3rd GEN Intel CPU Semi-Stable Testing *Preview* - 56K WARNING

[MSIRawZ]

Hi Guys,

First things first, this is a preview only and does not contain any comparison benchmarks of either Intel' or AMDs latest crop of CPUs.

This is only initial testing to gauge an idea of what the CPU could possibly do. Stability testing done for a short amount of time only - semi-stable.

This is the first Retail E1 Stepping 9 3rd GEN CPU I have tested. This is not an ES E0 Stepping 8. There will be differences between Retail and ES.

Test Setup:

Intel 3rd GEN i5-3XXXK 3.40GHz (Turbo frequency 3800MHz) Processor

MSI Z77A-GD55 Z77 Motherboard

GeiL EVO CORSA 4GB (2x2GB) DDR3 2133MHz 10-11-11-30 1.5v @ stock

Corsair H100 Water-Cooling CPU Cooler @ medium fan speed

Microsoft Windows 7 Premium 64-Bit SP1

Overclocking Test Summary - LinX 0.6.4 (5 passes):

4.20GHz semi-stable - 1.1v - PASS *TEMPS OK*

4.30GHz semi-stable - 1.1v - PASS **TEMPS OK*

4.40GHz semi-stable - 1.13v - PASS *TEMPS OK*

4.50GHz semi-stable - 1.17v - PASS *TEMPS OK*

4.60GHz semi-stable - 1.23v - PASS *TEMPS WARNING*

LinX 0.6.4 Testing Screenshots:

4.20GHz semi-stable - 1.1v - PASS *TEMPS OK*



4.30GHz semi-stable - 1.1v - PASS **TEMPS OK*



4.40GHz semi-stable - 1.13v - PASS *TEMPS OK*



4.50GHz semi-stable - 1.17v - PASS *TEMPS OK*



4.60GHz semi-stable - 1.23v - PASS *TEMPS WARNING*



LinX Analysis:

Quite surprised about the lack of voltage needed for the same Overclocks as Sandy Bridge "K" SKUs. What I wasn't expecting was the temperatures. If you try and put more than 1.3v (on this particular CPU), temps seems to skyrocket. However thinking about it, 22nm, smaller die size and a lot of voltage would seem to suggest high temps. I just wasn't expecting to see that with just 1.25-1.3v.

OCCT 4.2.0 Testing:

As LinX is a killer when it comes down to torchering the CPU under load, many people choose not to use this application while looking for stability. You won't find any other application that pushes a CPU so hard as LinX which is why some people prefer to use Prime95 or OCCT to measure stability as 95% of people will never push their CPUs so hard like LinX does. OCCT for example gives a better "real-world" scenario when it comes to load stability testing.

As I was hitting the top end temperatures with LinX, I couldn't push anymore without causing harm to the CPU. Below is a better real-world experience with load testing & temperatures under OCCT.

Overclocking Test Summary - OCCT 4.2.0 (15 minutes):

4.50GHz semi-stable - 1.22v - PASS *TEMPS OK*

4.60GHz semi-stable - 1.22v - PASS **TEMPS OK*

OCCT 4.2.0 Testing Screenshots:

4.50GHz semi-stable - 1.22v - PASS *TEMPS OK*



4.60GHz semi-stable - 1.22v - PASS*TEMPS OK*



OCCT 4.2.0 Analysis:

A lot better this time around with regards to temps under load. LinX usually adds a good 10-15C over Prime95 and OCCT so this was expected. However after this initial testing, it does appear in these early stages that 4.60GHz (multi wise) is the max stable OC of this particular chip. I did try for 4.7 & 4.8GHz but the chip was found to be unstable within the first 10 minutes of OCCT. Increasing some of the voltage settings within the BIOS did not help stabilize the higher OCs and temps skyrocketed again.

General Short Summary:

In short, these 3rd GEN Intel CPUs are a world away from Sandy Bridge Overclocking. You can't simply chuck in 1.3 - 1.35v and look for early stability indications like with Sandy Bridge. Doing so and you'll find yourself hitting between 90-100C (under LinX). Low voltage seems to be the key here along with a very decent cooling setup. Regarding cooling, I'm going to have to suggest to those guys Overclocking on air to be careful or stick with Sandy Bridge if your CPU is capable of a decent stable 4.7-4.80GHz OC on air cooling.

Those people who have a sub £35 air cooler may need to stick to a low OC or invest in a better cooling method. You'll notice I was using the top end Corsair H100 and if those are the temps I was getting, you will struggle of a regular air cooler. You will need to look at getting a H100 or H80, or a top end air cooling solution like the be quiet Dark Rock Pro or Noctua NH-D14.

For now I can only really suggest the 3rd GEN CPUs to serious Overclockers using high-end cooling methods like Phase, DICE or LN2. Those with decent custom WC loops will also benefit from these CPUs.

I not trying to put you off these 3rd GEN CPUs as I have benchmarked this and it's a nice (albeit small) improvement over Sandy Bridge, but you will need some better cooling for high OCs.

 

[SieB]

Nice one RawZ, very informative.

Temps are a bit of a shocker, I was expecting cooler then SB even when overclocked. To be honest I think I will stick with my 2500k, the only thing tempting me to go IB at the moment is the up to 50% performance boost with Lucid MVP.

 

[FTLN]

Thanks for the info Rawz,,,, On the ball as always ,,,,

Oh and please start shipping to France ASAP ... Prices are ridiclious out here......,...

 

[MSIRawZ]

Nice one RawZ, very informative.

Temps are a bit of a shocker, I was expecting cooler then SB even when overclocked. To be honest I think I will stick with my 2500k, the only thing tempting me to go IB at the moment is the up to 50% performance boost with Lucid MVP.

Lucid MVP is a very nice perk, and with time as the software is fine-tuned and matures with updates & driver releases, 6 months from now it could be the very "in" thing to have for a gamer.

 

[PetrolHead247]

Awesome, cheers Rawz! Very shocked at the temps, even under the cooling power of a H100 the temps are very high.

To be honest I think I will stick with my 2500k, the only thing tempting me to go IB at the moment is the up to 50% performance boost with Lucid MVP.

Same here, i'm sticking with my 2500K too, though i admit Lucid MVP looks awesome and i'd love to have it. Gonna wait until Haswell to make a full upgrade.

 

[DWC]

but its interesting when LNC comes into pic

, but Cheers Rawz!

 

[Vicey]

I think the reason for the higher temperatures at the same volts as Sandy Bridge is probably the reduction in die size which would mean less surface area for the heat to conduct to the integrated heat spreader and then to the heatsink.

That would be my educated guess anyway. Nice post!

 

[DWC]

I think the reason for the higher temperatures at the same volts as Sandy Bridge is probably the reduction in die size which would mean less surface area for the heat to conduct to the integrated heat spreader and then to the heatsink.

That would be my educated guess anyway. Nice post!

but then it would be kinda hard for IB with 6 cores

 

[Vicey]

but then it would be kinda hard for IB with 6 cores

If you've seen die shots of Sandy Bridge-E they stick the 8 cores on the edge of the die with a shared L3 cache in the centre. I imagine that Ivy Bridge-E will be a similar situation. Sticking the cores a little far apart and sticking a huge cache in the centre should help with thermal transfer.

In any eventuation I'm sure Intel have thought about this as they've gone through quite a few manufacturing shrinks over the years. There is also the lower TDP combined with higher efficiency which could mitigate any loss in MHz achievable through overclocking.

 

[Firat]

Does lucid virtu when connected to On board Gpu, decrease the wattage a Gpu would use in idle? So at the moment at idle I pull 170watts would my idle wattage usage decrease? As my Gpu is my main watt user

 

[nepas]

errrrrrr,OLD version of realtemp? 3.70 only has Sb-E,may need to wait for the next?

But seeing as Ivy is a revision of Sandy,would it make much difference?

 

[Master&Puppet]

Interesting! Glad I run under water

 

[SieB]

Does lucid virtu when connected to On board Gpu, decrease the wattage a Gpu would use in idle? So at the moment at idle I pull 170watts would my idle wattage usage decrease? As my Gpu is my main watt user

As far as I know there is a power saving mode that will save power on the GPU by using the IGP. Not sure how much it saves but I think it only works for non intensive things like web browsing and such, not 100% sure on it.

This is what I am interested in from Lucid though

http://www.tweaktown...dge/index1.html

 

[MSIRawZ]

Posted by another Overclocker - same stepping (E1). This time, looks to be the i7-3770K instead. Unsure what stability tests we're done - perhaps only using Cinebench to measure performance and temps.

 

[PCTwin]

Nice preview Tom

Warm chips

 

[nepas]

Nice preview Tom

Warm chips

errrrrr,wut?this is nothing to do with Tom.

Thanks RawZ,that looks to support your benchies,BUT what was used to check temps?is there an offset problem or are these things mini nukes?I remember when realtemp used to report 10-15 over everything else with my old Q66,so unless you guys are using a new *beta* of realtemp or will show a different program I am gonna laugh(that is a perfect way to shut me up and use a core temp program that is known to work with IB)

These figures do not make sense....

Ivy is either hotter than the surface of the sun or the temp programs need a big update(thinks back to the rumour with Intel saying the next gen WILL need water as standard)

 

[ObscureParadox]

errrrrr,wut?this is nothing to do with Tom.

Thanks RawZ,that looks to support your benchies,BUT what was used to check temps?is there an offset problem or are these things mini nukes?I remember when realtemp used to report 10-15 over everything else with my old Q66,so unless you guys are using a new *beta* of realtemp or will show a different program I am gonna laugh(that is a perfect way to shut me up and use a core temp program that is known to work with IB)

These figures do not make sense....

Ivy is either hotter than the surface of the sun or the temp programs need a big update(thinks back to the rumour with Intel saying the next gen WILL need water as standard)

[font=Verdana, sans-serif, FreeSans]errrrrr,wut? Rawz real name is Tom

[/font]

 

[yassarikhan786]

Thanks for the info RawZ, very nicely done.

 

[killablade]

Nice results fella, guess the Phanteks TC14PE I'm going to get will actually come in handy!

As for Lucid VirtuMVP, it seems that at high resolutions (above FullHD) the performance actually decreases. Don't remember where I read the article but with a little help from google you should be able to find it should you wish to read it.

 

[nepas]

[font=Verdana, sans-serif, FreeSans]errrrrr,wut? Rawz real name is Tom

[/font]

Learn something new everyday!!