Intel Haswell Core i7-4790K vs. i7-4770K Comparison

[MSnyder]

Intel released the Intel Core i7-4790K, which created quite the stir with Intel Core i7-4770K owners. Does the Intel Core i7-4790K outperform the Intel Core i7-4770K in every way or does it fall a little flat? If you already own a 4770K, is the upgrade worth it?

Show full review

 

[brandonwh64]

So its like 1%? benches look to be not much of a fraction off a 4770

 

[revin]

Thanks for the review. Do any of the adjustments involve the Core Current Limit ? As in how many A allowed to use?
That was the key for my 2600K to reach 5-5.2 on air. All voltage were teawked to no avail, but after using XTU was able to allow the CCL from 97 up to 110 and bingo, all other voltages are stock, except the ram is slightly under volted even at 19xx.

 

[LAN_deRf_HA]

The fact that the temps aren't any better just highlights what others have found, that the problem was never the quality of the paste but the fact that it is paste and paste does poorly with gaps, like the one created by the epoxy layer. That's why you'll see people with no temp improvement after delidding but then after they go back and scrape off more epoxy they finally get a good temp drop. I suspect solder does much better with a gap and that's why the performance is so different.

Which all begs the question why did Intel bother to spend money on new paste to begin with? Did they really just take our word for it that paste was the problem? It would make sense since the importance of the gap only became common knowledge after they had finalized the design, but surely their engineers could have figured that out on their own, right?? I mean if they're going to spend money it should be on solder or some new IHS adhesion method that leaves no gap, not some meaningless new TIM just for the sake of flaunting it in a press release.

 

[Prima.Vera]

You should have added a 3770K there just for the sake of it...

 

[FreedomEclipse]

you should of added a 3930k just for the fun of it

 

[Sasqui]

Other than idle power consumption... meh. Waiting for Broadwell.

 

[Hilux SSRG]

Skip Broadwell and go straight to Skylake.

 

[MSnyder]

Thank you all for the incredible feedback! I think you will all find the newest generation Haswell-E review very interesting. In particular I actually wondered how many people would seriously consider going i7-4790K roughly 4 months before LGA2011-3. Probably not many unless they were in a rush. The pace of the CPUs seems to be advancing faster than before.

 

[Ed_1]

I don't see why it was faster than the 4770@4.8 other than the fact bus was OC to 102 .

Hmm, why was superPi slower with OC CPU's .

 

[Sasqui]

Hmm, why was superPi slower with OC CPU's .

Look again, lower is better.

 

[yogurt_21]

The newer Intel Core i7-4790K features the exact same die layout as the Intel Core i7-4770K. The processor is also built on a 22 nm process, but with a 88 W maximum TDP. Given we previously covered the specifications, there is no need to state the same information over and over again. What Intel does tout, however, is the latest 4.0 GHz base frequency. While it is nice, there is only a difference of 100 MHz overall.

Compared to the base frequency of the 4770k it is a 500MHZ boost, not 100MHZ. The same applies to their turbo speeds.

 

[crmaris]

Welcome on board and many thanks for your effort and the review!

 

[MikeMurphy]

I think your thermals graph is incorrect if you're saying the 4790k has better thermals.

 

[MSnyder]

^ The 4790K had better thermals but was outstepped in that area by the 4770K OCed. As people are mentioning, the actual space between the heatspreader and the chip inside is possibly the reason.

 

[Ed_1]

Look again, lower is better.

hehe, you look again , the OC have more time in SuperPI .

http://www.techpowerup.com/reviews/Intel/Core_i7_4790K/4.html

 

[MSnyder]

Compared to the base frequency of the 4770k it is a 500MHZ boost, not 100MHZ. The same applies to their turbo speeds.

Thanks for pointing that out. It's a typo and will be fixed.

 

[MSnyder]

hehe, you look again , the OC have more time in SuperPI .

http://www.techpowerup.com/reviews/Intel/Core_i7_4790K/4.html

Ideally overclocking them should have lead to lower SuperPi numbers, but it seems latency issues stepped up as well. Could be particular CPUs themselves but the test averaged out 3 times to give those numbers. In any case the other tests are more revealing than the SuperPi

A lot of people have a hard time getting the 4790K to 4.7GHz - I got both to 4.8GHz. Can be possible that latency stepping is an issue when it comes to SuperPi array

 

[Sasqui]

hehe, you look again , the OC have more time in SuperPI .

http://www.techpowerup.com/reviews/Intel/Core_i7_4790K/4.html

Yep, you are correct. That is a big wow.

Ideally overclocking them should have lead to lower SuperPi numbers, but it seems latency issues stepped up as well. Could be particular CPUs themselves but the test averaged out 3 times to give those numbers. In any case the other tests are more revealing than the SuperPi

A lot of people have a hard time getting the 4790K to 4.7GHz - I got both to 4.8GHz. Can be possible that latency stepping is an issue when it comes to SuperPi array

Memory timings changed with the OC?

 

[MSnyder]

Memory was kept the same timing as the default RAM but it seems that turning the overclock up on those two processors leads to a stability issue - so timings were in fact loosened to get it both to 4.8GHz - yes - which would explain the 3/5 second differences compared to their stock clocks which SuperPi seems to be sensitive towards (Stock was tested at the loosened timings and showed better response). I did contact wPrime systems as well and they pretty much confirmed that if the BCLK requires memory timings to change for stability, then this can impact SuperPi results depending on the severity of the overclock and specific CPU. They also mentioned that overclocking does not always necessarily lead to lower SuperPi timing computation.

 

[Ed_1]

Memory was kept the same timing as the default RAM but it seems that turning the overclock up on those two processors leads to a stability issue - so timings were in fact loosened to get it both to 4.8GHz - yes - which would explain the 3/5 second differences compared to their stock clocks which SuperPi seems to be sensitive towards (Stock was tested at the loosened timings and showed better response). I did contact wPrime systems as well and they pretty much confirmed that if the BCLK requires memory timings to change for stability, then this can impact SuperPi results depending on the severity of the overclock and specific CPU. They also mentioned that overclocking does not always necessarily lead to lower SuperPi timing computation.

I was going to say SuperPI is extremely sensitive to memory timing and only time I see it go down, is loosening the timing .
That makes sense now, did know it was changed .

Yes, they seem to hit a wall around 4.7 .
I still am surprised the OC 4790@4.8 is faster than 4770@4.8 , I would think it be very close .
I guess to check chip we would need to go down to 4.7 and keep all settings the same to see how each chip does with same exact settings .

 

[cadaveca]

I'll just offer my own opinion...24/7 stable will never be much more than 4.7 GHz. However, if you like to run benchmarks only, more is possible. This wasn't really meant to be much more than an improvement for overclockers that like to run benchmarks, IMHO.

 

[warhammer23]

Just for my curiosity, may i ask what is the real difference between the 4790k at stock vs a 2700k at stock?
How better is it in gaming ? 10-15-20% ?
Thank you.

 

[newtekie1]

Considering the price difference of $5 right now, there isn't much sense in buying a 4770K. However, I can see the usefulness of this review in informing people that already have a 4770K that it isn't really worth the upgrade. Nice review.

 

[eidairaman1]

Pretty big disappointment for intel camp...