• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Intel's Haswell Architecture Analyzed in detail

I always said back when I got my Bloomfield that Haswell would was the earliest CPU I could envisage wanting to upgrade to. However, although lower heat output and power usage would be nice, I have no reason to upgrade still.

Maybe Intel shot themselves in the foot with the i7-9xx series. ;)
 
IB-E in Q3 of this year is what I'm getting, Haswell doesn't sound like its going to be much of a tock.

8 core at hopefully around 4ghz, will be significant enough to make it worthwhile to upgrade.
 
ULP said:
IB-E in Q3 of this year is what I'm getting, Haswell doesn't sound like its going to be much of a tock.

8 core at hopefully around 4ghz, will be significant enough to make it worthwhile to upgrade.
Click to expand...

bulldog147 said:
I hope Haswell will bring in stock speed of 4200MHz with 8 cores.
Click to expand...

There's no precedent to suggest that this will happen. For desktop components Intel have been systematically reducing the power consumption, and increasing the computational power of the overall CPU/GPU.

The Socket 2011 components appears intrinsically linked to the server Xeon developments - therefore it follows that suggestions that the next development we will see will be components like the Xeon E5 2690 (8 cores, 16 threads) making their way to the extreme desktop platform - see: http://www.xbitlabs.com/news/cpu/di...xt_Gen_Xeon_Core_Extreme_Microprocessors.html
 
triggerthat said:
What was the performance jump from SB to IB? Can I assume that the jump from IB to Haswell will be similar?
Click to expand...

SB to IB was "up to" 10% boost in per clock performance. The "tock" cycle is usually more than "tick" but with Intel shifting focus to onboard GPU development I'm not expecting as much as in the past.

To be honest, From Sandybridge onwards there's very little reason to upgrade simply for for day-to-day performance. They're plenty powerful enough for all current desktop uses, especially when overclocked. That doesn't account for the "upgrade bug" though :)
 
I haven't bought a new board/cpu since 2007, but I am itching to buy new and shiny and as it is Intel are giving me very little incentive to with haswell.

Performance wise, my most demanding usage is gaming and some folding for night time heater fun, so middle of the pack requirement.
So things being like they look like they will be (processor releases and console hardware/game direction), if I were to give in and buy new, I see myself dropping money on a steamroller hex or maybe even an apu when it comes.

(Every board change so far has been accompanied with more cores, so I would like to continue that progression for no other reason than I would like to. But I see no reason to pay a premium for another quad with marginal performance benefit or silly money for hexy+ goodness).
 
Can somebody with the current Intel HD4000 graphics run the dirt3 at those settings and see what sort of FPS you get so we can get some kind of metric.
I mean if the HD4000 runs it as a juddering mess then it gives us some idea, whereas if it manages 25FPS or so. anyway you get the idea.
 
Martini1991 said:
If that's anywhere close to the truth, surely that's 7660D beating performance?
Click to expand...

Id say that's almost a certainty if it kept up with a GT 650m.

According to the Dirt 3 benchmarks over on notebookcheck a GT 650M DDR5 gets 80 FPS at high quality settings, and the HD4850 desktop card gets 84 O.O


bru said:
Can somebody with the current Intel HD4000 graphics run the dirt3 at those settings and see what sort of FPS you get so we can get some kind of metric.
I mean if the HD4000 runs it as a juddering mess then it gives us some idea, whereas if it manages 25FPS or so. anyway you get the idea.
Click to expand...

According to the aforementioned benchmarks the Intel HD 4000 gets 22FPS. That's a serious improvement in Intel's GPU :O

If it does turn out to be on par with the GT 650m DDR5 (a card that's roughly on par with a freaking 8800GTX) then Intel have come a very long way very fast considering its only half a decade ago that review sites were mocking their products for being unable to outperform a Voodoo 2 Banshee :P
 
Last edited:
I actually made a post over on the graphics section about the GT3E:

http://forums.overclockers.co.uk/showthread.php?t=18477674

They are using a desktop chip with a desktop motherboard.

ubersonic said:
Id say that's almost a certainty if it kept up with a GT 650m.

According to the Dirt 3 benchmarks over on notebookcheck a GT 650M DDR5 gets 80 FPS at high quality settings, and the HD4850 desktop card gets 84 O.O




According to the aforementioned benchmarks the Intel HD 4000 gets 22FPS. That's a serious improvement in Intel's GPU :O

If it does turn out to be on par with the GT 650m DDR5 (a card that's roughly on par with a freaking 8800GTX) then Intel have come a very long way very fast considering its only half a decade ago that review sites were mocking their products for being unable to outperform a Voodoo 2 Banshee :P
Click to expand...

They are running a desktop chip with the GT3E version with embedded RAM. This is meant to be a large MCM. It appears the GT2 and GT1 versions will be the common ones found in most of the Haswell CPUs and lack the embedded RAM.

You do also know that shadows have been set to medium and the GT650M comes in GDDR3 versions too??

Moreover,is it running in DX9 or DX11 mode??

The desktop HD7660D runs DiRT3 at around 40FPS at 1920X1080 with shadows on high in DX11 mode according to HC. In the second they tested the HD4000 could muster 24FPS at 1920X1080 on high.

Moreover,where is the framerate meter in the video?? None. So the GT650M could be producing 70FPS and the GT3E 35FPS for all you know??

Why are they not allowing people to actually play the game instead of a pre-done time demo??

I have run DiRT2 on an IGP and above 30FPS looks similar to any of those videos.

Here is the new AMD Temash SOC running DiRT Showdown at 1920X1080:

http://forums.overclockers.co.uk/showthread.php?t=18477685

It could be smoother,but the SOC in total only has a 5.9W TDP and at least people were allowed to play the game. I need to find what settings were used though.
 
Last edited:
CAT-THE-FIFTH said:
Moreover,where is the framerate meter in the video?? None. So the GT650M could be producing 70FPS and the GT3E 35FPS for all you know??
Click to expand...

You need to read between the lines "Intel wouldn't let us report performance numbers, but subjectively the two looked to deliver very similar performance." is basically saying that the performance numbers seen were very similar but their not allowed to report them due to NDA.


CAT-THE-FIFTH said:
Why are they not allowing people to actually play the game instead of a pre-done time demo??
Click to expand...

They did, "Note that I confirmed all settings myself and ran both games myself independently of the demo."
 
Interesting article about DiRT3 on HardOCP:

http://www.hardocp.com/article/2011/06/13/dirt_3_gameplay_performance_review/3

Shadows : The Shadows option determines the resolution and smoothness of cast dynamic shadows. This setting has the single largest impact on performance of any setting, outside of resolution. The highest setting ("Ultra") shows precise and smoothed shadows, while the lowest setting ("Ultra Low") shows no dynamic shadows at all.
Click to expand...

Now it is clear why they set shadows to medium instead of high for the demo.

CAT-THE-FIFTH said:
I actually made a post over on the graphics section about the GT3E:

http://forums.overclockers.co.uk/showthread.php?t=18477674

They are using a desktop chip with a desktop motherboard.



They are running a desktop chip with the GT3E version with embedded RAM. This is meant to be a large MCM. It appears the GT2 and GT1 versions will be the common ones found in most of the Haswell CPUs and lack the embedded RAM.

You do also know that shadows have been set to medium and the GT650M comes in GDDR3 versions too??

Moreover,is it running in DX9 or DX11 mode??

The desktop HD7660D runs DiRT3 at around 40FPS at 1920X1080 with shadows on high in DX11 mode according to HC. In the second they tested the HD4000 could muster 24FPS at 1920X1080 on high.

Moreover,where is the framerate meter in the video?? None. So the GT650M could be producing 70FPS and the GT3E 35FPS for all you know??

Why are they not allowing people to actually play the game instead of a pre-done time demo??

I have run DiRT2 on an IGP and above 30FPS looks similar to any of those videos.

Here is the new AMD Temash SOC running DiRT Showdown at 1920X1080:

http://forums.overclockers.co.uk/showthread.php?t=18477685

It could be smoother,but the SOC in total only has a 5.9W TDP and at least people were allowed to play the game. I need to find what settings were used though.
Click to expand...


ubersonic said:
You need to read between the lines "Intel wouldn't let us report performance numbers, but subjectively the two looked to deliver very similar performance." is basically saying that the performance numbers seen were very similar but their not allowed to report them due to NDA.
Click to expand...

You are reading way into it. Anyone can say that two videos look similar,but in reality it does not mean they are??

ubersonic said:
"Note that I confirmed all settings myself and ran both games myself independently of the demo."
Click to expand...

Supposedly,and again without FRAPs or anything how does he know the difference between the two?? Its just opinion not facts.

Its like taking a GTX560 and a GTX660TI and saying they feel subjectively the same in DiRT3:

http://images.anandtech.com/graphs/graph6276/49739.png

However,it does not mean they are the same in actual performance.


Firstly,I linked to the GT3E results before anyone here on OcUK so I know what he said clearly. This is not the first time a big deal about Intel graphics has been made?? Remember the HD4000 in Ultrabooks?? In reality they were not as good as the Intel hyped and demoed as they barely got out of Turbo,so until we get solid numbers I am not going to get too excited.

Do you think it should be able to at least match an HD7660G or exceed it? I don't see why it should not be able to,TBF, especially with a whole load of eDRAM and a faster CPU backing the IGP,but again we need reviews to confirm this. However,the AMD mobile IGPs do have much lower clockspeeds than their desktop versions. The GT3E part has at least twice the number of shaders as the GT2 and is a MCM with a whole load of stacked eDRAM. It is a big package indeed and Intel is very quiet about the GT1 and GT2.

However,will it be as fast as a GT650M GDDR5?? I would extremely suprised if it did with image quality settings ramped up especially in newer games.

If this was a reference laptop even running it I would be more excited TBF.

So lets wrap up:
1.)Do we have an indication of actual framerates??

Nope. Just a subjective evaluation of them. Its the same with things like stutter.

2.)Is this DiRT3 under DX9 or DX11?? Why not DiRT Showdown which is more intensive??

We don't know.

3.)Shadows set to medium - why not high if all other settings are at high??

As HardOCP indicates shadows are the major image quality setting which can crater performance outside resolution.

4.)What version of the GT650M?? GDDR3?? GDDR5??

5.)There are no videos of people actually playing it.

6.)Was the laptop running on battery or mains power??

7.)Use of a desktop chip for the Intel system.

The laptop is running a lower clocked laptop CPU.

What are the specifications of the laptop used??

8.)They supposedly have a Haswell laptop too and yet use a desktop to run the demo??
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom