Intel’s “Iris” wants to change how you feel about integrated graphics | Hacker News Reader

Intel’s “Iris” wants to change how you feel about integrated graphics

arstechnica.com

45 pointspedrocr43 comments

Support OpenCL on something other than Windows and we can talk about how I feel. For now I feel I've wasted a lot of money on expensive silicon die area that I can't put to use.

Edit: I mean OpenCL that uses the GPU. The last I looked Intel had OpenCL for Linux, but it only used the CPU. Its a big world, someone may even have a use for that.

Good news friend. This[1] was announced a few weeks ago.

[1] http://cgit.freedesktop.org/beignet/tree/README.md

…you may expect serious problems and GPU hangs.

Only IVB is supported right now. Actually, the code was only run on IVB GT2. You may expect some issues with IVB GT1.

The run-time is far from being complete.

In the future there may be good news.

Intel have had support for OpenCL on Linux for a while now. For example this page, which describes how to install it for Ubuntu, dates from May 2012:

http://intel-opencl-sdk-for-ubuntu.blogspot.co.uk/

The official download page with links for both Windows & Linux versions is here:

http://software.intel.com/en-us/vcsource/tools/opencl

Anyone else think that "Iris" is a fairly obvious reference to "retina" displays, indicating that Intel expects devices which come with Haswell/Iris to include retina-standard displays?

(It has been years since I've commented on a hardware thread, I had forgotten how much fun this kind of codename Kremlinology can be)

We can assume the new MacBooks will come with Haswell. Perhaps the Iris graphics will be driving the Retina display? Right now the Retina MacBook comes with Ivy Bridge and a separate GPU although I don't know if it uses integrated graphics to drive the display.

rMBPs have switchable graphics and will use both the IVB GPU and the nVidia GPU to drive the display. The IVB GPU is noticeably choppy on some types of scrolling when you've got it set to 1680x1050 equivalent or higher, so the Haswell improvements will be welcome for newer buyers.

The 13" doesn't have switchable graphics though, right? That's what has stopped me from upgrading so far. Hoping that June brings good news.

Right, 15" only.

The 13" Retina MBP uses integrated graphics only.

Optimistic news sites and forum posters seem to expect that the Air will get Retina treatment this year. That seems extremely unlikely to me. Even it if were cost/tech viable, it would significantly erode the differentiation between the Air and Pro lines.

Is a safe bet for both the new 11" and 13" Airs to get the non-Iris GT3 part? Or is the 11" in weedy GT2 territory?

Intel seems to be applying the ^TM to Iris, so I'd think it's intended to be a product name rather than a code name. Intel is so obsessive with trademarks that it seems likely that they acquired the trademark from the ruins of Silicon Graphics' line of graphics workstations.

Intel usually chooses seemingly-random geographic features for their code names. Coordinating the two deals (Apple and the Iris trademark) for the secret pleasure of having a subtle code name seems uncharacteristically subtle.

They're not seemingly random geographical features: they're usually geographical features associated with the Intel lab that did most of the lifecycle of development. With tick-tock, they have a new chip each year. I believe every chip development takes 6 years, so they have three labs concurrently working on things at various stages.

Yonah (Core Solo and Core Duo) - Intel Israel

Tejas (planned Pentium 4 successor; cancelled) - Intel Bangalore

Willamette (Pentium 4) - Intel Oregon

Nehalem - Nehalem River in Oregon

Here's a whole bunch of etymologies http://en.wikipedia.org/wiki/List_of_Intel_codenames

OK, so Intel codenames are seemingly random geographic features usually within 1000 miles of the design team. But the point is the codenames show no evidence of coordination with the features of the chip itself.

Someone needs to start selling "Sclera" computer cases.

Don't you think that's kind of cornea?

Showing my age here, but when I think "Iris" I think of SGI, not Apple.

https://upload.wikimedia.org/wikipedia/commons/thumb/3/3f/SG...

I guess we can't expect a benchmarking yet but it feels like an intentional oversight to mention AMD once and only in context of standalone graphics chips.

According to Anandtech, performance is on par with GTX 650m which is really impressive considering intel's past stance on integrated graphic. AMD APU's graphic advantage's been eaten by intel bit by bit, not a good news.

You saw that AMD announced hUMA? So AMD APU's will have the advantage again if they land about the same time.

Well for a start, Intel would have to make an IGP that supports dual-link (resolutions over 2048x1536)

Because as of right now, none of them can

What is IGP?

For example HD4000 can drive a 2550x1440 display, but it doesn't work over DVI, only DisplayPort on the integrated mobos. I've never understood if that's related to a mobo limitation (that is just present on all Sandy Bridge mobos) or if the CPU can't pinout to Dual Link DVI but can to DisplayPort?

Computing wise DisplayPort looks a lot like an ethernet. DVI/HDMI has all kind of weird TV spec stuff that a computer display shouldn't really need to deal with.

The graphics output from the CPU is sent over Intel's own FDI (Flexible Display Interface), so the limitation lies in the PCH (chipset) that converts FDI to the various display interfaces.

I looked at one datasheet [0] for the 7-series chipsets and it didn't mention Dual Link DVI at all (although DVI was obviously mentioned). It did say "DisplayPort/HDMI/DVI 2560x1600 at 60Hz" in one of the tables, but your guess is as good as mine if that implies support for Dual Link DVI or not.

Edit: added link to referenced datasheet. See chapter 5.28.

[0] http://www.intel.com/content/dam/www/public/us/en/documents/...

Haswell's iGPU will support 4K resolutions.

http://www.anandtech.com/show/6926/intel-iris-iris-pro-graph...

I'm presuming this will lead to cheaper laptops :) not to mention higher res panels ably powered by the on board GPU

This announcement doesn't really seem targeted toward cheaper laptops - the parts themselves will likely cost just as much as their previous-generation counterparts, with the addition of a few "premium" options for people who want more than the standard internal graphics but less than a discrete GPU.

The market fragmentation issue is important, and unfortunately it's not clear that there is a fragment that is specifically targeted toward the budget end of the market.

I think it's fair to assume that perf/$ will increase, and that will lead to cheaper devices with acceptable perf.

It's AMD's fusion chips that have changed my feelings about integrated graphics. Intel GPUs? Not so much.

I have an cheap Nvidia graphics card from 2009.

2009.

I haven't met anything it can't run even if it isn't the fastest.

Intel's integrated garbage can't keep up. And it's not just the speed either, crap drivers, screen artifacts, and in 3d (what it's for), texture issues, lighting issues, shader problems, etc.

Without even knowing what's under the hood of a particular machine, you can almost feel the crappiness of the intel integrated graphics chip, haphazardly trying to present a workable frame of some game you're playing.

I have no faith that these parts will match up.

edit here's a demo https://www.youtube.com/watch?v=Otcge1cn8Os

It's all fun in games until the nvidia's binary blob locks up. This happens to me monthly on linux. In terms of stability, Intel's much better than nvidia based on what I've seen.

On Linux. On anything else, Nvidia have the best drivers.

Even on linux, I've found nvidia's drivers the best by far.

It seemed to me that the open source ones tend to be the most stable, regardless of vendor.

Shrug; again, very much not my experience over the course of several nvidia (and a couple of ati/amd and intel) cards.

Honestly, with the creation of retina displays and so on, games are the least of concern here. The 13" Retina MBP was criticised heavily for its everyday graphics performance- if an internal chipset can deliver my retina desktop without slowing down, I'll be happy.

Publish an implementable spec for high-performance opensource drivers and I'll change how I feel.

As they have been for most of the last decade. Here's hoping this time they do. :)

They already came a long way. I recently got an i5 with a HD4000 and was blown away by the fact that it runs Mass Effect 3 smoothly at 1600 x 900. Having good hopes for the Haswells.

looking at the power consumption the performance is incredibly good.

Nvidia should take a leaf from their book, then again I guess Inte's got a power consumption edge on nvidia in terms of die size

Can't wait to get a Haswell macbook!

I wonder how this will filter down to Intel based mobile phones?

Almost surely still going to be utter shit compared to a decent GPU like a Radeon HD 7970 or GeForce GTX 680.

Have you seen the size of those cards? I have a GTX 670 and it is a foot long and uses close to 200 watts of power. You cant compare high end graphics cards running on that much power and size which are costing over $400 for an integrated GPU

Probably the GTX650 and HD 7750 would be better to look at, which are now under $100 and don't pull that much power.

Whenever I look at them and they don't have enough discrete memory, I just wonder if they can really run fast enough due to that single bottleneck. If the onchip graphics had 512m or 1G of memory, they'd probably perform well, but then you lose the power efficiency. I read something that intel was possibly doing some new tech for onchip memory to get lower power, so it's possible it'll happen.

It's probably still in the future, though.

Reply on news.ycombinator.com