Wednesday, October 22nd 2014
Intel Core i7 "Broadwell-E" HEDT Chips Arrive in 2016
Intel is beginning to put out the first details of its next high-end desktop (HEDT) processors, internally. Codenamed "Broadwell-E," the company's next Core i7 HEDT chips will be built in the existing LGA2011v3 package, and will be compatible with existing motherboards based on Intel's X99 Express chipset (with BIOS updates). Much like "Ivy Bridge-E" was to "Sandy Bridge-E," these chips will introduce only incremental updates, and nothing major, in terms of architecture.
To begin with, Core i7 "Broadwell-E" will be built in the 14 nanometer silicon fab process, and will feature 6 to 8 cores based on the "Broadwell" micro-architecture. These cores will be cushioned with up to 20 MB of L3 cache. The chip is pin-compatible to "Haswell-E," and so its I/O will be identical, featuring a quad-channel DDR4 integrated memory controller. One difference is that Intel may can the 28-lane PCIe approach with the entry-level part; or at least it doesn't find mention on the slide. If it's true, all parts based on this silicon, will feature 40-lane PCIe interfaces. The TDP of these chips will be rated at 140W. Intel is expected to launch the Core i7 "Broadwell-E" in 2016.
Source:
VR-Zone
To begin with, Core i7 "Broadwell-E" will be built in the 14 nanometer silicon fab process, and will feature 6 to 8 cores based on the "Broadwell" micro-architecture. These cores will be cushioned with up to 20 MB of L3 cache. The chip is pin-compatible to "Haswell-E," and so its I/O will be identical, featuring a quad-channel DDR4 integrated memory controller. One difference is that Intel may can the 28-lane PCIe approach with the entry-level part; or at least it doesn't find mention on the slide. If it's true, all parts based on this silicon, will feature 40-lane PCIe interfaces. The TDP of these chips will be rated at 140W. Intel is expected to launch the Core i7 "Broadwell-E" in 2016.
59 Comments on Intel Core i7 "Broadwell-E" HEDT Chips Arrive in 2016
The jig is up. Shrinking and x86 are done.
I just need a new GPU and more RAM
Large-scale deployments benefit purely from the performance increase and power savings, which together, make the upgrade to faster chips a savings, rather than an expense.
Haswell-based designs also make tablets and such much more attractive performance-wise, from that added power savings, too.
These two cpu's are aimed at different markets. I think the extra heat(tdp) is justified.
Just like I don't need my smartphone, I don't need broadwell-e. None the less, I will continue to drool over it and would be a happy owner of one. I would also put it to 100% use 24/7 for WCG!
With what you are saying, the clock speed an dperforamnce is equal, and that's not likely. There's also this wonderful thing called cache size and speed that makes AMD look so poor with the low TDP/performance ratio.
That will be top platform 15 months at least and than Broadwell E will support X99 chipset and 2011-3.
I would like to see Rampage 5 Extreme Full Black immediately before New Year at least some pictures.
This paradigm works both ways; it's not an anti-AMD bias. You forget that Intel's single core Pentium 4 Prescott processors even today still have a reputation for being hot and power hungry, yet they were only 95W-130W TDP CPUs. Modern HEDT processors consume more power than the Prescott Pentium 4 and yet very few people complain about power consumption of these HEDT CPUs because the performance is there to match.
6-8, 10-12, and 14-18.
The 5960x could be a fully enabled 8-core, or a cut-down 12 core. The lower core count CPUs are a way of managing yields. For the same reason a 12-core Xeon could be a fully enabled part, or a cut-down of the 18 core die.
If they were good and smart, they would launch software capable to fully utilise the 18-core CPUs, to flood the market with them, and we, to be happy already with gaming at Ultra HD resolutions and beyond...
But hey no, it is a matter of fact of life to share the same planet with bastards.
That said, Intel also knows that different die configurations have different performance characteristics and has shown no evidence that they sell different die configurations under the same model number. They'll always pick the smallest die possible even if yields are low because once yields improve, you would have to either release a new SKU or gimp chips that would otherwise be fully functional.
OK 1000W example in gaming is not normal but I talk about power consumption of premium Radeon and GeForce graphics during last few years, I never complain.
It's bad when AMD give half performance of Intel Performance in every day situations for double more power consumption and offer similar performance only if some application use all cores and if AMD CPU is overclocked and Intel not. I think overclocked AMD CPU need more power than GTX970 SLI... That's not normal situation because single core is almost half of Intel.
He waste power and today on earth that is something worse but they have special treatment because anti monopoly and nobody want situation where Intel no competition.
It would be interesting to someone compare power need for GTX980 SLI reference in gaming vs FX-8350 on 4.8-5.0GHz in Prime95 or FX-9590 overclocked.
Calm the duck down.
AM3+ CPUs also don't have the PCI-E root complex on the CPU, that's more power. SB-E has 4 memory channels where AMD has 2, that's more space and power. AMD processors are also made on an SOI process which has traditionally produced more leakage than Intel's HKMG process for any given manufacturing node.
With respect to transistors:
Likewise, a 140 watt TDP isn't unreasonable. Neither AMD nor Intel can fight the laws of physics, so jamming more components into a design will invariably increase thermal dissipation needs. As nodes get smaller and smaller this is compounded, because more switching occurs over a smaller area. Yes, voltages may decrease with shrinks. On the same token, increasing component count more than compensates for that reduction.
As far as AMD, they've announced that they aren't competing with Intel for the HEDT market. While they might have a position to defend, they've publicly acquiesced to Intel. Their best silicon is dedicated to APUs, so we haven't seen anything genuinely good come from AMD in years. Thuban was the last real competitor, and it only existed because it could OC to the moon and back. Releasing an 8 thread processor at 125 watt TSP and a 12 thread processor at 140 watt TDP isn't really unreasonable. With Intel throwing everything onto the CPU that's very reasonable heat generation.
In short, Broadwell-e is treading water. It's a minor incremental improvement, and changing the low end offering to something with all the PCI-e complex unlocked makes no sense. Intel would have to juggle the top end offerings to compete with it, and the high end consumer level offering would have huge competition. If the H/P/Z based PCH offerings don't have DDR4 pricing to differentiate themselves Intel will be removing the high-end sales there for low end enthusiast boards. I highly doubt Intel is that stupid.
At any rate, Broadwell-e is shaping up to be a joke. Like IB-e, there's no compelling reason to pay for a minor upgrade. Like SB-e to Haswell-e, the performance increase isn't likely to be justified by the large initial cost. Broadwell-e isn't making me want to give up on SB-e, despite any shortcoming I've experienced. Minor performance gains in multiple generations means no compelling reason to give Intel any of my money.