Intel Core i7-4770K 'Haswell' HD Graphics 4600 GPU Performance

[Delta6326]

Wow I must be getting negative scaling on Grid 2 I'm only getting about 32fps... 2x 4850

Performance is still lacking, but it will be very interesting to see BGA performance.

 

[Recus]

yeah and the same as a million other chips since way back in 2006 with the quad-core Kentsfield QX6700, ok so this wasn't fully integrated but my point is its been 7 years since the first Intel x86 quad-core, its time to step it up now. 8-core should have been the mid-range default by now, we are not really pushing frequencies up like we used to back in the old days and we seem to have stopped increase core-count too, and Intel wonders why people aren't buying as much and its profits are now?!

Give us something to be excited about and we'll put our hand in our pockets!

The QX6700 was made using 65nm, we have had 45nm, 32nm and now 22nm, QX6700 was over twice the size of the 4770k even taking into account the massive on-board GPU. Couple this with the massive improvements in thermal management and we should have moved on in the last 7 years.

Give us octa-core already!

 

[Thefumigator]

Haswell is an improvement. As simple as that. I like the power consumption improvement, and I still dislike its 3D performance despite being better than ever for intel.

The simple fact there are other products around like trinity with better 3D performance in one chip spoils a bit the whole product but on the other side, if you get haswell for its x86 performance, which is great performance indeed, I don't really get why is the GPU integrated into the chip (apart from making quicksync work). I don't really get it, I mean, top of the line FX processors lacks GPU because you don't want these. So I still believe intel should take that thing out in my opinion.

Anyway the Intel beast gets closer on the GPU side, and closer every day, someday it may catch.

 

[lordz]

Why are the 5800k and 6800k being ran on 1600mhz ram when testing their IGP?
Its well known that these boards support much faster ram and the integrated graphics on these chips perform drastically better on 1866 or 2133mhz.

If you were building any of these systems for their IGP value you would run the AMD's on the correct speed ram and if this had been done then there would be another 10fps on the AMD chips making them far superior than the intel's HD4600.

 

[lordz]

I should clear it up a little bit, i understand they used 1600mhz ram for both systems to be fair, but really they should have ran them BOTH at 1866 or 2133. so that the IGP is actually showing what i can do.

Its kind of like comparing a titan and a gtx 560 with both systems running a celeron. Its going to be a limitation before the GPUs hit their limits.

 

[Thefumigator]

I should clear it up a little bit, i understand they used 1600mhz ram for both systems to be fair, but really they should have ran them BOTH at 1866 or 2133. so that the IGP is actually showing what i can do.

Its kind of like comparing a titan and a gtx 560 with both systems running a celeron. Its going to be a limitation before the GPUs hit their limits.

I agree, but the radeon on 1600 is still better than the intel IGP, so I can imagine that at 2133 the radeon will improve like 10 to 20% its performance, so should do the intel IGP, but that's just an hypothetical argument, it would be nice to compare IGPs by its ram speed and see what happens to both.

 

[lordz]

Agreed, although seeing the intels rated to 1600mhz ram i dont know that there will be much improvment.
It appears amd still are king at IGP

 

[Cheeseball]

I'd be interested to know if they finally got 23.976 fps refresh rates right this time. That was one of the biggest roadblocks to using an Intel chip for a HTPC in previous generations.

This might be of interest to you and me.

EDIT: LOL, I just saw that you commented on that article. Heh.

 

[Casecutter]

This whole "gaming exercise" is just redeeming futility.... sure fun to know they can now match/beat the low-end card from 2 years ago. (I'll ask was that 6450 sporting DDR5 or DDR3?) It's just bizarre to explore the GPU performance of an unlock (84W) $340 CPU/IGP, for its' modern gaming merits! What amounts to Intel's benevolent offering of a GPU/IGP, just so you can run it without your discrete graphics card for either initial set-up or diagnose without it a card. Sure it make sense if this was an i3 box to see if your pre-teen might play something, but again these titles, it is a lesson in futility. While who today would buy a new OEM box and wouldn’t end-up pairing it with a 1920x1080 monitor, and while sure you could set the resolution lower, where’s the fun in that.

 

[Renald]

I see no point in this test ....

People who used these chips to play, don't put AA and games with a lot of shaders. They play CS or WoW without High graphics. Surely, ICP will fail on hard requirments, no need to test it on 20 games.

The question of these IGP are : "Can I play most of current games in medium settings ?" not "Am I going to reach 1235692 FPS on Tomb Raider 5760*1080 with TressFX activated ?".

Most of the test here are often useful and well-driven, but this one is totally useless. It totally misses the point of IGP.

 

[REDHOTIRON2004]

APU and GPCPU is the future not a discrete CPU/GPU!

I see no point in this test ....

People who used these chips to play, don't put AA and games with a lot of shaders. They play CS or WoW without High graphics. Surely, ICP will fail on hard requirments, no need to test it on 20 games.

The question of these IGP are : "Can I play most of current games in medium settings ?" not "Am I going to reach 1235692 FPS on Tomb Raider 5760*1080 with TressFX activated ?".

Most of the test here are often useful and well-driven, but this one is totally useless. It totally misses the point of IGP.

Looking at the future of computing. I will say that you are wrong about the IGP's or ALU's from Intel and AMD respectively.
I believe that with the miniaturization of the complete system. The GPU and CPU's would be one component that would handle both games and general purpose requirements in a heterogeneous way.
So, for example in case you want to upgrade your CPU or GPU. Then you would just upgrade to a APU which would deliver the same performance as a discrete level GPU presently.

Also, in the future APU are going to be cross-fired/SLI-ied to increase the overall performance 2twice or thrice.
Integrated graphics on CPU are not strong at present. But, with the introduction of heterogeneous computing(which is gaining momentum). Separate GPU's and CPU's would become obsolete!
And only the GPCPU/APU would be, what anyone else would look out for!

 

[pjl321]

Looking at the future of computing. I will say that you are wrong about the IGP's or ALU's from Intel and AMD respectively.
I believe that with the miniaturization of the complete system. The GPU and CPU's would be one component that would handle both games and general purpose requirements in a heterogeneous way.
So, for example in case you want to upgrade your CPU or GPU. Then you would just upgrade to a APU which would deliver the same performance as a discrete level GPU presently.

Also, in the future APU are going to be cross-fired/SLI-ied to increase the overall performance 2twice or thrice.
Integrated graphics on CPU are not strong at present. But, with the introduction of heterogeneous computing(which is gaining momentum). Separate GPU's and CPU's would become obsolete!
And only the GPCPU/APU would be, what anyone else would look out for!

I agree up to a point, you only have to look at some of the handbrake or Adobe numbers that are optimised for OpenCL to see the future of high performance computing is heterogeneous but they will never put a heavily overclocked 8-core x86 (+200w) plus a state of the art GPU (+300w) onto one tiny little die. Ivy Bridge and Haswell has shown us as we shrink transistors we create more heat problems than we solve, and that is at a poxy 84w! Ok so the problems don't start at 84w but you can't overclock any 22nm chip very much without massive heat density issues. This will only get worse as we shrink further.

 

[REDHOTIRON2004]

I agree up to a point, you only have to look at some of the handbrake or Adobe numbers that are optimised for OpenCL to see the future of high performance computing is heterogeneous but they will never put a heavily overclocked 8-core x86 (+200w) plus a state of the art GPU (+300w) onto one tiny little die. Ivy Bridge and Haswell has shown us as we shrink transistors we create more heat problems than we solve, and that is at a poxy 84w! Ok so the problems don't start at 84w but you can't overclock any 22nm chip very much without massive heat density issues. This will only get worse as we shrink further.

Just some days back, I was going through some articles that mentioned the kind of miniaturization that is being targeted by companies like Intel, IBM etc. And you might be surprised to know that Intel is already having prototypes for less than 10nm process. And we would start to see 14nm shortly within 6months from now. And 10nm(by using new materials like graphene, silicate etc capable of running in several Terahertz frequency instead of GHz as we have now) in 2014. They are targeting, to take it down to 2nm. And according to some findings/calculations even if these companies would reach 1nm process. Even then there would be a huge opportunity for reduction as the atoms are still much smaller than 1nm(probably quantum computing would kick-in .
And at 2nm the number of transistors that can be accommodated on these chips would theoretically increase the performance by 300times when compared to 32nm process.
So, a powerful CPGPU can easily be made at that level of miniaturization. And more power can be gained through SLI/cross-firing those CPGPU's.
The heat issue would be tackled by using more efficient materials such as the one I mentioned above at lower manufacturing process. And of-course we would see some more innovations during these years which would improve the architecture/efficiencies even more.

According to me, the GPU concept would vanish in a couple of years from now. As the new breed of much powerful APU's(with huge number of cores) would take care of both CPU and GPU(games) tasks with the capability for scaling simultaneously/parallel(as GPU) and sequentially(as CPU)!

 

[pjl321]

Just some days back, I was going through some articles that mentioned the kind of miniaturization that is being targeted by companies like Intel, IBM etc. And you might be surprised to know that Intel is already having prototypes for less than 10nm process. And we would start to see 14nm shortly within 6months from now. And 10nm(by using new materials like graphene, silicate etc capable of running in several Terahertz frequency instead of GHz as we have now) in 2014. They are targeting, to take it down to 2nm. And according to some findings/calculations even if these companies would reach 1nm process. Even then there would be a huge opportunity for reduction as the atoms are still much smaller than 1nm(probably quantum computing would kick-in .
And at 2nm the number of transistors that can be accommodated on these chips would theoretically increase the performance by 300times when compared to 32nm process.
So, a powerful CPGPU can easily be made at that level of miniaturization. And more power can be gained through SLI/cross-firing those CPGPU's.
The heat issue would be tackled by using more efficient materials such as the one I mentioned above at lower manufacturing process. And of-course we would see some more innovations during these years which would improve the architecture/efficiencies even more.

According to me, the GPU concept would vanish in a couple of years from now. As the new breed of much powerful APU's(with huge number of cores) would take care of both CPU and GPU(games) tasks with the capability for scaling simultaneously/parallel(as GPU) and sequentially(as CPU)!

I possibly didn't make my point clearly, I am not saying that in the future you won't fit the power of today's top CPUs and GPUs onto one small chip as I know the pace tech evolves. What I am saying is there will always be limiting factors to integration, most likely this will be heat but soon we will see things like quantum tunneling being an issue.

But any new tech that comes out will be generally available for all. A CPU company will already push it's chips to the max but at the say time a GPU company will always push its chips to the max also. You can't expect to combine two things at their limits without something giving, i.e. making a compromise. If you have two separates then there is no compromise, therefore a much faster system and therefore separates will never completely die.

P.S.

Don't get too excited about the next big thing as it almost never actually comes and when it does its always a massive let down. The Pentium 4 was meant to ship at speeds of over 10GHz, Sandy Bridge was going to be an octa-core and Hitachi is 3 years late on its pledge on 5tb hdd!

http://www.geek.com/chips/intel-predicts-10ghz-chips-by-2011-564808/

http://www.theregister.co.uk/2008/07/04/hitachi_5tb_hdd_2010/

 

[George_o/c]

Such a pitty they limited the GT3 and made it unavailable for desktops... Maybe they could offer some more competition at AMD. Thanks a lot for the review!