Intel Haswell i7-4770K IPC and Overclocking Review

Intel's clock keeps ticking and today lands on a "tock" in the development cycle. The new desktop Haswell processor represents a new microarchitecture built on the tried and true 22nm process technology that we have come to know and love with Intel's current Ivy Bridge microarchitecture. But what does Haswell mean for the computer enthusiast?

continued...

Haswell Overclocking Results

We used a couple of different ASUS motherboards for our overclocking over the last couple of weeks. I primarily used the ASUS Z87-A because we are also in the middle of the review process for it and that review will hopefully be completed for publication Monday. I did also have a solid day overclocking the Core i7-4770K on the ASUS Z87-Deluxe. Both of these motherboards returned very similar if not almost identical results so I will just speak to the overall experience. Overclocking results here are not with the motherboard "tuned down" either. All controllers and features are turned on and the motherboard is in a fully functional state. In fact I went out of my way to use many different software programs on the test system while overclocking to see if there were any hiccups.

Overclocking tests were performed with Corsair Vengeance memories; sets rated at 2400MHz and 1866MHz. For the overclocks shown here I used 16GB (4x4GB) of Vengeance at 1866MHz (9-10-9-27-2T). As reflected in the previous benchmarks, I was able to push the full 16GB to 2000MHz at 4.8GHz. A few different video solutions were used for testing; a GTX TITAN, GTX 470, and GTX 7600GT. A Koolance Exos 775 system with 370 series water block (now replaced by the model 380I) was used for cooling. The ambient temperature was kept at 72F.

Our results with overclocking the i7-4770K were nothing short of excellent, but as we went into detail about on the previous page, we do not expect these results to be common.

Article Image

Here we see our i7-4770K running at a very stable 4.5GHz/1866MHz with a "very low" 1.136v Core Voltage. Our per-core temperatures of 63C to 67C, are staying well within manageable parameters. Worth keeping in mind here is that it has been reported to us that some 4770K processors have required 1.2v to maintain stability at 4.3GHz.

Article Image

Above we see the settings used in our 4.8GHz/2000MHz benchmarks with a set 1.26v vCore. This setting was "mostly stable" for this specific 4770K even using a high heat generating synthetic test like Prime95. I say "mostly" because after a couple of hours this setting would finally crash running Prime95. Conversely, you could run Aida64 Stress Test with 100% long term success. While we would see core temperature spikes into the 80C range, our Koolance unit would generally keep this configuration in the upper 70s. Things start to get dicey in terms of stability once you get into the upper-80C ranges.

Article Image

If you really want to get crazy, which we did, we pushed to see if we could get anything stable at the 5GHz mark. What you cannot see above is the system set to 50x100, but you can see the extremely high 1.472v vCore (1.45v set in BIOS). Using Prime95 to stress this setting (we are only using Aida64 in this screenshot to monitor the temperature and load), you can see the insanely high core temperatures and the throttling of the CPU clock. Getting this particular system stable at 5GHz with this load seemed to be luck after dealing with plenty of BSODs at 4.9GHz.

Tests are Tests, Right?

ASUS warned us against using synthetic stress tests like are shown above.

Unvalidated stress tests are not advised ( such as Prime 95 or LinX or OCCT, Intel Burn Test or other comparable applications ). For high grade CPU/IMC and System Bus testing Aida64 is recommended along with general applications usage like PC Mark 7. Aida has an advantage as it is stability test has been designed for the Haswell architecture and test specific functions like AES, AVX and other instruction sets that prime and like synthetics do not touch. As such not only does it load the CPU 100% but will also test other parts of CPU not used under applications like Prime95. Other applications to consider are SiSoft 2013 or Passmark BurnIn. Additionally this generation has a more specialized point of consideration for synthetic stress tests. When using an adaptive vid voltage control will be automatically controlled by the iVR when a complex concurrent AVX load is initialized from Applications like Prime95 or Aida Or LinX more voltage will be supplied than has been defined/requested.

So we did in fact go back and test some with Aida64, which does have a 30 day trial version. (And also as ASUS told us above, we did see elevated vCore voltages as well using Prime95.)

Article Image

Using the Aida64 Stress Test at 5GHz we had good results. We are still seeing temperatures bump into the mid-90c range, but these are quick spikes and we only see a tiny bit of CPU clock throttling. As of writing this I am still running long term testing at 5GHz, and I have seen a bit more throttling as the system gets heat saturated, but honestly I do not truly find value in this Aida64 "stress test." The fact of the matter is that even though Aida64 shows my system to be "stable," this is not true. All you have to do is start Handbrake, and use it to encode a Bluray rip at 1080P and the system will hard-lock. Your mileage may vary, but I am going to stay old school in terms of how we decide if hardware is "stable."