Core i7-4770K: Haswell's Performance, Previewed

[falchard]

The pariah of Intel Integrated GPUs is finally here. And it turns out its still bad.

 

[downhill911]

Worth it or not, I prefer to sell my hardware every year on ebay and buy new one if good deal is available rather then keeping it until it ages and has almost no value.

 

[ubercake]

...I think this is a good related article:
http://www.anandtech.com/show/6670/dragging-core2duo-into-2013-time-for-an-upgrade

I think the article would have been better if they chose an extreme-edition 775 processor to begin with. They used the best processors for the other platforms (Sandy-E and Ivy). This way we could actually compare performance differences in architecture. Why did they choose an E6400 in this article, but to prove a point we hope would be the truth?

I bet they'd hate to prove a 775 extreme edition with a GTX 580 could offer as good a gaming experience as a Sandy-E with a GTX 580, huh? I don't think this is possible, but the processor choice in that article begs a couple of questions.

 

[de5_Roy]

Frankly, I'm upgrading to Haswell by the end of the year, so that any teething troubles are sorted and prices fall. I'm thinking of getting the 4770K instead of the 4670K simply because i'm not sure what'll happen after this. If my Q8400 lasted this long, I'm pretty certain that a 4770K would outlast the next console generation (the PS4/Xbox 720 one).

better to wait for motherboards with revised pcb (higher than 1.0) and revised lynx point chipsets. revised pcb and mature bios will take care of early problems and revised chipset will solve the usb 3.0 glitch. or you could go for higher end mobos which have 3rd party usb 3.0 controllers.
it could easily take months to launch new chipsets, remember cougar point. rumor: intel is considering slow roll out, so prices could be high for a while.

 

[InvalidError]

[citation][nom]ericjohn004[/nom]Hopefully by next year or the year after we'll start getting some really solid 10-15% increases in performance from one gen to the other.[/citation]
We are unlikely to ever see such high annual performance growth rates again for multiple reasons:
- power-effective architectural improvements have been tapped out
- mass-production process has hit a practical performance-per-watt brick wall at ~3.5GHz
- mainstream software that makes significant use of more than two cores (or even more than one single core in many cases) is still largely missing in action, more than 10 years after multi-cored/multi-threaded CPUs became standard
- adding unneeded cores increases arbitration overheads across the CPU which means performance and power penalties for lightly threaded software which account for the majority of most people's uses
- extremely heavy emphasis on power-efficiency to accommodate rapidly growing laptop and tablet sales along with pressure for green datacenters
- receding desktop sales across all performance segments further emphasizing the need to focus on lower-power non-desktop platforms
- growing customer base still using their 5+ years old PCs because they are still more than good enough for everyday use due to lack of mainstream killer apps
- etc.
- and most importantly: if increasing performance was so easy, why is AMD failing to catch up? The simple answers: it is harder than it sounds and is not cost-effective so AMD chose to quit trying

Enthusiasts, gamers and high-compute professionals/students/research are only ~10% of Intel's market so it is only natural for Intel to focus on the other 90% who benefit a lot more from power efficiency and integration.

 

[Novuake]

So, I've been thinking about upgrading from a Core2 E8400. When can we expect Haswell?

Should I sit tight for Haswell i5's or just get a 3570k?

Frankly, I'm upgrading to Haswell by the end of the year, so that any teething troubles are sorted and prices fall. I'm thinking of getting the 4770K instead of the 4670K simply because i'm not sure what'll happen after this. If my Q8400 lasted this long, I'm pretty certain that a 4770K would outlast the next console generation (the PS4/Xbox 720 one).

better wait for motherboards with revised pcb and revised lynx point chipsets. revised pcb and mature bios will take care of early problems and revised chipset will solve the usb 3.0 glitch. or you could go for higher end mobos which have 3rd party usb 3.0 controllers. it could easily take months to launch new chipsets, remember cougar point. rumor: intel is considering slow roll out, so prices could be high for a while.

Actually from what I have read a simple chipset revision is not going to fix the USB 3 issue on native 1150 chipset(Z87 for example).

 

[JonnyDough]

[citation][nom]Novuake[/nom]Ivy runs VERY hot when OCd. So he is not wrong.[/citation]

So does a Pentium 4, an Athlon 64, etc. Ivy Bridge is ALREADY OC'd. My i5 runs at 3.8ghz on turbo by DEFAULT.

 

[Novuake]

[citation][nom]Novuake[/nom]Ivy runs VERY hot when OCd. So he is not wrong.[/citation]

So does a Pentium 4, an Athlon 64, etc. Ivy Bridge is ALREADY OC'd. My i5 runs at 3.8ghz on turbo by DEFAULT.

Highlighted is a VERY noobish comment, I won't explain the obvious reason why.

Yeah when using apps running on less than all 4 cores. If it is not well cooled and using more cores, that clock rate will drop to 3.4GHz.

Anyway. What he meant was that he is hoping that Haswell runs more similar to Sandy when OCd.
Keep in mind that Sandy's IHS is soldered onto the CPU die, but Ivy's just has a sort of thermal paste instead(this was cost saving on Intels part I assume). The heat transfer is much less efficient on Ivy thanks to this.
So yeah, we are all hoping that Haswell uses the soldering method(so effectively it runs cooler than which Ivy did).

Makes sense?

 

[ojas]

Hey Chris, just had a sudden idea:

I don't remember seeing real-world encryption tests.

How about re-running your 7-zip and WinZip (OpenCL enabled) tests with AES-256 encryption? For future benchmarks, of course.

 

[ojas]

[citation][nom]ubercake[/nom]I think the article would have been better if they chose an extreme-edition 775 processor to begin with. They used the best processors for the other platforms (Sandy-E and Ivy). This way we could actually compare performance differences in architecture. Why did they choose an E6400 in this article, but to prove a point we hope would be the truth?I bet they'd hate to prove a 775 extreme edition with a GTX 580 could offer as good a gaming experience as a Sandy-E with a GTX 580, huh? I don't think this is possible, but the processor choice in that article begs a couple of questions.[/citation]
I don't disagree with you, but that's probably all they had? And notice they're not all the best, there was an i3 in there too...but yeah, look on page 5 of the news comments to find my benchmarking comparison with a Q8400.
Actually, here it is again: http://bit.ly/15VXs8b
[citation][nom]jesot[/nom]So, I've been thinking about upgrading from a Core2 E8400. When can we expect Haswell?Should I sit tight for Haswell i5's or just get a 3570k?[/citation]
Haswell, IMO. As Chris wrote, last generation which is known to be an open platform, we don't know what'll happen going forward. Haswell chips will cost the same, so why not? At least, that's what i'm doing.
[citation][nom]de5_Roy[/nom]better to wait for motherboards with revised pcb (higher than 1.0) and revised lynx point chipsets. revised pcb and mature bios will take care of early problems and revised chipset will solve the usb 3.0 glitch. or you could go for higher end mobos which have 3rd party usb 3.0 controllers. it could easily take months to launch new chipsets, remember cougar point. rumor: intel is considering slow roll out, so prices could be high for a while.[/citation]
Yeah that's what i'm going to do, wait and watch. It's not like i'll be able to afford an i7-4770K with a dual-LAN mobo like i plan to before January anyway...
[citation][nom]Novuake[/nom]Actually from what I have read a simple chipset revision is not going to fix the USB 3 issue on native 1150 chipset(Z87 for example).[/citation]
Hmm? I thought it was as simple as a firmware fix...at least, that's what i read. Anyway, i rarely ever put my desktop to sleep, and IIRC you'll just have to restart the crashed application to get it working again. So no big deal there, really. At least, that's what i hope. June couldn't come soon enough!

Excellent points by InvalidError. Though i must say, 10 to 15% improvements each year over the last 5 have resulted in a huge increase over the Core 2 generation, and arguably over the first Core ix generation as well.

 

[Onus]

I just looked over the CPU hierarchy chart again. I would suggest everyone do so, just for a little perspective. Over the years, a given CPU will "find itself" in lower and lower tiers. Game review articles all show yesteryear's CPUs beginning to struggle in those titles that are CPU-bottlenecked. Many games are not though, and even where they are, it's now usually due to lack of cores as we're finally seeing games (and applications) better optimized for more cores / more threads.
I understand those who always want to be on the leading edge, even if only for bragging rights, or personal pride; you're going to upgrade whether you "need" it or not. If your goal is to play games though, there is nothing about Haswell that is the least bit compelling if you already have a CPU in the top three or even four tiers.

 

[ojas]

Here're my Cinebench results:

In case the BBcode fails:
http://bit.ly/WzwosC

EDIT: Worth noting that the 3632QM has a 3GHz Turbo which it can hold for a very long time (even under prime95!). The U-series i7...can't remember how much or how long it held its Turbo.

The QM was in a WAIO E14A and the U-series was the soul of the XPS 12. The P8600 resides in a MacBook Pro from late 2009.

 

[ojas]

[citation][nom]Onus[/nom]I just looked over the CPU hierarchy chart again. I would suggest everyone do so, just for a little perspective. Over the years, a given CPU will "find itself" in lower and lower tiers. Game review articles all show yesteryear's CPUs beginning to struggle in those titles that are CPU-bottlenecked. Many games are not though, and even where they are, it's now usually due to lack of cores as we're finally seeing games (and applications) better optimized for more cores / more threads.I understand those who always want to be on the leading edge, even if only for bragging rights, or personal pride; you're going to upgrade whether you "need" it or not. If your goal is to play games though, there is nothing about Haswell that is the least bit compelling if you already have a CPU in the top three or even four tiers.[/citation]
Obviously not the top 3, even Tom's notes that unless you're jumping more than three tiers the upgrade would be sort of parallel. However 5 or 6 tiers...night and day. Same applies to GPUs...i went from a 9600GT to a GTX 560...lol night and day.

 

[Novuake]

Hmm? I thought it was as simple as a firmware fix...at least, that's what i read. Anyway, i rarely ever put my desktop to sleep, and IIRC you'll just have to restart the crashed application to get it working again. So no big deal there, really. At least, that's what i hope. June couldn't come soon enough!

Excellent points by InvalidError. Though i must say, 10 to 15% improvements each year over the last 5 have resulted in a huge increase over the Core 2 generation, and arguably over the first Core ix generation as well.

It could be a bad rumour, but this seems to be related to power stepping on the Haswell chips. So as far as i can gleam, this can not be simply fixed by a simple chipset fix as stepping is inherently on CPU and only controlled by the chipset.
http://www.techspot.com/news/51895-new-hardware-required-to-fix-haswell-usb-30-connection-bug.html
Then again I could be wrong in my logic.
http://techreport.com/news/24494/report-new-chipset-revision-to-fix-haswell-usb-3-0-bug
What do you think?

I just looked over the CPU hierarchy chart again. I would suggest everyone do so, just for a little perspective. Over the years, a given CPU will "find itself" in lower and lower tiers. Game review articles all show yesteryear's CPUs beginning to struggle in those titles that are CPU-bottlenecked. Many games are not though, and even where they are, it's now usually due to lack of cores as we're finally seeing games (and applications) better optimized for more cores / more threads.
I understand those who always want to be on the leading edge, even if only for bragging rights, or personal pride; you're going to upgrade whether you "need" it or not. If your goal is to play games though, there is nothing about Haswell that is the least bit compelling if you already have a CPU in the top three or even four tiers.

Although I agree with you on 3 tiers as games just are still VERY GPU boudn dispite improved multithreading, it is a very fine balance of threads vs per core performance when considering an upgrade in my opinion. 4 is pushing it a bit.

 

[omarccx]

So, I've been thinking about upgrading from a Core2 E8400. When can we expect Haswell?

Should I sit tight for Haswell i5's or just get a 3570k?

I'm building a 3570k rig next week. I can't wait another month.

 

[Novuake]

[citation][nom]Onus[/nom]I just looked over the CPU hierarchy chart again. I would suggest everyone do so, just for a little perspective. Over the years, a given CPU will "find itself" in lower and lower tiers. Game review articles all show yesteryear's CPUs beginning to struggle in those titles that are CPU-bottlenecked. Many games are not though, and even where they are, it's now usually due to lack of cores as we're finally seeing games (and applications) better optimized for more cores / more threads.I understand those who always want to be on the leading edge, even if only for bragging rights, or personal pride; you're going to upgrade whether you "need" it or not. If your goal is to play games though, there is nothing about Haswell that is the least bit compelling if you already have a CPU in the top three or even four tiers.[/citation]
Obviously not the top 3, even Tom's notes that unless you're jumping more than three tiers the upgrade would be sort of parallel. However 5 or 6 tiers...night and day. Same applies to GPUs...i went from a 9600GT to a GTX 560...lol night and day.

NO NO NO!!! LOL... 3 tiers on GPUs is what I have the patients for. Especially if you consider that from 9600GT you got a jump from DX9 to DX11.

 

[cangelini]

[citation][nom]Novuake[/nom]better wait for motherboards with revised pcb and revised lynx point chipsets. revised pcb and mature bios will take care of early problems and revised chipset will solve the usb 3.0 glitch. or you could go for higher end mobos which have 3rd party usb 3.0 controllers. it could easily take months to launch new chipsets, remember cougar point. rumor: intel is considering slow roll out, so prices could be high for a while.[/quotemsg]Actually from what I have read a simple chipset revision is not going to fix the USB 3 issue on native 1150 chipset(Z87 for example).[/citation]
A chipset revision actually should fix the USB 3.0 issue. There are issues with C1. Couple of different memory modules in thumb drives are problematic, but it's not expected to affect a lot of people. C2 will fix it, but that's going to be July/August, I hear.

 

[Novuake]

[citation][nom]Novuake[/nom]better wait for motherboards with revised pcb and revised lynx point chipsets. revised pcb and mature bios will take care of early problems and revised chipset will solve the usb 3.0 glitch. or you could go for higher end mobos which have 3rd party usb 3.0 controllers. it could easily take months to launch new chipsets, remember cougar point. rumor: intel is considering slow roll out, so prices could be high for a while.

Actually from what I have read a simple chipset revision is not going to fix the USB 3 issue on native 1150 chipset(Z87 for example).[/citation]
A chipset revision actually should fix the USB 3.0 issue. There are issues with C1. Couple of different memory modules in thumb drives are problematic, but it's not expected to affect a lot of people. C2 will fix it, but that's going to be July/August, I hear.[/quotemsg]

Horrible formatting there Chris.
Anyway... That is very good to hear. If it were to effect more people i would have hoped they would at least delay the launch for that month or two.
Thanks!

 

[Bulat Ziganshin]

> 7z with -core-avx2

people, are you know assembler?? is it hard to realize that sse/avx can't improve non-vector code?!

 

[ojas]

[citation][nom]Novuake[/nom]It could be a bad rumour, but this seems to be related to power stepping on the Haswell chips. So as far as i can gleam, this can not be simply fixed by a simple chipset fix as stepping is inherently on CPU and only controlled by the chipset.http://www.techspot.com/news/51895 [...] n-bug.htmlThen again I could be wrong in my logic.http://techreport.com/news/24494/r [...] sb-3-0-bugWhat do you think?[/citation]
:| looks like something similar to the Sandy Bridge fiasco...will they issue a recall? I don't know...maybe they're risking it with desktops, they'll say "if you're having this problem send it back for replacement"...whoever doesn't use sleep with a USB 3.0 device attached will never see the problem. Sounds like a chipset revision thing...in retrospect a firmware fix never made too much sense

I think it's better to wait till December if one doesn't want to risk it...oh wait i see Chris has replied...he'd know far better so i'm just going to listen to him.

[citation][nom]Novuake[/nom]NO NO NO!!! LOL... 3 tiers on GPUs is what I have the patients for. Especially if you consider that from 9600GT you got a jump from DX9 to DX11.[/citation]
Hahaha but what about the monies!? :lol:

The 9600GT supported DX10 btw. But it would suck at it, outside of Halo 2, which didn't use it much anyway it seemed to me...it was a good card. Sold it to a friend who's found it adequate...

6-tier cycles for the CPU and GPU seem more justifiable to me, partly because i almost continuously keep adding or changing other components...my 560's already hit the 6th tier if you count single-GPUs...7th if you count Titan...by next year it'll be upgrade time again, i suppose! :|

 

[Novuake]

[citation][nom]Novuake[/nom]It could be a bad rumour, but this seems to be related to power stepping on the Haswell chips. So as far as i can gleam, this can not be simply fixed by a simple chipset fix as stepping is inherently on CPU and only controlled by the chipset.http://www.techspot.com/news/51895 [...] n-bug.htmlThen again I could be wrong in my logic.http://techreport.com/news/24494/r [...] sb-3-0-bugWhat do you think?[/citation]
:| looks like something similar to the Sandy Bridge fiasco...will they issue a recall? I don't know...maybe they're risking it with desktops, they'll say "if you're having this problem send it back for replacement"...whoever doesn't use sleep with a USB 3.0 device attached will never see the problem. Sounds like a chipset revision thing...in retrospect a firmware fix never made too much sense

I think it's better to wait till December if one doesn't want to risk it...oh wait i see Chris has replied...he'd know far better so i'm just going to listen to him.

[citation][nom]Novuake[/nom]NO NO NO!!! LOL... 3 tiers on GPUs is what I have the patients for. Especially if you consider that from 9600GT you got a jump from DX9 to DX11.[/citation]
Hahaha but what about the monies!? :lol:

The 9600GT supported DX10 btw. But it would suck at it, outside of Halo 2, which didn't use it much anyway it seemed to me...it was a good card. Sold it to a friend who's found it adequate...

6-tier cycles for the CPU and GPU seem more justifiable to me, partly because i almost continuously keep adding or changing other components...my 560's already hit the 6th tier if you count single-GPUs...7th if you count Titan...by next year it'll be upgrade time again, i suppose! :|

And yes the 9600GT used to be a great card. Still have one in one of my systems. LOL...
And its a Biostar reference boards, shocked that its still running(albeit a bit glitchy at times).

Nah I only consider the single GPU's.

And yeah, I never use sleep either. My machines are always doing something or are off. So I could not care less so long as I can fix it later with a BIOS update.

 

[ojas]

[citation][nom]Novuake[/nom]And yes the 9600GT used to be a great card. Still have one in one of my systems. LOL...And its a Biostar reference boards, shocked that its still running(albeit a bit glitchy at times).Nah I only consider the single GPU's.[/citation]
Hmmm mine was a Leadtek...

Sorry, i meant if you *only* consider single GPU cards. Man, i need sleep. I knew this Haswell article would keep me obsessed and awake till 2 AM the moment i saw it here

 

[Novuake]

[citation][nom]Novuake[/nom]And yes the 9600GT used to be a great card. Still have one in one of my systems. LOL...And its a Biostar reference boards, shocked that its still running(albeit a bit glitchy at times).Nah I only consider the single GPU's.[/citation]
Hmmm mine was a Leadtek...

Sorry, i meant if you *only* consider single GPU cards. Man, i need sleep. I knew this Haswell article would keep me obsessed and awake till 2 AM the moment i saw it here

HAHA!!! I know the feeling. Cheers mate, see you around.

 

[ODuron]

LGA1156? How does it compare against the LGA2011? I'd love to see a test against the i7 3930K.

 

[Novuake]

LGA1156? How does it compare against the LGA2011? I'd love to see a test against the i7 3930K.

Research it maybe? There are many benches for Ivy VS 2011. So add 7-13% to Ivy's results and viola, you have your comparison.