Haswell Debuts: Intel Core i7-4770K Review

AMD CPUs are different, a lot of people want to only compare it to the 3770k because amd puts it at its top of the line desktop processor so yea that's not a great tdp when soon that comparison. But yes power per core it's also better than people give credit, it may not be the best CPU, but it does its job and its good at it.

In terms of straight performance, it does a good job, overclocks fairly easily, and it's at a low price point. It costs the Same as a 3570k while being above it. All these chips from both lines have their places, and each fill the roles nicely.
 
A 225w CPU would be instant death to the silicon after a short while, followed by AMD filing for bankruptcy.

AMD CPU's (higher end) have always run hotter than Intel's since the Core series. The FX 8350 has a TDP/ACP of 125w. Phenom started with flagships with 130w TDP's, and since then, Intel has gone from 95w (SB), to 77w (IB), to 84w (Haswell). AMD's problem are their engineers, and/or lack of money for R&D. I can only guess, because we all know AMD is not happy with second place, so something is up.

AMD needs to get an efficient architecture on a low process technology. Not one or the other. Adding more cores/modules, higher frequency memory, higher core clocks to make up for a poor architecture is a band-aid. If AMD wants to be taken seriously, they need to step their game up. Rory seems to be on the right track, but I don't see anything really great coming from AMD (CPU's) for another couple years at least.

True.

And what exactly would that be, power consumption per core?

Are you smiling at the inefficiencies of AMD CPU's, because they use less power per core? Come on, you are trying to picture grass growing in a desert, because you choose not to live in a pasture.
Unfortunately, that is the case for AMD. The Intel Core i3 is basically equivalent in performance to the FX4300 while using a ton less power.
 
AMD CPUs are different, a lot of people want to only compare it to the 3770k because amd puts it at its top of the line desktop processor so yea that's not a great tdp when soon that comparison. But yes power per core it's also better than people give credit, it may not be the best CPU, but it does its job and its good at it.

In terms of straight performance, it does a good job, overclocks fairly easily, and it's at a low price point. It costs the Same as a 3570k while being above it. All these chips from both lines have their places, and each fill the roles nicely.

and how much do you pay for power in your house ghost ryder?
 
and how much do you pay for power in your house ghost ryder?

Not much, my power usage is not that high. My video cards use more power than my processor and I still don't see any issues with it. It's not like draining the neighborhood of power lol.

The 4770k is interesting, I want to see the heavy overclocks on this and see if it could hit sandy bridge levels.
 
Is there any way we could get gaming tests with two 7970's/GTX 680's or similar or higher class hardware?
 
I'm wondering why you think chips should be compared by the power they consume regardless of performance.
Passmark: (I like comparing Passmark benchmarks as well)
That is a larger gap in performance ratio (24:18) compared to the power ratio (26:25) you provided. I think the idea is to compare CPU power consumption ratios with CPU's that perform similar not miles apart.
Even that is disingenuous to an extent. The 3930K is a salvage part (actually closer to a twentieth tier salvage part) . It's kind of like comparing GPUs and putting up the GTX 770 against the HD 7870 XT for a comparison of architectures.
The 3930K has a 12MB L3 cache (the full die is 20MB), and six cores (the full die has eight). The full die also features (aside from the 5GT/sec I/O) two 8GT/sec interconnects which remain unused on the i7. Both turbo to 3.8GHz. The only downside (aside from the artificially inflated price) is that the 3930K is a 130 watt part, while the full die is 135 watt (Xeon E5-2690).
 
So jc713 if I'm reading what you said correctly/posted. The 4770k has actually less overclocking headroom than the 3770k???

Ummm now I'm officially lost because I'm not really seeing the point if this chip. Like I said, haswell sounds more like a mobile chip and it may shine there.

I'd like to personally see ivy-bridge e
 
Even that is disingenuous to an extent.
Not that I would have known any better, I wasn't the one that selected the two CPU's for comparisons. But since they were selected, why not look at a few popular details?

One thing is for sure DBZ. During my time on TechSpot, I have learned not to argue with you. :) Half the time I don't know what the crap you are saying, and I'm considered a PC Guru by everyone that personally knows me. I guess what I'm trying to say is, thanks for all the extra voluntary data. Maybe just maybe, one day it will all make sense.
 
Just for an apples-to-apples comparison, I'd note that early comparisons with mature products aren't always the best indicator.
For instance:
Asus' testing of initial 2600K overclocking ability
1. Approximately 50% of CPUs can go up to 4.4~4.5 GHz
2. Approximately 40% of CPUs can go up to 4.6~4.7 GHz
3. Approximately 10% of CPUs can go up to 4.8~5 GHz (50+ multipliers are about 2% of this group)
From early-mid 2012, the number of D2 revision 2600K's able to do 4.7 or higher comfortably is a pretty healthy percentage (my own 2600K will do 5GHz under duress)
From Asus' testing of initial 4770K overclocking ability
Of the processors Asus has tested, 70% hit 4.5GHz, 30% reached 4.6GHz, and 20% made 4.7GHz. Only 10% were stable at 4.8GHz.
So, the numbers from initial launch to initial launch aren't too dissimilar. Overclocking headroom obviously isn't the stopping point, its the amount of heat generated and needing to be dissipated from such a small die.
Not that I would have known any better, I wasn't the one that selected the two CPU's for comparisons.
Ah, I was dealing with an incomplete data set. That's what you get for putting people on ignore :eek: (not you obviously!).
Just saying that this thread seems to have become an architecture vs architecture discussion rather than the pro's/con's of the 4770K in particular...and if the discussion is architecture then it probably helps to use an apples-to-apples evaluation ( fully enabled CPU in both cases). Sorry for any misunderstanding.
 
So it is just batches really that determines your OCing ability?
There is a school of thought that different batches can yield better a better chance of a good overclocker. That is why there are so many threads devoted to tallying FPO batch numbers
The method does have merit. Not all wafers are created equal, and there is said to be a correlation between where on the wafer the die comes from and its performance. Not that an FPO will tell you that.
 
Depends on the quality of the part, just like with anything, if the batch of parts for the chip is slightly better than the previous one, you could get better overclockability.

Chips generally stick to a certain overclocking area and then depending on your chip, you can go a little further. In this regards, at the launch the 4770k seems like its almost a step backwards in terms of performance because it overclocks less, uses more power under load, and it has a higher tdp. I don't see it as a bad chip, but other than mobile market with idling, this does not seem like much of an improvement over the 3770k and was not worth making a whole new socket which isn't a good thing. If they had at least made it the same socket, it would have been at least a feasible thing to later plop one onto your board and try it out for a little improvement.
 
If they had at least made it the same socket, it would have been at least a feasible thing to later plop one onto your board and try it out for a little improvement.

I'm still hoping that Intel isn't really insane :)D) and that the socket change was done on purpose. If the socket remained the same and people threw down $300 on a new top-end CPU and received <10% performance boost, there would be an outcry.

However, this way the chip is only limited to brand-new builds, where there is definitely an improvement overall. Think OEM machines with no discreet GPU needed.
 
So it is just batches really that determines your OCing ability?
There actually is some truth to this. It's story time kids.
1000 years ago when Bloomfield was king I was coming/upgrading from an AM3 720 X3 (unlocked and overclocked to 3.5Ghz on a 770-G45) then the i7 950 released (as just a bump in the clock speed for the i7 9XX series basically). I just about bought one but countless forums and posts from people I trust were not happy with the overclock results and many builders recommended the 930 over the 950. So I listened.
Well my chip turned out to be a damn good one. I even had it running @ 4.4Ghz HT/on 24/7 reliably on air (push/pull) for a little but I backed it down to 4 even (24/7). It pushes my 670 to a P9605 3DMark11 (nothing great but decent for my old rig) and I still have OC headroom on my GPU and CPU if needed.
 
There actually is some truth to this. It's story time kids.
1000 years ago when Bloomfield was king I was coming/upgrading from an AM3 720 X3 (unlocked and overclocked to 3.5Ghz on a 770-G45) then the i7 950 released (as just a bump in the clock speed for the i7 9XX series basically). I just about bought one but countless forums and posts from people I trust were not happy with the overclock results and many builders recommended the 930 over the 950. So I listened.
Well my chip turned out to be a damn good one. I even had it running @ 4.4Ghz HT/on 24/7 reliably on air (push/pull) for a little but I backed it down to 4 even (24/7). It pushes my 670 to a P9605 3DMark11 (nothing great but decent for my old rig) and I still have OC headroom on my GPU and CPU if needed.

Interesting.
 
Yeah, not sure if batch and stepping are similar, but stepping D0 on the i7 920's was the beastly stepping that allowed higher OCs.
 
Some chips do better at overclocking than others, it just depends on sometimes the combination of parts that sometimes allows for things to survive better at different frequencies (or stable).

It runs luck of the draw it seems in most cases, but at least in this day and age we still can see at least a decent overclock range reachable on a giant variety of chips. We can all at least see something more out of every chip we get (Unlocked of course).
 
It's not just consumer CPU chips either. To quote a specific case, the Google Nexus 4 phones have a wide variety of binned CPU's in them which allow for very slightly better performance.
 
Yeah, not sure if batch and stepping are similar, but stepping D0 on the i7 920's was the beastly stepping that allowed higher OCs.
Technically D0 would be a process revision since the initial Bloomfield parts were C0. A change in letter usually means that the CPU has undergone an extensive metal layer respin, while a number change (using the same start letter) usually means an alteration to only some parts/levels of the die.

Batch numbers are more important where the production process improves but the revision stays the same (I.e. all production 2600K's are D2 revision. D0/D1 were reserved for pre-production silicon - usually seen as engineering samples)
 
Nothing ever lives up to the hype, that is the nature of hype - overstating reality - so I'm not disappointed with the new architecture. The potential is there, only waiting for the software to catch up. I'm also not mad that it doesn't blow my Ivy Bridge chip away - maybe I would be mad if it ran circles around my 3570K system, having just built it 6 months ago, but it doesn't, so I can probably clock mine higher and keep up with it. The extra SATA 6GB/s ports would be nice, but aren't essential (only SSDs really need the bandwidth), and my Z77 board has enough USB 3 ports. I wanted a good reason to upgrade, but this isn't it...
 
How come there are no 6-core & 8-core chips out for the 2011 socket??
 
There is:

Intel Xeon E5-2680 is an 8 core 2011 chip

Intel I7 3930K is an 6 core

There are also others on the 2011 socket.
 
Back