Intel Core i7-4790 Incrementally Faster than i7-4770K: Review

[btarunr]

TweakTown went ahead and posted the first formal review of Intel Core i7-4790, the fastest Core "Haswell" Refresh quad-core processor. Tested on an unnamed Z97 Express motherboard, with discrete graphics, the chip was found to be only incrementally faster than the i7-4770K, and just what you'd expect from a 100 MHz clock speed bump. The i7-4790 is clocked at 3.60 GHz, with a maximum Turbo Boost speed of 4.00 GHz, compared to the 3.50 GHz and 3.90 GHz clocks of the i7-4770K. It lacks an unlocked base-clock multiplier, and so its overclocking potential is severely limited, and close to non-existent; and so the i7-4790 is really a chip for those who want the best gaming performance, and don't intend to overclock their CPUs. Enthusiasts may want to hold out for "Devil's Canyon." Find the full review at the source.



View at TechPowerUp Main Site

 

[Blazer]

Obviously this benchmark does not utilize a 6 core CPU with 12 threads

Of corse it doesn't, both the 4770K and its predecessor the 3770K had only 4 cores, so why would this have more?

 

[HumanSmoke]

What a waste of time. The only real point of interest being voiced (for the top end SKUs) is how well it overclocks and whether the revised thermal interface material is any better than the previous two generations.

Also quite weird that AMD's sole representative is the (relatively) old FX 8150.

TT reaffirming their relationship with the cutting edge

 

[matar]

Of corse it doesn't, both the 4770K and its predecessor the 3770K had only 4 cores, so why would this have more?

I made a mistake I thought I saw i7-4960 I deleted my post

 

[The Von Matrices]

Of corse it doesn't, both the 4770K and its predecessor the 3770K had only 4 cores, so why would this have more?

If you look at the raw benchmarks they also test 6-core CPUs as well, although I am left to wonder why btarunr chose to use the memory bandwidth charts as the figure in the original post. Of all the benchmarks that is probably the one with the least correlation to real world performance.

Although looking at the charts does lead to the question as to why the processors with the 4-channel memory controllers have 50% less bandwidth than the ones with the 2-channel memory controllers.

 

[mouacyk]

I don't even wan to think about cleaning up my CLU just to upgrade from 4770K to 4790K for that 100MHz... hope these are better binned so they can help extend Z8/97's life.

 

[HumanSmoke]

Although looking at the charts does lead to the question as to why the processors with the 4-channel memory controllers have 50% less bandwidth than the ones with the 2-channel memory controllers.

Well, that's going to have to remain a mystery as long as TT keep their test system fit-outs secret. Maybe they outfitted their 4 channel board with a couple of sticks of DDR3-1066 !?!?!

 

[The Von Matrices]

Maybe they outfitted their 4 channel board with a couple of sticks of DDR3-1066 !?!?!

It's likely literally a couple of sticks, as in two instead of four.

 

[jmcosta]

Although looking at the charts does lead to the question as to why the processors with the 4-channel memory controllers have 50% less bandwidth than the ones with the 2-channel memory controllers.

because its an Ivy bridge?

 

[The Von Matrices]

because its an Ivy bridge?

Both the 3770K and the 4960X have Ivy bridge cores. The 4960X has a 4 channel memory controller while the 3770K has a 2 channel memory controller, yet suspiciously the 4960X has lower memory bandwidth.

 

[HumanSmoke]

It's likely literally a couple of sticks, as in two instead of four.

This is Tweaktown, it could just as easily be that they recycled the benchmarks from an old bench using a AIDA64 2.xx build.

 

[d1nky]

''What will be more interesting is the release of the K based processors. While they will carry the same speed bump, what we're really hoping for is better overclocking in both the core and memory department.''

(source tweaktown)

This is what may persuade me to buy one, and maybe a lot more enthusiasts!

 

[jmcosta]

Both the 3770K and the 4960X have Ivy bridge cores. The 4960X has a 4 channel memory controller while the 3770K has a 2 channel memory controller, yet suspiciously the 4960X has lower memory bandwidth.

slightly lower performance, maybe they only used 2sticks lol

 

[The Von Matrices]

This is Tweaktown, it could just as easily be that they recycled the benchmarks from an old bench using a AIDA64 2.xx build.

You're right, but it's even worse. If you look closely at the article it says:

Version and / or Patch Used: 1.00.1035BETA

So a 2010 era benchmark for a 2014 processor.

 

[HumanSmoke]

WRT to my last (half joking) comment about the possibility of Tweaktown recycling an old benchmark score using an old AIDA64 build. It actually looks as if that is exactly the case. The numbers are exactly the same as this November 2012 review.


EDIT: Out-Clouseau'ed by TVM. Well ninja'ed sir.

 

[Prima.Vera]

Forget those synthetic craps, check out the REAL benches:

and the others:
http://www.tweaktown.com/articles/6...h-cpu-and-z97-performance-preview/index5.html

As you can see there is 0 (zero) difference from this CPU and the older 3770K.
Yeah, thanks a lot intel, lol.

 

[Blazer]

TweakTown went ahead and posted the first formal review of Intel Core i7-4790, the fastest Core "Haswell" Refresh quad-core processor. Tested on an unnamed Z97 Express motherboard, with discrete graphics, the chip was found to be only incrementally faster than the i7-4770K, and just what you'd expect from a 100 MHz clock speed bump. The i7-4790 is clocked at 3.60 GHz, with a maximum Turbo Boost speed of 4.00 GHz, compared to the 3.50 GHz and 3.90 GHz clocks of the i7-4770K. It lacks an unlocked base-clock multiplier, and so its overclocking potential is severely limited, and close to non-existent; and so the i7-4790 is really a chip for those who want the best gaming performance, and don't intend to overclock their CPUs. Enthusiasts may want to hold out for "Devil's Canyon." Find the full review at the source.

I made a mistake I thought I saw i7-4960 I deleted my post

Ok, gotchya.



Source: TweakTown

T

 

[HumanSmoke]

Forget those synthetic craps, check out the REAL benches

It should come as no surprise to ANYONE that graphics content isn't CPU dependant for a single GPU in virtually any scenario barring (rare) core count coding issues or heavy usage of CPU calculated AI routines or physics.

It is why CPUs are a secondary consideration in a gaming machine unless you're needing to keep multiple GPUs occupied

 

[radrok]

It's Haswell refresh guys, what were you expecting? A new microarch?

We all know the real deal will be X99's 8 core CPUs.

 

[cadaveca]

although I am left to wonder why btarunr chose to use the memory bandwidth charts as the figure in the original post.

I suppose that you answered your own question.

It's Haswell refresh guys, what were you expecting? A new microarch?
We all know the real deal will be X99's 8 core CPUs.

So I guess I don't need to do a 4790 review then, eh? But then...why do I have one?

I mean, given that there was some obvious confusion about what is/was coming, I'm not surprised by the reactions. It's a non-K chip, not some flagship model. And then, with X99, we have DDR4 to deal with too...or so the rumour goes.

 

[radrok]

Dave I have the answer, are you willing to open the spoiler?

Spoiler

Intel deliberately refreshed Haswell with a clock bump to waste your time

Or our time maybe.

What bugs me is that now they are going to release yet another pointless (??) CPU, Devil's Canyon which is basically what delidders did one year ago and still continue to do.

Maybe it will target who didn't want to void their warranty.

I'm still skeptical it will bring higher clocks to the table, if it doesn't scale better with voltage you are left with small headroom compared to the first Hsw iteration.

After all Intel's fab process is kinda sensitive to voltage, you don't have to operate much more than 1.5v to degrade irreparably your chip.

 

[cadaveca]

Dave I have the answer, are you willing to open the spoiler?

Spoiler

Intel deliberately refreshed Haswell with a clock bump to waste your time

And give me a free chip or three? Why would I say no? So I can less CPUs to show in future reviews?

To me, it's useful to have as the example of "top-of-the-line" non-K chippery. Beyond that, there's still not much more I can say yet.

 

[Fiery]

This is Tweaktown, it could just as easily be that they recycled the benchmarks from an old bench using a AIDA64 2.xx build.

Most likely that's exactly what's happened. AIDA64 v2.xx used to have only single-threaded memory bandwidth benchmarks. Since v3.00, AIDA64 uses all available CPU threads which boosts memory scores considerably.

 

[RCoon]

There's only one processor I'm interested in these days.
I want to see the Pentium anniversary edition benchmarked against these things.

At 8Ghz.

I'll be laughing while my $60 unlocked chip is laying waste to all these quad core HT'd on liquid cooling in an ITX case.

 

[DarkOCean]

The pc hardware world, at least for me has become more boring than watching the grass grow.