News about next Iris Pro(GT4) or Lap/Mob/Desk Broadwell?

[Fjodor2001]

@witeken: I meant the four CPU generations prior to Skylake. I.e. SB, IB, HW and BW. SB->BW (desktop) will be January 9 2011 to 2015Q1, for three CPU generations jumps. Assuming BW's Q1 means February (mid Q1) we get (4*12 + 1) / 3 = 16.3 months between CPU generations on average.

 

[witeken]

Intel started volume production of 14nm in Q1, so why would it take (almost) a full year to release them?

 

[Fjodor2001]

Well that's the latest news I've heard at least. Some limited Broadwell U and Y models will be available in 2014Q4, the rest including desktop in 2015Q1 or later. Maybe they're still working on perfecting the yield and are prioritizing U and Y models when supply is limited, or they want to clear the sales channels of old Haswell and Ivy Bridge inventory first. Who knows.

 

[zlatan]

How does it really, officially, improve upon the old Gen?

The architecture is Gen8 and it has many improvements. It will get an MMU which will lead to better D3D12 performance, and also might support tiled resources, but it is hard to say now which tier.

Will it, officially, have more cores? How many more?

It will have more EUs compared to Haswell, but I don't know how many. At the prototype stage they usually send limited products.

Will the L4 be higher? (Since the 128MB didn't really show that much improvement, that I remember. That I remember!)

No. It will get the same eDRAM chip as the Haswell family. Actually it doesn't really need more memory. Even 32 MB is enough.

What API's will it support?

D3D12 support with tier1 tiled resurces and feature-level 11.1 is logical from this architecture.
It will also support OpenCL 2.0 with shader virtual memory, and nested parallelism. But it won't support C1x atomics and pipes by the hardware, so AMD Kaveri is more advanced at this point. But shared virtual memory is far the most important feature, and this will be a huge win for the PC.
I think AMD Kaveri and Intel Broadwell can execute the next gen console ports without any limitations.

Will they bring the highest-end iGPU to a desktop socket?(Or will there be a GT4e on Mobile/BGA?)

They have plans for this, but nothing official right now.

 

[MisterLilBig]

If I remember correctly Intel compares the jump of Gen7.5 to Gen8 as a bigger jump than a past Gen, 3 to 4 or 4 to 5, not sure.

How big was that jump in the past really? By how much did real world performance improve?

 

[witeken]

Quote:


In general terms, Intel’s Ben Widawsky, who works on Intel’s Linux graphics driver efforts, says that “Broadwell graphics bring some of the biggest changes we’ve seen on the execution and memory management side of the GPU… [the changes] dwarf any other silicon iteration during my tenure, and certainly can compete with the likes of the gen3->gen4 changes.”

By generation three and four, Widawsky is referring to Intel’s GMA GPUs (circa 2005-2006), which underwent a huge reworking between the 3000 series and X3000 series. While generation three GPUs only accelerated a few features in hardware (i.e. it leaned heavily on the CPU for support), generation four GPUs were much closer to the modern idea of a GPU, with lots of fixed function units and a handful of programmable execution units.

 

[MisterLilBig]

But the real world performance increase of such changes were big?

From 3000s to X3000 the bandwidth saw a 20% increase. EU's were the same. Dynamic memory increased by 50%. (And, all the Arch. changes.)

Didn't really find any performance comparison info in my quick search.

 

[Arachnotronic]

Intel started volume production of 14nm in Q1, so why would it take (almost) a full year to release them?

If they began volume production in Q1, that means they are building the Broadwells now. However, they have not qualified those parts for sale (which will happen by the end of Q2). Once those products are qualified for sale, Intel then needs to ship them to the OEMs which in turn need to build enough systems to support a launch (let's call it 3 months to do that).

And that's why Broadwell systems show up in 4Q.

 

[mikk]

GT4 is coming for Skylake by the looks of it.

http://chinese.vr-zone.com/111256/w...el-skylake-platform-confirm-in-2015-05042014/

 

[IntelUser2000]

But the real world performance increase of such changes were big?

From 3000s to X3000 the bandwidth saw a 20% increase. EU's were the same. Dynamic memory increased by 50%. (And, all the Arch. changes.)

Didn't really find any performance comparison info in my quick search.

I would say that engineer is giving a really bad example, considering how widely varied that part(X3000) was in performance depending on games.

Depending on the game, it could have next-generation iGPU performance, or it might perform at 1/3 to 1/2 the previous generation iGPU. It had a terrible bottleneck that wasn't fixed until Gen "5.75" Ironlake 4 years later.

Now maybe he wasn't alluding to that. But nothing seems to indicate that the change is big as moving from GMA 950 to GMA X3000. I mean, they moved from fixed function units to programmable!