like there's a lot to be learned here

Yeah, we overclock the servers, for a very specific purpose. We're totally aware of the fact we're not using ECC RAM, along with the inherent limitations of not going for something like multi-cpu Xeons with QPI and a bunch of cores - basically we have an application that's single threaded and has to run *super* fast - so fast that multi-threading isn't even an option. It's not about throughput, it's all about latency. Power consumption is immaterial, as is whether the chips burn out in 12 months - by then we'll have moved on to the next Intel processor tree.

 

[yottabit]

Yeah, we overclock the servers, for a very specific purpose. We're totally aware of the fact we're not using ECC RAM, along with the inherent limitations of not going for something like multi-cpu Xeons with QPI and a bunch of cores - basically we have an application that's single threaded and has to run *super* fast - so fast that multi-threading isn't even an option. It's not about throughput, it's all about latency. Power consumption is immaterial, as is whether the chips burn out in 12 months - by then we'll have moved on to the next Intel processor tree.

It sounds like you have an incredible job! Haha. So the 2600k is just for the extra 2mb cache over the 2500k then?

 

[Anubis]

i cant comment on the 5GHz question but the move from a 8600 -> I5 2500 will be night and day

i went from 8400 running at 4.4 to a I7 920 running 4.0 over a year ago and it was night and day, the I5s are faster then the I7s

 

[ThankMeLater]

It sounds like you have an incredible job! Haha. So the 2600k is just for the extra 2mb cache over the 2500k then?

Yep. We try and keep our application in chip memory as much as possible - we don't even try and use system RAM.

 

[DrBoss]

I run my 2600K @ 4.4Ghz - 24/7

Additional Turbo Voltage = +0.02
Offset = -0.05

Load temperatures top out around 62C
Build thread for my rig

I did run the computer for a while at 5.0Ghz, but temps got up to 70C... "homey don't play dat".

LOAD


IDLE

 

[supra2jzgte]

Sorry no screen shots, I have had my 2600-K for a few months now, and to be honest, its really easy to push 4.4GHz or over with very little changes in the BIOS and it will be stable.

One thing I recommend, liquid cooling

 

[Idontcare]

They really let you overclock the servers? Are you the head IT there? I guess I don't understand what kind of application it is... I'm awfully surprised you're even using non-ECC memory, let alone overclocking. Why not use Xeon parts? And do they not care about power consumption? How do you test stability?

I don't mean to sound like a jerk, I'm just very surprised to see this being used in a server environment and I hope you understand why I'd be skeptical. I feel like there's a lot to be learned here

Keyphrase by the person was:

financial services organisation

They are trading, time to transaction is key. HFT (high-frequency trading), etc.

Also don't be so surprised to find out that people in professional organizations intentionally take their computers, and their computations, into harms way.

Even Via (Cyrix/Centaur) uses overclocked server farm for their work.

The systems that were in cases were water cooled Core i7s, overclocked to 5GHz. There are two folks at Centaur who build each and every one of these machines, and overclock them. You and I know that overclocking both Nehalem and Sandy Bridge results in much better performance for the same dollar amount, but this is the first time I've seen overclocking used to speed up the simulation of microprocessors.

Source

If it is not a matter of life and death, and merely a reduction to a matter of opportunity cost, then the elevating your risk of losing some money owing to errant computation results (no ECC, OC'ed cpu, etc) is a simple cost-benefits analysis and you will find that in certain situations you come out ahead on average by taking the gamble.