Broadwell has amazing iGPU performance. Intel afraid of Zen+HBM APUs?

We probably all have realized by now that Intel has been largely stagnant over the last few years due to the lack of competition from AMD. They have still been making progress with each iterations, but none have really been anything noteworthy until now. My guess as to what happened is that Broadwell is Intel's response to the rumor of Zen APUs with HBM. If Intel didn't show much progress on the iGPU front, then everyone would keep saying that AMD is going to slaughter their iGPU with their APUs. This makes things much more interesting. However, the really interesting bit is that in the benchmarks shown in the link below, the Iris Pro 6200 has performance very similar to an r7 250. The reason that this is very interesting to me is that both the 7850k and the r7 250 have the same amount of stream processors and roughly the same clock speeds (725 vs ~800 in favor of the 250 obviously), and the tests shown there are with a 860k paired with the 250 which is basically the same cpu as what the 7850k has. So what we are seeing here is that 6200 performance around the same as what the 7850k would perform if it didn't have anything holding it back. The likely bottleneck here is the memory. DDR3 is just not good for these tasks. So if the 7850k had HBM, it would perform roughly the same as the Broadwell chips in-game.

This gives me hope for the APUs in the future. If Zen is everything that we expect it to be, and they do effectively implement HBM with their APUs, then we should see some serious integrated graphics performance. I expect that the APUs will still be ahead in-game whenever they release, compared to Broadwell and likely Skylake as well. However, the superior cpu performance of Intel's parts makes this a very interesting upcoming battle. Is the trade off worth it? Can AMD get their CPU performance up enough to make the trade off between iGPU and CPU swing in their favor? AMD is finally marketing to the "casual" gaming market (MOBAs and the like), and it looks like Intel might just take that market away frmo them if they don't step up. Zen needs to be a huge step forward, and it needs to come fast because right now, Intel is taking away every possible nook and cranny that AMD could hide itself in. The tech market place is likely to be an interesting place for the next few years, to say the least.,4169-6.html

we always knew that 128mb of esram kicked arse on crystalwell, it did even on the previous gen... its just the cost was always too high.

Double the cost for 20% more igpu performance doesnt really seem worth it.

I think hw wise amd gcn is still way more powerfull hbm will only be viable on apu's depending upon its cost / performance ratio.

The only company that should be concerned about the increases in integrated graphics (by both amd and intel) are nvidia

For the mainstream market, integrated graphics will soon become more than enough. However, the main drawback that I see with it is that graphics performance needed increases dramatically every few years, while the cpu requirements are teh same (or likely to decrease with better apis like DX12 and Vulkan). Because of that, it is a good thing to have the gpu and cpu separate so that you can swap out the gpu every few years while keeping the same cpu. With integrated graphics, that isn't feasible. Especially considering that the cost of a new cpu with better igpu is twice as much as the price of a gpu with the same performance. What I mean here is that igpus will become a limiting factor after a few years and will either need to be replaced with a new apu/cpu or supplemented with a new gpu. Either way though, the increase in availability of decent graphics is going to be good for the mainstream MOBA-esque market, letting them increase the visual fidelity in coming years.

But yeah, the increase in performance is likely largely resultant from the eDRAM which is nothing compared to HBM, and Intel's graphics units are likely nothing compared to GCN, so yeah, the upcoming Zen APUs should be real killers.