Intel Announces 'Iris' as Top Tier Graphics for Haswell

[ChillyWilly007]

I seriously doubt Intel's integrated graphics solution will hold a candle to AMD's solution.
If they claim just 2x the current iGPU, then it's crap.
Truth be told, Intel's chips are more powerful and expensive, but lack good iGPU, vs. AMD's less powerful, less expensive, but very capable APU.
Any serious gamer will still purchase a discrete graphics card.

 

[Wisecracker]

Go Intel.
Not sure why anyone would buy an i7 4770K for the graphics, though, and no one plays 3DMark.
I look forward to seeing some game benchies against the AMD APUs (with some fast RAMs).

 

[CrArC]

I remember being impressed with the output from the Core-i7 3615QM, which seemed to churn away at Bioshock 2 nicely. I mean it was still a bit crap, but more than I was expecting. So, pleased to hear about this alleged doubling of performance for future chips.

AMD's predicted response: "Intel is trying to make graphics which don't suck, lads. Hah, I know right? Just whack in some more Unified Shaders and roll out a new model, no rush or anything."

 

[sonofliberty08]

Because, Bill Campbell, some of us don't want to buy a graphics card if we do not have to. Some of us do not exactly need 1337 graphics because we aren't gaming. I personally used the HD graphics on my i7-3770K because i just need it to display documents for school on two screens. No need for cutting edge gpus, just enough to get it to show computations on the screen.

u don't even need i7 for that

 

[sonofliberty08]

But...can it run Crysis?
Seriously, can it? I'd like to know.

hmmm...... maybe can, but will lag big time, intel graphic is a joke

 

[InvalidError]

Why do they waste their time, and the end user's money? (R&D costs the end user in the end)

Because over 90% of people and businesses who buy PCs do not play intensive games or use intensive 3D apps so GPUs are a waste of money for their use if they can get an IGP that is just good enough with sufficient display output count.

Most devices that will integrate BGA -R models (netbooks, ultrabooks, laptops, all-in-ones, NUCs, pico-ITX, etc.) won't have a PCIe slot to put discrete graphics on anyhow.

 

[Ro0ster]

To be taken seriously, Intel is going to have to put their graphics solutions up against discreet cards. Until they can show their new archetecture equates to even a VC thats three gens old, I'll be using discreet solutions.

 

[sarinaide]

I remember being impressed with the output from the Core-i7 3615QM, which seemed to churn away at Bioshock 2 nicely. I mean it was still a bit crap, but more than I was expecting. So, pleased to hear about this alleged doubling of performance for future chips.

AMD's predicted response: "Intel is trying to make graphics which don't suck, lads. Hah, I know right? Just whack in some more Unified Shaders and roll out a new model, no rush or anything."

You think the brains trust at AMD are just nonchalant enough to be that naive in its process development. HSA and APU's AMD takes a lot more serious than what you are making it out to be, by the same token the APU is not just some fuddy duddy chip with a GPU fused on it. The APU represents the point where hardware and software interact to improve seemless performance, The APU represents the future and is evolving well, this is certainly not low end technology. The APU is quite contrary the most sophisticated chip out and it is testament to AMD that nobody can match it or hold a candle to it.

The real kicker is AMD have expressly termed Trinity an infant in its APU evolution and Kaveri will be testament to that.

 

[technoholic]

1) GT3/GT3e is going to be very expensive relative to a APU to achieve about the same theoretical performance as the top end HD7660D

2) 1300mhz core clocks on the iGPU to achieve performance, thats a near 50% clock bump.

3) Richland Desktop is around 20-40% faster than Trinity on the iGPU side on a mere 44mhz bump and mobile Richland has touted 50-60% graphics improvement. Trinity is over a year old but AMD timed Richland to nullify Haswells effect. On price to performance Richland will be much faster in gaming on integrated solutions.

4) Kaveri based on Steamroller core architecture along with unified memory space and on die GDDR5 will take AMD's APU's to a completely new level. This factoring in AMD's drive to HSA will also yield HPC gains. Early rumors on Kaveri's graphics performance is probably on or better than a HD7770 which will open the game significantly. With broadwell a no go it will be mid 2015 before intel can address it, and by then excavator is out.

Sorry but these look like good wishes to me and not the ultimate facts. I mean your thoughts are true for now but everyone who has some knowledge about CPU/GPU market is aware of the fact that Intel is a wild beast. They came to a point where their CPUs are superior in any way compared to AMD's and they caught a good GPU power in a very short time. With this speed, Intel will catch up and surpass AMD's iGPU advantage in a short time. Only a matter of time. Intel has many resources that AMD does not. Process advantage, own fabs, huge R&D resources, huge budget and brand superiority. AMD tries to close the huge gap by efforts in software side and by selling their inferior products for much cheaper. Unified memory, HSA and on die GDDR5 is all cool but: 1-) They are having to sell this products with little profit 2-) Intel will for sure answer these in a short time. They have the budget and the superior process advantage and they can almost put a refrigerator in their CPU if they wanted )) I wish AMD could surprise us though

 

[InvalidError]

To be taken seriously, Intel is going to have to put their graphics solutions up against discreet cards.

Most people and businesses who are interested in IGPs are not interested in paying $80 for a discrete GPU no matter how much better the discrete GPU might be because the IGP is already good enough for their needs.

As for "Intel having to put its IGP against discrete", they already compare GT3e to HD6670/GT640. There will be plenty of pitting everything against everything else to see whether or not Intel's claims are true after products become available on market next month.

 

[deksman]

Comparing Haswell to Richland is not a 'fair game'.
Richland is a mere 'refresh' of Trinity, while Haswell features an architectural change.
In the desktop space, Haswell should be directly compared to Steamroller, while in the mobile space, Kaveri will be based on (fixed/improved) Steamroller cores and feature a new iGP architecture (will be released later this year).
But if the 3dMark benchmarks I saw indicated anything, then Haswell's igp will only be 7% better than Richland igp.
Of course, it will still have an advantage in CPU performance, but the overall jump in clock per clock from Ivy Bridge to Haswell seems to be mere 10%.
Jumping from Sandy Bridge to Ivy Bridge was mere 5% improvement in cpu clock per clock.
Nehalem to SB was biggest 'improvement' from architecture, which resulted in roughly 20% clock per clock jump.
Overall, that's a 35% clock per clock from Nehalem to Haswell in CPU performance (which is not exactly impressive).
AMD on the CPU side, if it fixes its single-threaded performance (resulting in AMD's estimates of 30% to 40% gains in clock per clock), then it will be equally comparable to Intel offerings in CPU power alone, and the iGP is 'expected' to gain at least 30% in performance (if not more).
We can only wait and see how it will turn out though.

 

[sarinaide]

1) GT3/GT3e is going to be very expensive relative to a APU to achieve about the same theoretical performance as the top end HD7660D

2) 1300mhz core clocks on the iGPU to achieve performance, thats a near 50% clock bump.

3) Richland Desktop is around 20-40% faster than Trinity on the iGPU side on a mere 44mhz bump and mobile Richland has touted 50-60% graphics improvement. Trinity is over a year old but AMD timed Richland to nullify Haswells effect. On price to performance Richland will be much faster in gaming on integrated solutions.

4) Kaveri based on Steamroller core architecture along with unified memory space and on die GDDR5 will take AMD's APU's to a completely new level. This factoring in AMD's drive to HSA will also yield HPC gains. Early rumors on Kaveri's graphics performance is probably on or better than a HD7770 which will open the game significantly. With broadwell a no go it will be mid 2015 before intel can address it, and by then excavator is out.

Sorry but these look like good wishes to me and not the ultimate facts. I mean your thoughts are true for now but everyone who has some knowledge about CPU/GPU market is aware of the fact that Intel is a wild beast. They came to a point where their CPUs are superior in any way compared to AMD's and they caught a good GPU power in a very short time. With this speed, Intel will catch up and surpass AMD's iGPU advantage in a short time. Only a matter of time. Intel has many resources that AMD does not. Process advantage, own fabs, huge R&D resources, huge budget and brand superiority. AMD tries to close the huge gap by efforts in software side and by selling their inferior products for much cheaper. Unified memory, HSA and on die GDDR5 is all cool but: 1-) They are having to sell this products with little profit 2-) Intel will for sure answer these in a short time. They have the budget and the superior process advantage and they can almost put a refrigerator in their CPU if they wanted )) I wish AMD could surprise us though

Wild beast get hunted for sport, nothing about Haswell has been impressive and GT3 has been rumor for the best part of 16 months now with nothing concrete other than its "beastly" 1300mhz+ graphics clock. HD4600's prowess which was exposed is at times struggling against HD4000 iterations, there is nothing to suggest that this part will be the alpha code to graphics immortality.

What is fact, Intel does not produce a scalable graphics solution, AMD have Radeon cores which far outstrip the compute performance of any Intel Xeon or Extreme processor. The HD7970GE made legacy x86 look slow, I really don't see a Intel graphics solution within the next 5 years hitting that high water line, it is such that HSA is a pipe dream for Intel. HSA is progressive albeit still a good year away from anything substantial but already the few instances of pure HSA support see's low level Trinity parts out draging Intel top end parts. The point where software and hardware are in unison the results are substantial.

As before Richland comes out soon in Desktop and Mobile trim, the numbers are looking around 20-40% depending on the nature of the application, while mobile is over 50% faster than HD4000 mobile, which is what GT3 was touted at. I expect Intel to have parity with trinity but on a significantly higher cost and on much higher frequency albeit shortlived as Richland with its aging VLIW based core will still be the fastest iGPU solution.

 

[smokeybravo]

I don't see this surpassing the Trinity A10. Anyway, Richland is due to be released early summer, and Kaveri later this year. Intel will not be beating AMD in IGP performance any time soon.

 

[alfaalex101]

Slideshow ready technology!

 

[marshal11]

If one of those CPUs land in a notebook for around 600$, I'm totally sold. Right now my lappy is a i3 370m with INTEGRATED 1ST GEN GRAPHICS. It's terrifyingly bad.

 

[lordstormdragon]

Aaaan... no comparisons to AMD or Nvidia, here. 3x the graphics power of previous Intel HD graphics? That's still what, 1/100th of the weakest A-series? Pathetic.

 

[InvalidError]

3x the graphics power of previous Intel HD graphics? That's still what, 1/100th of the weakest A-series? Pathetic.

HD4000 was about half as fast as AMD's Trinity A10. GT3e is supposed to be ~3X as fast as HD4000 (2X for mobile variants) so that should make it about even with AMD's newest APUs if Intel's drivers are up to the task.

As far as comparisons go, there was a chart in earlier leaks this week that lined GT3e with HD6670 and GT640.

 

[lpedraja2002]

I've been waiting for "nice" graphics performance since I bought my first laptop almost 6 years ago, it's still working (surprisingly) I can't believe how much performance has improved over the years, I think in about 3 or 5 years we will be able to game with high settings on 720p displays using only the IGP on the CPU.

 

[InvalidError]

I think in about 3 or 5 years we will be able to game with high settings on 720p displays using only the IGP on the CPU.

From what people have seen with Intel's hands-on GT3e demo, GT3e can already handle some games with most details maxed out at 1080p.

Sounds like more than enough for whatever little gaming I might want to do on a laptop.

 

[Matsushima]

I only play Team Fortress 2. This new graphics engine will be able to max it out.

 

[sarinaide]

I think in about 3 or 5 years we will be able to game with high settings on 720p displays using only the IGP on the CPU.

From what people have seen with Intel's hands-on GT3e demo, GT3e can already handle some games with most details maxed out at 1080p.

Sounds like more than enough for whatever little gaming I might want to do on a laptop.

The metrics being used are simplified, software and games don't distinguish potential of a part in that manner. Which games was it supposedly maxing out at 1080P, because there are graphics cards like a 7770 and 650ti that struggle to do that. I think these slides are benchmarketing, the numbers are changing week by week and yet nothing concrete has been disclosed. Lets also bear in mind it will be competing against a 5 year old VLIW4 part and a 14 month old APU by that time. When it comes to graphics, Richland's performance is looking very impressive and I expect it to be the alpha part once the dust settles, I also have more faith in AMD producing graphics competence which is I am assuming down to "Radeon" technology, which just maybe is a little bit better than any graphics solution Intel will ever come up with.

Gt3 is being build up to fail, people are now getting expectations of GTX640, 650 level of performance which was about 180% faster than HD4000. The other issue is again costs involved and flexibility, it will be a BGA which leaves no expansion or upgradeablity. AMD's platforms are future proofed, upgradeable and have dual graphics for ancillory support, there are more SFF FM2 setups out now with full hosting of goodies.

 

[smeezekitty]

Only some of the Haswell CPUs will be BGA.

 

[Matsushima]

I think this is being targeted at the wrong market... An i7 will most likely have a graphics card paired with it. Intel would get a lot of sales if these were used on Pentiums or i3s.

 

[InvalidError]

Gt3 is being build up to fail, people are now getting expectations of GTX640, 650 level of performance which was about 180% faster than HD4000. The other issue is again costs involved and flexibility, it will be a BGA which leaves no expansion or upgradeablity.

It has been said for over a year that the top-end Haswell IGP will be ~300% as fast as HD4000 and the slides reiterate what has been rumored since the Haswell rumor mill started rolling. I do not see much of a reason to doubt Intel's ability to deliver that 3X HD4000 at least on the hardware side. Drivers could be a different tar ball.

As for BGA's lack of upgradability, I think that fear is grossly exaggerated. The vast majority of people these days will never upgrade a CPU in their lifetime and even if they could, PCs these days are good enough for 5+ years to most people so by the time they may want to upgrade, Intel's current chips by then would have changed sockets twice already, DDR5 will be just around the corner and just about every interface will have been refreshed with their next higher-speed iteration.

Many Sandy enthusiasts are choosing to skip Ivy and possibly Haswell/Broadwell too. If even enthusiasts are choosing to skip generations and sockets, what are the chances that a mainstream consumer who chose the right CPU for his foreseeable needs will ever bother even thinking about replacing a Haswell CPU before the platform is outdated and becomes effectively non-upgradable (not worth upgrading) anyhow?

I would still be using my C2D-E8400 if I could stuff 16GB RAM in it but 16GB DDR2 cost nearly as much as the i5 + h77 + 16GB DDR3 I ended up upgrading to instead, so upgrading the C2D made absolutely no sense.

Long story short: as long as people buy the right CPU for their needs up-front, it will last them long enough that it won't be worth upgrading by the time they may actually need to do so, rendering socket vs BGA moot for most of the intended markets.

 

[sarinaide]

I was hearing for a long time around 2x the performance, which makes sense as its about double the resources and 50% higher clocks so yeah its achievable. Now its 3x, tomorrow its 14x, then its matching a GTX680. I don't question intels process, but I question what they call a graphics component, I also question the leak for the slides the source is not given and unlike coolaler whom often land the money shot it seems to be photoshoped to appear to the market centric community. Obviously telling people the GT3 will smash a relatively low spec'd Trinity is the same old benchmarketing of old. If it fails Intel won't worry as they have the sale. This goes completely against what the constituant reviewers where projecting.