Anandtech Haswell review is up!

[lopri]

It's relevant because it threatens your apparent need to justify owning a 2500K. Relax your ego, and evaluate Haswell objectively.

The first sentence can't be further from the truth and I have no internet creed to maintain. But you're right on that I don't know much about this new chip. I'll bow out.

 

[jpiniero]

Applies all in partially loaded situations, and at same TDP, which isn't what a 84W 4770K is versus a 77W 3770K.

Wonder if Intel raised the TDP more to save money by selling "leakier" parts for the 84W models.

 

[Abwx]

Applies all in partially loaded situations, and at same TDP, which isn't what a 84W 4770K is versus a 77W 3770K.

It wont improve the picture much , efficency at high TDP
is worse and increase as power usage decrease , so far
the TDP level at wich it start to be more efficient that IB
should be quite low , in fact this will be the case
only in mobile CPUs.

 

[TuxDave]

I would like to see a 4.30 GHz quad-core anything with higher Linpack score than I posted above.

Linpack? lol.

You may want to consider focusing on a different benchmark because linpack is the one that Haswell will dominate due to the simple fact of having FMA units.

<-- personally not a fan of linpack.

 

[Maximilian]

*Yawn*

 

[Grooveriding]

http://www.newegg.ca/Product/Product...82E16819116899

 

[LogOver]

It wont improve the picture much , efficency at high TDP
is worse and increase as power usage decrease , so far
the TDP level at wich it start to be more efficient that IB
should be quite low , in fact this will be the case
only in mobile CPUs.

Anandtech:
Here I’m showing an 11.8% increase in power consumption, and in this particular test the Core i7-4770K is 13% faster than the i7-3770K. Power consumption goes up, but so does performance per watt.

xbitlabs:
8-13% performance increase in x264 for 5% higher power consumption.

What did you say about efficiency?

 

[Revolution 11]

Whatever happened to Intel's "2% rise in performance for every 1% rise in power" golden rule?

If power is that important, why does the power rise almost as much as the performance? If performance were important, why worry about power or a low TDP at all?

 

[Abwx]

Haswell and USB 3.0 seeing big issues with USB drives
http://www.tweaktown.com/news/30730...ith-usb-drives/index.html#SFefabMRoMesyMym.99

Your attempt to save anything and everything Intel is pathetic just as these new chips.

It wasnt launched a saturday by chance.....

 

[Homeles]

Whatever happened to Intel's "2% rise in performance for every 1% rise in power" golden rule?

That is for Atom. Their flagship architecture is held to a 1:1 standard.

If power is that important, why does the power rise almost as much as the performance? If performance were important, why worry about power or a low TDP at all?

Because the architecture is about power in a mobile environment. Desktop performance isn't important.

 

[SunRe]

Anandtech:
Here I’m showing an 11.8% increase in power consumption, and in this particular test the Core i7-4770K is 13% faster than the i7-3770K. Power consumption goes up, but so does performance per watt.

xbitlabs:
8-13% performance increase in x264 for 5% higher power consumption.

What did you say about efficiency?

This?

 

[Lepton87]

Where are some benches with 4770R at fixed frequency against both 3770K and 4770K?No turbo boost please.
I want to see how much that cache affects gaming performance so get a discreet card in there, either an overclocked 780 or a Titan. 4770R should see another 5% more IPC over 4770k or maybe even more, especially in games which interest me the most yet no one tested that. I want that checked. It's a bummer that we won't get 4770RK LGA, but I guess it would cost almost as much as 6 core CPU, but it still might be worth it depending on the results. 25% more IPC then SB would have been awesome.
I just hate Intel market segmentation, the best haswell won't be available for LGA and won't be unlocked either. I don't care about iGPU, only about those 128MB of eDRAM which effectively act like L4. So if there was 4770RK LGA with soldered IHS I would probably upgrade, I have to see more tests. Right now I'm going to wait for IVY-E or even Haswell-E.

 

[blackened23]

As a desktop SKU, i'm indifferent. The overclocking could be better, but it is still the best part to purchase for a new desktop - this isn't really a compelling option for those who already have IB and were considering yearly upgrades.

As a mobile SKU, however, Haswell is going to be VERY exciting. I can't wait to see what products PC makers come up with, they're finally competing with Apple on quality (and will FINALLY have higher screen resolutions) and not just on price. Additionally, this opens up the door for a lot of Haswell based tablets and form factors that just weren't suited to intel products prior to Haswell. In this respect, Haswell will be a game-changer IMO.

The desktop part isn't extremely exciting, and the desktop user in me is somewhat disappointed. Overall, it was expected though - intel stated from the outset that mobility/efficiency was the goal. And here we are, no surprises.

 

[moonbogg]

I'm still unclear on gaming advantages Haswell has over 3930k. I guess i'll have to keep digging, or realize that the benefit must be so small that no one has posted a huge banner announcing it.

 

[Abwx]

Anandtech:
Here I’m showing an 11.8% increase in power consumption, and in this particular test the Core i7-4770K is 13% faster than the i7-3770K. Power consumption goes up, but so does performance per watt.

xbitlabs:
8-13% performance increase in x264 for 5% higher power consumption.

What did you say about efficiency?

You did misread what i wrote...

Efficency in nowhere close to expectations ,
besides , anand got the lower numbers of the whole web
particularly iddle power number.

Efficency is at best equal to IB or even worse ,
it wont show clear superiority only at low TDPs ,
as in mobile CPUs and surely it will still be much
less than awaited by the enthusiast crowd.

 

[blackened23]

I'm still unclear on gaming advantages Haswell has over 3930k. I guess i'll have to keep digging, or realize that the benefit must be so small that no one has posted a huge banner announcing it.

What gaming advantage? You're better off with the 3930k. All of the performance advantages with Haswell involve a small 10% IPC increase with the rest being related to AVX2. Unfortunately, nothing uses AVX2 yet.

 

[blackened23]

You did misread what i wrote...

Efficency in nowhere close to expectations ,
besides , anand got the lower numbers of the whole web
particularly iddle power number.

Efficency is at best equal to IB or even worse ,
it wont show clear superiority only at low TDPs ,
as in mobile CPUs and surely it will still be much
less than awaited by the enthusiast crowd.

The desktop part intentionally has a higher TDP than the mobile parts. The desktop part was never meant to be more efficient, because do you really care if you idle at 20W instead of 22W on a desktop? The mobile ULV Haswells start at a mere 2W IIRC - the mobile parts WILL be far more efficient than ivy bridge.

You can read Anand's article detailing all of this, of course. The design goal with Haswell from the outset was mobility, so the results aren't that surprising. Of course the overclocking for the desktop part is disappointing, and I wish it had been better in this respect.

 

[zebrax2]

Looks "meh" upgrade for those with previous gen. The idle power consumption looks sweet though.

What I'm most curious about is how the lack of TSX would affect the "K" chips in terms of performance in comparison to non-k chips as programs start adopting it.

 

[LogOver]

This?

What is your point?

What about this

or this

or this

 

[Face2Face]

I think this would be a good time for current C2D, Phenom II or Core i7 first gen owners to upgrade. 2nd and 3rd gen owners can probaly skip this gen from what I am seeing. I am the only guy that is tired of going from a quad to another quad then another quad

 

[leper84]

Has anyone seen a review containing mobile 5100 graphics?

GT3e cleans up trinity, and probably richland too, pretty damn well, but at ~$650!!!! it damn well better. I would really like to see the performance without the embedded ram, I wonder how 5100 would compare to trinity/richland.


I think AMD would be absolutely infinitely stupid if they haven't already planned to follow suit and release a top kaveri chip with embedded ram too. Would make things pretty interesting.

 

[Abwx]

The desktop part intentionally has a higher TDP than the mobile parts. The desktop part was never meant to be more efficient, because do you really care if you idle at 20W instead of 22W on a desktop? The mobile ULV Haswells start at a mere 2W IIRC - the mobile parts WILL be far more efficient than ivy bridge.

You can read Anand's article detailing all of this, of course. The design goal with Haswell from the outset was mobility, so the results aren't that surprising. Of course the overclocking for the desktop part is disappointing, and I wish it had been better in this respect.

I pointed that it s a mobile chip stretched up to DT usage ,
but you should look elsewhere that at Anand , most sites
measured higher power comsumption at stock frequency
and all benches have not the neccsary boost in perfs
to negate the effect on power efficency.

 

[Concillian]

Whatever happened to Intel's "2% rise in performance for every 1% rise in power" golden rule?

If power is that important, why does the power rise almost as much as the performance? If performance were important, why worry about power or a low TDP at all?

In short, I believe that Haswell was designed to perform most efficient at a lower voltage. As a result, they have to push it harder just to get the same stock speed as IB. I expect the lower power versions will be significantly more efficient for a given computational level, as we already have a taste of by seeing significantly lower idle power when compared to IB.

 

[Abwx]

What is your point?

What about this

or this

or this

Well choosen , WPrime doesnt stress the FP units , it use
mainly integer , while CB is not the most stressful of all tests ,
why not use your beloved linx instead..?...

 

[blackened23]

I pointed that it s a mobile chip stretched up to DT usage ,
but you should look elsewhere that at Anand , most sites
measured higher power comsumption at stock frequency
and all benches have not the neccsary boost in perfs
to negate the effect on power efficency.

Well, after deciphering your post, you're telling me the completely obvious. Yes, a desktop 84TDP part has less efficiency than a desktop 77W part. Hold the press, thank you for telling me something we already knew MONTHS IN ADVANCE from the specs alone.

I'm speaking strictly of the mobile ULV parts, where the primary design consideration of haswell went. The mobile parts are far more efficient than their ivy bridge counterparts with nearly double the battery life. The desktop part isn't as efficient - It wasn't made more efficient by design, for which you can simply look at the TDP. Desktop users don't care about unplugged battery life.