Intel Skylake 15W processors to include Iris 6100 graphics

[HEXUS]

Thin and light devices with Skylake processors were on show at IDF Shenzhen.

Read more.

[yeeeeman]

Man, they are crazy about the thinness of laptops. They probably want to make them like a postage stamp or something. To run Windows 10 on a nail. We want reliability, acceptable performance and battery life. Huge battery life. A week, not 12 hours! And because they can't acomplish that, the focus is on reducing size, to give the customers a "reason" to upgrade.
Well, I'll stick with my "heavy" HP6930p with can last as long as a current thin laptop on battery with acceptable performance.

[mudpie]

if Skylake Desktop cpus do not offer at least 20% improvement over ivy bridge, I'm not upgrading.

[shaithis]

if Skylake Desktop cpus do not offer at least 20% improvement over ivy bridge, I'm not upgrading.

TBH, I wouldn't even bother upgrading for 20%. In 90% of situations your never going to notice a difference.

My IVB chip only gets maxed in the rare occasion I do some transcoding. It got maxed regularly in BF4 until they nerfed a load of the physics (about 9 months ago). Otherwise it's still a tad overkill even for a power user.

[Ryrynz]

if Skylake Desktop cpus do not offer at least 20% improvement over ivy bridge, I'm not upgrading.

Then I can tell you right now you're already not upgrading. the CPU will be on average about 10% the GPU will be quite a bit faster though likely above 20%, remember though the 6100 is going to be limited to 15W combined with the CPU so even though it's the same part, it likely won't achieve the same performance as the 28W Haswell.

You know it doesn't really cost you that much to upgrade, (providing you pick up the new one at a good price) either you do it now or later on.. Either way depreciation is going to affect your current part.. so really it's just a question of when do you want something new? If GPU and CPU performance is fine on your current system then I wouldn't bother.. pop an SSD in there instead (I'm assuming it's a mobile part)

[anselhelm]

Though Skylake will doubtlessly be great for mobile computers and for new builds/new PCs, it doesn't currently really excite me.

In terms of tech I'm actually actively looking forward to, I really have to say I'm more interested in seeing where AMD ends up in late 2016 / early 2017 with its new architecture, Zen.

Why am I so interested in it? Many reasons, but one in particular: Jim Keller (of K7 "Athlon" and K8 "Athlon 64" fame) is the lead architect on the project. Furthermore, we're looking at AMD closing the gap on Intel by moving to a 14nm process and hopefully finally abandonning CMT (clustered multi-threading) in favour of SMT (simultaneous multi-threading).

Don't get me wrong: I've few issues with my FX 8320 as I do a lot of video encoding and have yet to find issues with gaming on it. That said, it'd be nice to be able to migrate my system to something new around Q4 2016 / Q1 2017, if only to keep relatively up-to-date with technology.

[koperta]

I can't not agree with Yeeeeman above. Everyone is driving to achieve unnecessary thinness forgetting about reliability, lasting power supply and comfort of use.

[2001]

OK, Intel... Let's see a soldered unlocked enthusiast CPU with no GPU at all. And, we don't need thousands of cores. Just make a quad without hyperthreading. Replace the stupid stuttery Anniversary Pentium with it. Four very fast pure core threads should be enough, with enough clockspeed.

[2001]

hopefully finally abandonning CMT (clustered multi-threading) in favour of SMT (simultaneous multi-threading).

It doesn't seem totally clear that CMT is inferior to SMT.

[DanceswithUnix]

OK, Intel... Let's see a soldered unlocked enthusiast CPU with no GPU at all. And, we don't need thousands of cores. Just make a quad without hyperthreading. Replace the stupid stuttery Anniversary Pentium with it. Four very fast pure core threads should be enough, with enough clockspeed.

Unfortunately there is a problem with that. A design like a CPU needs a lot of pads to get PCIe, memory etc connectivity to the outside world and those pads can't be too small. So once you have a rectangle of pads you work out how to put the bits inside. If you look at a die shot of a modern CPU, you need to find *something* to put in that rectangle in the middle. Using it for L3 cache may help in some cases, but at 14nm even Intel's huge cores start to look a bit lost hence I think they are padding things out with graphics.

Just take a look at a small die:



I doubt they can make it much smaller than that, so if you lose the graphics part then what are you going to put in that empty area? I think 6 cores and bigger L3 cache is the smallest they could do these days, but then it wouldn't work in laptops and for small servers would require external graphics so actually quite limiting in use.

It doesn't seem totally clear that CMT is inferior to SMT.

I think it is clear that in a single threaded case SMT can give that thread a stupid amount of hardware to run on and that helps it go faster. CMT is only better if you can give it enough threads and keep it fed with data, and that is proving hard. Still, as software is becoming increasingly threaded moving to an old single thread biased system seems a shame rather than ironing the kinks out (after a couple of years owning an FX8350 it doesn't feel like it has aged, almost the opposite like the software is catching up). Plus, first gen SMT is likely to have wrinkles for AMD.

[shaithis]

People just want that silicon USED or they don't want the cost associated with half the chip that they aren't using.

Technically, its a breeze for Intel. Financially, they won't do it because it will eat into workstation parts. Hell, looking at the map you could have 6 cores and a small gfx core, ideal for workstations and a lot of servers.

[DanceswithUnix]

People just want that silicon USED or they don't want the cost associated with half the chip that they aren't using.

That's the problem though with being pad limited. If you have ~1150 connections to go to the outside world, then you need that many pads. That sets the circumference of the chip, and hence the area inside that has to be filled. If you want a smaller chip, you have to remove pins because you are paying for the area inside those pads whether you make use of it or not.

[shaithis]

That's the problem though with being pad limited. If you have ~1150 connections to go to the outside world, then you need that many pads. That sets the circumference of the chip, and hence the area inside that has to be filled. If you want a smaller chip, you have to remove pins because you are paying for the area inside those pads whether you make use of it or not.

We know the technicalities behind it but that doesn't stop people feeling cheated by Intels design decisions though.

Theoretically they could shrink it by having all the GPU pins be around the edge of the socket. They could have created Sockets 1150A and 1150B. They could have had dual-purpose pins with active switching based on CPU inserted. They could have only released GPU-embedded chips on the mobile sockets only.

But of course all those ideas would have eaten into their profit margin which isn't worthwhile because they have no competition to really worry about

I know over the years we have always subsidised silicon via the binning process but now we seem to be subsidising laptop users.

[mudpie]

if Skylake Desktop cpus do not offer at least 20% improvement over ivy bridge, I'm not upgrading.

Then I can tell you right now you're already not upgrading. the CPU will be on average about 10% the GPU will be quite a bit faster though likely above 20%, remember though the 6100 is going to be limited to 15W combined with the CPU so even though it's the same part, it likely won't achieve the same performance as the 28W Haswell.

You know it doesn't really cost you that much to upgrade, (providing you pick up the new one at a good price) either you do it now or later on.. Either way depreciation is going to affect your current part.. so really it's just a question of when do you want something new? If GPU and CPU performance is fine on your current system then I wouldn't bother.. pop an SSD in there instead (I'm assuming it's a mobile part)

It's a desktop part, I just like desktops, and I do streams occasionally. Well, if Skylake is not at least 20% better, I'll try and get a new Z68 MB, although there seem to be hard to find, prolly more so later this year. My current MB seems to have some issues. Got SSDs too,2 of them, very nice improvement over HDD.
Edit:
I just want the upgrade cost to justify itself :/ is all.

[DanceswithUnix]

I know over the years we have always subsidised silicon via the binning process but now we seem to be subsidising laptop users.

I think it is worse than that, a lot of laptop users want discrete graphics as well.

So yes, I would rather see more cache and cores, less graphics. A little is OK, I have use cases for that.