Intel Ivy Bridge-E slated for September 2013 launch

Jesse

Posts: 358   +42

The launch window for Intel’s Ivy Bridge-E processor line has been narrowed down to September 2013. Information previously available targeted Q3, but VR-Zone has revealed a leaked slide from Intel giving a more precise date for the new HEDT (high-end desktop) line of processors, the i7-4820K, i7-4930K, and i7-4960X.

Ivy Bridge-E CPUs are based on a 22nm lithography and are a die shrink of their 32nm Sandy Bridge-E predecessors. Current SB-E owners will be happy to hear that these new CPUs use the same LGA 2011 socket.

intel ivy bridge-e hedt

Comparisons between the Core i7-4960X and the Core i7-3970X show the new generation performing between 5-10 percent faster across the board in benchmarks. Performance gains are purely from improvements in efficiency, as the number of cores has not increased.

VR-Zone reports that the top tier i7-4960X actually contains 20MB of L3 cache and eight cores, but has one-fourth of its resources disabled for the consumer version of the product. A full eight core/20MB cache processor will be sold under the Xeon brand, but will undoubtedly go for a much higher price than $999. 

If Intel maintains the same pricing structure as SB-E, the three units will sell for around $1000, $600, and $300-350 respectively. The feature list for the new lineup includes 40 PCI-Express lanes (Gen 3.0), DDR3-1866 memory support, Hyper Threading, and Turbo Boost 2.0. All three processors come unlocked out of the box.

Intel Ivy Bridge-E HEDT Lineup

CPU Cores/Threads Base Clock Turbo Clock L3 Cache TDP
Core i7-4960X 6/12 3.6 GHz 4.0 GHz 15 MB 130W
Core i7-4930K 6/12 3.4 GHz 3.9 GHz 12 MB 130W
Core i7-4820K 4/8 3.7 GHz 3.9 GHz 10 MB 130W

Along with information about Ivy Bridge-E comes a few leaked details about the Haswell-E HEDT platform, codenamed “Lituya Bay.” Haswell-E is slated for the first half of 2015, will be based on the namesake 22nm architecture, and will again remain compatible with LGA 2011 boards. However, a new chipset will be implemented in motherboards to allow for native DDR4 memory support.

intel ivy bridge-e hedt

Permalink to story.

 
Well, since AMD is pretty much no longer competing, Intel can go back to business as usual. That is, a refresh of the same old chip with 5-10% performance optimizations every 6 months or so, at a premium price! Don't expect any huge performance jumps from Intel in the foreseeable future. They have no competition now to put the squeeze on them. Intel has its old monopoly of the x86 performance market back and they have no reason to jump ahead of themselves. It makes a lot more sense to "trickle" these small improvements to the consumers and keep charging them a premium price each time.
 
Well, since AMD is pretty much no longer competing, Intel can go back to business as usual. That is, a refresh of the same old chip with 5-10% performance optimizations every 6 months or so, at a premium price! Don't expect any huge performance jumps from Intel in the foreseeable future. They have no competition now to put the squeeze on them. Intel has its old monopoly of the x86 performance market back and they have no reason to jump ahead of themselves. It makes a lot more sense to "trickle" these small improvements to the consumers and keep charging them a premium price each time.

Did you miss the Sandy Bridge to Ivy Bridge to Haswell jump in performance? Combined it's been a 50% increase in IPC.

Stop trolling.
 
Sometimes I see Intel as pure blood sucker. They have the kind of technologies they can bring to the market but choose not to. Instead, they make tiny steps, tiny changes - all to make sure they get most of your money with the minimum of effort. That's technology monopoly at its worst, that's why I want to see them burn someday soon, probably by ARM.

For example, DDR4 has been ready for mass production for a while now, by most memory manufacturers. They just seat quiet because this is what Intel wants, giving us a promise of 2015 sometime - that's just so bad.

Also, this is the first time ever Intel decides to release Extreme edition of processors a whole generation behind the current one. Haswell products will be on the market starting from June, and in September we get the all new E-Ivy Bridge - that's nice. How should PC enthusiasts feel about cashing out 1K for a top CPU that's already one generation behind?
 
Sometimes I see Intel as pure blood sucker. They have the kind of technologies they can bring to the market but choose not to. Instead, they make tiny steps, tiny changes - all to make sure they get most of your money with the minimum of effort. That's technology monopoly at its worst, that's why I want to see them burn someday soon, probably by ARM.

For example, DDR4 has been ready for mass production for a while now, by most memory manufacturers. They just seat quiet because this is what Intel wants, giving us a promise of 2015 sometime - that's just so bad.

Also, this is the first time ever Intel decides to release Extreme edition of processors a whole generation behind the current one. Haswell products will be on the market starting from June, and in September we get the all new E-Ivy Bridge - that's nice. How should PC enthusiasts feel about cashing out 1K for a top CPU that's already one generation behind?

I encourage you to read this article: http://arstechnica.com/business/2013/04/the-rise-and-fall-of-amd-how-an-underdog-stuck-it-to-intel/

Intel isn't sticking it to anyone. The reason they are so successful is due to their Tick-Tock strategy. AMD caught Intel by surprise with the Athlon and x86-64 instruction set and since then, Intel has refused to let it happen again.

So while you may look at Intel's year-over-year 10%-15% improvement and scoff, AMD has been unable to offer anything close. Intel has offered that performance improvement even without a competitive AMD.

As for ARM, ARM will never be competitive in the x86 space. ARM is an architecture and instruction set. Where ARM is important is in challenging Intel in the mobile and power usage space. ARM will challenge Intel to lower TDP, as we see with Haswell. The ARM instruction set and architecture is not really very competitive however, as Intel's Mobile Atom absolutely destroys anything ARM has to offer with only a single core. Intel will dominate the mobile space within a generation or two.

Disclaimer: I own both Intel and ARM stock.
 
I encourage you to read...
Ok, thank you for the info, but when you quote somebody, you usually answer the questions/points made there, which you did not, except for the ARM part.

And I wouldn't write off ARM as a competitor to Intel, they are developing at a much faster pace than Intel now, which means it is only a matter of time before they catch up. This Q3 will see first product using 8-core 2.3 mobile ARM CPU-s, with more to come. And 64-bit ARM should enter the market very soon, to make it another trouble for Intel. This is also why we know AMD has turned its eye to ARM.
 
Sometimes I see Intel as pure blood sucker. They have the kind of technologies they can bring to the market but choose not to. Instead, they make tiny steps, tiny changes - all to make sure they get most of your money with the minimum of effort.
Intel is built/repurposed five fabs to produce 22nm wafers. The price tag for each runs to around $5-6 billion dollars per (D1C and D1D development fabs, Fab 28, Fab 12, and Fab 32) . Given that Intel takes almost all of the capacity at those fabs, where would you expect Intel get the money to upgrade and build for every process node (and by extension, microarchitecture) ?
And that is only for 22nm. Intel has plunged a further $5+ billion into Fab 42, $4bn into Fab 24, and $4bn into D1X (with another $2bn earmarked to add a 450mm wafer extension) for 14nm.
Return on investment.
That's technology monopoly at its worst, that's why I want to see them burn someday soon, probably by ARM.
Unlikely.
1. Software ecosystem (esp in enterprise) has a very long lead-in, and a high degree of built in inertia. x86-64 isn't going away anytime soon, and if it were;
2. Intel has the ability to buy an ARM licence just like everyone else...and would be welcomed with open arms if it did.
For example, DDR4 has been ready for mass production for a while now, by most memory manufacturers.
DDR4 has been shown. That isn't the same as ready for volume production.
DDR4 specification calls for 1.2v. Bandwidth range DDR4-2133 to DDR4-4266
DDR3 already has 1.25v modules in circulation, and Haswell (in conjunction with an ASRock Z87 motherboard) has already demonstrated DDR3-3322 with four DIMM's populated:
ASRock-Z87M-OC-Formula.jpg


DDR4's specification adds little other than the promise of 1.05v operation at DDR-4266 bandwidth "sometime in the future"
Also, this is the first time ever Intel decides to release Extreme edition of processors a whole generation behind the current one. Haswell products will be on the market starting from June, and in September we get the all new E-Ivy Bridge - that's nice. How should PC enthusiasts feel about cashing out 1K for a top CPU that's already one generation behind?
The mainstream and HEDT platforms aren't really in competition. Depends on what kind of workload you use the system for. There would also be a school of thought that anyone interested in Ivy Bridge-E might actually be aware of Haswell already in the marketplace when it is released. Likewise, Haswell's Z87/Z85 is offering basically offering the same features as Z77/Z68, albeit in a different socket. Haswell would require a new motherboard, while Ivy Bridge-E will be compatible with existing X79/C600/C602/C606 boards- which may, or may not interest SNB-E owners.
 
@dividebyzero
Thanks for a constructive reply ;)

On the first issue - it would require looking at overall profit versus expenses figures, year by year, not just expenses in one area, to be able to judge fairly. On a better day I would even dig into Intel's financials for an argument, but not today :)

About DDR4. Here's from a credible source: http://news.micron.com/releasedetail.cfm?ReleaseID=727730
From there you can read that Micron had DDR4 positioned as product of the year in 2012, with mass production slated for 2013.

And then this is what really happened: http://news.softpedia.com/news/DDR4-Will-Not-Touch-the-PC-Market-in-2013-296622.shtml

Bummer.

Software ecosystem (esp in enterprise) has a very long lead-in, and a high degree of built in inertia...
This inertia has been decimating at a progressive rate. The IT market is becoming more flexible and open to alternative solutions. Evidence to this we see everywhere, but with most progress in the mobile market. Companies find it justifiable today to offer full alternatives to entire ecosystems, like Mozilla with Mobile FireFox, Blackberry with their own mobile OS, etc, because the quality of the underlining fundamentals (like Linux kernel) is really good now.

What it teaches us most of lately is to be quick at changing over to what makes most sense, not to be fixed on any particular tool or platform. I get the sense of it everywhere. Myself, I did C++ for 15 years, then I was a hardcore .NET developer for 7 years, and to my own disbelief, after all that, today I'm moving into Python, because even financial organizations, as my main clients, see it as one that makes most sense to use today. What do you know... :)

If the big companies aren't willing to keep pace, smaller businesses will pick it up, as they usually do. This is why we see today things like Rspberry Pi, MineCraft (which I personally hate :)), and God knows what else is being transpired on kickstarter and alike :)

I hope I haven't wandered off completely :)

Cheer'o!
 
@ mevans336. "Did you miss the Sandy Bridge to Ivy Bridge to Haswell jump in performance? Combined it's been a 50% increase in IPC." I see what you did there. Any chance you can source that claim?
 
VitalyT
On a more metaphysical note you could argue that the more the consumer has, the more the consumer wants, and when that unrealistic expectation isn't met it translates into disappointment.

There seems to be a perception that every µarch and platform should being about a profound leap in feature set and performance, but with the exception of the first Core 2 Duo/Extreme (say E6700 and X6800), Bloomfield and Sandy Bridge it has always been a case of incremental advance in the modern era. A quantum leap in performance -say Sandy Bridge for example, cannot be sustained with every platform. The pressure to perpetuate the leaps in performance would cripple any company from an R&D perspective.

With regards the DDR4 timeline:
About DDR4. Here's from a credible source: http://news.micron.com/releasedetail.cfm?ReleaseID=727730
From there you can read that Micron had DDR4 positioned as product of the year in 2012, with mass production slated for 2013.
Micron actually jumped the gun with DDR4. JEDEC only ratified the DDR4 specification six months ago. So while I don't doubt that DDR4 memory modules wouldn't be too problematic in producing, there is still the issue of the DDR4 memory controller which is a logic block within the CPU to design and validate.

The arguments surrounding Intel's cadence to me seem at odds with each other. One group lament the frequent socket changes since LGA 775 (even though very few boards actually supported every LGA 775 CPU, and the chipsets offered little actual feature set advancement from chipset-to-chipset) even though VRD requirement and higher pin counts as features expanded or moved to the CPU package necessitated at the very least a motherboard change.

The other group wish for a faster rate of change- which to me comes under the heading "Be careful what you wish for..."
1. Faster implementation of new features means a compressed timescale for correct debugging
2. Faster implementation of new features means that the company with the larger R&D budget is hugely favoured. Even more so if the company has checks and balances to ensure a strict time-to-market cadence....and,
3. What would be the purpose of adding all these extra features ? Quad channel DDR4, all-SATA 6GB, Thunderbolt, 5+GHz at the press of an button etc. Sounds good for the 1%, but the vast majority of people don't use their present systems to its fullest extent - either in hardware fit-out or computationally heavy software.

The only all-sector advancement I could see would be the lowering of power requirement- which takes us back to process shrinks. Even that has its own issues, as even a lower power CPU on a smaller node means a much smaller surface area available to remove waste heat.

The mobile sector isn't that unlike the early years of CPU and GPU design. The fact that anyone can grab an ARM license -either off the shelf or an architectural license, makes it fundamentally different from the x86-64 market- which is where Haswell resides. In the low power segments maybe Intel falls, maybe it flourishes- but in any event, we're a long way from RISC being a catch-all solution in computing.
@ mevans336. "Did you miss the Sandy Bridge to Ivy Bridge to Haswell jump in performance? Combined it's been a 50% increase in IPC." I see what you did there. Any chance you can source that claim?
A direct IPC measurement could be partially taken from a compute heavy benchmark like SHA2-256
sandra-cryptography.png

...which shows a 33.7% speedup, but that doesn't take into account Haswell's AVX2 instruction set. From Dave Kanter's excellent Haswell article (quote from page 6):
Overall, we estimate that a Haswell core will offer around 10% greater performance for existing software, compared to Sandy Bridge. For workloads using the new extensions, the gains could be significantly higher. In theory, AVX2 and FMA can boost performance by 2×, but the impact on most vectorizable workloads will be much lower. Research from AMD has shown that lock elision gains 30% for the right workloads, although the benefits depend strongly on the actual concurrency.
 
'nough said :)

Striving for better is in human nature, it got nothing to do with the real need :) we are caught in the age of a technological loop where computers are used to create better computers, and with technology getting as good, the human brain, as the weakest link in that chain knows only one thing - "I want more, I want better!" :)

p.s. I was just talking to myself, by the way... :)
 
@ mevans336. "Did you miss the Sandy Bridge to Ivy Bridge to Haswell jump in performance? Combined it's been a 50% increase in IPC." I see what you did there. Any chance you can source that claim?



Look at the IPC improvement from Nelahem allt he way up to ivy then just add the numbers.

We were told Haswell should also be a 10% boost in IPC so you can add that aswell.
 
Probably it's a noob question, but, can anyone estimates how much longer does LGA 2011 will last? Because intel tends to shift between socket types quite quickly and consistently in mainstream, like 1156 to 1155 then will be replaced by 1150 in this june..
..so, will LGA 2011 had same fate as LGA 1156/1155? replaced sooner than we expected?
 
Probably it's a noob question, but, can anyone estimates how much longer does LGA 2011 will last? Because intel tends to shift between socket types quite quickly and consistently in mainstream, like 1156 to 1155 then will be replaced by 1150 in this june..
..so, will LGA 2011 had same fate as LGA 1156/1155? replaced sooner than we expected?

I would guess it will be late 2014 at the earliest before Intel replaces Ivy Bridge-E. The 2011 platform is really a scaled back server platform and they are not on the same release cadence as the mainstream parts in 1155/1150, Intel is not going to release IVB-E and scrap it a couple months later they'll get at least a year out of it before it is replaced.. Haswell-E is rumored to use DDR4 which will make it incompatible with SNB-E/IVB-E, if that's true anyone buying 2011 will need a platform swap to go above IVB-E.
 
Back