Intel Ivy Bridge i5-3570K & i7-3770K Review

CPU by leeghoofd @ 2012-04-24

Time seems to fly. Just over one year ago Intel introduced Sandy Bridge to the world. Packing a high performance CPU, with mega overclockability for the K skews, yet keeping power consumption and heat to a bare minimum. The best part was that Sandy Bridge was affordable and even a big threat to Intel's flagship, the socket 1366. World records were smashed at HWBot, as this little socket 1155 CPU overclocked to 5.8Ghz and beyond. The instructions per clock were way faster then anything we were accustomed too. End of last year, it was time to replace the aging socket 1366 by Sandy Bridge-E. The socket 2011 has big potential with it's quad channel RAM support and multi GPU excellence via 40 PCI-E lanes. But the overall clock speeds of the SB-E were disappointing. Most CPU's don't even go over 5500mhz on cold. Today we are gonna introduce the "Die" shrink of the little affordable Sandy Bridge CPU's : Codename Ivy Bridge

  • prev
  • next

Overclocking

There has been a lot of talk, due to leaked numbers, on various overclocking forums. Big discussion on how the new 22nm technology would overclock. Wild speculations on maximum MHz, power consumption, heat,... Well the numbers are out and it has been proven that for those that put great value on 5GHz E-Peen : Ivy bridge will not cut it on air. 4600-4800MHz is easily doable with our ES models. But that's where it ends on air or water. We were spoiled with Sandy Bridge with some silicon doing daily 5000MHz on air. Benching on air or water at +5600MHz was no issue with the SB generation.

Ivy Bridge overclocking on air and water is limited by heat that can't be properly transferred to the IHS. So even if you opt to upgrade your watercooling loop with extra fans, bigger rads, it will not help. As the conduction of heat is inferior due to two values: the smaller die size means less contact surface and apparently the thermal paste used between the Die and the IHS isn't that good as a soldered version. But the latter is pure speculation as I never removed the IHS myself. We conducted a quick test by disabling the fan on our air cooler and the heatsink didn't get hot at all. The coretemps however went up to 85°C at stock clocks.

As you can see in the chart below the CPU multiplier has been upped to 63X. Plus the new high RAM dividers will allow the overclockers to smash old world records. Life is again beautifull !

 

 

When we prime tested both the CPU's with the custom 12-12K set we directly spottted very high temp readouts in the Coretemp program. We settled with 4800Mhz for the i5-3570K and 4700Mhz for the i7-3770K. Not far from the magical 5GHz number everybody is aiming for. But as mentioned before if the goal is to reach the magic number 5.  Do yourself then a favour and buy a SB. Burn candles and pray to the lords of overclocking that the SB can do them clocks. You will need them to beat the scores we obtained from our IB CPU's.

If we briefly compare the quadcores without Hyperthreading : at stock clocks we hardly see a difference in consumption and temperature (take note that the i5-3570K runs at higher clocks and our GB board runs all cores at full Turbo mode when stressed ) At 4500mhz we already spot at 12°C difference between both CPU's in average core temperature. Nevertheless the IB 3570K draws less power. At 4800mhz we reached our own safe limit of 85°C when priming. In the graphs below are the results of the AIDA64 stress tests. Looking at the temps and voltages our CPU reached it's sweetspot at 4700mhz at 1.27Vcore. More acceptable temperatures and power draw in comparison to the extra heat generated for them last 100mhz.

 


Our Hyper-threaded versions tell a similar tale. Our i7-2700K is able to run daily at 5GHz at 1.49Vcore and 5600Mhz for benching purposes. Initially our i7-3770K didn't seem to be a stellar sample as it topped at 4700Mhz prime stable. But looking at other results obtained it ain't that bad :p

 

 

While priming our i7-3770K at 4700Mhz we noted core temps of 90°C. By downclocking 100MHz to 4600Mhz we were able to drop the temperatures by 15°C and were also able to reduce the voltage with 0.1Vcore. It doesn't take a rocket scientist to find the sweetspot. Common sense and logic will get you there. Now 4600Mhz is a bit further from 5000Mhz. However remember that clock per clock the Ivies got a decent advantage, that allows them to make up for the lost 200-300MHz.

Conclusion for air and watercooled IB PC's : look for the sweetspot and maybe it's not a bad idea to keep the voltages around 1.3ish Vcore max.

But that's for 24/7 setups. Overclockers all around the globe are welcoming Ivy Bridge. Finally we have a CPU that scales with voltages and more importantly with cold. Liquid Nitrogen galore again. No more will your scores be beaten by air or watercooled guys as they had access to that single silicone gem. Most IB's scale easily up to 6GHz and beyond. Time to get that Dewar filled lads. The Cascade guys can reach up to 5.8Ghz too, but to get more MHz you will have to cool down these babies. We had some fun with our ES samples at the Intel Resellers Conference. Our ES i5-3570K was a fun CPU topping at 6400Mhz with 2800MHz RAM speed. Sadly the i7-3770K had IMC problems once we went colder then -120°C. So it topped at 6ghz with only 2600mhz RAM speed. Are we seeing a SB-E flashback, where most K skews were better than the eXtreme version ?

Some snippies from the Conference demo days :

 

 (Click to enlarge )

  • prev
  • next
Comment from Teemto @ 2012/04/27
How about the issues with the heat spreader?
Maybe better to wait on buying one till Intel addresses this issue and switches back to the paste used in SB. If they plan to do that at all?
Other option would be to tear off the heat spreader but I'm not going to risk that on a 300€ CPU
Comment from leeghoofd @ 2012/04/27
Maybe Intel will correct it but I doub t it as for 24/7 there's no problem,
These CPU's are screamingly fast for daily usage and gaming. For benchers LN2 is the only way to go. Removing IHS for LN2 benching is too risky in my book...
Comment from Stefan Mileschin @ 2012/04/27
I am waiting to see some retail CPUs benched on air with the HSF removed
Comment from nigel @ 2012/05/01
except for the litle issue with the ihs and such these look just great.

I just hope to see more results with retail samples and modified ihs. Like lapping, no ihs, remounting ihs...

But nonetheless this is my next upgrade


Also nice write up once again

 

reply